Anda di halaman 1dari 483

Gender and person agreement in Cicipu discourse

A dissertation submitted to the School of Oriental and African Studies, University of London, in partial fulfilment of the requirements for the degree of Doctor of Philosophy

Stuart John McGill September 2009

I, Stuart John McGill, confirm that the work presented in this thesis is my own. in the thesis.

here

information has been deri!ed from other sources, I confirm that this has been indicated

Abstract
"he #icipu lan$ua$e %&ain'i, (enue)#on$o* of northwest +i$eria has the kind of robust noun class system characteristic of (enue)#on$o lan$ua$es ,
G-+.-/

a$reement is

found on a $reat many a$reement tar$ets inside and outside the noun phrase. 0or a number of these tar$ets, $ender a$reement is in competition with a separate paradi$m, that of
1-/S2+

a$reement. "he dissertation focuses on the distribution of this alternation

with respect to sub'ect prefi3es, ob'ect enclitics, and pronouns, based on a corpus of 42,000 clauses of spoken lan$ua$e. "he alternation pro!es to be comple3 to describe, in!ol!in$ a constellation of le3ical, phonolo$ical, morphosyntactic, semantic and discourse)pra$matic factors. In particular, both animacy and topicality are
#2+.I"I2+S

%#orbett 2005* on a$reement.

hile inanimate or animal participants normally tri$$er $ender a$reement, if they are topics then they may tri$$er person a$reement. 6ikewise while human nouns typically tri$$er person a$reement, this is not always the case, and $ender a$reement is more likely if the referent is of incidental importance to the discourse. 0urthermore it is ar$ued that this alternation is sensiti!e to discourse topic %e.$. .ooley 2007* rather than sentence topic %e.$. 6ambrecht 4998*. (oth $ender and person sub'ect prefi3es are ambi$uous a$reement markers accordin$ to the typolo$y of (resnan and Mchombo %4997* and Siewierska %4999*, since both can take part in $rammatical or anaphoric a$reement. "hus the #icipu data supports #uly:s %2000* contention that topicality is an independent dimension for the classification of a$reement markers, rather than deri!ati!e of the $rammatical !s. anaphoric a$reement distinction, and leads us to re)e!aluate the common assumption that dependent person markers %Siewierska 2008* cannot !ary with respect to their discourse function. Since #icipu is otherwise undescribed, a ma'or part of the dissertation consists of a phonolo$ical and $rammatical sketch.

Acknowledgements
<u$e thanks are due to 1eter =ustin, my super!isor. 0rom the be$innin$ he has been a source of $ood ad!ice and encoura$ement, with an ama>in$ knowled$e both of theoretical lin$uistics and of data from throu$hout the world:s lan$ua$es. <e has been patient with my lack of pro$ress or understandin$, without ne$lectin$ to administer a kick up the backside when needed. 1erhaps abo!e all I am $rateful that he has been so enthusiastic about the #icipu lan$ua$e , it is impossible for a 1h. student to o!erstate the amount of encoura$ement that this attitude brin$s, when often there seems to be no) one else interested? "hanks too to the other members of my committee, 0riederike 6@pke and 6ut> Marten. It has been $reat to ha!e specialists on both est =frican and (antu lan$ua$es who were willin$ to read and comment on !arious papers or chapters. I was also able to make many impro!ements because of informal discussions with them. I am !ery $rateful to the <ans /ausin$ -ndan$ered 6an$ua$es =cademic 1ro$ramme for their financial support durin$ the writin$ of this thesis, in the form of a 1h. (ursary. =dditionally, two field trips to +i$eria were made possible by $rants from the <ans /ausin$ -ndan$ered 6an$ua$es .ocumentation 1ro$ramme %2005, 0"G0402* and from the Ani!ersity of 6ondon #entral /esearch 0und %2009*. I would also like to thank my #icipu consultants, in particular Markus Mallam Babani who threw himself into this work with enthusiasm. here there is insi$ht into the lan$ua$e in this thesis it is often as much due to him as to me. I am also $rateful to Mohammed Musa, especially for arran$in$ inter!iews with =kula elders and for his <erculean transcription efforts. My #icipu lan$ua$e teachers Musa .an'uma Mai An$uwa and his brother Ibrahim $enerously sacrificed their farm work to help me. 2thers who assisted with lan$ua$e work include =mos (ako "imothy, =yuba Sani, Ishiaku Ibrahim, Mohammed Mallam, Sani Cantanyi, Ishaya =udu, and Makawo Madu$u. Many others from Galadima and the surroundin$ !illa$es not only helped me to understand their lan$ua$e, but also treated me with kindness and offered friendship beyond the reDuirements of hospitality. Special thanks are due to Can'uma Galadima, the Mai An$uwa of (a>ama, and (aba Babani Galadima , for allowin$ me to li!e with them in their !illa$e, and for their ad!ocacy of the work I was doin$. hile carryin$ out fieldwork in +i$eria I had the help of many people from 8

!arious local institutions. I recei!ed a $reat deal of assistance from #=1/2 members, especially Joshua &adon, who twice helped me obtain a !isa. I am $rateful for the opportunity to work with Israel ade, whose enthusiasm for #icipu is inspirational. I hope we $et the chance to $o back to$ether. /euben and Eictoria =cheson:s friendship and hospitality were more than I could ha!e dreamed of. Many others helped me in !arious ways, includin$ (ro Busuf and Mama ->ra, (ro James, (ro =demola, Sister Joke, Sister Stella, and not least (ro =li and Sister (ose 2kpo. 1rofessor Sati Amaru 0watshak and .r John +en$el from the Ani!ersity of Jos kindly pro!ided letters of introduction and ha!e offered to archi!e the lan$ua$e materials resultin$ from this pro'ect. 1rofessor .oris (o>imo of =hmadu (ello Ani!ersity went out of her way to photocopy Mathews: Historical and Anthropological Report on the =cipu for me. I would like to thank the &ebbi State #ommissioner for 6ocal Go!ernment and #hieftaincy =ffairs for writin$ me a further letter of introduction. I am especially $rateful to the possible to carry out this research. I ha!e many reasons to be $rateful for the hospitality and ad!ice %both lin$uistic and practical* of se!eral lon$)term lin$uists and missionaries in +i$eria. 6i> and .a!id #ro>ier ha!e been of enormous personal encoura$ement and administrati!e help to me, ri$ht from the conception of this pro'ect in 200; until the present day. I:d like to thank .a!id especially for sharin$ his data and thou$hts on #ishin$ini and its relationship with #icipu. I hope that this thesis will be as useful to others as his was to me. .a!id <eath has also helped me in numerous ways, but in particular by $enerously allowin$ me the use of his house and office in "un$an Ma$a'iya. /o$er (lench has shared his work, time, and e3pertise on &ain'i lan$ua$es, as well as bein$ Duick to comment on and impro!e any documents I ha!e sent him. Just findin$ someone else in the world with an acti!e interest in &ain'i lan$ua$es would ha!e been nice in itself, but /o$er:s enthusiasm and his concern for these lan$ua$es and their documentation has been a $reat moti!ation for me as well. I am $rateful to Ste!e and Sonia .ettweiler for their hospitality, and for sharin$ their thou$hts and lan$ua$e data. Mike /ueck lent me his brand)new G1S which pro!ed in!aluable both for mappin$ the =cipu area and for allowin$ me a certain amount of independent tra!el throu$h the head)hi$h $uineacorn? 0inally, I benefited from the hospitality of the Guessler and 0uller families in "un$an omo of &orisino, whose support made it

Ma$a'iya on se!eral occasions. I ha!e been fortunate to study at one of the best places for learnin$ <ausa in the world. (arry (ur$ess and =bba "i'ani ha!e not only been e3emplary lan$ua$e teachers, they also listened with patience to my parochial and often wide)of)the)mark Duestions about <ausa $rammar and its influence on #icipu. If it were not for I!an 6owe:s inspirational teachin$ at SI6 A&, I probably would ha!e ne!er taken up lin$uistics. =lon$ the way I:!e been pri!ile$ed to learn from many othersG with respect to this thesis, my understandin$ of topicality has benefited $reatly from the teachin$ and writin$ of (ob .ooley. Many other friends, collea$ues, and lin$uists ha!e had direct or indirect input into this dissertation throu$h the con!ersations and discussions I had with them, or the materials they ha!e been kind enou$h to pro!ide me with. "hese include Mary #hambers, Sophie Salffner, 1ete (udd, #haithra 1uttaswamy, Ser$e Sa$na, 6ameen Soua$, 2li!er (ond, Monik #harette, -li "iman, James Mac.onell, (ecky 1aterson, Gre!ille #orbett, (ir$it <ellwi$, 0rank Seifart, and 6udwi$ Gerhardt, as well as e!eryone else from /;04. (ernard <oward, "om #astle, and .a!id +athan from -6=/ deser!e particular thanks for their help with electrical and audio eDuipment. =nd a bi$ thank you to =lison &elly for sanity)restorin$ chats and all her administrati!e help, and also to her predecessor Hara 1ybus. I:d like to thank my friends in .idcot, especially Graham, Ga!in, -lli, -ileen, Jill and =ndrew for some $reat times o!er the last few years while not writin$ this thesis. <a!in$ friends like this to come back to made the daily commute bearable. =nd of course Mum, .ad and 0iona for your forbearance, lo!e, and timely mini)breaks? 0inally merci beaucoup to my wife &yoko for her encoura$ement, and for keepin$ me in my place by lau$hin$ at my attempts at prose? "his thesis is dedicated to her. I can think of no better way to round up than to echo Ishaya =udu:s prayer followin$ his own contribution to the documentation of #icipu. Wan Kungwa tugwede vu ana yaawana tu a wumpa ulenji. Ana tiyaana kampa komisooni, naha iv n kati i kaavu iri yina tidamana i yyuu yi ka'al arka a ahula ha!eesu. "ipata vu po iri yina tiyaana yina yilu a daidai, naha iv n kati i kaavu i sungusanu. Ki ke lee# Lord God we thank you that you've brought us to this day. Now we've prepared this story, may you yoursel bless the things we've told in the name o !esus. "e ask you whatever we've done that's lacking, may you bring it to ul ilment. Amen# 5

Conventions
Transcription
"he transcription used for the #icipu e3amples in this thesis is based on the recommendations that were made in the recently)proposed ortho$raphy statement4, with additional tone)markin$. "he symbol y stands for the palatal appro3imant IjJ, $ and j for the affricates It%J and Id&J respecti!ely, and the apostrophe ' for the $lottal stop I'J. = double !owel aa indicates a lon$ !owel Ia(J. "one and nasality are marked on the first * but apply to the full !owel %Ia( J* , contour tones are realised o!er both letter only %aa !owels to$ether. "he followin$ accents and abbre!iations are used for toneK a a) a* a+ , < 6 <i$h 6ow

<6 0allin$ 6< /isin$ .ownstep occurs after this point

.ownstep occurs both within and between words. In $eneral only word)internal downstep is indicated, and this has conse$uences or the interpretation o the tone on ob%ect N&s %see L;.8.7 fn. 2F*. Intonation contours are indicated in some e3amples, especially in chapter 9. "he system used here is based on #hafe %4998K3iii*K , . M N ... a terminal contour which is not sentence)final a sentence)final fallin$ pitch a yes)no Duestion terminal contour intonation unit boundary %could be any of the abo!e three* intonation unit continues beyond what was transcribed

"ranscriptions enclosed in IsDuare bracketsJ use standard I1= symbols. 1honetic transcriptions !ary in detail dependin$ on the distinctions in Duestion. Anintelli$ible speech is indicated by 333, hesitations by %.*. "he start of a para$raph is indicated by O. #icipu phonolo$y in!ol!es a $ood deal of underspecification, and the symbols and . indicate underspecified consonants and !owels in phonemic representations, in addition to their normal use in indicatin$ syllable structure. A indicates an affi3 !owel whose Duality is determined by rules of !owel harmony %L;.F*. / stands for a nasal
4 httpKNNwww.cicipu.or$NpapersN2009)09)4;Portho$raphyPproposalP!02.pdf.

homor$anic with the followin$ consonant. "hree lines are pro!ided for the ma'ority of e3amples , the transcription, a morpheme)by)morpheme $loss, and then a free translation in -n$lish pro!ided by the author. 2ccasionally morphophonemic processes obscure the boundaries between morphemes, in which case an e3tra morphophonemic line has been added on topK %4* * ) n)tiivi * n)0taa1vi )

4S)shootQ/6SR;S.1/2

' shot him

I#ross)referenceK easm004.00FJ

-ach e3ample is cross)referenced to its source, as shown abo!e. See L4.8.; for the con!entions used. -3ample numberin$ restarts at the be$innin$ of each chapter. = note on the $lossin$ of nouns is in order here. =s will be seen later, nouns in #icipu are comprised of a prefi3 plus root e.$. ka)0taari SstoneT, a)0taari SstonesT. "hrou$hout the thesis these prefi3es will be indicated in the $losses !i>.
+#2)stone. +#4)stone,

"his has been done for the reader:s con!enience and is not intended as a

statement concernin$ the psycholo$ical reality of the prefi3)root distinction. hen conte3t is supplied for an e3ample, descriptions of the conte3t are in normal font. 1araphrases of the sentences immediately before the e3ample are $i!en in italics. %2* I#onte3tK .escription of a huntin$ party. ' didn't see (ako in the grass. "hen he stood up suddenly...J * n)0taa
4S)shootQ/6S

' shot In a few -n$lish e3amples


SM=66 #=1S

are used to indicate sentence accent. Bold te3t,

whether in #icipu or -n$lish, is not used for accent or emphasis. Instead it is intended to draw the reader:s attention to the features of the e3ample under discussion.

English paraphrases
"he analysis in chapter 9 in!ol!es the consideration of a number of len$thy sections of te3t. /ather than clutter the chapter with pa$es of interlinearised te3t, lon$ e3amples are paraphrased in -n$lish, with the rele!ant $rammatical features marked up in bold. Interested readers are encoura$ed to !iew and listen to the ori$inal te3ts at httpKNNwww.cicipu.or$Nte3ts.html. "he followin$ key may be useful in the interpretation of these paraphrasesK 9

;S.1/2 =G4)1/2 ;S.2(J ;S)came =G4)came ;s.12SS

Independent personal pronoun Independent $ender pronoun %also =G2)1/2, and so on* 1erson)marked ob'ect clitic 1erson sub'ect prefi3 on !erb Gender sub'ect prefi3 on !erb 1ossessi!e pronoun

In-text examples
hen #icipu words are discussed in the main body of the te3t they are printed in bold followed by the -n$lish $loss in in!erted commas e.$. dukwa S$oT. to the standard ortho$raphy i.e. unmarked for tone or !owel len$th. ords from <ausa

and other lan$ua$es are printed in italics. <ausa words are $enerally spelled accordin$

Abbreviations
"he followin$ abbre!iations are used %based on the 6eip>i$ Glossin$ /ules, (ickel et al. 2008*K 4 2 ;
=G =+"I# =116 =/" #=AS #+0#" #21 0A" G-+ <=( IM1 I// &.2. 62# 6#EH/

4st person 2 person ; person a$reement anti)causati!e applicati!e article causati!e counterfactual copula future $eniti!e habitual imperati!e irrealis kind of locati!e locati!iser
rd nd

6 +# +-G +M6H 1 1=/" 10E 16=# 12SS 1/2 1/2< /-.A1 /-6 /-S /6S S S-1 E-+"

loanword noun class ne$ati!e nominaliser plural particle %unknown meanin$* perfecti!e pluractional possessi!e pronoun prohibiti!e reduplicated relati!iser resultati!e realis sin$ular separati!e !enti!e

#icipu does not ha!e ob'ect a$reement, so all a$reement markers on !erbs are sub'ect a$reement markers. 9

Table of Contents
Abstract............................................................................................................................3 Acknowledgements..........................................................................................................4 Conventions......................................................................................................................7 "ranscription.................................................................................................................7 -n$lish paraphrases......................................................................................................9 In)te3t e3amples............................................................................................................9 =bbre!iations................................................................................................................9 Part 1 Introduction..................................................................................................... C!a"ter 1 Introduction............................................................................................ 3 4.4 Moti!ation for the study and research Duestions..................................................2; 4.4.4 /esearch Duestions........................................................................................2F 4.2 6an$ua$e situation................................................................................................25 4.2.4 "he lan$ua$e name........................................................................................25 4.2.2 .emo$raphic and ethno$raphic notes...........................................................25 4.2.; .ialects..........................................................................................................;2 4.2.8 Multilin$ualism.............................................................................................;8 4.2.F -ndan$erment...............................................................................................;F 4.; /e!iew of literature on est &ain'i, &ambari, and #icipu..................................;5 4.;.4 est &ain'i....................................................................................................;7 4.;.2 &ambari.........................................................................................................;9 4.;.; =nthropolo$ical and lin$uistic literature on #icipu......................................80 4.8 0ieldwork settin$ and data....................................................................................8; 4.8.4 .ata collection methodolo$y........................................................................8F 4.8.4.4 "opic)stimulation sessions.....................................................................89 4.8.2 #orpus imbalance..........................................................................................F0 4.8.; #ross)referencin$ con!entions......................................................................F4 4.8.8 #ontributors...................................................................................................F4 4.F 2!er!iew of the thesis..........................................................................................F; C!a"ter #esearc! conte$t.....................................................................................%4 2.4 +oun classes and $ender.......................................................................................F8 2.4.4 (enue)#on$o noun class systems.................................................................F8 2.4.2 #oncepts and terminolo$y.............................................................................F7 2.4.2.4 Gender and noun class lan$ua$es..........................................................F7 2.4.2.2 =fricanist terminolo$y...........................................................................F9 2.4.2.; #ontroller and tar$et $enders................................................................F9 2.2 =$reement.............................................................................................................58 2.2.4 .omain..........................................................................................................58 2.2.2 Mechanism....................................................................................................57 2.2.; Syntactic and semantic a$reement................................................................57 2.2.8 Eariation in a$reement..................................................................................59 2.2.8.4 1re)reDuisites and conditions.................................................................59 2.2.8.2 Eariation and topicality .........................................................................72 2.2.8.2.4 2ptional a$reement........................................................................72 2.2.8.2.2 #hoice of controller.......................................................................7; 2.2.8.2.; #hoice of feature paradi$m............................................................75 2.2.F "ypolo$y of a$reement markers....................................................................79 40

2.2.F.4 =naphoric a$reement and topicality......................................................9; 2.; "opic......................................................................................................................99 2.;.4 Sentence topic...............................................................................................90 2.;.2 .iscourse topic..............................................................................................92 UVWVUVXYZ[\]^.....................................................................................................92 2.;.2.2 .ooley...................................................................................................9; 2.;.2.; 0ormal si$nals of discourse topic..........................................................95 2.;.2.8 1ara$raphs.............................................................................................97 2.;.2.F #odin$ wei$ht.......................................................................................99 2.;.2.5 =ccess and inte$ration functions of topics..........................................404 2.;.; Summary.....................................................................................................404 2.8 #hapter summary................................................................................................402 Part II P!onological and grammatical sketc!.......................................................1&4 C!a"ter 3 A "!onological sketc! o' Cici"u..........................................................1&% ;.4 Syllable and root structure..................................................................................40F ;.4.4 Anambi!alent syllables...............................................................................40F ;.4.2 6abialisation and palatalisation...................................................................407 ;.4.; 6on$ consonants..........................................................................................407 ;.4.8 6on$ !owels and diphthon$s......................................................................409 ;.4.F 1renasalised stops and affricates.................................................................409 ;.4.5 =ppro3imants..............................................................................................444 ;.4.7 1refi3es and ideophones..............................................................................442 ;.4.9 /oot structure..............................................................................................44; ;.4.9 Summary.....................................................................................................448 ;.2 #onsonants..........................................................................................................44F ;.2.4 1honemic in!entory.....................................................................................44F ;.2.2 =llophones and $eneral phonetic rules........................................................44F ;.2.; .istributional restrictions and e3amples.....................................................445 ;.2.8 6en$th..........................................................................................................449 ;.; Eowels.................................................................................................................428 ;.;.4 1honemic in!entory....................................................................................428 ;.;.2 =llophones and $eneral phonetic rules.......................................................428 ;.;.; +asal !owels...............................................................................................42F ;.;.8 6on$ !owels and diphthon$s......................................................................42F ;.;.F .istribution.................................................................................................427 ;.8 "one.....................................................................................................................429 ;.8.4 "one in!entory.............................................................................................429 ;.8.2 .owndrift, downstep, and upstep................................................................429 ;.8.; Spreadin$.....................................................................................................4;0 ;.8.8 1olar tone....................................................................................................4;2 ;.8.F 6e3ical tone in nouns..................................................................................4;2 ;.8.5 Grammatical tone on !erbs.........................................................................4;; ;.8.7 #omplement tone perturbation on nouns....................................................4;7 ;.F Eowel harmony...................................................................................................4;9 ;.F.4 .istribution of !owels in #E#E noun roots...............................................480 ;.F.2 6oanwords...................................................................................................480 ;.F.; =ffi3es and clitics........................................................................................484 ;.F.8 #ross)lin$uistic comparisons......................................................................48; ;.5 +asalisation.........................................................................................................488 ;.5.4 1honemes affected.......................................................................................488 44

;.5.2 .irection......................................................................................................48F ;.7 Morphophonemic processes................................................................................48F ;.7.4 #oalescence and elision..............................................................................48F ;.7.2 u)anticipation...............................................................................................489 ;.7.; i)anticipation...............................................................................................489 C!a"ter 4 A grammatical sketc! o' Cici"u..........................................................1%1 8.4 Introduction.........................................................................................................4F4 8.2 "ypolo$ical o!er!iew..........................................................................................4F4 8.; #lause)le!el synta3.............................................................................................4F; 8.;.4 Grammatical relations.................................................................................4F; 8.;.4.4 .ouble)ob'ect constructions................................................................4F8 8.;.4.2 /efle3i!es............................................................................................4FF 8.;.2 #onstituent order.........................................................................................4F5 8.;.2.4 #onstituent)frontin$............................................................................4F7 8.;.2.4.4 0ocus............................................................................................4F7 8.;.2.4.2 "opic............................................................................................4F9 8.;.; +on)!erbal clauses......................................................................................454 8.;.;.4 1redicate nominals...............................................................................454 8.;.;.4.4 +ouns...........................................................................................454 8.;.;.4.2 1ersonal pronouns........................................................................45; 8.;.;.2 1redicate locati!es...............................................................................458 8.;.;.; -3istential clauses...............................................................................45F 8.;.;.8 1resentational clauses..........................................................................455 8.;.8 /elati!e clauses...........................................................................................455 8.;.F +e$ation......................................................................................................470 8.;.F.4 #lausal ne$ation..................................................................................470 8.;.F.2 #onstituent ne$ation............................................................................474 8.;.5 _uestions.....................................................................................................472 8.;.5.4 BesNno Duestions..................................................................................472 8.;.5.2 wh)Duestions........................................................................................47; 8.;.7 #lause coordination and subordination.......................................................475 8.;.7.4 #oordination........................................................................................475 8.;.7.2 Subordination......................................................................................479 8.8 +ominals.............................................................................................................490 8.8.4 =d!erbs.......................................................................................................494 8.8.4.4 .emonstrati!e ad!erbsNpredicates.......................................................494 8.8.4.2 "ime ad!erbs.......................................................................................49; 8.8.4.; Ideophones...........................................................................................49; 8.8.2 1roper nouns................................................................................................498 8.8.; 1ronouns......................................................................................................498 8.8.;.4 Independent personal pronouns...........................................................49F 8.8.;.2 Independent noun class pronouns........................................................49F 8.8.;.; 2b'ect clitics........................................................................................49F 8.8.;.8 .emonstrati!e personal pronouns.......................................................49F 8.8.8 #ount and mass nouns.................................................................................495 8.8.F "he noun phrase..........................................................................................497 8.8.F.4 "he associati!e %S$eniti!eNpossessi!eT* construction..........................497 8.8.F.4.4 =ssociati!e %Spossessi!eT* pronouns............................................490 8.8.F.2 .emonstrati!e modifiers.....................................................................492 8.8.F.; =rticle..................................................................................................498 42

8.8.F.8 +ominal coordination..........................................................................499 8.8.F.F 1ronominal function of modifiers.......................................................200 8.8.F.5 +1)internal synta3...............................................................................204 8.8.5 +ominalisin$ suffi3es.................................................................................20; 8.8.5.4 Stati!e )ni.............................................................................................208 8.8.5.2 6ocati!e )tu..........................................................................................20F 8.F 1repositions.........................................................................................................20F 8.5 Eerbs and the !erb phrase...................................................................................209 8.5.4 Eerbal word template..................................................................................209 8.5.2 Mood...........................................................................................................240 8.5.2.4 /ealis...................................................................................................240 8.5.2.2 Irrealis..................................................................................................244 8.5.2.; Imperati!e............................................................................................242 8.5.2.8 #ounterfactual.....................................................................................24; 8.5.; =spect and tense..........................................................................................248 8.5.;.4 1erfecti!e.............................................................................................248 8.5.;.2 0uture...................................................................................................248 8.5.;.; <abitual...............................................................................................24F 8.5.;.8 .ependent imperfecti!e.......................................................................245 8.5.;.F 1ro$ressi!e..........................................................................................247 8.5.;.5 #ontinuous...........................................................................................249 8.5.;.7 1erfect..................................................................................................249 8.5.8 Ealence........................................................................................................220 8.5.8.4 #ausati!e.............................................................................................224 8.5.8.2 =pplicati!e...........................................................................................222 8.5.8.; =nticausati!e.......................................................................................22; 8.5.F 2ther !erbal suffi3es...................................................................................228 8.5.F.4 Eenti!e.................................................................................................228 8.5.F.2 /esultati!e...........................................................................................22F 8.5.F.; Separati!e............................................................................................225 8.5.F.8 1luractional..........................................................................................227 8.5.5 (orrowed !erbs...........................................................................................229 8.5.7 =u3iliary and aspectual au3iliary !erbs......................................................229 8.5.9 2rder of affi3es...........................................................................................2;4 8.5.9 #o)occurrence.............................................................................................2;4 8.7 =d'ecti!es............................................................................................................2;2 8.7.4 /eduplication..............................................................................................2;; 8.9 +umerals.............................................................................................................2;8 8.9.4 #ardinal numbers........................................................................................2;8 8.9.2 =ttributi!e numerals....................................................................................2;F 8.9.; 2rdinal numbers..........................................................................................2;F 8.9.8 S=d!erbialT numerals...................................................................................2;5 8.9.F .istributi!e numerals..................................................................................2;5 8.9 _uantifiers...........................................................................................................2;5 Part III (!e noun class system o' Cici"u................................................................ 3) C!a"ter % (!e Cici"u noun class system.............................................................. 3* F.4 2!er!iew.............................................................................................................2;9 F.2 Genders...............................................................................................................284 F.2.4 Gender 4N2...................................................................................................287 F.2.2 Gender 8NF...................................................................................................287 4;

F.2.; Gender 5NF...................................................................................................289 F.2.8 Gender 7N9...................................................................................................289 F.2.F Gender 9N2...................................................................................................289 F.2.5 Gender 9N;...................................................................................................2F0 F.2.7 Gender 9N2...................................................................................................2F0 F.2.9 Sin$le class $enders....................................................................................2F0 F.2.9.4 Gender 4..............................................................................................2F0 F.2.9.2 Gender ;a.............................................................................................2F0 F.2.9.; Gender ;b............................................................................................2F4 F.2.9.8 Gender 8..............................................................................................2F4 F.2.9.F Gender F..............................................................................................2F4 F.2.9.5 Gender 5..............................................................................................2F4 F.2.9.7 Gender 7..............................................................................................2F2 F.2.9.9 Gender 9..............................................................................................2F2 F.2.9.9 Gender 9..............................................................................................2F2 F.2.9 InDuorate $enders........................................................................................2F2 F.; .eri!ational use of noun class markers..............................................................2F; F.;.4 .iminuti!e %$ender 8NF*..............................................................................2F; F.;.2 =u$mentati!e %$ender 4N2*..........................................................................2F8 F.;.; =bstract Duality %$ender 5*..........................................................................2F8 F.;.8 6ocati!e and de)ad'ecti!al abstract nouns %$ender 7*.................................2FF F.;.F -thnic $roups..............................................................................................2F5 F.;.5 1re)prefi3es.................................................................................................2F5 F.8 .e!erbal nominalisation.....................................................................................2F7 F.8.4 Infiniti!es and locati!es %$ender 7*.............................................................2F7 F.8.2 /epeated action %$ender 9*..........................................................................250 F.8.; #ontinuous action %$ender 8*......................................................................252 F.8.8 =$ent %$ender 9N2*.......................................................................................25; F.8.F Indi!iduati!e %$ender 4N2*...........................................................................258 F.8.5 Miscellaneous..............................................................................................258 F.F 1refi3es and allomorphs......................................................................................25F F.F.4 Eowel)initial roots.......................................................................................25F F.F.2 Eowel harmony............................................................................................255 F.F.; "one.............................................................................................................257 F.F.8 #lasses ;a and ;b........................................................................................259 F.F.F #lass F.........................................................................................................259 F.F.5 #lass 5.........................................................................................................270 F.F.7 #lass 9.........................................................................................................270 F.F.9 -penthesis and reduplication.......................................................................27; F.5 6oanwords...........................................................................................................278 C!a"ter + Agreement.............................................................................................. 7) 5.4 =$reement prefi3 allomorphs.............................................................................279 5.4.4 Eowel harmony............................................................................................279 5.4.2 a$F allomorphs............................................................................................279 5.4.; a$9 allomorphs............................................................................................290 5.2 =$reement tar$ets...............................................................................................29; 5.2.4 1refi3 tone...................................................................................................298 5.2.2 +umerals.....................................................................................................29F 5.2.; =d'ecti!es....................................................................................................297 `VUVaY^b^^c[Yde^fgh.................................................................................................297 48

5.2.F wh)words.....................................................................................................297 5.2.5 +e$ati!e copula...........................................................................................299 5.2.7 .emonstrati!e ad!erbs................................................................................299 5.2.9 =ssociati!e construction..............................................................................299 5.2.9 1ossessi!e pronouns....................................................................................299 5.2.40 =rticleNrelati!iser.......................................................................................290 5.2.44 .emonstrati!e modifier.............................................................................294 l dn^eopqrh...........................................................................................294 5.2.42 )ijkm Y 5.2.4; Sub'ect.......................................................................................................294 5.2.48 #opula.......................................................................................................292 5.2.4F +oun class pronouns.................................................................................292 5.; =ntecedentless a$reement morpholo$y and semantic a$reement.......................29; 5.8 +eutral a$reement...............................................................................................;0F 5.8.4 a$9 neutral a$reement.................................................................................;0F 5.8.2 +eutral $ender assi$nment..........................................................................;05 5.F #omple3 +1s and !ariation in a$reement..........................................................;09 5.5 Gender resolution................................................................................................;40 Part I, -ender and "erson agreement in Cici"u..................................................31 C!a"ter 7 -ender and "erson agreement on "erson markers...........................313 7.4 Independent personal pronouns..........................................................................;48 7.4.4 Morpholo$y.................................................................................................;48 7.4.2 Synta3..........................................................................................................;4F 7.2 Independent noun class pronouns.......................................................................;45 7.2.4 Morpholo$y.................................................................................................;45 7.2.2 Synta3..........................................................................................................;47 7.; 2b'ect clitics.......................................................................................................;49 7.;.4 1honolo$y...................................................................................................;49 7.;.2 Morpholo$y.................................................................................................;49 7.;.; Synta3..........................................................................................................;29 7.8 1erson sub'ect prefi3es.......................................................................................;29 7.8.4 1honolo$y...................................................................................................;29 7.8.2 Morpholo$y.................................................................................................;;4 7.F Gender sub'ect prefi3es.......................................................................................;;4 7.F.4 1honolo$y...................................................................................................;;4 7.F.2 Morpholo$y.................................................................................................;;; 7.5 Syntactic status of the sub'ect a$reement prefi3es.............................................;;; 7.5.4 SMulti)representationT.................................................................................;;; 7.5.4.4 1erson sub'ect prefi3es........................................................................;;8 7.5.4.2 Gender sub'ect prefi3es.......................................................................;84 7.5.4.; Summary of prereDuisites for sub'ect a$reement................................;8; 7.5.4.8 =$reement conditions..........................................................................;89 7.5.2 +umber of ar$uments encoded on the !erb................................................;89 7.5.; /eferentiality...............................................................................................;89 7.5.8 .escripti!e content......................................................................................;F2 7.5.F (alance of information................................................................................;F2 7.5.5 2ther tests....................................................................................................;F2 7.5.5.4 _uestioned sub'ects.............................................................................;F; 7.5.5.2 /elati!ised constituents.......................................................................;FF 7.5.7 =re sub'ect prefi3es obli$atoryM..................................................................;F5 7.5.9 #omparison with #entral &ambari.............................................................;F9 4F

7.7 1erson marker summary.....................................................................................;50 7.9 0eatures in!ol!ed................................................................................................;54 7.9 #hapter summary................................................................................................;58 C!a"ter ) -ender and "erson agreement in Cici"u discourse...........................3+% 9.4 Introduction.........................................................................................................;5F 9.2 .ata analysis.......................................................................................................;59 9.; 1articipant reference and marked topics in #icipu.............................................;74 9.;.4 1articipant reference....................................................................................;74 9.;.2 Marked and contrasti!e topics....................................................................;78 9.8 =nimacy and discourse topicality as a$reement conditions...............................;75 9.8.4 +oun class 9................................................................................................;77 9.8.2 = Sminimal pairT showin$ contrast in topicality..........................................;79 9.8.; Inanimates...................................................................................................;94 9.8.;.4 Inanimates, non)topical.......................................................................;94 9.8.;.2 Inanimates, topical...............................................................................;92 9.8.8 =nimals.......................................................................................................;9F 9.8.8.4 =nimals, non)topical............................................................................;95 9.8.8.2 =nimals, topical...................................................................................;95 9.8.F <umans.......................................................................................................;99 9.8.F.4 <umans, non)topical............................................................................;99 9.8.F.2 <umans, topical...................................................................................;90 9.8.F.; S+aturalT $ender...................................................................................;94 9.8.5 Sentence and discourse topic......................................................................;9; 9.8.7 #onflict with speakersT intuition.................................................................;98 9.8.9 Ane3plained data.........................................................................................;9F 9.F 1ro$ressions........................................................................................................;97 9.F.4 ithin)para$raph pro$ressions...................................................................;99 9.F.4.4 -3amples of pro$ressions....................................................................800 9.F.4.2 "he boundary between $ender and person a$reement.........................805 9.F.4.; -3ceptions to the $ender s person directional constraint....................807 9.F.2 "e3t)le!el pro$ressions................................................................................848 9.5 =lternati!e e3planationsM....................................................................................84F 9.5.4 =nimacy......................................................................................................84F 9.5.2 "ransiti!ity..................................................................................................847 9.7 /e!ised flowchart...............................................................................................847 9.9 Gender, person and codin$ wei$ht......................................................................849 9.9 "he $enderNperson alternation on other a$reement tar$ets.................................828 9.9.4 "he article....................................................................................................82F 9.9.2 .emonstrati!es............................................................................................829 9.9.; )t^uYdvp[wph................................................................................................8;; 9.9.8 #opula.........................................................................................................8;; 9.40 #hapter summary..............................................................................................8;5 C!a"ter * Conclusion.............................................................................................43) 9.4 Main findin$s......................................................................................................8;9 9.2 #ontribution to lin$uistics...................................................................................884 9.; =reas for further study........................................................................................885 A""endi$ A (wo Cici"u te$ts...................................................................................44* "he bloodthirsty sword %samy002*...........................................................................889 "he loss of "ikula %sa$b004*.....................................................................................8F; A""endi$ B Contributors........................................................................................4%) 45

A""endi$ C Person and gender agreement on "erson markers re'erring to !umans. animals. and inanimates..............................................................................4+& #e'erences.....................................................................................................................4+3

47

Index of tables
"able 4K #icipu dialectsNsettlements and their <ausa names...........................................;; "able 2K Structure of the interlinearised corpus used for this study................................88 "able ;K .istribution of noun roots accordin$ to syllable structure..............................44; "able 8K .istribution of !erb roots accordin$ to syllable structure...............................448 "able FK #onsonant phonemes.......................................................................................44F "able 5K /oot)initial consonant phonemes in nouns......................................................447 "able 7K /oot)initial consonant phonemes in !erbs.......................................................449 "able 9K #icipu !owel)in!entory...................................................................................428 "able 9K 2ral and nasal !owel contrasts........................................................................42F "able 40K 2ral and nasal !owel contrasts......................................................................425 "able 44K .iphthon$s.....................................................................................................427 "able 42K .istribution of noun roots accordin$ to tone pattern.....................................4;; "able 4;K Eowel co)occurrence restrictions in #E#E noun roots where E4 and E2 are both short oral !owels %E4 down the left, E2 alon$ the top*........................................480 "able 48K <armonisin$ affi3es.......................................................................................482 "able 4FK u) anticipation before )w= suffi3es...............................................................489 "able 45K 0ocused independent personal pronouns.......................................................458 "able 47K #icipu wh)Duestion words.............................................................................47; "able 49K <ausa subordinators and their #icipu eDui!alents.........................................479 "able 49K -3ample of intra)para$raph tail)head linka$es..............................................490 "able 20K Independent personal pronouns.....................................................................49F "able 24K Independent noun class pronouns in #icipu..................................................49F "able 22K 2b'ect clitics..................................................................................................49F "able 2;K .emonstrati!e pronouns................................................................................495 xnjfqYUayYzqie^{orno[\qY|re^e}^{Y~qr[\q~YreiYYd|qr{e^h....................................495 "able 2FK 1ossessi!e pronoun paradi$m with $ender a$reement prefi3es....................494 "able 25K .emonstrati!e modifiers...............................................................................49; "able 27K "one on the habitual prefi3 plus root, accordin$ to root structure.................24F "able 29K Sources of pluractional markin$....................................................................227 "able 29K -3amples of ad'ecti!es..................................................................................2;2 "able ;0K #ardinal numbers in #icipu...........................................................................2;8 "able ;4K Ani!ersal Duantifiers......................................................................................2;7 "able ;2K #icipu noun classes........................................................................................280 "able ;;K .istribution of #icipu $enders %sin$ulars down, plurals across*...................288 "able ;8K Semantic content of the lar$er #icipu $enders..............................................287 "able ;FK -thnic $roup paradi$m..................................................................................2F5 "able ;5K Miscellaneous de!erbal nominalisations.......................................................25F "able ;7K +oun class prefi3es occurrin$ with !owel)initial roots.................................25F "able ;9K .istribution of class ; prefi3es accordin$ to $ender.....................................259 "able ;9K -3amples of nc9 consonant)len$thenin$ prefi3............................................272 "able 80K /eduplicated plurals with root)initial lon$ consonants.................................27; "able 84K 1roportion of loanwords in $enders...............................................................27F "able 82K 1ossible a$9 allomorphs accordin$ to tar$et type.........................................292 "able 8;K =$reement prefi3es found before consonant)initial and !owel)initial stems 298 "able 88K =$reement tar$ets..........................................................................................29F "able 8FK 1ossessi!e pronoun paradi$m with $ender a$reement prefi3es....................290 "able 85K .emonstrati!e modifiers with $ender a$reement prefi3es............................294 "able 87K Summary of #icipu person markers..............................................................;4; 49

"able 89K Independent personal pronouns.....................................................................;48 "able 89K Independent noun class pronouns in #icipu..................................................;45 "able F0K 2b'ect clitics..................................................................................................;49 "able F4K 0reDuency of reduced forms as opposed to normal ob'ect clitics..................;27 "able F2K 1erson sub'ect prefi3es %before consonantsNbefore !owels*..........................;29 "able F;K Eerb prefi3 tone patterns with a ;ps sub'ect..................................................;29 "able F8K Gender sub'ect prefi3es.................................................................................;;4 "able FFK 1artial ambi$uity of person prefi3es with respect to $ender prefi3es, before consonant)initial stems..................................................................................................;;2 "able F5K "otal ambi$uity of person prefi3es with respect to $ender prefi3es, before !owel)initial stems.........................................................................................................;;2 "able F7K .istribution of sub'ect +1s co)occurrin$ with ;ps sub'ect a$reement..........;;5 "able F9K .istribution of the !i) and ) allomorphs of the a$9 sub'ect prefi3 before #) initial !erb stems, accordin$ to animacy of referent.....................................................;;9 "able F9K Sub'ect a$reement prefi3es tri$$ered by nc9 controllers..............................;;9 "able 50K =$reement possibilities before #)initial !erb stems......................................;F7 "able 54K Summary of properties of #icipu person markers.........................................;50 "able 52K Summary of tests for syntactic status of the sub'ect prefi3es........................;54 "able 5;K Sub'ect a$reement prefi3es for !erbs in the realis mood %sin$le flat paradi$m* .......................................................................................................................................;52 "able 58K Sin$le paradi$m with person and class as a subset of person........................;5; "able 5FK /e!ised sin$le paradi$m with person and number features but not $ender. .;5; "able 55K 1ersonNnumber paradi$m...............................................................................;58 "able 57K +oun class paradi$m......................................................................................;58 "able 59K 1roperties coded for person markers.............................................................;59 "able 59K .istribution of a$reement pro$ressions accordin$ to animacy.....................;99 "able 70K .istribution of $ender s person a$reement pro$ressions accordin$ to animacy and discourse topicality.................................................................................................;99 "able 74K +o. of occurrences in the corpus of $ender)marked and person)marked pre) nominal articles, accordin$ to the noun class of the controller.....................................827 "able 72K +o. of occurrences in the corpus of $ender)marked and person)marked pre) nominal articles accordin$ to the noun class of the controller %e3cludin$ nouns contributin$ more than fi!e tokens*...............................................................................827 "able 7;K +o. of occurrences in the corpus of $ender)marked and person)marked pre) nominal articles accordin$ to the animacy of the controller %e3cludin$ nouns contributin$ more than fi!e tokens and nc9 nouns*......................................................829 "able 78K +o. of occurrences in the corpus of $ender)marked and person)marked pre) nominal articles accordin$ to the animacy of the controller %e3cludin$ nouns contributin$ more than fi!e tokens, nc9 nouns, and nouns with human referents*.......829 "able 7FK +oun class pronouns plus copula %repeated from L5.2.4F*...........................8;8 "able 75K 1ersonal pronounNcopula comple3es %repeated from "able 45*.....................8;8 "able 77K Su$$ested historical deri!ation of the person)marked copula forms.............8;8 "able 79K Summary of properties of #icipu person markers %repeated from "able 54* 8;9

49

Index of figures
0i$ure 4K 6ocation of =cipuland within +i$eria %map courtesy of Ani!ersity of "e3as 6ibraries, httpKNNwww.lib.ute3as.eduNmapsN*....................................................................27 0i$ure 2K 6ocation of #icipu dialects and ma'or towns...................................................29 [}rqYWyYfonrYr)|bY[^YonrbY\[ffnqYe^Yer[{[^e...................................................29 [}rqYayYxrn~[o[e^nf{ogfqYp}oY[^YbY\[ffnqYe^Yer[{[^e........................................;0 [}rqYyYr[{[^eYwp[f~rq^Yon[^Y|nroY[^YopqYuiuYq{o[\nf.......................................;4 0i$ure 5K Eiew from Makuku lookin$ east, with &orisino hill on the left......................;2 0i$ure 7K /eported structure of #icipu dialects...............................................................;; 0i$ure 9K #icipu and nearby lan$ua$es %reproduced from Gordon 200F*.......................;7 0i$ure 9K #lassification of (enue)#on$o, by kind permission of /o$er (lench %May 2009*................................................................................................................................;9 0i$ure 40K Subclassification of est &ain'i, by kind permission of /o$er (lench %May 2009*................................................................................................................................;9 0i$ure 44K /e!ised subclassification of &ambari, (lench %n.d. aKF*...............................;9 0i$ure 42K Markus Mallam Babani, "irisino speaker......................................................F2 0i$ure 4;K Musa Can'uma Mai An$uwa, "irisino speaker, and Mohammed Musa, "ikula speaker..................................................................................................................F2 0i$ure 48K "he $ender system of /umanian %taken from #orbett 4994K4F2*..................50 0i$ure 4FK &ambari noun class system accordin$ to =fricanist tradition %based on .e olf 4974K50*. <offmannTs %495;* numberin$ is $i!en in brackets...............................52 0i$ure 45K &ambari $ender system accordin$ to #orbettTs methodolo$y %based on .e olf 4974K50*. <offmannTs %495;* numberin$ is $i!en in brackets...............................5; 0i$ure 47K Schematic representation of the "se> a$reement system...............................74 [}rqYXyYwpqino[wYrq|rq{q^ono[e^YeYopqYn}inrbYnrqqiq^oY{g{oqi.........................75 0i$ure 49K Schematic representation of the #entral &ambariN#icipu a$reement system79 0i$ure 20K Informal comparison of 6ambrecht:s sentence topic %abo!e* and .ooley:s discourse topic %below*....................................................................................................9F [}rqYUXyYZ[\]^{YXWY|pe^efe[wnfYwe~[^Yvq[poY{wnfq.........................................99 0i$ure 22K a!eform of )^noYd{inffY{|[~qrh............................................................420 0i$ure 2;K a!eform of n)^nYdwevh............................................................................420 0i$ure 28K a!eform of )f]ewbYdo[iqh.....................................................................424 0i$ure 2FK a!eform of l)fiYd^e[{qh.........................................................................424 [}rqYU`yYn\qeriYeY)~iY)nYdopqYver~Yvein^h.............................422 l YdeYoev^{h..................................................................42; [}rqYUyYn\qeriYeYu)r)rqm [ [}rqYUyY[owpYornwYerYpuqvtYdopqgYrq}{q~hYrqnf[{................................................4;5 [}rqYUyY[owpYornwYerYpqvtYdopqgY{pe}f~Yrq}{qhY[rrqnf[{....................................4;5 0i$ure ;0K #icipu $enders accordin$ to noun prefi3es..................................................28; 0i$ure ;4K #icipu $enders accordin$ to a$reement prefi3es.........................................28; [}rqYWUyYn\qeriYeYqni|fqYWWyYvb^~YwuYY Y~[~^oY{qq....................................;2F [}rqYWWyYn\qeriYeYqni|fqYWyYYvb^~YwwuYY Y~[~^oY{qqYge}..........................;2F [}rqYWayYn\qeriYeYqni|fqYaayY^pY||nYdfqoYiqYfqn\qYge}Y{VYpqrqh........;27 0i$ure ;FK 0lowchart showin$ which sub'ect +1s tri$$er $ender and person a$reement .......................................................................................................................................;88 0i$ure ;5K /e!ised flowchart modellin$ $enderNperson alternation.............................;89 0i$ure ;7K /elati!e positions of person and $ender a$reement in #entral &ambari and #icipu on Siewierska:s $rammaticalisation scale..........................................................;F9 0i$ure ;9K Eariation in a$reement accordin$ to inherent and discourse topicality.......;57 0i$ure ;9K <ow animacy affects the #icipu $enderNperson a$reement alternation.......;94 20

0i$ure 80K 0inal re!ision of flowchart modellin$ $enderNperson alternation................849 0i$ure 84K /elati!e positions of person and $ender a$reement in #entral &ambari and #icipu on (resnan and MchomboNSiewierska:s a$reement marker typolo$y %repeated from 0i$ure ;7*..............................................................................................................8;9 0i$ure 82K Eariation in a$reement accordin$ to inherent and discourse topicality, repeated from 0i$ure ;9................................................................................................880

24

Part 1 Introduction

Chapter 1 Introduction
1.1 Motivation for the study and research questions

#icipu, a (enue)#on$o lan$ua$e spoken in northwest +i$eria, has a robust noun class system of the kind found in many parts of sub)Saharan =frica, particularly the (antu lan$ua$es spoken in the south and east of the continent. 1redicates, pronouns, and most kinds of noun modifier all a$ree in
G-+.-/

with their sub'ects, antecedents, or head

nouns. "he #icipu e3amples below show two nouns from different $enders followed by a demonstrati!e, relati!iser, and a relati!e clause predicate, each of which a$rees in $ender with the head nounNsub'ect. %4* ka0llu)
/C1)hun$er

ke0lle)

A-1)that

ka0na)

A-1)/-6

ka)0ya*a0na)

A-1)arri!eQ/6S)10E

that hunger which arrived Isayb004.F85J %2* ma)0waa me0lle)


/C4)do$ A-4)that

ma0na) ma)0si 0 hu)na)


A-4)/-6 A-4)<=()kill

that dog which kills Itats004.004.0F0J #icipu differs from the (antu lan$ua$es, howe!er, in that se!eral of these tar$ets of $ender a$reement ha!e the potential to a$ree in
1-/S2+

instead. "hus there is a series of

competin$ paradi$ms across a number of different a$reement Star$etsT %#orbett 2005*. "his can be seen in the two e3amples below. "he referent in both cases is the same, a camel %ka)0ra)kumi , +#4*, but in %;* the !erb dooho SdisappearT takes a gender a$reement prefi3 ko0, while in %8* it takes a person a$reement prefi3 u0. %;* ko)0dooho)
A-1)disappearQ/6S

it disappeared Itats004.002.0;;J %8* u)0dooho)


30)disappearQ/6S

it disappeared Itats004.002.0;FJ Similarly the camel could be referred to by the $ender)marked pronoun k0e or by the ;1S pronoun evi , and demonstrati!es, the interro$ati!e Duantifier 0e)ne SwhichT, the copula, 2;

and the article all offer this choice of a$reement feature. "he factors influencin$ the alternation are comple3 to unra!el but of considerable theoretical interest. "his study has three main aims. "he most specific is to $i!e an account of the alternation between these two different paradi$ms of a$reement morpholo$y, one in!ol!in$ the feature G-+.-/ and the other 1-/S2+. "his account is based on a Dualitati!e analysis of te3ts from different $enres, and identifies a ran$e of rele!ant le3ical, phonolo$ical, morphosyntactic, semantic and discourse)pra$matic factors. "he second $oal is wider in scope but necessarily narrower in depth, and that is to pro!ide a comprehensi!e description of the #icipu noun class system. "he third $oal is simply to e3pose the #icipu lan$ua$e to the academic world by pro!idin$ a sketch $rammar. +aturally each of these $oals build on the others, with the sketch $rammar in 1art II pro!idin$ the base for the noun class description in 1art III, which in turn informs the detailed in!esti$ation of the a$reement alternation in 1art IE. "he rest of this section summarises the theoretical conte3t rele!ant for this thesis and outlines the research Duestions that it attempts to address. "he treatment here is brief, since the research conte3t is set out in much more detail in chapter 2, in particular L2.2 on a$reement and L2.; on topicality. 1art IE of this thesis is lar$ely concerned with sub'ect markers on !erbs. In the last two decades the nature of such markers has been hotly)debated in the lin$uistics literature. In particular a distinction has been made between
G/=MM="I#=6 =G/--M-+",

where the !erb ar$ument is e3pressed as a le3ical +1 in addition to the a$reement marker e.$. ka)0ra)kumi ko)0dooho) Sthe camel disappearedT, and
=+=1<2/I# =G/--M-+",

where there is no le3ical +1, as in e3amples %;* and %8* abo!e. "his latter kind of a$reement marker has been associated with topical referents, especially by (resnan and Mchombo %4997*. <owe!er !erb a$reement markers can differ alon$ other dimensions apart from the $rammatical !s. anaphoric one. In chapter 2 we will see that (resnan and Mchombo:s typolo$y of a$reement markers makes no predictions about what mi$ht influence the alternation of the two sets of #icipu a$reement markers when they are both found in the same syntactic en!ironment, as is the case for %;* and %8*. "his is the central analytical task with which the thesis is concerned. "o illustrate the sort of Duestions that will be addressed in 1art IE, ima$ine a te3t

28

in!ol!in$ a camel as one of the participants. If the speaker wants to say that the camel disappeared, under what circumstances mi$ht we e3pect to find %;*, with a sub'ect prefi3 inflected for instead for
1-/S2+M G-+.-/,

and under what circumstances would %8* appear, inflected

Is the distribution of the two kinds of markers dependent on

topicality , and if so, what kind of topicalityM Is the concept of Sdiscourse topicT of any use in describin$ the alternationM <ow important is the initial referrin$ e3pression in determinin$ future a$reement markersM Is it possible to discern any kind of discourseM =bstractin$ away from particular usa$e e!ents, what sorts of referents %or what kind of nouns* are more likely to tri$$er $ender a$reement rather than person a$reement, and !ice !ersaM Is animacy rele!antM =re nouns from certain $enders more likely to tri$$er person a$reement than other $endersM %so if the camel was referred to usin$ the <ausa loanword da a) Sanimal, results in a correlation with $enderM =s mentioned abo!e, it turns out that le3ical, phonolo$ical, morphosyntactic, semantic, and discourse)pra$matic factors are all rele!ant to the alternation between $ender and person a$reement, and it is therefore necessary to consider all of these areas of lin$uistics for a comprehensi!e description. "his is true not only for sub'ect a$reement markers, but also for other a$reement tar$ets such as pronouns and the article. "opic does pro!e to be important, but not the concept of Ssentence topicT to which $enerati!e theories of synta3 are, by their nature, restricted. Instead, a coherent analysis of this aspect of #icipu a$reement reDuires a theory of topic that takes into account both the way in which topics inte$rate a te3t, and the intrinsic interest which they hold for the speaker.
+#9T, 1/2G/-SSI2+

in!ol!in$ the two different kinds of a$reement, as the referent is tracked throu$h the

would this make any differenceM*. =nd

if so, is this a true $ender effect or is it an underlyin$ semantic factor which merely

111

!esearch "uestions
hat are the a$reement tar$ets in #icipuM hich of these a$reement tar$ets inflect for more than one a$reement feature paradi$mM 2F

1art III and particularly 1art IE will address the followin$ research DuestionsK

=re the syntactic en!ironments in which the different paradi$ms occur mutually)e3clusi!eM If not, what are the factors that influence the a$reement paradi$m when there is a choiceM If topicality is one of these factors, what kind of topicality is itM Is the data better e3plained by theories of Ssentence topicT or of Sdiscourse topicTM

"he rest of this chapter is structured as followsK L4.2 pro!ides some sociolin$uistic back$round, L4.; re!iews the anthropolo$ical and lin$uistic works which refer to #icipu, and L4.8 describes the fieldsite and the data and methodolo$y used for this research. 0inally L4.F $i!es an o!er!iew of the thesis.

1. 1#1

!anguage situation The language name

"he )thnologue %Gordon 200F* has an entry for #icipu under S estern =cipaT, with the IS2 5;9); code awc. "he name S estern =cipaT is unfortunate for two reasons. 0irst, it su$$ests the lan$ua$e is a close relati!e of the nearby lan$ua$e known as S-astern =cipaT. =s we will see in L4.;, this is not the case. Secondly, S=cipaT is not a word in any lan$ua$e. It is an =n$licised contraction of the <ausa ethonym Acipawa, or of the correspondin$ term for the lan$ua$e Acipanci. "he autonyms are -i pu) %one person,
+#9*,

A)0$i pu) %the people, +#2*, and -i 0 ) $i pu) %the lan$ua$e, +#5*. =lthou$h the etymolo$y

of these terms is unknown, they ha!e no ne$ati!e connotations, and since so little has been written about S estern =cipaT there seems little point in continuin$ to use a confusin$ and inaccurate name4. Incidentally, #icipu was entered as SSa$amukT in the 'nde* o Nigerian Languages %#ro>ier and (lench 4992* and its accompanyin$ lan$ua$e map. "he location on the map is correct but the name is not. Instead it seems to be the autonym for the )astern =cipa lan$ua$e %.ettweiler and .ettweiler 2002K;F*.

1##

$emographic and ethnographic notes

#=1/2 %499FK490* $i!es an estimate of 20,000 #icipu speakers. It is impossible to e!aluate this fi$ure with any de$ree of certainty, since there is no reliable census data,
4 = reDuest to chan$e the name to S#icipuT was accepted in January 2009 by the IS2 5;9); $o!ernin$ body, and the new name will trickle down into the )thnologue, probably in 204;.

25

the birth rate is hi$h, and se!eral dialects are endan$ered. Gunn and #onant %4950KFF* Duoted a fi$ure of ;,F72 adults from a 4985 census, but a$ain this is uncertain. "he lan$ua$e is spoken in a conti$uous area di!ided between Sakaba 6ocal Go!ernment =rea in &ebbi State, and &onta$ora 6ocal Go!ernment =rea in +i$er State.

+igure ,- Location o Acipuland within Nigeria .map courtesy o /niversity o 0e*as Libraries, http-11www.lib.ute*as.edu1maps12 "he area is remote and much of the re$ion can be inaccessible by car in the rainy season , partly because of swollen ri!ers, and partly because there is no path throu$h the corn wide enou$h. "he map in 0i$ure 2 shows some of the ma'or routes %sin$le)track dirt roads*, !illa$es, and appro3imate locations of the se!en dialects and the surroundin$ lan$ua$es. =s .ettweiler and .ettweiler %2002* obser!ed, most =cipu are farmers and do not li!e in towns , these are populated by <ausas %and also by people who are ethnically =cipu but no lon$er ha!e #icipu as their mother ton$ue*. Instead the =cipu li!e in smaller !illa$es and hamlets. -3o$amous marria$e is rare, and althou$h =cipu !illa$es are interspersed with those of the =!ai &ambari in particular, mi3ed)tribe !illa$es do not seem to be common. = si$nificant number of =cipu ha!e mi$rated far into +i$er State to the south)west, since farmland is plentiful there, and in the !illa$e I stayed in appro3imately a Duarter of the youn$ men had done this. It is not known to 27

what e3tent these mi$rants form a cohesi!e community, nor whether they continue to speak #icipu in their new settlements.

+igure 4- Location o 5icipu dialects and ma%or towns Ko)ri si )no*o hill %<ausa 3arisen* is marked on the map in 0i$ure 2. "his is the seat of the Wm SchiefT %also referred to as 20daa Skin$T*, who li!es on the hi$hest point of the hill, accompanied only by his wife and his constant companionNad!isor the 3a0llu* SteacherT %<ausa mallam2. 0irst and foremost the kin$ is the head of the 4)ri si )no*o di!ision of the =cipu, but he also seems to ha!e precedence in certain matters o!er the leaders of the other di!isions. &orisino used to be inhabited in li!in$ memory, althou$h the old people I spoke to had mo!ed to the !alley when they were youn$. -le!en !illa$es %includin$ two wu0u5o) SshrineT* are still maintained on the hilltop, each correspondin$ to specific settlementsNclans in the !alley. Generally they are only used durin$ festi!als, althou$h an old man may decide to return there to die, in which case his wife will accompany him, and family members will continue to !isit and care for

29

them. "he other di!isions each ha!e their own hilltop settlements, althou$h they are not as lar$e and well)maintained as &orisino. "he only other hill I !isited was 6)0kula), which belon$s to the A)kula) di!ision and is located 'ust north of Maburya. "he =kula ha!e almost all con!erted to Islam, and houses are no lon$er maintained on the hill. 2nly the elderly continue to carry out reli$ious obser!ances there.

+igure 6- Altar %ri 0 ) pi 7i * ) 7897:;<=>?@7A8BB>CD7E97FE?8G89E "he =cipu are mostly peasant)a$riculturalists. "hey $row $uineacorn for subsistence, and sell any e3cess alon$ with cash crops includin$ beans, $roundnuts, and soya beans. 0armin$ is $enerally done by hand, althou$h plou$hin$ with o3en is becomin$ more common. "he most common li!estock are chicken and $oats, and sheep are also kept. Some =cipu do own a few cows, which are $enerally tended by 0ulani. #amels are a common si$ht around har!est time %.ecember*, when they are brou$ht down from the north by <ausas in return for a share of the har!est. <orses and pi$s are not kept, althou$h the former seem to ha!e been more common in the past. = small amount of rice is $rown by women, and some dry)season $ardenin$ is done. #rops are $rown in the fields ne3t to and radiatin$ out from the !illa$es, and so li!estock are tethered durin$ the rainy season. "here are occasional confrontations o!er land use with 0ulani pastoralists, whose settlements are interspersed with those of the =cipu, but usually relations are peaceful. Some men en'oy huntin$, which is done at ni$ht with locally)

29

made $uns. Most of the bi$ $ame has now disappeared, but monkeys are a fairly common si$ht in more remote areas and occasional si$htin$s of leopards and hippos are reported. "here are crocodiles on &orisino, but these are considered sacred and are not to be molested. "raditionally huts are round, althou$h these are now $i!in$ way to rectan$ular huts. "hatched roofs still seem to be the norm, but wealthier indi!iduals are be$innin$ to purchase >inc roofin$ sheets. Guest huts %<ausa Haure* are round, with opposin$ doorways to allow a cross)bree>e. #ompounds are fenced rather than walled, and can be entered without $oin$ throu$h the Haure. Men:s and women:s $ranaries are still built usin$ the traditional method described in 1ra>an %4977*.

+igure JK7L?>M8<8E9>BNG<OBD7PQ<7897:;CRS@7A8BB>CD7E97FE?8G89E "he ma'ority of the =cipu are Muslims, althou$h abo!e a certain a$e the traditional reli$ion %ko)0ri nno)* is still dominant, particularly in the "irisino dialect area. Iaigiro worship %"emple 4922* is still obser!ed %o)0ki iso, sin$ular k0ki iso*, as it is by seemin$ly e!ery est &ain'i people $roup. -ach =cipu di!ision has its own festi!als held on their particular hill. =t &orisino there are fi!e ma'or festi!als, the most spectacular of which is the Ka)077eme festi!al held towards the end of the dry season. =s part of their initiation into adulthood, youths of about thirteen2 endure beatin$s by a line of youn$er boys armed with spiked sticks. Some youths end up with hea!ily)lacerated backs,
2 =nd older youths and men, if they choose.

;0

althou$h serious in'ury seems to be rare.

+igure TK7U?8G89E7VP8BM?D97<>W89C7X>?<7897<PD7F;YYZ[Z7\DG<8A>B Muslims also attend festi!als on &orisino, althou$h they do not take part in reli$ious ceremonies. "he influence of the <ausa culture lon$ pre)dates actual con!ersion to Islam, as can be seen from the chieftancy titles listed in Mathews %4925b*, and from the fact that the !ast ma'ority of people $o by a <ausa name. #hristians make up a small minority, and most are relati!ely youn$. "hey will ha!e nothin$ to do with the traditional reli$ion, and this is a ma'or source of conflict and persecution within families. I was not able to elicit any kind of creation myth, but the story of the initial settlement of the =cipu hills is well)known. "he outline of the story, recounted by Mathews %4925b*, is that the first ruler of the =cipu, .amasa son of .amerudu son of the ma$ician)kin$ &isra, fled from his ori$inal home far to the east after war with the 1rophet. hen they had tra!elled as far as the foot of &orisino, .amasa:s wife was pre$nant and they decided to settle on the hilltop. "he StrueT =cipu %talakawa, <ausa for ScommonersT* in!ited .amasa to become their kin$ and &orisino was founded. .amasa:s people are therefore the ancestors of the royal clan, the 4)0dondo. "he details of the story %particularly names and kinship relations* !ary from dialect to dialect, both in Mathews: day and the present. 2ther nearby $roups tell similar stories %e.$. "sureshe, =$amalafiya 2009*, and the names of .amerudu and &isra also appear in Boruba and ;4

(ussawa folk histories %Mathews 4925b*.

+igure ]- ^iew rom Iakuku looking east, with 3orisino hill on the le t In more recent times, the =cipu claim to ha!e fou$ht and defeated the notorious 49 th century sla!e raider +a$wamatse of &onta$ora. +e!ertheless the impact of sla!ery on the =cipu was no doubt se!ere and it accounts for the fact that e!en within li!in$ memory they li!ed e3clusi!ely on the inaccessible hilltops. =t the time of Mathews: %4925b* report hillside terraces were bein$ abandoned in fa!our of farmin$ in the !alley below. =s mentioned abo!e the hills are now almost deserted. 0or further ethno$raphic details see Gunn and #onant %4950* and #=1/2 %499F*.

1#%

$ialects

.ettweiler and .ettweiler %2002* identified four =cipu SclansT, althou$h Sdi!isionsT is a better term, since they each number hundreds or thousands of people and themsel!es consist of smaller clans. "he names recorded by the .ettweilers were S=riseneT, S=kumbasiT, S=kulaT, and S=>akacihunT %lit. Speople of &akihumT*, correspondin$ to the first and the last three di!isions in 0i$ure 7. "he =cipu I spoke to reco$nised se!en sub) di!isions within the ethnic $roup, which they said corresponded to different dialects. "he names of the dialects are $i!en in "able 4, while 0i$ure 7 is an attempt to represent their internal structure;.
; =lthou$h these names do turn up in discussions about dialects, they do not seem to be as entrenched as the names for other lan$ua$es such as "i 0 ) va8i ) %"su!ai* or "i )0hwi 'i %#:6ela*. It is probably more

;2

0able ,- 5icipu dialects1settlements and their Hausa names .ialect autonym "i 0 ) ri si )no*o "i 0 ) di )po "i 0 ) 7o)ori yo) "i 0 ) ddi )m* "i 0 ) kum asi "i 0 ) kula) * "i 0 ) $uhuu =cipu settlement Ko)0ri si )no*o Ko)0di )po Ko)07o)ori yo) K)0ddi )m* 6)0kum asi 6)0kula) * Kwe)0$uhuu <ausa name of settlement 3arisen13arishen13arishin13arissen 3adonho IaHarko 3adedan 3umbashi Iaburya 3akihum "irisino "idipo

#icipu

"i>oriyo "idodimo

"ikumbasi "ikula "icihun +igure _- Reported structure o 5icipu dialects "he branchin$ in this dia$ram is based on discussions with nati!e speakers 8, but I ha!e not yet been able to carefully check if these $roupin$s are supported by lin$uistic e!idence. I ha!e first)hand knowled$e of "irisino, "idipo, and "ikula, and certainly the first two are closer to each other than either is to "ikula. "he most ob!ious difference between "irisino and "idipo is the loss of nasalisation in many words in the latter, e.$. I'u*J !s. "irisino I'u*J Sthere, far offT, Ipo*J !s. "irisino Ipo*J SallT. Support for at least the first and last $roupin$s in 0i$ure 7 can be found in the n\n[fnjfqYver~f[{o{VYx[w}pY[{Yq^qrnffgYrqwe^[{q~Yn{Yjq[^YopqYie{oY~[\qrq^oY~[nfqwo
common to use a phrase such as ka)0dama ki 0 ive) Stheir speechT, or e.$. ka)0dama ka0aKula) Sspeech of the =kulaT. "he word ke)0re'e) Ston$ueNlan$ua$eT cannot be used to refer to these dialects. 8 "his information was first pro!ided by Ibrahim asako, a "idipo speaker, and independently corroborated by .o$o "imbidii, a "ikula speaker, and Markus Mallam Babani and Musa Can'uma Mai An$uwa, both "irisino speakers.

;;

and it can be seen from the data presented by .ettweiler and .ettweiler %2002* that there is a sound correspondence between eY [^Yx[w}pYn^~ Yo in the other dialects. 0or e3ample, the $reetin$ poopo) ShelloT %sannu in <ausa* is pronounced Ipe(pe) Y[^Yx[w}p and doonu SsitT is pronounced Ide(nuJ. In fact this correspondence is a stereotype of x[w}pY{|qqwpYrqwe^[{q~YjgYeopqrYw[|}V +otwithstandin$ these differences, the dialects are actually !ery close to each other, and there is no doubt that they belon$ to a sin$le lan$ua$e. "hey are all mutually intelli$ible, and the le3icostatistical e!idence points to the same conclusion , "irisino pn{YnYfq[wnfY{[i[fnr[ogYeYnoYfqn{oYYvpq^Ywei|nrq~YoeYx[}ijn{[Yx[w}pYx[~[|e "i>oriyo, and "ikulaF. "hese dialects are clearly distinct from the closest non)#icipu !ariety, "su!ai, which has 50 le3ical similarity with #icipu %McGill n.d.*. "irisino is the presti$e dialectK e!en speakers from other dialects will admit that "irisino is the SbestT form of #icipu.

1#&

'ultilingualism

.ettweiler and .ettweiler %2002K4;* wrote thatK "he e3tent of <ausa bilin$ualism in the estern =cipa lan$ua$e community needs to be e!aluatedG limited bilin$ualism is reported, but this should not 'ust be assumed as true. In the nine months I spent in the "irisino dialect area I did not find anyone who could not understand <ausa. "he oldest women are at least able to understand it, e!en if they rarely speak it. It is my impression that the =cipu are somewhat less isolated and more Se3posedT than their =!ai nei$hbours, some of whom it appears really do not ha!e much command of <ausa. Greetin$s amon$st =cipu men, e!en within the same family, appear to be mainly in <ausa. Members of other tribes rarely speak #icipu well, e!en the closely)related =!ai, and in mi3ed company <ausa is the norm. -!en without the presence of outsiders, youn$ men and boys often speak <ausa amon$st themsel!es. Greetin$s amon$st women seem to be in #icipu, as do $reetin$s between men and women. In addition to <ausa, many =cipu speak one or more of the nei$hbourin$ est &ain'i lan$ua$es. 0rom my obser!ations in the "irisino dialect area this is most likely to be "su!ai, followed by #:6ela. 0ulfulde is not $enerally learnt apart from $reetin$s and
F (ased on the 229)item wordlist collected by .ettweiler and .ettweiler %2002*. +o data e3ists for "idodimo.

;8

other !ery basic communication. In the course of nine months: fieldwork I encountered only two or three fluent -n$lish speakers amon$st the =cipu. -n$lish is of course the medium of instruction in secondary school, which is attended by some of the =cipu from the !illa$e I stayed in. Ansurprisin$ly this is often to little purpose.

1#(

)ndangerment

=ccordin$ to (lench %4999* there are rou$hly 2F0)800 lan$ua$es in the +i$erian Middle (elt, dependin$ on how broadly the area is concei!ed. /ou$hly half of these ha!e less than 800 speakers. ith appro3imately 20,000 speakers %L4.2.2*, #icipu is far in e3cess of the median. 0or +i$eria as a whole, Gordon %200F* ranks #icipu as the 490th most populous out of the F40 li!in$ lan$ua$es. <owe!er numbers aren:t e!erythin$, and e!en nearby lan$ua$es with lar$er populations can be considered at risk5, the main dan$er, for the lar$er lan$ua$es like #icipu at least, comin$ from shift to <ausa. "he reality is that for #icipu the le!el of endan$erment is different for each dialect, and correlates with the de$ree of con!ersion to Islam. =ccordin$ to (lench %4999*, there is a stron$ correlation in the Middle (elt between switch to <ausa and con!ersion to Islam, when this con!ersion has taken place in fairly recent times , which is the case for all of the =cipu. "he 2risino, who until recently ha!e been almost wholly animist, ha!e maintained their lan$ua$e, and children in e!ery !illa$e I !isited spoke #icipu as their mother ton$ue. In contrast, the =kula ha!e almost entirely con!erted to Islam, and "ikula is possibly the most endan$ered of the dialects, with only a few children in a couple of !illa$es still learnin$ #icipu. "he "ikumbasi dialect may be similarly endan$ered, and others likely fall between these two e3tremes. "he influence of #hristianity on lan$ua$e maintenance has yet to be seen, since the few con!erts are mostly recent. .espite the potential positi!e effect of (ible translation and associated literacy pro$rammes, it should be noted that in =cipuland <ausa is the lan$ua$e of both the mosDue and the church. General factors that fa!our lan$ua$e maintenance include endo$amous marria$e practice and the hi$h birth rate in the area , at least for those dialects where children are still learnin$ the lan$ua$e. Speakers: attitudes to their own lan$ua$e are mostly positi!e, and initial runs of #icipu storybooks and cassettes were recei!ed with enthusiasm. "he
5 0or e3ample the est &ain'i lan$ua$e "sureshe %(lench 4999*.

;F

=cipu also seem to be proud of their cultureG there is a sayin$ amon$st them that there isn:t a land in the entire world which hasn:t heard of &orisino , a belief no doubt sustained by the !isits of curious estern scholars e!ery few years. "here is no written literature in #icipu, apart from a small number of recently) produced storybooks based on oral narrati!es. =n ortho$raphy DuestionnaireNproposal has been de!eloped based on the lin$uistic analysis carried out durin$ this pro'ect, but this is still to be circulated.

1."

#evie$ of literature on %est &ain'i( &ambari( and )icipu

-!er since the =cipu first came to the attention of scholars in the early twentieth century, they ha!e been linked to both the &ambari people and the &amuku people. <owe!er no lin$uistic e!idence has e!er been published to support a connection between #icipu and any other lan$ua$e, and it seems that in pre!ious lan$ua$e classifications cultural similarity and purported shared ethnic ori$in has outwei$hed the lin$uistic e!idence. hen the lin$uistic e!idence is considered %McGill n.d.*, it is clear est &ain'i that #icipu should be re$arded as part of the &ambari $roup. In this section I will re!iew the a!ailable lin$uistic literature on the $roup as a whole %L4.;.4*, on the &ambari cluster %L4.;.2*, and on #icipu %L4.;.;* , and for #icipu I will also briefly mention the rele!ant anthropolo$ical sources. "he $eo$raphical location of the lan$ua$es concerned is $i!en in 0i$ure 9. "o $uard a$ainst confusion it is worth stressin$ that #icipu %which, as mentioned in L4.2.4, has also been called S estern =cipaT* and -astern =cipa are not close relati!es, despite the names and the $eo$raphical pro3imity of their homelands.

;5

+igure `- 5icipu and nearby languages .reproduced rom Gordon 4aaT2

1%1

*est +ain,i
est &ain'i lan$ua$es are !ery di!erse both in their est &ain'i lan$ua$es est

(lench %n.d. a* has obser!ed that

le3icon and morpholo$y. #onseDuently the $enetic unity of the

was not reco$nised until (ertho %49F2*, who used S&ambriT as a co!er term for the $roup. Greenber$Ts %49FF* 1lateau 4a $rouped to$ether what are known today as &ain'i $roup on the same le!el as 1latoid by Gerhardt %4999*. illiamson and (lench %2000* contains the most recent published classification of (enue)#on$o, based lar$ely on iso$losses delimitin$ the !arious sub$roups. It places &ain'i in the #entral +i$erian branch of -ast (enue)#on$o. (lench:s latest workin$ classification is $i!en in 0i$ure 9K &ain'i lan$ua$es. 1lateau 4a and 4b %-ast &ain'i* were then promoted to a separate

;7

1roto)(enue)#on$o

#entral +i$erian

Akaan M (antoid)#ross

&ain'i

1lateau

Jukunoid #ross /i!er (antoid .elta #ross

+orthwest 1lateau

#entral 1lateau (eromic

"arokoid Apper #ross 6ower #ross 2$oni

S- 1lateau

+orth "ikar .akoid Mambiloid "i!oid

South (endi M

(uru +yan$ (eboid 0uru cluster -koid Grassfields +demli /in$ Menchum Momo -astern = $roup (antu includin$ Jarawan +arrow (antu

+igure b- 5lassi ication o (enuec5ongo, by kind permission o Roger (lench .Iay 4aab2 "here are twenty)fi!e est &ain'i lan$ua$es accordin$ to Gordon %200F*, althou$h new

lan$ua$es are still comin$ to li$ht %Spencer 2009, McGill 2009*. "he best known lan$ua$es are the lar$er onesK .uka %or <un)Saare*, #T6ela %also known as .akarkari*, and #entral &ambari, althou$h the lar$est of all, "su!ai with 4F0,000 speakers %Gordon 200F*, is yet to be the sub'ect of any lin$uistic publication. Se!eral are moribund %Gordon 200F, McGill 2009* and others still are endan$ered %(lench 4999*.

;9

/ eshe 6aru 6opa &ambari ( asa &amuku 1on$u Gurm ana (aushi Ara 1on$u cluster c6ela <un)Saare M aTin Gwamhi) uri .amakawa IMJ #icipu Apper +i$er

6ake

Proto12est 3ain4i

( asa)&am uku #entral

+orthwes t
/M( May 2009

+igure ,a- dubclassi ication o "est 3ain%i, by kind permission o Roger (lench .Iay 4aab2 (lench:s latest workin$ subclassification of est &ain'i is reproduced in 0i$ure 40, and

is a$ain based on le3ical inno!ations delimitin$ the !arious sub$roups %see (lench n.d. a*. =s mentioned abo!e, McGill %n.d.* demonstrates that #icipu should be re$arded as belon$in$ to the &ambari $roup, rather than (asa)&amuku %as was the case in Gerhardt 4999, illiamson and (lench 2000, and Gordon 200F*, and 0i$ure 40 reflects this re!ised classification. "he structure of the &ambari branch is presented belowK
tsiGai tsuVai ciShingyini ciBaangi Proto-Kambari tsiGaushi tsiKimba tsiWnci Kambari II Cicipu Kambari I

+igure ,,- Revised subclassi ication o 3ambari, (lench .n.d. a-T2

1%#

+ambari

/emarkably both &ambari and &amuku feature in &oelleTs %495;* &olyglotta A ricana, first published in 49F8. &ambari appears as &-mbri %II.-.9*, and .alby %4958* ;9

identifies this with the modern term &ambari. <offmann %495F* confirmed that the forms in &oelleTs &ambari list were almost identical to those in the #entral &ambari wordlist he collected. Johnston %4949* briefly mentions S&ambaliT. <e notes that it resembles the (asa) &amuku lan$ua$es Gurmana, &amuku, and (asa, but that apparently...IdoesJ not make use of prefi3es , at any rate for determinin$ sin$ular and plural %4949K785*. It is not clear what the source of Johnston:s data was, and no known &ambari lan$ua$e lacks noun class prefi3es today. &ambari is also listed in Meek %492F !ol. 2K4;7* and estermann and (ryan %49F2*, and /owlands %4952* contains a 482)item wordlist and discusses its relationship with &amuku, .uka and #T6ela. 0or #entral &ambari <offmann pro!ided a description of the noun class system %495;* and a F00)item wordlist %495F*. 2n the basis of <offmann:s work, .e olf %4974* showed how the #entral &ambari noun class system mi$ht be deri!ed from his 1roto)(enue)#on$o reconstruction. "he most important source for &ambari, howe!er, is #ro>ierTs %4998* unpublished 1h. thesis on the discourse $rammar of #entral &ambari. "his includes a detailed account of the lan$ua$eTs noun phrase, !erb phrase, and clause and sentence structure. More recently, trilin$ual &ambari)<ausa)-n$lish dictionaries ha!e been published for three &ambari lan$ua$es7 , "sikimba, #ishin$ini, and "sishin$ini %Stark 200;, 2008a, 2008b*9. Anfortunately for the lin$uist, the dictionaries omit important phonolo$ical details and contain no <ausa or -n$lish finder lists.

1%%

Anthropological and linguistic literature on Cicipu

=nthropolo$ists ha!e historically paid more attention to the =cipu people than lin$uists. =s well as $eneral ethno$raphic works on the +i$erian Middle (elt, se!eral articles ha!e been written about the artefacts housed in the Wm:s palace at &orisino, said to ha!e belon$ed to the le$endary ma$ician)kin$ &isra %Mathews 4950, Ste!ens 497F, Stewart 4990*. =s far as I know, the =cipu people were first mentioned in print by .uff
7 It is misleadin$ to speak of &ambari as a sin$le lan$ua$e, and the )thnologue treats it as a cluster of si3 lan$ua$es. In this thesis I will use S&ambariT as a co!er term when discussin$ the cluster as a whole %minus #icipu*, and use indi!idual lan$ua$e names %e.$. #entral &ambari* otherwise. 9 #onfusin$ly, Stark %2008a* and #ro>ier %4998* use the term S#ishin$iniT to refer to two different lan$ua$es. Stark uses it to mean estern &ambari, while #ro>ier uses it to mean #entral &ambari %which Stark has arbitrarily called S"sishin$iniT %2008b*, apparently to a!oid ha!in$ two dictionaries with the same title*. 0or clarity I will stick to #entral &ambari and estern &ambari throu$hout this thesis.

80

%4972I4920J*9. <e comments on the Iaigirro %o)0ki iso* reli$ious practices %4972K55)57* and notes that the &ambari are descended from the =cipu and the &atsinawa %4972K;0*. "he .akarkari %6elna* are also said to be descended from the =cipu, and .uff notes that the surroundin$ tribes %Hamfara, (an$awa, .ukawa, and .akarkari* acknowled$e the Wm as the SfatherT of the whole country, althou$h they owe him no political alle$iance. "emple %4922K;0);4* includes a section on the =cipu %S=tsifawaT*, apparently deri!ed from the same source as .uff %4972*. Mathews %4925b* is an unpublished 80)pa$e historical and anthropolo$ical report on the =cipu %S=chifawaT*, based on a two)month !isit. <e includes a few #icipu words in the main te3t, and also in an appendi3 on lin$uistics. In two places he reports that the =cipu deny any relationship with the &amberri, presumably a response to elicitation based on "emple:s report. Gunn and #onant %4950* contains a chapter on the =cipu %S=chipawaT*, lar$ely based on Mathews %4925b*. 6ikewise #=1/2 %499F* includes a similar section on the S=chipawaT, but also contains ori$inal research on the traditional reli$ion and the distribution of #hristians amon$st the people. =s well as the #icipuN&ambari link, early sources also mention a connection between the lan$ua$e referred to here as #icipu, and a lan$ua$e identifiable as -astern =cipa from the &amuku sub$roup. "emple %4922K;0, 207)209* stated that S=chipanchiT was also spoken to the east, in the &amuku area. <e writes that the two $roups know of no connection with each other, but at the same time records that the eastern $roup mi$rated there from the western area a lon$ time a$o , this is supported by the local traditions reported by .ettweiler and .ettweiler %2002K9*, whose informants did reco$nise the link between the two peoples. Mathews %4925b* noted that "here are also =chifawa in the central Makan$ara hills of the +i$er 1ro!ince Ii.e. -astern =cipaJ, whither they mi$rated from the parent body many years a$o, and where =chifanchi is still said to be spoken but this may 'ust be "emple:s remark rephrased. Gunn and #onant %4950KFF* state that the -astern S=chipawaT %autonym S"ochipoT* ha!e denied any connection with the =chipawa of &onta$ora Ii.e. on earlier reports. estermann and (ryan were the first lin$uists %as opposed to anthropolo$ists* to mention the lan$ua$e spoken by the =cipu, commentin$ %49F2K408* that
9 .uff refers to notes on the =cipu taken by #. -. (oyd, which I ha!e been unable to locate.

estern =cipaN#icipuJ, but a$ain this may well be based

84

It is not known what lan$ua$e or dialect is spoken by the Achi awa %Achipawa* amon$ the eakarkari and euka Ii.e. #icipuJ. "hey appear not to be the same as the Achipawa or Achi anchi 3amuku Ii.e. -astern =cipaJ. It has already been mentioned that (ertho %49F2* was the first to reco$nise the $enetic unity of the est &ain'i $roup. <e di!ided the ele!en est &ain'i lan$ua$es for which est &ain'i he had data into three sub$roups. (ased on 42F)item wordlists, he placed #icipu %S=tshefa)nshiT* to$ether with #entral &ambari and "su!ai in one of three sub$roups. "his $roupin$ is si$nificant, since althou$h it differs from published classifications, it is supported by the e!idence presented in McGill %n.d.*. (ertho:s other two S&ambriT sub$roups correspond to the S6akeT and S+orthwestT clusters of est &ain'i, and he did not include &amuku in his &ambri $roup at all. Instead he assi$ned it to his S(aoutchiT $roup, correspondin$ to what are now known as 1lateau lan$ua$es, based on e!idence from noun class prefi3es. In fact there is a Duestion mark as to the accuracy of (ertho:s sources for est &ain'i. Anfortunately he did not pro!ide the actual lan$ua$e data on which he based his comparisons, and so it is impossible to know how accurate the material was. <e made one specific comment about =tshefa)nshiK that around fifty percent of the nouns had lost their class prefi3es. "his is surprisin$, $i!en that #icipu has a robust noun class system. <e made similar errors concernin$ "su!ai and #entral &ambari %<offmann 495;K452*40. +e!ertheless the one S=tshefa)nshiT word he did $i!e %kacteri SboneT* is likely to be #icipu %ke)0tere in "irisino*, and there are no other ob!ious candidates for the lan$ua$e name S=tshefa)nshiT in the &ambari cluster as it is currently understood. In any case, whate!er errors there were in his data, he seems to ha!e $ot the classification of #icipu ri$ht. "he ne3t lin$uist to mention S=cifanciT was /owlands, who wrote thatK "he &amuku proper are di!ided into two sectionsMy informant called the other section Ii.e. the section he didn:t belon$ to , S.M.J by the <ausa name S=cifawaT and their dialect S=cifanciT %/owlands 4952K90*. It is probable that /owlands: informant was referrin$ to -astern rather than autonyms, may ha!e influenced the GerhardtN(lench classification of -astern =cipa to$ether under &amuku44.
40 See L4.;.2 for the same error by Johnston %4949*. 44 /o$er (lench %p.c. 2009* has obser!ed that cultural differences set the &ambari apart from the =cipu and &amuku, and this may also ha!e been a factor.

estern

=cipa. /owlands did not make this distinction, and this fact, to$ether with the similar estern and

82

=s mentioned in L4.2.4, .ettweiler and .ettweiler %2002* studied the relationship between S esternT and S-asternT =cipa. "hey did this by collectin$ 229)item wordlists from three estern =cipa dialects and two -astern =cipa dialects, and then performin$ a le3ical comparison between them. "hey found that the le3ical similarity between any two representati!es of the two $roups was always less than twenty percent, and proposed a re!ised classification of est &ain'i with estern =cipa remo!ed from &amuku and allocated a branch of its own. 0inally, McGill %2007* pro!ided an initial description of the #icipu noun class system. "he material in that paper is further de!eloped in 1art III.

1.*

+ield$or, setting and data

"he analysis in this thesis is based on data collected durin$ field trips to the =cipu area from September 2005 to March 2007, and from January to =pril 2009. "he resultin$ annotated corpus consists of appro3imately si3 hours of transcribed, translated, and interlinearised audio!isual recordin$s, to$ether with elicited recordin$s of se!en hundred words from the SI6 #omparati!e =frican ordlist %Snider and /oberts 2008*. "he le3icon compiled from these sources amounts to around two thousand le3emes, and the corpus contains appro3imately twel!e thousand clauses. .urin$ these field trips I stayed in the "irisino)speakin$ !illa$e of Galadima in Sakaba 6G=, between Makuku and &orisino %see 0i$ure 2*. Galadima consists of ten family compounds, nine of which belon$ to =cipu. "he remainin$ compound is inhabited by /euben =cheson, a +i$erian missionary from &aduna State, and his <ausa) and -n$lish)speakin$ family, and it was with them that I stayed. In total there are appro3imately 200 inhabitants, most of whom are children. =ll the =cipu in the !illa$e speak both #icipu and <ausa, and none of them speak -n$lish. .urin$ the first two months, elicitation sessions were conducted mainly in <ausa. (y this time my #icipu was no worse than my <ausa, and so we switched to the !ernacular from then on. 2utside of elicitation sessions I tried to communicate in #icipu at all times, and li!in$ in a #icipu !illa$e naturally allowed me to obser!e many thin$s about the lan$ua$e and culture which would not ha!e been possible otherwise. 0or e3ample, with reference to this study, I noticed that speakers did use $ender sub'ect a$reement in the absence of sub'ect +1s, and that they did use person a$reement with non)human controller referents, contrary to what I was repeatedly told in elicitation 8;

sessions? In $eneral, I ha!e tried to a!oid usin$ data from elicitation sessions when $i!in$ e3amples in the chapters that follow, preferrin$ instead to draw on the corpus. <owe!er there are times when I ha!e found elicited e3amples useful, either because they more clearly elucidate some point, or because the construction under discussion did not occur in the te3t corpus. My methodolo$y is therefore corpus)informed rather than corpus) dri!en %Mc#arthy 2004K22*, and only limited use is made of Duantitati!e methods42. "he e3amples are mainly taken from "irisino speakers, althou$h if a "ikula or "idipo e3ample makes the point better then I ha!e not hesitated to use it. "he use of such e3amples should not be taken to mean that the property or phenomenon bein$ illustrated is absent from "irisino , if that is the case, then it will be stated in the te3t. "he corpus contains data from se!eral different $enres as can be seen in "able 2. 0able 4- dtructure o the interlinearised corpus used or this study (e$t ty"e 6bser!ed -enre #on!ersations Sermons Son$s4; 0ta$ed 0olktales <istorical narrati!esNinter!iews 6ocal e!ent narrati!es /iddles 1rayers Stimulated "opic)stimulation 0ish 0ilm 1ear 0ilm 7licited 5uration in minutes F 28 280 90 480 7 40 2 92 40 20 nNa

"he te3t)types in "able 2 are based on <immelmann:s %4999* typolo$y. <e defines
2(S-/E-.

communicati!e e!ents as communicati!e e!ents in which e3ternal

interference is limited to the factthat the on$oin$ e!ent is bein$ obser!ed %4999K27*.
42 0reDuency data is $i!en for $enderNperson a$reement pro$ressions in L9.F, but no statistical analysis is attempted. 4; =ll the son$s recorded are of the Scall)and)responseT type common to sub)Saharan =frica. 2ften a sin$le line is repeated for se!eral minutes, and so the actual contribution of son$s to the corpus in terms of clauses is minimal. =dditionally the translation of son$s is !ery difficult, since the meanin$ may be obscure e!en to the sin$ers.

88

+aturally such e!ents are difficult to record, and the lin$uist has no control o!er the lin$uistic data. Sta$ed communicati!e e!ents, on the other hand, are communicati!e e!ents which are enacted for the purpose of recordin$. I ha!e subdi!ided these into e!ents proper and
S"IMA6="-. S"=G-.

e!ents. "he difference between these two lies in the

amount of control the lin$uist has o!er the proceedin$s. #ommunicati!e e!ents of the former kind are less sub'ect to the control of the lin$uist. -3amples mi$ht be an e!enin$ of folktales or$anised by nati!e speakers because they knew the lin$uist would want to record them, or simply someone askin$ the lin$uist to record them speak. In stimulated e!ents, on the other hand, the speakers may be asked to perform a particular task by the lin$uist, for e3ample to carry out a photo)matchin$ e3periment, to retell the story of a !ideo they had 'ust watched, or to simply talk about a referent of the lin$uist:s choosin$. "he analyses and freDuency data in 1art IE are based on obser!ed, sta$ed, and stimulated data, but not on elicited sentences.

1&1

$ata collection methodolog-

2ne of the ma'or problems when in!esti$atin$ the effects of discourse factors on a$reement is $ettin$ hold of the ri$ht kind of data for testin$ hypotheses. In particular, elicitation has not pro!ed to be a !ery useful method of obtainin$ rele!ant data for e3aminin$ the alternation between $ender and person a$reement, and it seems that the phenomenon in Duestion is not amenable to analysis by nati!e speaker introspection. Moreo!er, e!en in areas of information structure where relati!ely simple paradi$ms can be set up %e.$. for ar$ument focus*, elicited sentences may pro!e a bad $uide to what actually happens in obser!ed communication %e.$. (iber 4998, (earth 4999*. #onseDuently I ha!e relied mainly on the analysis of te3ts rather than disco!eries made in elicitation sessions. =s mentioned abo!e, there is a middle $round between the collection of SauthenticT te3ts and traditional elicitation. = well)known e3ample is the specially)made stimuli de!eloped by the Ma3 1lanck Institute for 1sycholin$uistics %e.$. .an>i$er 499;*. "his techniDue allows the lin$uist to focus on a particular area of the lan$ua$e, for e3ample askin$ Duestions about a set of subtly different photo$raphs or models may illuminate how the lan$ua$e encodes spatial relations, ideally in a less artificial settin$ than a strai$htforward elicitation session. Most such stimuli seem to be desi$ned for the study 8F

of semantics rather than discourse, althou$h the 1ear 0ilm %#hafe 4990* and 0ro$ stories %(erman and Slobin 4998* are well)known e3ceptions %see also Skopeteas et al. 2005*. If the a$reement alternation under discussion here was sensiti!e only to the information structure rele!ant to the form of a sin$le sentence %in other words, if the alternation could be accounted for usin$ 6ambrechtTs theory of sentence topic discussed in L2.;*, then e3periments could be desi$ned to stimulate the rele!ant data. If, howe!er, discourse topicality %L2.;* is in!ol!ed, then the prospects are less $ood. "he problem is that when workin$ with an alien stimulus such as a film prepared by a forei$n researcher, it is difficult to control, or e!en to predict with any de$ree of certainty which participants will be construed as more central at any one time. In fact, nati!e speakers may well ha!e no clear idea themsel!es, and this is bound to decrease the naturalness of the te3t. "here is also a more $eneral problem with e3periments in!ol!in$ stimuli that are intended to $enerate narrati!e, identified in 0oley %200;*. 0oley used the 0ro$ story book mentioned abo!e with a speaker of atam %1apua +ew Guinea*, in order to atam narrati!es, a distortion $enerate an oral narrati!e based on the pictures and story therein. "he resultin$ features of the te3t were markedly different to those of traditional which 0oley attributed to the !isual stimulus of the book and its illustrations IwhichJ undoubtedly lead to a $reater refle3i!e awareness of the participants and e!ents of the story %200;K9F*. "his resulted in a normalised literate te3t, despite its oral modality. #onseDuently much of what is interestin$ about the $enre of narrati!e in to surface in the e3periment. .espite my reser!ations, I did prepare some M1I)style photo)matchin$ e3periments since the data I collected in my first field trip only had a few e3amples of the $enderNperson alternation %apart from the ones I had bullied out of my lan$ua$e consultants*. I desi$ned an e3periment based on the usual M1I photo)matchin$ task %e.$. .an>i$er 499;*, with two modifications. 2ne was that the ScallerT could not see all the photos at once, but picked them one)by)one from a face)down pile. "he other modification was that the caller was told beforehand that the SmatcherT had e3tra photos %this was not in fact true*. "hese modifications were put in place to ensure that the callers described the scenes as carefully as possible, rather than takin$ shortcuts and atam failed

85

relyin$ on lo$ical deduction on the part of the matcher 48. (y usin$ this e3periment in con'unction with #hafe:s %4990* 1ear 0ilm I hoped that I would stimulate e3amples of the alternation for referents of !aryin$ animacy %human, animal, and inanimate*, and in sufficient number to discern a pattern. "he e3periment used se!en sets of twel!e photos each. "he sets fell into two basic cate$ories , in some sets the Sfi$ureT referent was kept constant and photo$raphed a$ainst a !ariable S$roundT, while in the other sets the fi$ure was !aried. Earious other controls were implemented %e.$. some sets contained referents from the same noun class, others contained referents from a mi3 of noun classes*. "he caller in the e3periment was instructed to carefully describe to the matcher how the ob'ects are situated. (ased on pre)fieldwork trials carried out on nati!e -n$lish speakers, and on what I already knew about #icipu, I e3pected the e3periments to elicit constructions in!ol!in$ sub'ect a$reement prefi3es , in particular !erbs indicatin$ possession, orientation, and location. It was thou$ht that the sets with a constant referent mi$ht encoura$e the caller to use person %StopicalT* a$reement, and that for the mi3ed sets the caller mi$ht be more likely to use $ender %Snon)topicalT* a$reement , the lo$ic bein$ that an e3tended task in!ol!in$ only one fi$ure would increase the likelihood of that referent bein$ percei!ed as topical, no matter how inanimate it was. "he instructions made it clear that it was the same referent %e.$. bird* in each of the twel!e pictures in the constant)fi$ure sets. = catalo$ue of problems, some of which are discussed below, meant that the data was useless for what I had intended it for. "o my mind, these problems ar$ue a$ainst the use of these kind of stimuli in this field site, and perhaps others like it. Many of the problems were linked to sub'ects: unfamiliarity with the e3perimental mediaK physical photo$raphs, and !ideos shown on a laptop. Most of the participants were under ;0 and relati!ely Se3posedT , they would all ha!e seen films before, and many of them had completed at least two years of primary school. <owe!er they are certainly not used to performin$ the tasks they were set in these e3periments. 1eople would hold the photos upside)down when describin$ them 'ust as often as they held them upri$ht, thus renderin$ useless oppositions that depended on orientation. +o matter how many times I ShelpedT by arran$in$ the cards carefully in a ;38 $rid, after a couple more $oes they
48 ithout these modifications, the caller:s attention, on !iewin$ two similar photos, would be drawn to the ground %the part that differed* rather than %as desired* the igure %which in half the sets remained constant*.

87

would be scattered around in a manner most unconduci!e to comparison. "here were further problems with the identification of e!eryday local ob'ects in the photo$raphs. =nother une3pected problem was participants: inability to distin$uish between mirror ima$es. Se!eral of the sets included two photos which were mirror)ima$es of each other, for e3ample a bird facin$ to the left, and the same photo with the bird facin$ to the ri$ht. hen I trialled this in -n$land, speakers had no problem differentiatin$ the two and seein$ the difference as si$nificant with respect to the task they had been $i!en. <owe!er #icipu speakers either stru$$led to see that there was anythin$ different about the two pictures, or e!en when they did realise, they stru$$led to communicate this to the other participant. =s well as unfamiliarity with the media, the fact that speakers are not used to takin$ part in e3periments also causes problems. .espite ha!in$ been $i!en precise instructions, the participants did not understand that in many of the tasks each of the photos was supposed to show the same ob'ect in different situations. Instead of !iewin$ the pro$ression of photos as relatin$ to a sin$le referent increasin$ in discourse topicality, with each new photo they would say somethin$ like =nd here:s another bird. #onseDuently there was not e!en the opportunity for StopicalT person a$reement to arise. = lot of psycholin$uistic e3periments are repetiti!e and reDuire the participant to perform the same basic tasks o!er and o!er a$ain, somethin$ which pro!ed more difficult than I had ima$ined, perhaps because sustainin$ this o!er se!eral minutes seemed rather pointless to those takin$ part. In one case the participant $a!e an online hi$her)le!el analysis of the e!ents, rather than concentratin$ on the task at hand? "he 1ear 0ilm e3periments cannot be said to ha!e been a complete failure with respect to my purpose in carryin$ them out, since the transcriptions do contain a few e3amples of both $ender and person a$reement with the same referent. <owe!er se!eral incidents occurred which illustrate 'ust how alien the task of retellin$ the e!ents of a !ideo is in the #icipu culture. 2ne of the speakers recounted thatK hen he fell, the thin$s then fell. <is friends came and helped him clear up, and they lifted him back on to the bike. "hen he went and he fell a$ain. "hen they a$ain $athered the fruit and they a$ain put him on the bike Itapf002.00;.045J. =t first I found her account confusin$, since there is no such repetition in the film. <owe!er when the film had been shown, the laptop crashed and part of the !ideo had to 89

be replayed. "hat the speaker interpreted this as a continuation rather than a repetition of the action demonstrated a nai!ety with respect to !ideo that I had not %althou$h perhaps should ha!e* anticipated. =nother speaker be$an his account with the words we saw a li$ht, by which he meant the li$ht bein$ emitted from the laptop screen itself. "he effect of these and other misunderstandin$s is limited in scope, but they do show that there can be problems e!en when the stimulus in!ol!ed is one as carefully desi$ned as the 1ear 0ilm4F. It should be stressed that many of these issues arose when I tested the materials on -n$lish)speakin$ +i$erian $raduates li!in$ in the =cipu area. It was fascinatin$ to see them try to cope with the difference between mirror ima$es, and 'ust like the =cipu they ended up usin$ the words east and west rather than le t and right. Since they too held the photos upside)down they would ha!e stru$$led whiche!er terms they used? Such SrecalcitrantT beha!iour could perhaps be countered by $i!in$ more trainin$ and detailed instructions before the start of the e3periment, but there is a dan$er of imposin$ so many constraints that the discourse produced under these conditions becomes less and less like what the participants would actually say in e!eryday life, so that the e3periment becomes a test of what speakers can be made to say. =nd of course if instructions are too complicated they may simply be i$nored, as was the case for the photo)matchin$ e3periments.
1&11 Topic.stimulation sessions

(ecause of the failure of the stimuli e3periments, I adopted a less sophisticated approach to in!esti$atin$ the $enderNperson alternation. "wenty)four te3ts were recorded, all of which are monolo$ues, with occasional inter'ections by the inter!iewer. "he speakers were asked to tell the inter!iewer e!erythin$ that they knew about particular topics %some human, some animals, some inanimate*, and the Duestion was phrased with the topic noun in sub'ect position e.$. please tell me e!erythin$ that tobacco does. "he Duestion was asked in <ausa to a!oid makin$ a choice between $ender and person sub'ect a$reement45, but the speaker was asked to reply in #icipu. I was aimin$ for 2); minutes for each te3t , if the speaker stopped before then %which
4F See also L9.;.4 for differences with respect to the li$ht sub'ect constraint between these and more traditional narrati!es. =nother difference is that the dependent imperfecti!e aspect %L8.5.;.8* was far more freDuent in the 1ear Stories, perhaps because participants are $i!in$ careful attention to situations that they would normally $loss o!er %e.$. by reportin$ them with the perfecti!e*. 45 In <ausa the pre)!erbal 1erson)=spect)#omple3 is marked for both $ender and person.

89

was rare, most te3ts are rou$hly that len$th, some are si$nificantly lon$er*, then they were prompted %this time in #icipu* to say more. It is normally recommended that in the case of a forei$n director, any interaction with the sub'ects should be carried out by an appropriately)trained nati!e speaker. "herefore I be$an by askin$ a lan$ua$e consultant to do the inter!iews, but this actually seemed to e3acerbate rather than miti$ate the obser!er parado3, and so later I resorted to conductin$ the inter!iews myself. I found people $a!e more comprehensi!e accounts to an outsider than when asked to $i!e an account to someone who already knows e!erythin$ that is $oin$ to be said. "his is doubtless because they could identify a moti!ation for takin$ part in the e3periment , to instruct the forei$n researcher, someone already cast as a learner and well)known for his i$norance. = side benefit of tryin$ to ensure that participants are moti!ated is that it helps to stren$then relationships between the researcher and the community. Speakers can ha!e a hard enou$h time understandin$ why a fieldworker is present in a community without bein$ $i!en silly thin$s to do. "he downside of this approach is ob!iousG the sub'ects are addressin$ themsel!es to a cultural outsider and a non)nati!e speaker , the te3ts that arise are of a kind that would not normally occur outside of such sessions, since people are not, as far as I can tell, e!er instructed in such a manner %while the 1ear 0ilm is also artificial, people do at least tell each other stories or recount somethin$ they saw happenin$, presumably in all cultures*. I admit that this is a problem with my methodolo$y, but I do not see how it could ha!e been miti$ated. "he decision to carry out the e3periments is, to my mind, !indicated by the e3tent to which the patterns that emer$ed in the e3periments could also be found later, albeit more disparately, in te3ts which had been collected under less artificial circumstances. "his will be demonstrated in chapter 9.

1&#

Corpus imbalance

"he corpus used for this study re$rettably suffers from an imbalance in two main areas. In terms of $enre, con!ersation is under)represented. =lthou$h there are many e3amples of interaction between speakers, such as inter!iews, riddles, and folktales, and a smaller number of short StrueT con!ersations nested within non)con!ersational $enres, I ha!e no te3ts which are primarily con!ersational. #onseDuently the potential effect of turn transitions on $enderNperson a$reement pro$ressions has not been ri$orously studied. F0

1erhaps a more serious failin$ is the under)representation of women:s speech in the corpus. "he $enderNperson alternation and the a$reement pro$ressions discussed in chapter 9 ha!e been obser!ed in female speech, but the o!erwhelmin$ ma'ority of them are in male speech. In particular it would be interestin$ to test whether the male tendency to refer to women with $ender a$reement and men with person a$reement is replicated or re!ersed in the speech of women. "he redress of these imbalances should be a hi$h priority for future #icipu documentation pro'ects, re$ardless of any theoretical aims related to this study.

1&%

Cross.referencing conventions

"he e3amples in this dissertation are cross)referenced to the source of the data. Most cross)references consist of a te3t identifier followed by an utterance number within that te3t. "he te3t identifier contains the followin$ componentsK
"-")"B1-

, M2.=6I"B , S1-=&-/ , "-")+AM(-/.

So for e3ample, the cross)reference e-a-my-015.004 refers to the 'ourt! %004* utterance in an audio %a* recordin$ of an elicitation %e* session, the 'i'teent! %015* such recorded te3t with 8arkus 9abani %my* as the principal speaker. "he !alues used in the first slot are e for 7licitation, t for s(imulated e!ents, s for 0taged e!ents, and o for 6bserved e!ents. "he !alues used in the second slot are a for audio and v for video. If the e3ample is in a dialect other than "irisino then this is indicated alon$ with the cross)reference. "he te3ts and accompanyin$ metadata are in "oolbo347 format and ha!e been archi!ed at the -ndan$ered 6an$ua$es =rchi!e at the School of 2riental and =frican Studies. "hey are also a!ailable from www.cicipu.or$. 2ther cross)references are in the form yyyy)mm)dd.3, where the first part is the date and the * stands for the *th fieldnote made on that day. =ll such e3amples should be assumed to be the product of unrecorded elicitation sessions unless otherwise stated.

1&&

Contributors

"he names of =cipu who contributed to the corpus used in this study are pro!ided in =ppendi3 (, alon$ with their dialect, se3, and appro3imate a$e. "hree people in particular were of $reat assistance in the painstakin$ task of transcribin$ and translatin$
47 httpKNNwww.sil.or$NcomputIn$Ntoolbo3N

F4

te3ts, as well as pro!idin$ lan$ua$e consultancy durin$ elicitation sessions. "hese were Markus Mallam Babani and Musa Can'uma Mai An$uwa, both "irisino speakers li!in$ in Galadima !illa$e, near &orisino, and Mohammed Musa, an -n$lish)speakin$ #ipu from the =kula di!ision at Maburya, who arran$ed, conducted, and transcribed two lon$ inter!iews with members of the =kula chieftancy. 2thers who pro!ided formal lan$ua$e consultancy were "idipo speakers =yuba Sani and Ishiaku Ibrahim, and "irisino speaker Ibrahim Can'uma Mai An$uwa.

+igure ,4- Iarkus Iallam fabani, 0irisino speaker

+igure ,6- Iusa gan%uma Iai /nguwa, 0irisino speaker, and Iohammed Iusa, 0ikula speaker

F2

1.-

.vervie$ of the thesis

"his thesis is made up of four parts. 1art I consists of this introductory chapter to$ether with chapter 2, which presents !arious theoretical preliminaries necessary for the later partsK noun class systems %L2.4*, a$reement %L2.2*, and topicality %L2.;*. 1art II %chps. ;)8* pro!ides a sketch $rammar of the #icipu lan$ua$eG chapter ; is concerned with phonolo$y and chapter 8 with morphosynta3. 1art III %chps. F)5* is a description of the #icipu noun class system. #hapter F presents the noun classes and $enders %i.e. pairin$s* and discusses their semantic structure and the deri!ational use of class prefi3es. #hapter 5 is concerned with the multiple a$reement tar$ets found in the lan$ua$e, and includes sections on neutral a$reement and the use of antecedentless a$reement morpholo$y. 1art IE %chps. 7)9* focuses on the research Duestions raised in L4.4 and seeks to unco!er the !arious factors $o!ernin$ the alternation between $ender and person a$reement. #hapter 7 e3amines the phonolo$ical, morpholo$ical, and syntactic properties of the !arious pronouns and a$reement markers, and identifies the a$reement prereDuisites which apply to each of the competin$ paradi$ms. <a!in$ secured this foundation, we are in a position in chapter 9 to in!esti$ate the distribution of $ender and person a$reement markers in discourse, in an attempt to determine the rele!ant a$reement conditions for those syntactic en!ironments which do allow a choice. 0inally, chapter 9 summarises the findin$s of the thesis and considers their theoretical implications.

F;

Chapter # !esearch context


1art III of this thesis is a description of the
+2A+ #6=SS

system of #icipu. "his

description assumes certain notions and terminolo$y from both the =fricanist and typolo$ical traditions, and so these are introduced in L2.4 as a preliminary. 1arts III and IE are both concerned with the $rammatical phenomenon of
=G/--M-+",

introduced in L2.2. 2f particular concern are the domain of a$reement,

!ariation in a$reement, and the typolo$y of a$reement markers. "o anticipate the findin$s of chapter 9, we will see that the semantic and pra$matic notions usually employed in treatments of !ariations in a$reement do not fully account for the #icipu data, and instead we must also consider
"21I#=6I"B. .IS#2A/S- "21I#=6I"B

as opposed to

S-+"-+#-

"herefore this chapter will also introduce the notion of discourse topic %L2.;*.

.1

/oun classes and gender


+2A+ #6=SS,

It is not strai$htforward to define what is meant by the e3pression

in part

due to the comple3ity of such systems, and in part due to the inconsistent terminolo$y used by authors from different fields of lin$uistics. "o $et a feel for what is in!ol!ed, it is perhaps best to be$in with some practical e3amples, and so in L2.4.4 I will set out some prototypical properties of (enue)#on$o noun class systems. Section 2.4.2 e3plains in more detail the concepts and terminolo$y used in both typolo$ical and =fricanist research.

#11

/enue.Congo noun class s-stems


olf

+oun class systems are found in the !ast ma'ority of (enue)#on$o lan$ua$es %.e

4974K4F*, most famously in the (antu lan$ua$es of southern and eastern =frica. "hese systems classify nouns accordin$ to the different sin$ular and plural affi3es that they take, and usually reDuire other constituents to a$ree with the noun. "he e3act definition of a Snoun class systemT is problematic and is deferred to the ne3t section, and so for the moment, I will build on the abstract characterisation I ha!e 'ust $i!en by introducin$ and e3emplifyin$ some prototypical properties of (enue)#on$o noun class systems. +ouns in (enue)#on$o lan$ua$es usually occur with both sin$ular and plural affi3es, which are attached to the noun stem as in the followin$ e3ample from #icipuK

F8

%4*

ka)0 ara

/C1)oldPman

a)0 ara
/C

)oldPman

old man

old men

Ieamd045.;9;J

"he sin$ular form of the noun for Sold manT is composed of the noun class prefi3 ka)0 followed by the noun root ara, while the plural form of the noun is a)0 ara, containin$ a different noun class prefi3 a)0. "his in itself simply shows that #icipu, alon$ with most other (enue)#on$o lan$ua$es, is typolo$ically unusual in markin$ both sin$ular and plural in its noun morpholo$y. hat makes it a noun class lan$ua$e is the fact that ka)0 and a)0 are 'ust two out of a number of different prefi3es, which can be paired in different ways. 0or e3ample, the #icipu noun root di ya ShareT takes a sin$ular prefi3 ma)0 and a plural prefi3 n)0. %2* ma)0di ya
/C4)hare

n)0di ya hares

/C%)hare

hare

Ieamd020.4045J di ya is said to belon$ to a different noun class %in both its sin$ular and its plural forms* from ara because it takes a different pair of prefi3es. Most (enue)#on$o noun class affi3es are prefi3es, althou$h suffi3es and %rarely* infi3es also occur. #onser!ati!e %i.e. unreduced* (enue)#on$o systems can ha!e as many as twenty different noun class n[q{YqVVY{q\q^oqq^Y[^Yp[wpqnYYwpeijeYUay`V In addition to the system of noun class affi3es, there is usually a correspondin$ system of a$reement affi3es which mark other constituents within the noun phrase. 1rototypically both the head noun and the modifiers are marked. =$reement within the noun phrase occurs in all e3ceptionK %;* ka)0 ara
/C1)oldPman A-1)that

est &ain'i lan$ua$es studied to date, and #icipu is no ma)0di ya me0lle)


/C4)hare

ke0lle)

A-4)that

that old man

that hare Ieamy00F.004, eamd0;2.0F9J

<ere the head noun $o!erns the form of the a$reement prefi3 on the modifier. +oun class a$reement is not usually limited to the noun phrase. Sub'ect %and sometimes ob'ect* noun phrases determine the a$reement affi3 used on predicates, and anaphoric pronouns a$ree with their antecedent. =ll three types of a$reement %+1)

FF

internal, predicate, and anaphoric* can be seen in the followin$ e3amples from the (antu lan$ua$e -'a$ham, which in!ol!e the same head noun in parallel structuresK sin$ular in %8* %with class F affi3es* and plural in %F* %with class 9 affi3es*K %8* e)0yu)
/C%)yam

e)j0a*

A-%)your

j09t

A-%)one

n)j0i

A-%)this

n)j0i )

A-%)which

n)na)me'',

4SNbou$ht

nj09ne) e0 hi )p
A-%)it

A-%N10E)bad

0his one yam o yours that ' bought, it is spoiled. I-'a$ham, %F*
/C))yam

atters 2000K202J

i0 ) yu)

i )h 0a*

A-))your

i0 ha'e m) 0i
A-))two

A-))this

m) 0i )

A-))which

n)na)me'',

4SNbou$ht

m 09ne) i 0 hi )p
A-))they

A-)N10E)bad

0hese two yams o yours that ' bought, they are spoiled. I-'a$ham,

atters 2000K202J

ithin the front)shifted noun phrase, the head noun %hNOi SyamT or jkNOi SyamsT* tri$$ers a$reement on four other constituentsK the possessi!e pronoun l SyourT, the numerals mn< SoneT or bhopq StwoT, the demonstrati!e r SthisT, and the relati!e pronoun s SwhichT. In the main clause the anaphoric sub'ect pronoun mn9h S;rd person pronounT a$rees with its antecedent yt, while the !erb is marked with the appropriate sub'ect a$reement prefi3. In the -'a$ham data 'ust $i!en, the noun prefi3es and a$reement prefi3es differ si$nificantly in their phonetic realisation, and there are a number of different sets of a$reement prefi3es %possessi!e, numeral, demonstrati!e, sub'ect a$reement marker*. "his is not always so for (enue)#on$o lan$ua$es, and in fact the te3tbook case is for all affi3es to be the same or similar in form, a phenomenon known as
=G/--M-+" =66I"-/="IE-

, illustrated in this old chestnut from &iswahiliK ki0kapu


/C7)basket

%5*

ki0ku wa ki0moja ki0lianguka


A-7)lar$e A-7)one A-7)fell

one large basket ell I&iswahili, elmers 497;K474J =s well as showin$ that noun class a$reement operates both inside and outside the noun phrase, these e3amples show how !aried the different hosts of a$reement affi3es can be. #hapter 5 pro!ides a structured account of these hosts, or a$reement "=/G-"S, for #icipu.

F5

#1#
#1#1

Concepts and terminologGender and noun class languages

0or the o!er!iew 'ust $i!en a ri$orous definition of terms was not necessary. <owe!er partly due to the comple3ity inherent in many noun class systems, and partly due to the different traditions of !arious $roups of lin$uists, there is a dan$er of confusion if terms are not made e3plicit. "herefore in this section I will introduce and define a number of terms rele!ant for the study of noun class systems. 0irst of all, the ob!ious Duestions areK what Dualifies as a Snoun class lan$ua$eTM =nd how is one different from a S$ender lan$ua$eTM #orbettTs work on $rammatical $ender has been hi$hly influential in recent decades, and he follows <ockettTs %49F9K2;4* definition of $enderK Genders are classes of nouns reflected in the beha!ior of associated words. #oncernin$ the difference between $ender and noun class #orbett %4994K485* writes that there is little point in maintainin$ a strict separation between S$enderT and Snoun classT since similar systems are described as $enders in one family and as noun classes in another. Indo)-uropean S$ender systemsT are based on what is sometimes called Snatural $enderT i.e. masculine, feminine, and neuter. In contrast, (enue)#on$o Snoun class systemsT $enerally ha!e a lar$er number of distinctions, and the primary semantic feature in!ol!ed is animacy, rather than natural $ender. =ikhen!aldTs %2000* typolo$y of noun cate$orisation de!ices contrasts noun classifiers with noun class systems. =lthou$h both systems cate$orise nouns, in a noun class system some constituent outside the noun itself must a$ree in noun class with a noun %2000K20*. "hus by insistin$ on the presence of a$reement she too eDuates Snoun classT with #orbettTs S$enderT, althou$h she continues to write of Indo)-uropean S$ender lan$ua$esT and =frican Snoun class lan$ua$esT as a concession to the respecti!e traditions. Most lin$uists workin$ on (enue)#on$o lan$ua$es do not howe!er take the terms S$ender lan$ua$eT and Snoun class lan$ua$eT to mean the same thin$. .e %4974K;F*K In order to ha!e a true noun class system it is necessary that all nouns in a $i!en lan$ua$e be M=/&-. for the cate$ory of $ender by means of prefi3es, infi3es, suffi3es or some of these combined, such that this $ender cate$ory be a selecti!e F7 olf writes

one and, regardless of whether there e3ist types of class concord or class a$reement in modifiers andNor substitutes such that these modifiers andNor substitutes be marked for $ender with the result that $ender constitutes an obli$atory, inflectional cate$ory for these substitutes and modifiers. Imy italics , S.M.J hile #orbett and =ikhen!ald ha!e ri$htly emphasised the similarities between Indo) -uropean $ender lan$ua$es and =frican noun class lan$ua$es, there is nonetheless an important difference in the way that nouns themsel!es are marked in these systems. #orbett %4994K52* uses the term
2E-/" G-+.-/

to refer to $ender which is e!ident from

the form of the nouns themsel!es. It appears that all (enue)#on$o lan$ua$es ha!e noun prefi3es indicatin$ their noun class, as in %4* and %2* abo!e, and so all (enue)#on$o $ender lan$ua$es display o!ert $ender. If, on the other hand, the $ender is not shown by the form of the noun, the lan$ua$e is said to ha!e
#2E-/" G-+.-/.

German is such a

lan$ua$e , there is no way of knowin$ from the form of the noun Hand ShandT whether it is masculine, feminine, or neuter4. In (enue)#on$o lan$ua$es, Snoun class belon$in$T can therefore be realised both throu$h noun affi3es and throu$h a$reement affi3es. "hese may or may not e3ist in a one)to)one correspondence and may or may not share the same phonetic forms, and as a result the typolo$y of (enue)#on$o noun class systems is fairly comple3 %see Maho 4999K427)482*. #icipu, howe!er, has strai$htforward o!ert $ender %see 1art III*.
#1## Africanist terminolog-

"he first lar$e)scale comparati!e study on (enue)#on$o noun classes was carried out by ilhelm (leek %4952, 4959*. <e de!ised a numberin$ system for (antu noun classes, which was later re!ised by another German philolo$ist #arl Meinhof %49;2 I4999J*, and became known as the (leek)Meinhof %(M* numberin$ system. In this system each noun class affi3 %and the associated set of a$reement markers* is $i!en a number, ori$inally ran$in$ from 4)49. So each noun belon$s to two noun classes %with certain e3ceptions such as mass and abstract nouns*, one for the sin$ular affi3 and one for the plural affi3. "o capture the o!erall beha!iour of a noun, =fricanists usually refer to the sin$ularNplural pairin$s as $enders2. So $i!en the #icipu e3ample below, repeated from %4*, an =fricanist mi$ht speak of the KA noun class or the A noun class, but the KA0:
4 Hubin and &pcke %4995* ha!e shown that $ender assi$nment in German is less random than was once thou$ht. +e!ertheless, for many German nouns it is not possible to predict the $ender from the phonolo$ical form. 2 #onfusin$ly, some authors refer to a noun class on its own as a $ender. "his will be a!oided here.

F9

A0

$ender. %7* ka)0 ara


/C1)oldPman

a)0 ara
/C

)oldPman

old man

old men

"erminolo$ical confusion resultin$ from the two traditions may arise if a lan$ua$e has sin$ularNplural noun prefi3es but no a$reement. "hus .e olf writes that the 1lateau lan$ua$e /ukuba has Sse!en two)class $endersT %i.e. sin$ularNplural pairin$s* %4974K409*, at the same time as statin$ that the lan$ua$e, as far as is known, Sis without concordT %4974K409*. "his usa$e of S$enderT is in conflict with mainstream lin$uistics. It would perhaps be better to use the phrase
+2A+ #6=SS 1=I/I+G

for lan$ua$es such as

/ukuba, and reser!e the term S$enderT for lan$ua$es where an a$reement relation e3ists. In any case it will be clear from what follows that #icipu is unambi$uously a S$ender lan$ua$eT %since it has a$reement* as well as a Snoun class lan$ua$eT %since it has o!ert noun class prefi3es*. "he phrase
#6=SS -12+-+"

%.e

olf 4974* is a con!enient

umbrella term for any kind of class marker %noun or a$reement affi3* which si$nals $ender and number. #lass e3ponents may be nullG this is the case for #icipu class 9 %LF.F.7, L5.4.;*. = SI+GA6=/
#6=SS

is a noun class whose e3ponents attach to, or a$ree with, nouns


&=
;

which are sin$ular in number.


16A/=6 #6=SS

in e3ample %7* abo!e is a sin$ular class of #icipu. = is a noun class

is a noun class whose e3ponents attach to, or a$ree with, nouns which are
=

plural in number.

is therefore a plural class of #icipu. = SI+G6-

#6=SS

with members which do not participate in the sin$ularNplural alternation of count nouns. "ypically mass, liDuid, and abstract nouns are found in sin$le classes in (enue)#on$o lan$ua$es. "here is ob!iously potential for confusion with the term Ssin$ular classT, and ideally Sunpaired classT would be a better term. <owe!er Ssin$le classT seems to be the most widely)used term, and so I will use it here.
#1#% Controller and target genders
#2+"/266-/ G-+.-/S

#orbett makes an important distinction between a$reement

and

"=/G-" G-+.-/S

%4994K4F0)450*. "ar$et $enders are the $enders which are marked on particular
"=/G-"S

%e.$. ad'ecti!es, !erbs, demonstrati!es*, while controller $enders are

; 0or non)(antu lan$ua$es it is often inappropriate to use the (M numberin$ system, and so noun classes of indi!idual lan$ua$es may sometimes be referred to usin$ SM=66 #=1S. hen the #icipu system is dealt with in detail a different %non)(M* numberin$ system will be introduced.

F9

the $enders into which nouns are di!ided accordin$ to their composite beha!iour %similar to the =fricanist use of S$enderT to mean a noun class pairin$*. "he difference between controller and tar$et $enders can perhaps best be illustrated usin$ #orbettTs own e3ample of /umanian. #onsider e3amples %9)40* from #orbett %4994K4F0)4F4*K %9* %a*
man.the

ar atul e

is $ood

un0;

%b*

men.the

ar atii si *nt
are

$ood

un0i IM=S#A6I+-J

the man is good %9* %a* <ata e un0a

the men are good %b* <etele si *nt


are

$irl.the is $ood

$irls.the

$ood

un0e I0-MI+I+-J

the girl is good %40* %a* s$aunul e un0;

the girls are good %b* s$aunele si *nt


chairs.the are

chair.the is $ood

$ood

un0e I+-A"-/J

the chair is good

the chairs are good

/umanian ad'ecti!es take one of two different suffi3es %0; or 0a* when a$reein$ with sin$ular nouns, and one of two different suffi3es % 0i or 0e* when a$reein$ with plural nouns. "he problem is caused by neuter nouns like scaun SchairT which ha!e no uniDue a$reement forms. "hey share the masculine form for sin$ular nouns, and they share the feminine form for plural nouns. "he situation can be summarised usin$ an affi3 net8K sin$ular a m. n. f. plural i e

+igure ,J- 0he gender system o Rumanian .taken rom 5orbett ,bb,-,T42 #orbettTs terms allow us to capture the fact that althou$h there are three separate ways of di!idin$ up nouns %masculine, feminine, and neuter*, ne!ertheless the morpholo$y is simpler than the case of 6atin say, which has three independent $enders. controller $enders.
8 =ffi3 nets can be used to represent either a system of nominal affi3es, or as in this case a system of a$reement affi3es. "he lines represent the possible sin$ularNplural pairin$s.

e can say that

/umanian has two tar$et $enders in the sin$ular and two in the plural, but three

50

Maho %4999K482)48F* pro!ides a helpful discussion on the applicability of #orbettTs methods to (antu noun class systems. "he main thrust of his ar$ument is that neither controller $ender nor tar$et $ender corresponds to an eDui!alent (antuist term. =t first $lance it seems that indi!idual noun classes correspond to tar$et $enders, and that (antuist $enders, or noun class pairin$s, correspond to controller $enders. <owe!er this is only true in the case of a strai$htforward one)to)one correspondence between nominal affi3es and sets of a$reement markers. It is often the case that two or more noun classes with distinct noun affi3es share a set of a$reement affi3es, as do the #entral &ambari classes 2a and 2b %<offmannTs 495; numberin$*. #lass 2a has a prefi3 consistin$ of a !owel .) %the e3act !alue of . is determined by !owel harmony rules*, with identical a$reement prefi3es. #lass 2b shares the same a$reement markers, but has a different nominal prefi3 n=/0 %where / is a nasal homor$anic with the followin$ consonant*. 0or e3ample =0n=s=(r= S-uropeansT %2a* has a different prefi3 from n=m0 =s=r= ShawkT %2b*, but they both tri$$er the same set of a$reement markers. In #entral &ambari these two subclasses enter into distinct noun class pairin$s, which is an e3tra reason to treat them separately. "he followin$ affi3 nets show that we are not 'ust dealin$ with a terminolo$ical difference, since the two approaches result in conflictin$ number of $enders.

54

sin$ular AA >A@ ii >C@ mA >E@ /:m9 >F@ uu >D@ -:v9 >?a@ ; >? @ ts9 >G@

plural aA >Ba@ aB >B @ ii >C@ /:m9 >F@

-:v9 >?a@

+igure ,T- 3ambari noun class system according to A ricanist tradition .based on ee "ol ,b_,-]a2. Ho mannus .,b]62 numbering is given in brackets In 0i$ure 4F it can be seen that there are ten noun classes and twenty)two $enders %thirteen sin$ularNplural pairin$s and nine sin$le class $endersF , the latter represented with underscores*. "he important thin$ to note is that the plural classes 2a and 2b take the same a$reement markers, but their prefi3es are different and so they are treated as separate noun classes. "he same is true for the sin$ular noun classes 9a and 9b. +ow consider the same system dia$rammed accordin$ to #orbettTs approachK

F 2bser!e that classes ;, F, and 9a can occur with either sin$ular or plural nouns. In addition most #entral &ambari noun classes also occur as sin$le classes. .otted lines indicate an inDuorate $ender %#orbett 4994K470)47F* with !ery few members.

52

sin$ular AA >A@ ii >C@ mA >E@ /:m9 >F@ uu >D@ v9 >?@ ts9 >G@

plural a >B@ ii >C@ /:m9 >F@

v9 >?@

+igure ,]- 3ambari gender system according to 5orbettus methodology .based on ee "ol ,b_,-]a2. Ho mannus .,b]62 numbering is given in brackets "his dia$ram is similar to 0i$ure 4F, the only difference bein$ that classes 2a and 2b ha!e been mer$ed into a sin$le class 2, and classes 9a and 9b ha!e been mer$ed into class 9. In #orbettTs terms, for the set of count nouns there are si3 tar$et $enders in the sin$ular and four tar$et $enders in the plural, while for the set of mass nouns there are a further ei$ht tar$et $enders. "here are ei$hteen controller $enders %ten sin$ularNplural pairin$s and ei$ht sin$le class $enders*, four less than the twenty)two $enders shown in 0i$ure 4F. 0urthermore, some information about class pairin$s has been lost in 0i$ure 45. 0or e3ample, it is no lon$er apparent that class 2b is only found as the plural class for class 9b, and not for 9a. hile it is clear that #orbettTs terminolo$y does not eDuate with standard (antu terminolo$y, the difference between the two approaches pro!es to be minimal with respect to #icipu, with only two classes %;a and ;b* bein$ conflated under the former %LF.4)F.2*. +e!ertheless this discussion of #entral &ambari has ser!ed as a useful illustration of how #orbett:s analytical framework applies to (enue)#on$o lan$ua$es. (efore lea!in$ this section, a note is reDuired on the terms Sa$reementT and SconcordT. "hey are usually taken to mean the same thin$, with concord bein$ the preferred term in the (antu literature. Some authors do distin$uish the terms but do so 5;

in idiosyncratic ways , see #orbett %200;aK440)442* for a summary. <ere I will stick to Sa$reementT.

Agreement

In the pre!ious section we saw that the presence of a$reement is a definitional property of $ender systems %in the terminolo$y of mainstream lin$uistics* , this is what sets them apart from both classifier systems and reduced noun class systems. #orbett %2005K8* notes that a$reement is hard to define satisfactorily, and se!eral recent te3tbooks %#orbett 4994, 2005, Siewierska 2008* ha!e adopted Steele:s %4979K540* characterisation of a$reement as some systematic co!ariance between a semantic or formal property of one element and a formal property of another as a $ood workin$ definition. I will be$in by discussin$ the boundaries that ha!e been proposed between a$reement and other phenomena, in other words the
.2M=I+

of a$reement %L2.2.4*. "he

ne3t two subsections briefly discuss the mechanism of a$reement %L2.2.2* and the distinction between syntactic and semantic a$reement %L2.2.;*. #hapters 7 and 9 will in!esti$ate in detail a specific case of !ariation in a$reement, namely the $ender and person alternation found on se!eral a$reement tar$ets in #icipu. =s a preliminary then, L2.2.8 co!ers the different ways in which a$reement can !ary on a $i!en tar$et in a $i!en lan$ua$e, and what kinds of conditions tri$$er this !ariation. Much of chapter 7 is concerned with the syntactic status of the #icipu sub'ect a$reement markers, and so the typolo$y of a$reement markers and pronominal affi3es is also discussed here %L2.2.F*.

##1

$omain
.2M=I+

Many lin$uists ha!e attempted to limit the scope of a$reement %its

accordin$ to

the terminolo$y used in 0er$uson and (arlow %4999* and #orbett %2005**. .ifferent theorists draw the di!idin$)line between a$reement and related phenomena in different places, and their definitions of a$reement !ary accordin$ly. In particular, there is disa$reement as to whether the
=+"-#-.-+")=+=1<2/

relation should count as true

a$reement. (resnan and Mchombo %4997K78;* obser!ed that many current syntactic frameworks could account for differences between $rammatical and anaphoric a$reement %because entirely different mechanisms are postulated in these frameworks for $rammatical a$reement and pronominal incorporation...* 58

"he same is true of more recent $enerati!e theories of a$reement %e.$. #hun$ 4999*. (arlow %4992K4;8)4F2*5 comes to the conclusion that there is not a stron$ enou$h moti!ation for a cut)off point at the boundary between local a$reement and anaphoric pronouns. .espite the differences between local %or S$rammaticalT* a$reement and anaphoric a$reement %see L2.2.F*, inter)sentential a$reement often in!ol!es the same controllers, tar$ets, forms, properties, and conditions as intra)sentential a$reement. Siewierska %2008K420)424* obser!es thatK In most mainstream work on a$reement this notion includes within its scope the determination of the form of independent person markers more commonly called anaphoric pronouns. "he domain of a$reement is therefore not restricted to the clause, or e!en sentence, but may be a lar$er discourse unit such as a thematic para$raph. = further reason to admit anaphors as a$reement tar$ets is that they form a natural e3tension to #orbett:s =$reement <ierarchy, a lan$ua$e uni!ersal concernin$ the opposition between syntactic and semantic a$reement %L2.2.;*. #orbett and (arlow:s !iew will be adopted here. hile anaphoric a$reement thus Dualifies as a$reement under #orbett:s approach, it is not ScanonicalT a$reement accordin$ to the framework which has come to be known as #anonical "ypolo$y %e.$. #orbett 200Fa, 2005*. #orbett uses a number of criteria to characterise the ways in which !arious kinds of a$reement depart from a central canon %loosely defined as the best and clearest e3amples 2005K9*, and one of these criteria %#)4F* is that a Slocal domainT is more canonical than a Snon)local domainT. 2ne conseDuence of adoptin$ a fairly rela3ed !iew of the domain of a$reement is that in the case of non)canonical a$reement in!ol!in$ anaphoric pronouns, controllers can be a considerable distance from their a$reement tar$ets %pointed out by #orbett %2005K84* and 0er$uson and (arlow %4999K8)40**. <owe!er it does not necessarily follow that controllers can therefore operate at unlimited distances. "he further the distance between the tar$et and its antecedent, the harder it is to maintain that the form of the tar$et is controlled solely by formal properties of the antecedent word. Instead it becomes difficult to distin$uish between the influence of the controllin$ word, and the influence of the referent itself , in other words, there comes a point where anaphoric reference blurs into deictic reference. =s #orbett %4994K28;* puts itK
5 =ccordin$ to #orbett %2005K24* this is the only e3tended discussion of the issue.

5F

"he fact that there is no determinable ma3imum distance between antecedent and pronoun means that sometimes it is not fully clear which noun phrase is the antecedent of a particular pronoun. =nd when a pronoun is widely separated from possible antecedents it may be bein$ used deictically rather than anaphorically. More typically, of course, we think of deictic reference as introducin$ a referent into a discourse for the first time, as in the followin$ #icipu sentenceK %44* >ka)0taari @
+#4)stone

ka0mpa)
A-11t!is

this .stone2 Isaid while pointin$ to a stoneJ "his could be said without the noun, in which case the ka0 prefi3 on the demonstrati!e is %to a certain e3tent* determined by the $ender of ka)0taari , the basic)le!el term for the concept S"2+- %see /osch 4979 and the discussion in #orbett 4994K288*. It is not clear to me whether pronouns used in this way can be said to Sa$reeT, since there is no te3tual controller for them to a$ree with. "o be sure, pronouns tend to show the same Sa$reementT features whether they are used deictically or anaphorically %(resnan and Mchombo 4997K789)7F2*. <owe!er they do not satisfy Steele:s definition, since there is only one SelementT in!ol!ed, the Star$etT. +ote too that there is no Ssystematic co!arianceT here. =lthou$h the basic)le!el term for a concept undeniably influences the referrin$ e3pression that the speaker will choose, ne!ertheless referents and referrin$ e3pressions do not e3ist in one)to)one correspondence. 2ften there are synonyms with different $enders, and failin$ that, e!ery referent belon$s to a superordinate cate$ory, the name for which may also belon$ to a different $ender. 0or e3ample in #icipu a smallish stone %ka)0taari , ka0mpa) %this,
=G;, =G4*, +#4*

may be pointed out e3ophorically usin$ any one of


=G8,

ma0mpa) %this,

based on ma)0taari Spebble,

+#8T*,

yi 0 mpa) %this,

based on the hypernym i 0 ) ri Sthin$, +#;T* or perhaps most likely of all e0mpe) %this,

;1S*. "he actual term chosen will depend on the te3t)e3ternal conte3t as well as the speaker:s communicati!e purpose. #orbett %2005* does not directly address deictic use, althou$h his #)4 criterion %2005K40* states that Scontroller presentT is more canonical than Scontroller absentT, implyin$ of course that Scontroller absentT a$reement still Dualifies as a$reement. In a recent paper on #anonical "ypolo$y, (ond %2009* ar$ues that a strict definition %SbaseT* is reDuired to determine the domain of in!esti$ation for phenomena in!esti$ated in the #anonical "ypolo$y framework. It is not clear whether these two positions can be reconciled without loosenin$ the definition of a$reement too much. In L9.9 I will 55

discuss the possible rele!ance of the distinction between a$reement and dei3is to the #icipu $enderNperson alternation.

###

'echanism

"he traditional $enerati!e understandin$ of a$reement is that it is a redundant and asymmetric syntactic operation, where features are copied from the controller to the tar$et. <owe!er, as hi$hli$hted by (arlow %4992K22)8F*, certain e3amples cannot be properly accounted for by a feature)copyin$ model. 0or e3ample in /ussian the second) person pronoun has only one form ty, but tri$$ers different a$reements dependin$ on the addressee:s se3K %42* %a* ty Hital0;
2S read.1S")M=S#

%b* ty Hital0a

2S read.1S")0-M

fou were reading Isaid to maleJ

fou were reading Isaid to femaleJ I#orbett 4994K429J

"he only way round the problem is to posit two homonymous pronouns ty, one marked for masculine and the other for feminine. In the li$ht of e3amples such as these, theories such as Generalised 1hrase Structure Grammar and 6e3ical)0unctional Grammar ha!e de!eloped Sfeature)mer$in$T or Sfeature)unificationT models7. In such approaches, features may occur independently on both controllers and tar$ets. 1ro!ided the feature !alues are compatible %i.e. not conflictin$*, then the sentence will be $rammatical without reDuirin$ the proliferation of homonyms. So for e3amples %42a)b* the second) person pronoun would be marked for person but not $ender, whereas the !erb would be marked for $ender and not person. =t some point in the representation of the sentence the two structures are unified, $i!in$ the !alues for both person and $ender.

##%

0-ntactic and semantic agreement

Steele:s definition of a$reement Duoted abo!e states that a$reement can be with either a semantic or formal property of the controller. #orbett distin$uishes between
=G/--M-+" SB+"=#"I#

and

S-M=+"I# =G/--M-+"

%4994K22F)250, 2005K205)2;7* , tar$ets showin$ the

7 #hun$ %4999* is an attempt to model a unification approach to a$reement within the 1rinciples and 1arameters framework. Anfortunately much of the moti!ation for her account of a$reement relies on theory)internal e!idence. 0or e3ample, e!idence for the e3istence of her S=ssociate relationT depends on a particular analysis of the deri!ation of ES2 word order in #hamorro , the sub'ect is $enerated within E1 %the SInternal Sub'ectT hypothesis*, raises to Spec I1 in order to $et case, and then lowered a$ain to an ad'unct of the !erb in order to produce ES2 word order. /e$ardless of the ad!anta$es of this approach within the 1rinciples and 1arameters framework, it is hard to e!aluate it $enerally because it relies on many theory)specific assumptions.

57

former a$ree with a formal property of the controller, those showin$ the latter a$ree with a semantic property. +ouns which offer a choice dependin$ on the a$reement tar$et are called
<B(/I. +2A+S

%#orbett 4994K49;)498*. 0or e3ample in &iswahili a

number of kinship nouns belon$ to morpholo$ical noun class 9 in the sin$ular and 40 in the plural. =s e3pected, these tri$$er $ender 9N40 a$reement within the noun phrase. <owe!er when it comes to sub'ect and ob'ect a$reement, they tri$$er $ender 4N2 %i.e. p}in^Ynrqqiq^oYe^YopqY\qrjYqfiqr{YXWyXYnonijnYUWyXXWVY ^Yp[wpqn 42N4; %diminuti!e* and 7N9 nouns denotin$ humans can tri$$er either syntactic %42N4; or 7N9* or semantic %4N2* a$reement, with the latter becomin$ more likely the further the separation between the controller and tar$et %#orbett 4994K289)2F0*. It should be stressed that, in contrast to the two types of a$reement tri$$ered by these (antu hybrid nouns, the alternation between #icipu $ender and person a$reement does not reduce to the syntacticNsemantic a$reement distinction. #icipu does ha!e hybrid nouns e.$. ma0ga)ji ) SpriestT %L5.;*, but this in!ol!es two different feature !alues for $ender rather than two completely different feature paradi$ms.

##&

1ariation in agreement

=$reement in noun class lan$ua$es is usually taken to be a syntactic phenomenon, and as such it is $enerally !iewed as obli$atory , if constituents ha!e the possibility of a$reein$ with the noun, then they do. "his is the o!erall picture of $ender a$reement that emer$es from introductions to =frican lin$uistics %e.$. elmers 497;K4F9)49;, atters 2000*, which understandably present a SnormalisedT account to their readers, and are not able to focus on the difficulties posed by indi!idual systems. Grine!ald and Seifart %2008* ar$ue that typolo$ists without first)hand e3posure to =frican noun class lan$ua$es ha!e been influenced by such simplified models. In fact, e!en more $eneral definitions may reDuire a$reement to be obli$atory, as in =ikhen!ald %2000K20*K +oun classes are defined syntactically. "hey constitute a closed obli$atory $rammatical system. +e!ertheless there are a considerable number of lan$ua$es where $ender a$reement is not obli$atory. #orbett implicitly acknowled$es the e3istence of optional9 a$reement when he introduces the term SenforcedT $ender a$reement %4994K20;*K If a particular tar$et type can mark a$reement in $ender then in many lan$ua$es it must. = 6atin ad'ecti!e, which distin$uishes masculine, feminine and neuter %in a $i!en case and number combination*, must do soK the endin$ cannot simply be
9 In the sense of Ssub'ect to pra$matic constraintsT.

59

omitted. "his may be called SenforcedT $ender a$reement. SAnenforcedT, or optional, $ender a$reement is rele!ant for this thesis, especially 1art IE where I deal with the alternation between $ender and person a$reement in #icipu. In preparation for those chapters, this section considers the different kinds of !ariation in a$reement that are found cross)lin$uistically, and the conditions that tri$$er them.
##&1 Pre.re"uisites and conditions

0er$uson and (arlow %4999K8* discuss both #2+06I#" and E=/I="I2+ in a$reement. "wo or more patterns of a$reement may be in conflict with each other, such as when two con'oined pronouns tri$$er a$reement in the predicate, or a plural pronoun is used to politely refer to a sin$ular referent. If in a case of conflict more than one a$reement pattern is $rammatical, then this results in alternati!e possibilities, or !ariation in a$reement. #orbett %2005* uses the term #2+.I"I2+S to refer to the factors that determine which of the a!ailable a$reement patterns occurs. -3amples of #orbett:s conditions include animacy, precedence %i.e. word order*, topic, and focus. <e is careful to distin$uish between conditions and 1/-)/-_AISI"-S %2005K49;*K "he essential difference is that prereDuisites specify what is necessary for a$reement,...while conditions affect the use of an a$reement form where the prereDuisites are met. 1re)reDuisites for a$reement are found at all le!els of lin$uistic structure %for e3amples see #orbett 4994K4;;)4;F, 2005K79)9F, 49;)498*. 0or e3ample with respect to synta3, in some lan$ua$es a$reement may be limited to, or e3cluded from, certain constructions. In &abyle %+orth (erber, 6aoust 4929K80, cited in =ikhen!ald 2000K;9* demonstrati!es a$ree in $ender when used as pronouns, but do not when they are used as modifiers. #on!ersely ad'ecti!es in German a$ree when used as modifiers, but do not when they occur as predicates %#orbett 4994K428*. =$reement may also interact with the cate$ories of definiteness, tense, person, number, case %#orbett 4994* and animacy %Siewierska 2008K4FF, Gi!n 4975 for &iswahili*. =lso, within a particular word class, some le3ical items fail to show a$reement. 0or e3ample, in some eastern (antu lan$ua$es there is a subset of ad'ecti!es %Sin!ariant wordsT* that do not take concords %Maho 4999K405)407*, often borrowed from other lan$ua$es. In chapter 7 we will see e3amples of #icipu a$reement pre)reDuisites in!ol!in$ both phonolo$y and morphosynta3. 59

#onditions, on the other hand, seem to be limited to the syntactic, semantic, and pra$matic le!els of lin$uistic structure %#orbett 2005K49;)498*. -3amples in!ol!in$ pra$matics are problematic operationally, howe!er, since in the absence of e3plicit formal codin$ of pra$matic relations it can be difficult to tell whether !ariation in a$reement is best analysed as a prereDuisite or a condition. Anlike features such as case or $rammatical relations, it seems to be relati!ely rare to find an independent formal marker of topicality that can be counted on for e!idence when it co)occurs %or does not, as the case may be* with a particular a$reement pattern. #omrie %4999K274* writes My impression is that it is e3tremely rare across lan$ua$es to find a formal de!ice that literally, in one)to)one correspondence, encodes some pra$matic distinction...9, and as .ooley %2007K400* points out, e!en well)known candidates such as the Japanese StopicT marker wa often seem to be codin$ the setting of a discourse unit rather than anythin$ of intrinsic interest, whether a referent %topic* or other kind of theme. So when lin$uists say that a particular case of a$reement is linked to the topicality of the controller, what they often mean is that there is a correlation between the occurrence of a$reement on the one hand, and on the other what they percei!e to be the discourse salience of the controller referent from their sub'ecti!e understandin$ of how the speaker intended the hearer to construe the te3t. Asually there is no independent markin$ of topic which can be relied upon, and so in such cases the !ariation with respect to a$reement could eDually well be !iewed as a pre)reDuisite or a condition. Ander the former analysis, a$reement is not possible unless the controller referent is topical. Ander the latter analysis, a$reement is possible with non)topical controllers, but not as likely. #orbett %2005K497)499* cites the case of lon$)distance a$reement with topic in "se> %1olinsky and #omrie 4999, #omrie 200;* as a clear instance of an a$reement condition. "he lan$ua$e has two kinds of $ender a$reement, with controllers of !aryin$ syntactic relation , either strai$htforward $rammatical a$reement with the absoluti!e ar$ument, or lon$)distance a$reement with an absoluti!e ar$ument in an embedded clause, as in %4;* where the !erb iy SknowT a$rees with the +1 magalu SbreadT rather than the e3pected absoluti!e ar$ument of the main clause %which in this case would be the entire subordinate clause*.

9 See also 6ambrecht %4998K449*.

70

%4;*

eni0r

mother).=" boy)-/G bread:AB0:3

Iu7+0aJ

magalu

#6;)eat)1S"1/")+M6H

0aJ$0'ru0KiL

C;3)know)1/S

0iy0Mo

0he mother knows that the boy ate bread

I1olinsky and #omrie 4999K447J

"he situation can be dia$rammed schematically as in 0i$ure 47, where # stands for controller, S for a$reement slot, and 1 for the feature paradi$m in!ol!ed. 0or "se> there is only one paradi$m %$ender* and one a$reement slot on the !erb, but there are two competin$ controllers. P C1 C2 P +igure ,_- dchematic representation o the 0seH agreement system S Verb

=ccordin$ to 1olinsky and #omrie this lon$)distance a$reement must occur when the referent of the absoluti!e noun phrase is the main internal topic of the embedded clause, and can be interpreted as a si$nal that the main topic of the embedded proposition is e3pressed by the absoluti!e +1 %1olinsky and #omrie 4999K422*. It seems then that 1olinsky and #omrie understand clause topic in "se> as a prereDuisite %usin$ #orbett:s terminolo$y* for lon$)distance a$reement rather than a condition. "his analysis is supported by what happens when the absoluti!e +1 is o!ertly)marked with either of the two topic particles found in the lan$ua$e. =ccordin$ to 1olinsky and #omrie %4999K42F*K If the absoluti!e ar$ument in the embedded clause is marked by either particle, 6.= Ilon$)distance a$reementJ is the only option...<owe!er if some constituent other than the absoluti!e is marked by a particle as topic, 6.= becomes impossible. =lthou$h it is not in doubt that "se> shows !ariation in a$reement which is dependent on discourse properties, there is an operational problem of decidin$ between pre) reDuisites and conditions. .espite this kind of difficulty, I will make use of #orbett:s terminolo$y in the description of #icipu person a$reement in 1art IE, since it pro!ides a useful point of departure for the analysis. In L9.8 I will show that in addition to morphosyntactic and semantic conditions, there is also a pra$matic dependency on sub'ect and pronominal person a$reement in #icipu in!ol!in$ topicality. "his 74

dependency will be analysed as a condition rather than a prereDuisite, since there are occasions when the controller referent is clearly topical but person a$reement does not occur.
##&# 1ariation and topicalit-

6ike "se>, many other lan$ua$es show !ariation in a$reement dependin$ on discourse properties. In e!ery case of optional a$reement in!ol!in$ discourse pra$matics that I am aware of, the occurrence of a$reement is associated with topical, identifiable, or specific referents, or definite noun phrases, while the absence of a$reement is associated with non)topical, unidentifiable, or non)specific referents, or indefinite noun phrases. "his point has been made by a number of theorists and typolo$ists. =ccordin$ to =ikhen!ald %2000K;24* +oun class a$reement often occurs only if the noun is topical, or definite. (arlow %4992K94)92* makes the same point for a$reement in $eneral, rather than 'ust $enderK "he classification of an ob'ect in terms of its position on a definiteness scale by !irtue of a$reement is Duite common. In $eneral, hi$h definiteness induces a$reement, whereas low definiteness is associated with absence of a$reement. .alrymple and +ikolae!a %200FK74* writeK #onditions on !erb a$reement are often assumed to be definable in purely syntactic terms. In some lan$ua$es, howe!er, the !erb shows more a$reement with topical ar$uments than with nontopical ar$uments. "hese theoretical statements can be supported by data from se!eral unrelated lan$ua$es, as shown below.
##&#1 2ptional agreement

"he best)known case of !ariation in a$reement in +i$er)#on$o is the optional cross) referencin$ of third)person ob'ects on the !erb in many (antu lan$ua$es %e.$. (resnan and Moshi 4990, Morimoto 2002*. ald:s %4979* study on &iswahili ob'ect a$reement identifies definiteness, animacy, and topicality as rele!ant factors. (resnan and wpeijeYXY~q{wr[jqY{}jqwoYn^~YejqwoYq^~qrYnrqqiq^oY[^Yp[wpqnYn^eopqr (antu lan$ua$e, and pro!ide an analysis framed in 6e3ical)0unctional Grammar. }jqwoYnrqqiq^oY[^Yp[wpqnY[{Yejf[noergYn^~Y[{Yn^nfg{q~Yn{Ynij[}e}{Yjqovqq^
G/=MM="I#=6 =G/--M-+"

where the a$reement marker is a redundant affi3 e3pressin$ the


=+=1<2/I# =G/--M-+",

$ender of a co)occurrin$ sub'ect noun, and 72

in which case the

a$reement marker is an incorporated pronominal with full ar$ument status, anaphorically bound to its antecedent. 2b'ect a$reement, on the other hand, is only possible with topical referents encoded by non)ar$ument +1s, and is therefore analysed as pure anaphoric a$reement. e will return to (resnan and MchomboTs analysis in the discussion of the typolo$y of a$reement markers in L2.2.F. "opicality is also rele!ant with respect to sub'ect a$reement in (antu. .emuth and Johnson %4999* ar$ue that the sub'ect a$reement prefi3es in Setawana are anaphoric |re^ei[^nf{Yvp[wpYe^fgY[^~qYoe|[wnfYrqqrq^o{Y}{oYf[qYopqYp[wpqnYejqwoYinrqr{V Morimoto %2009* proposes that in some (antu lan$ua$es !erb a$reement is controlled by the most topical +1 in the clause rather than the $rammatical sub'ect. 0urther afield, est Greenlandic pro!ides an e3ample in!ol!in$ person a$reement. "he !erbal morpholo$y depends on the status of the ob'ectG if it is definite, then person a$reement is present with both sub'ect and ob'ect as in %48* %the affi3 is a fused marker*. <owe!er in %4F*, where the ob'ect is indefinite, the !erb a$rees with the sub'ect only. %48* tuttu
caribou see);.SGG;.SGGI+.I#

taku0aa

He saw the caribou %4F* tutu0mik


caribou)I+S"/ see);.SGGI+.I#

taku0vuN

He saw a caribou I0ortescue 4998K95, cited in (arlow 4992K92J .efiniteness also plays a role in $ender a$reement in =rabic %#orbett 4994K42F* and Garifuna %+orthern =rawak, Munro 4997K88F*, and in person a$reement in &ambera %=ustronesian, &lamer 2009*. In Garifuna the !erb always a$rees with the sub'ect, but only a$rees with definite ob'ects. Similarly in &ambera indefinite ob'ects are not cross) referenced by a$reement clitics %&lamer 2009K298*. 0inally, the referentiality of the a$reement controller can also be rele!ant. In the ^~e^q{[n^Yfn^}nqYq^]q^uYn{nYe^fgYrqqrq^o[nfY}^~qreqr{Yor[qrYnrqqiq^o on transiti!e or ditransiti!e !erbs %=ustin 2008K44*.
##&## Choice of controller

In the e3amples so far %with the e3ception of "se>* the !ariation consists of an alternation between the presence of a$reement in the case of a topicalNdefiniteNreferential controller, and the absence of a$reement in the case of a 7;

controller lackin$ one of these properties. In each case the a$reement is optional, in the sense that it is possible to find $rammatical sentences lackin$ the rele!ant a$reement morpholo$y. "he ne3t few e3amples in!ol!e a different kind of !ariation , there is a choice of a$reement, with the controller bein$ selected accordin$ to its topicality. "he positi!e correlation obser!ed in the theoretical statements $i!en abo!e still holds , in each case the topical referent is the controller. xpqYine^[n^Yfn^}nqYn}inrbYpn|in^Yn^~Yzqrjg{p[rqYXXY[{Y|nro[w}fnrfg interestin$ because of the relationship that holds between the occurrence of a$reement and the
.IS#2A/S- "21I#=6I"B

of the controller. "he lan$ua$e displays a system with two

different paradi$ms markin$ the sub'ect as well as a small subset of the ad'ecti!es, in!ol!in$ oppositions from two distinct $ender systems. 2ne of these systems is based around natural $ender i.e. se3, the other around shape and other structural features of the referent. #hapman and .erbyshire %4994* use the terms S$enderT for the se3)based system and Snoun classT for the shape)based system, but this is simply a matter of con!enience , both would be re$arded as $ender systems in #orbett:s terminolo$y40. n}inrb{ Y d^e}^ Y wfn{{h Y n^~ Y dq^~qrh Y nrqqiq^o Y {g{oqi{ Y }{[^ Y pn|in^ Y n^~ .erbyshire:s terms* are found with different sets of controllers, althou$h there are some en!ironments where they can both occur simultaneouslyK namely the S ar$ument of intransiti!e !erbs, the 2 ar$ument of transiti!e !erbs, possessed nouns and a small subset of ad'ecti!es. "he four possibilities are shown below, first for the numeral SoneT, analysed by =ikhen!ald as an ad'ecti!e, and secondly for an intransiti!e !erbK %45* kavina
howlerPmonkey.M=S#.+2+)&= one)A-<8A0C=

hoara0na

one howler monkey %47* jomahi hoara0ni

I3AO-, /4/0KAJ

'a$uar.0-M.+2+)&= one)A-<>78=

one %aguar

I0-M, /4/0KAJ

40 erjqoo Y Uj Y er[[^nffg Y n^nfg{q~ Y n}inrb Y n{ Y n Y e}rq^~qr Y {g{oqi Y we^fno[^ Y pn|in^ Y n^~ .erbyshire:s Snoun classT and S$enderT systems , presumably resultin$ in the !alues masculine, feminine, &=) marked, and a SneuterT $ender to co!er the remainin$ nouns. =ikhen!ald %n.d.* ar$ues that it is better to !iew them as two independent systems, $i!en that they can both occur to$ether on the same a$reement tar$et in separate morphosyntactic slots, and they ne!er occur indi!idually in the same morphosyntactic slot , a !iew later accepted in #orbett %2007K2F5)2F7*.

78

%49*

kasi'i

crocodile.M=S#.&=

ka0hoara0na

A-<3A=)one)A-<8A0C=

one crocodile %49* kaira ka0hoara0ni

IM=S#, KAJ
A-<3A=)one)A-<>78=

$ua!a.0-M.&=

one guava %20* voroni0'i0ha ada kahami

I0-M, KAJ
.-M.=GIM=S#J

fall)=S1)"<.=GIM=S#J

palmPtree.M.+2+)&=

the palmctree ell down %24* ka0voroni0'i0hi ida ojoro

IM=S#, /4/0KAJ
.-M.=GI0-MJ

A-<3A=)fall)=S1)"<.=GI0-MJ

turtle.0.&=

the turtle ell down

I0-M, KA, =ikhen!ald n.d.K40)44J

"he a$reement patterns in transiti!e clauses are particularly interestin$ since the two kinds of a$reement beha!e differently with respect to the ar$ument functions of the !erb. +oun class a$reement is Spurely syntacticT, whereas $ender a$reement is Spra$matically determinedT %=ikhen!ald n.d. 4;*. (riefly, in transiti!e clauses noun class a$reement is only found with ob'ect +1s. Gender a$reement, on the other hand, marks the Spi!otT of the discourse %Swhat the story is aboutT , =ikhen!ald n.d. 44*, and can be found with either the = or the 2 ar$ument. %22* makha
snake.M.+2+)&= 4S)A-<3A=)thinkPwron$)#2//)=S1)(?.A-<>78=

'o0ka0vaniha0mani0'i0hi

ida

.-M.=GI0-MJ

li>ard.0-M.&=

a'dana

' mistook the liHard or a snake

IAikhenvald n.d. ABJ

In e3ample %22* ba'dana Sli>ardT is the pi!ot, or discourse topic, and so the !erb a$rees with it in $ender throu$h the chi feminine suffi3. "he two !erbal prefi3es are syntactic rather than pra$matic a$reement, with the kac prefi3 a$ain a$reein$ with ba'dana. "he system can be represented as in 0i$ure 49. =$ain 1 stands for feature paradi$m and the dotted line indicates a choice of a$reement controller, dependin$ on which of #3 or #y is the pi!ot of the discourse.

7F

P2 Cy C P1 +igure ,`K7vVPD[><8V7?DX?DGD9<><8E97E\7<PD7w>Q[>?@7>C?DD[D9<7GOG<D[ S1 Verb S2

= similar situation obtains in the 1apuan lan$ua$e Motuna %2nishi 4998, cited in =ikhen!ald 2000K;8*, where a !erb a$rees obli$atorily in person with its sub'ect and ob'ect, but also in $ender with the topical constituent %which may be either the sub'ect or ob'ect*. 0inally, in the .a$hestanian lan$ua$e .ar$i, when there are two third)person ar$uments the !erb a$rees with either the a$ent or patient, whiche!er one is topical %!an der (er$ 4999K454)45F*.
##&#% Choice of feature paradigm

e ha!e seen topicality %and related notions such as definiteness and referential status* functionin$ as an a$reement condition in two distinct ways in the abo!e e3amples. 0irst, we considered e3amples of optional a$reement where a$reement occurred only with topical controller referents. Secondly, we looked at e3amples of a$reement where different controllers are selected dependin$ on their discourse topicality. = third kind of pra$matically)conditioned a$reement is found in se!eral &ain'i and 1lateau lan$ua$es. Instead of two potential controllers competin$ for a sin$le a$reement slot, this time the competition is between two different feature paradi$ms. "he phenomenon is discussed by <offmann %495;* and #ro>ier %4998* for the est &ain'i lan$ua$e #entral &ambari44. <offmann linked the #entral &ambari sub'ect $enderNperson alternation to the definiteness of the sub'ectK = sub'ect noun may be shown to be indefinite or definite accordin$ to whether there is class a$reement in the finite !erb or not %495;K459*. <e $a!e the followin$ e3amples %<offmann 495;K457)459*K

44 /ecall from L4.; that #icipu is the most distant member of the &ambari branch.

75

%2;*

%a* ma0nun
+#8)bird

u0kuwete
30)die

the bird died %b* ma0nune me0kuwete


+#8)bird A-4)die

a bird died %28* %a* ts0al


+#5)meatPstrip

u0riyete

30)fallPdown

the strip o meat ell down %b* ts0ale


+#5)meatPstrip

tse0riyete

A-+)fallPdown

a strip o meat ell down In %2;a* and %28a* where the sub'ects are definite, there is no $ender a$reement, and instead the $eneral third person sub'ect prefi3es %sin$ular u0 or plural a0* must be used. In contrast, in %2;b* and %28b*, which ha!e indefinite sub'ects, the $ender a$reement prefi3 must be used instead of the person prefi3es. <offmannTs treatment of this alternation is !ery brief, and it is natural to wonder if there are other factors in!ol!ed, particularly $i!en the awkwardness of the -n$lish translations in the absence of any conte3tual information. =ccordin$ to #ro>ier %4998K24F)222*, althou$h <offmannTs analysis works most of the time, definite sub'ects do sometimes tri$$er $ender a$reement. "he determinin$ factor accordin$ to #ro>ier is the
"21I#=6I"B

of the sub'ect

referent. If the sub'ect is topical, then it will tri$$er person a$reement, otherwise it will tri$$er $ender a$reement. "his is true to some e3tent for #icipu as well, but as we will see in 1art IE it is only part of the story. 2ne ob!ious Duestion for these lan$ua$es is if the two sets of a$reement markers occur in the same morphosyntactic en!ironment, why do we set up two systems at allM "his Duestion will be addressed in L7.9, but for now we 'ust stress that person)markin$ is independent of $ender %i.e. all nouns can potentially tri$$er ;1S a$reement on the !erb, no matter what their $ender*, and also that the pre)reDuisites for a$reement differ for the two systems. "he #entral &ambariN#icipu)type system is dia$rammed in 0i$ure 49K

77

P2 C P1 +igure ,b- dchematic representation o the 5entral 3ambari15icipu agreement system S Verb

=lthou$h this alternation between $ender and person a$reement is best documented for #entral &ambari, it is apparently present in each of the ma'or branches of est &ain'i. =$amalafiya %2007* pro!ides an annotated folktale in the 6ake lan$ua$e "sureshe. (oth the ma'or participants in this story are marked, in different places, with $ender and person a$reement. Gender sub'ect a$reement is concentrated towards the be$innin$ of the te3t, whereas person sub'ect a$reement is the norm towards the end. (oth appear to be ambi$uous a$reement markers. "owards the end of the story a third, minor participant is introduced. .espite bein$ Son sta$eT for appro3imately one third of the story, it is ne!er marked with person a$reement, only $ender a$reement. become entrenched as discourse topics. In the +orthwest lan$ua$e ut)Ma:in %also known as 0akai* Smith %2007K94, 98* e3plicitly states that there is an alternation between $ender) and person)marked sub'ect prefi3es and ob'ect pronouns for human referents. She also mentions that either markin$ strate$y may be used to track referents throu$h discourse, which su$$ests that whate!er $o!erns the alternation, it is not definiteness. =nimacy can also be ruled out, since the alternation is only possible in the first place if the referent is human and third person. "opicality is an ob!ious alternati!e candidate $i!en the patternin$ in the lan$ua$es already mentioned. "he (asa)&amuku lan$ua$e 1on$u %also known as "arin, Mac.onell 2007* also has separate $ender and person paradi$ms on person markers. It is not clear from the a!ailable description when $ender a$reement occurs and when person a$reement occurs, but one inno!ation with respect to the other est &ain'i systems is the e3istence of what seem to be portmanteau sub'ect a$reement prefi3es, marked for both $ender and person. So, for e3ample, in addition to the $eneral ;1S sub'ect prefi3 u0 and the $ender sub'ect prefi3es i0 and u0, jP) and wa0 are also possible, for class i0 and u0 79 e mi$ht hypothesise that referents are more likely to be encoded by person a$reement as they

nouns respecti!ely %Mac.onell 2007KF4*. So as well as the indi!idual possibilities, 1on$u seems to allow a combination of these e3ponents. "he alternation is not limited to est &ain'i, and can be found in both the -ast &ain'i lan$ua$e =mo %=nderson 4990aK4F7)4F9G see also #orbett 4994K287* and the 1lateau42 lan$ua$e &a'e %Mc&inney 4979*. In =mo the alternation applies to sub'ect SpronounsT %probably prefi3es*, ob'ect pronouns, and possessi!e pronouns. =ccordin$ to =nderson, the $ender)marked sets can be used with either human or non)human referents, but the person)marked sets are limited to humans. +o information is $i!en as to what $o!erns the alternation for human referents. In &a'e the alternation is found on at least the sub'ect prefi3es. 0urther afield, in the Grassfields (antu lan$ua$e +$yembn %=nderson 4990b* plural nouns with human referents can tri$$er either $ender or person sub'ect a$reement on !erbs. 0inally, this competition between different feature paradi$ms can also be found outside =frica. "he =ma>onian lan$ua$e Miran Y q[nro Y U Y pn{ Y n Y |nro[w}fnrfg comple3 pronominal system whereby pronouns can be inflected for either person, specific noun class, or $eneral noun class %see Seifart 200FK2F9 for a te3tual e3ample in!ol!in$ all three*. "his section has illustrated the cross)lin$uistic correlation between the presence of a$reement and topicality. It has $lossed o!er the fact that the notion of StopicalityT is notoriously slippery and means different thin$s to different lin$uistsG I will take up this issue in L2.2.F.4 and L2.;. 0rom a different perspecti!e, topicality has also been linked to the syntactic status of a$reement markers, and it is to this topic which we now turn.

##(

T-polog- of agreement markers

"he status of !erbal affi3es co)inde3ed with +1s has been the sub'ect of much research in recent decades, with Siewierska %2008K424)427* and #orbett %200;b, 2005K99)442* pro!idin$ helpful summaries. "erminolo$ically, the Duestion of how to refer to such affi3es in a theory)independent manner is !e3ed. SIncorporated pronounsT, Spronominal affi3esT and e!en Sa$reement affi3esT all come with their own theoretical ba$$a$e and none of these are neutral terms. In the description of #icipu a$reement in 1arts III and IE I will use the terms Sa$reement markersT or Sa$reement prefi3esT when discussin$ the $ender and person sub'ect markers, but it should be borne in mind that they share
42 /ecall from L4.; that 1lateau and &ain'i are co)ordinate branches of #entral +i$erian.

79

properties with what many people ha!e called Sincorporated pronounsT or Spronominal affi3esT. "he central Duestion with re$ard to the classification of a$reement markersNincorporated pronouns is how to handle apparently SoptionalT !erb ar$uments. In a Spro)dropT lan$ua$e such as Italian a sentence can be eDually $rammatical with or without the sub'ectK %2F* >lui@
;S.1/2 speak);S

parl0a

he speaks I#ook and +ewson 4995KF7J In the lon$er sentence lui parlca most lin$uists would analyse the pronoun lui as the sub'ect and the suffi3 )a as a non)referential ;1S a$reement marker. (ut what about the shorter sentence parlcaM =ssumin$ %as many $rammatical theories do e.$. 6e3ical) 0unctional Grammar, /elational Grammar, G(NMinimalism* that e!ery clause has e3actly one sub'ect, the sentence parlca must also ha!e a $rammatical sub'ect. 2ne way to satisfy this constraint is to posit the e3istence of a phonetically)empty but semantically)referential pro, and this is the approach taken in G(, and also in qf}{ theory of a$reement %499;K;82);8;*. "his empty cate$ory then bears the sub'ect function, and the !erb is said to a$ree with the empty cate$ory. <owe!er many theorists are not content with the proliferation of co!ert ar$uments entailed by the abo!e approach. Jelinek %4998* su$$ested a radical alternati!e analysis of pro)drop lan$ua$es which a!oided null ar$uments alto$ether. <er analysis was based on the =ustralian lan$ua$e lan$ua$e. 0irst, arlpiri, and set out to account for three properties of the arlpiri allows null anaphora. Secondly, it is a non)confi$urational

lan$ua$e with relati!ely free word order. "hirdly, the lan$ua$e tolerates a $reat deal of discontinuity of apparent SconstituentsT such as the sub'ect +1 %e.$. the ad'ecti!e can be separated from the head noun*. She $a!e a unified e3planation for all three of these properties by assumin$ that in sentences correspondin$ to the pro)drop !ersion of %2F* the affi3 is the sub'ect. In other words, the affi3 is referential and bears the $rammatical function of sub'ect. "he three properties mentioned abo!e are accounted for by analysin$ all +1s as ad%uncts rather than as ar$uments of the !erb. Jelinek then e3tended her analysis to unrelated lan$ua$es such as Spanish and Italian, and refers to such lan$ua$es as S )typeT lan$ua$es. 90

/e$ardin$ (antu, two of the three properties she seeks to account for throu$h this analysis are $enerally absent %discontinuous e3pressions, and to a lesser e3tent free word order*. +e!ertheless she su$$ests that (antu lan$ua$es may be )type lan$ua$es, since in such lan$ua$es a !erb and its affi3es may stand alone in a sentence, the Ssub'ectT and Sob'ectT +1s bein$ optional %Jelinek 4998K70*. If this is so, then it means that the apparent sub'ect +1 is ne!er a true sub'ect , instead it can only be an ad'unct to the !erb. <owe!er the situation in (antu is complicated by the asymmetry between sub'ect and ob'ect %(resnan and Mchombo 4997, (resnan and Moshi 4990*, and Jelinek:s analysis does not strai$htforwardly apply. In particular, the ob'ect affi3 %under Jelinek:s )type analysis this would be the only possible candidate for the ob'ect ar$ument* is always optional in (antu lan$ua$es and so cannot be relied upon to bear the ob'ect function4;. =s was mentioned in L2.2.8.2.4, (resnan and Mchombo %4997* pro!ide an nfoqr^no[\qYn^nfg{[{YeYopqYn^o}Yfn^}nqYp[wpqnY[^YopqYq[wnf}^wo[e^nfYZrniinr rniqverV Y [i[fnrfg Y oe Y nrf|[r[ Y n Y p[wpqn Y \qrj Y v[op Y [o{ Y n[q{ Y wn^ Y jq Y n Y }ff sentence on its own, or alternati!ely nominals may occur simultaneously with nrqqiq^oYn[q{VYnopqrYopn^Yeffev[^Yqf[^qYjgYn^nfg{[^Yp[wpqnYn{YnYog|q lan$ua$e whose !erbal affi3es are always the true ar$uments of the !erb, they propose a cross)lin$uistic typolo$y of a$reement markers ran$in$ from pure
M=/&-/S =+=1<2/I# =G/--M-+"

on the one hand %which, as in Jelinek:s analysis of

arlpiri clitics, may not co) which are non)

occur with an ar$ument +1*, to pure

G/=MM="I#=6 =G/--M-+" M=/&-/S

referential and co)inde3ed to the true +1 sub'ect %as in the -n$lish sub'ect a$reement marker cs*. #rucially, in between these two e3tremes lie
=M(IGA2AS =G/--M-+" M=/&-/S ,

which may take part in either anaphoric a$reement %without a sub'ect +1* or $rammatical a$reement %with a sub'ect +1*, and thus ha!e a dual function 48. "he three types of a$reement marker form a $rammaticalisation cline %Siewierska 4999*K %25* =naphoric a$reement marker s =mbi$uous a$reement marker s Grammatical a$reement marker

4; "his would be problematic in many theoretical frameworks. 0or e3ample in G(, it would !iolate the "heta #riterion %#homsky 4994K;5*, which stipulates that the theta roles of the !erb must always be realised in syntactic structure. Similarly in 60G the #ompleteness #onstraint would be !iolated. 48 "his is made possible by the dual structure of 60G , $rammatical functions such as SA(J and 2(J are independently specified in the f)structure %functional structure* of a sentence rather than %as is the case in #homskyan frameworks* bein$ deri!ed from the c)structure %constituent structure*.

94

6ike Jelinek, (resnan and Mchombo e3tend their analysis to more familiar Spro)dropT lan$ua$es such as Spanish and Italian, and the )a suffi3 from %2F* would be analysed as an ambi$uous a$reement marker. Siewierska %4999K229ff* obser!ed that the distinction betweeen anaphoric and ambi$uous a$reement markers is not always clear)cut, since they may beha!e differently dependin$ on properties of the controller. 0or e3ample, she predicts from the +ominal <ierarchy %Sil!erstein 4975* that there should be instances of person a$reement with an independent person marker but not a nominal +1 %Siewierska 2008K4F2*. "his kind of a$reement system is apparently rare, but does e3ist and Siewierska $i!es the well)known e3ample of elsh sub'ect a$reement. Siewierska also emphasises %e.$. 4999K2;4)2;8, 2008K427* the independence of a particular a$reement marker:s morpholo$ical and syntactic statuses. "he typolo$y in %25* is based on whether the a$reement marker can co)occur with its controller in the same clause, not on its form. So for e3ample, clitics can function as syntactic a$reement markers, while affi3es can function as anaphoric a$reement markers %see Siewierska 4999 for specific e3amples*. SubseDuent to the de!elopment of this typolo$y, =ustin and (resnan %4995* re!iewed Jelinek:s claims about arlpiri %and by e3tension, most =ustralian lan$ua$es* and presented e!idence su$$estin$ that the !erbal clitics should be re$arded as ambi$uous a$reement markers rather than, as Jelinek ar$ues, anaphoric a$reement markers. =mon$st other ar$uments, they obser!e that the ability of pronominal affi3es to co)occur with indefinite or non)referential controllers results in complications for her analysis that are not suffered by the 60G account %=ustin and (resnan 4995K2;8)2;F*. =ustin and (resnan:s contention that the ability to co)occur with indefinite and non) referential controllers is a rele!ant factor in decidin$ between anaphoric and ambi$uous markers has been debated in recent years %e.$. (aker 4995K42F)429, -!ans 4999N2002, Mithun 200;*. -!ans %2002K85*, whose basic position is in a$reement with =ustin and (resnan, $oes so far as to contend, contrary to both JelinekN(aker and the usual 60G treatment, that an a$reement marker should be called S$rammaticalT when it loses its referentiality %i.e. it can be co)inde3ed with non)referential +1s*, independently of whether it can occur in the absence of a nominal ar$ument. #orbett %200;b* pro!ides a number of tests which can be applied to a$reement

92

markers in order to place them on this typolo$ical scale, and I will make use of these %and others* in the classification of #icipu a$reement markers in L7.5. e will see that while there is no difference morpholo$ically between #icipu $ender and person a$reement %both are affi3es*, and they can both be re$arded as ambi$uous a$reement markers, they appear to be at different sta$es of the $rammaticalisation cline, with $ender a$reement further towards the S$rammaticalT end. (oth $ender and person a$reement markers can co)occur with indefinite and non)referential sub'ect +1s, which makes them $rammatical a$reement markers accordin$ to -!ans: usa$e.
##(1 Anaphoric agreement and topicalit-

In 1art IE we will see that one of the differences between the anaphoric use of $ender and person a$reement markers in #icipu is that the latter are associated with topical referents. "his, of course, implies that the former are less likely to mark topics. <ow can this be, if, as is often assumed, incorporated pronouns always link to topical antecedentsM In this section I want to Duestion this assumption and show that incorporated pronouns do not ha!e to be topical. "he #icipu data presented in 1art IE will support this claim. (resnan and Mchombo %4997* assume that incorporated pronominals are anaphorically linked to topics. =lthou$h their paper starts by showin$ that floatin$ oe|[w{Y[^Yp[wpqnY{q^oq^wq{Ywn^Ye^fgYjqYf[wq^{q~YjgYn^Y[^wer|ernoq~Y|re^e}^Yopqg also assume the re!erse condition holds i.e. that incorporated pronouns always anaphorically link to a topic %e.$. p7F7 when the SMIsub'ect markerJ is used for anaphoric bindin$, its antecedent within the sentence has the rarely Duestioned. "akin$ a step back for a moment, it is important to stress that there is no a priori lo$ical connection between the incorporation of a pronoun and the topicality of the
4F .emuth and Johnson %4999K28* (M hypothesi>e that incorporated pronominals are topic)orientedG that is, they can only anaphorically link to items fillin$ the "21 function, Ayechi %4994K8;9* "he topical nature of the sentence initial noun phrase of the I+a!a'oJ bi)form follows directly from the pronoun incorporation analysis*, #orbett %200;K499* , if the marker is an incorporated pronoun, it will be linked !ia anaphoric a$reement to the StopicT of the sentence, Morimoto %2002K298* , I(antuJ ob'ect markin$ on the !erb appears only when it is topic)anaphoric %like -n$lish pronouns*, Morimoto %2002K29F* , the ob'ect marker functions only as a topic)anaphoric pronoun, bein$ in complementary distribution with a clause)internal, non)topical ob'ect +1., and Grine!ald and Seifart %2008K2F4* , I+i$er)#on$oJ ob'ect inde3ation is $enerally more discourse dependent and applies only if the ar$ument correspondin$ to the syntactic function ob'ect is topicali>ed amon$st others.
"21

function, p758 the

2M:s, used for anaphora to a topic*. "his association has often been repeated 4F but is

9;

pronoun:s referents. =fter all, different types of independent pronouns !ary with respect to their discourse functions %see Siewierska 2008K57)78 for e3amples*. Instead, the moti!ation behind the putati!e restriction of incorporated pronouns to topics is said to be their participation in an opposition with independent pronouns. (resnan and Mchombo %4997K758)75F* put it like thisK qYpn\qY{qq^YopnoYp[wpqnYpn{YoveY{qr[q{YeYn^n|per[wY|re^e}^{yYopqY{Y}{q~ for anaphora to a topic, and the independent ob'ect pronouns, used to introduce new topics or for contrast of ar$uments. &ameyama 499F has obser!ed that all lan$ua$es ha!e two kinds of pronominals that can be used anaphoricallyK those used for reference reco!erable from discourse, and those used for Scontrast, emphasis, or focusT. z[qrq^oYfn^}nq{Yq^we~qYopqYoveY{qo{YeY|re^ei[^nf{Y[^Y~[qrq^oYvng{VY ^Yp[wpqn pronominals may be free or bound, whereas in -n$lish the difference is between stressed independent pronouns and unstressed dependent pronouns. Siewierska %2008* makes a similar point. She is actually more e3plicit about the association between incorporated pronouns and topicality than (resnan and Mchombo, and her conclusion is more wide)ran$in$, since she is dealin$ with dependent pronouns as a whole %includin$ unstressed free pronouns in -n$lish, for e3ample* rather than 'ust incorporated pronouns. 6ike (resnan and Mchombo, she hypothesises %this time on the basis of empirical e!idence* that incorporated pronouns cannot !ary as to their discourse functionG they always encode topical referents. "he rele!ant passa$e in her book %2008K57* is as followsK In contrast to the typolo$ies of person markers based on morphophonolo$ical form and syntactic function, the typolo$y of their discourse function is Duite underde!eloped and under)in!esti$ated. Moreo!er, in comparison to the other two typolo$ies it is also much more restricted in scope, since only independent markers appear to !ary with respect to their discourse function. eependent markers invariably encode re erents which are highly cognitively accessible and topical within the discourse. Imy italics , S.M.J "here is an implicit assumption in these analyses that there is a neat bifurcation between pronominals that denote new or contrasti!e topics and those that denote non)contrasti!e topics. <owe!er if the notion StopicT is to ha!e any independent e3planatory power, then it must be reco$nised that topics are more than 'ust non)contrasti!e S$i!enT referents. #onsider the followin$ -n$lish e3ample concernin$ a teena$er:s first date, especially the e3pressions in bold typeK

98

%27*

hile our niece and her youn$ suitor wandered around the store, she reached into a bin of mms and sna$$ed her fin$er on somet!ing. hate!er it was, it cut her. <er suitor, wishin$ to be a $entleman, e3pressed concern, but also told her he didnTt like blood. She tried to shield her fin$er from him, but when the depth of her wound finally freaked her out a little, she showed it to him. "he boy fell. <ard. 2nto his face.45

=fter bein$ introduced with the indefinite pronoun something, the mystery ob'ect is twice referred to in the subseDuent sentence usin$ a definite e3pression, the pronoun it. +e!ertheless it is hard to maintain that any part of the abo!e discourse is SaboutT the thin$ that hurt the $irl. 0ree pronouns can be anaphoric, definite, and sub'ects without bein$ topics, at least accordin$ to pra$matically)based definitions of topic %L2.;*. So if the it in it cut her is not a topic, why then is it pronominalisedM =s we will note in L2.;.2.F, there is a distinction between Srecent)reference mentionT and topicality. "he sub'ect referent of it cut her is pronominalised simply because it has 'ust been mentioned, not because there is any te3tual span for which it is a matter of standin$ interest or concern %Strawson 4974K97*. "he passa$es Duoted abo!e from (resnan and Mchombo and Siewierska both conflate topicality with somethin$ else. &ameyama:s %499F* obser!ation is not the same as (resnan and Mchombo:s , not all instances of reference reco!erable in discourse are references to topics. Similarly, Siewierska:s hi$hly co$niti!ely accessible and topical do not mean the same thin$. /eferents can be hi$hly co$niti!ely accessible and reco!erable in discourse without the interlocutors takin$ the sli$htest amount of interest in them. If free pronouns do not ha!e to be topics, we mi$ht wonder why it should be the case that all incorporated pronouns are topicsM "here do not seem to be any pre) theoretical reasons why this should be so. In fact, $i!en that the classification of an affi3 as a$reement marker vs. incorporated pronoun is rarely a black)and)white affair, we mi$ht e3pect typical incorporated pronouns to be less stron$ly associated with topical referents than typical free pronouns. "he followin$ e3ample from #icipu in!ol!es an Sindirect anaphorT %i.e. a definite referrin$ e3pression without a te3tual antecedent , see Schwar>)0riesel 2007* and shows that incorporated pronouns can be used to mark referents which are decidedly non)topical. -3ample %29* is an e3tract from a 4F)minute
45 "aken from a blo$ entry at httpKNNsonmislocuras.comNpa$eN2N.

9F

te3t about the &oran, which is firmly established as the main discourse topic at this point. (y contrast, the Mallam is SintroducedT for the first time here, with 'ust the ;1S a$reement marker u0 ser!in$ as the indirect anaphor. %29* I#onte3tK i , say, ''m lying...J n)1u)0'ungQo)sRo)
whenR30)rise#=ASsQ/6S likePthat 4S)carryPonPheadQ/6S likePthat 4S)0A")fallQI//

hi nde)

n)0tangu)

hi nde),

mu0u0yu)wo)

when he Ithe MallamJ li ts Ithe &oranJ up like this and ' carry it on my head, ' will all Itats00F.004.4;0J In #icipu culture it is a $enerally)known fact that there will be a particular Mallam associated with the mosDue, and it is this that makes the e3ample acceptable. "he sentence in %29* is the limit of the Mallam:s in!ol!ement in the te3t, and no part of the te3t can be said to be SaboutT him47. If we accept that incorporated pronouns do not ha!e to be topical, a second Duestion becomes rele!antK whether incorporated pronouns can !ary with respect to their discourse functions. In other words, does it make sense to talk of a typolo$y of incorporated pronouns with respect to their discourse functionsM #uly %2000* pro!ides an interestin$ account of two different third)person ob'ect a$reement markers in the e3tinct +orth =merican lan$ua$e "akelma %see also =issen 200;*. 2ne of them, which is always null %in contrast to the non)null first) and second)person forms*, is analysed as an Santi)topicalT incorporated pronoun %#uly 2000K90*. "he other ob'ect marker, 0 kSwa, marks para$raph)le!el discourse topics. -!en when there is a local %sentence* topic referent such as the crows in %29*, the 0 kSwa o je$t marker is still used rather than ;0 to denote the dis$ourse topi$ -hi$ken0Tawk. %29* ta)(le(lakhU, me(l
listenPto crows land

tSka( mi )Vs teMepe)Vn, ka


one sayPso

that

$'ipi )n0kSwa

addressPto)!im

Hei I#hicken)<awkJ listened to them%, the 5rows% covering the land said so, that speech they% addressed to himi I-uly 2000K79J #uly su$$ests that some incorporated pronouns may be sentence topics %e.$. +a!a'o bicG Ayechi 499449* while others are topics at the le!el of the para$raph %e.$. "akelma
47 "he fact that incorporated pronouns may be used to introduce discourse participants should not really be seen as surprisin$, since in some lan$ua$es new referents may e!en be ScodedT by >ero)anaphora %6i and "hompson %4979K;49);49* for #hinese, (ickel 200; for (elhare*, in which case the hearer must reconstruct the reference entirely by inference from conte3t. See also Schwar>)0riesel %2007*. 49 =lthou$h Ayechi uses the term .iscourse "opic, it is clear from the discussion that she is talkin$ about what 6ambrecht %4998* and others ha!e called Ssentence topicsT %L2.;.4*

95

0 kSwa* or e!en story. #omparin$ the function of different dependent markers across
lan$ua$es is tricky, howe!er. hen comparin$ the abo!e)mentioned analyses of +a!a'o and "akelma, it must be remembered that Ayechi was workin$ e3clusi!ely with what are almost certainly elicited sentences. (y contrast, #uly was usin$ te3ts with no possibility of elicitation. Similarly, #uly:s contention %2000K79)90* that in (resnan and Mchombo:s e3ample %;0* the corn is not topical is !alid, but only if we assume %unlike (resnan and Mchombo* that discourse topic is the rele!ant notion, rather than sentence topic. %;0* Wi *si
hyena ate

anadya $hi manga. A0ta0$hi 0 dya anapi ta ku Oan Wran$is$o


corn@7A he)S-/)it@7A)eat he.went to S.0.

0he hyena ate the corn. Having eaten it, he went to dan +rancisco. p[wpqnYrq{^n^Yn^~YwpeijeYXK789J hether researchers talk about Ssentence topicT or Sdiscourse topicT depends as much on the $rammatical framework in which they work and the data that they use as e!idence, than on any principled distinction between the two notions within a sin$le lin$uistic theory. "he analysis applied in 1art IE a!oid this problem by applyin$ the same methodolo$y to an intraclanguage case of !ariation in discourse function. "his proposal that topicality is an independent dimension rele!ant for the classification of incorporated pronouns conflicts with Siewierska %2008K57*, who as we noted abo!e uses the more inclusi!e term dependent markerK "here may be different dependent markers for intra) as opposed to inter)sentential antecedents, but there do not appear to be distinct dependent markers solely for different information statuses of their referents within the discourse. 0his ollows largely rom the act that languages tend to have only one type o dependent person marker or a given syntactic unction. Iitalics mine , S.M.J 6an$ua$es which do ha!e more than one type of dependent person marker for a $i!en syntactic function are of course an ideal testin$ $round for the claims made by (resnan and Mchombo, Siewierska, and others. e ha!e already seen two lan$ua$es with two types of dependent marker for the same syntactic function, "akelma %ob'ect* and #entral &ambari %sub'ect, L2.2.8.2.;*49. /ecall that #ro>ier %4998* analysed person sub'ect prefi3es as showin$ topic a$reement, in contrast to $ender sub'ect prefi3es. "hese lan$ua$es thus seem to pro!ide counter)e!idence to Siewierska:s statements, the scope
49 #uly %2000* su$$ests a further e3ampleK the +a!a'o Sfourth personT pronominals are %in one of their uses* story le!el topic markers, in contrast to the sentence topic marker bic.

97

of which is the dependent person marker %which includes $rammatical and anaphoric a$reement markers*. #icipu, of course, also has more than one type of dependent person marker for a $i!en syntactic function, and this, as in the case of "akelma and #entral &ambari, makes it possible for different series of dependent markers to be associated with different pra$matic functions. In 1art IE we will see that both $ender and person a$reement markers are capable of takin$ part in anaphoric a$reement, with the latter far more likely to mark discourse topics. "hus the #icipu data pro!ides not only counter) e!idence a$ainst Siewierska:s characterisation of dependent markers, but also support for #uly:s contention that topicality should be re$arded as an independent dimension alon$ which incorporated pronouns can be placed. It should be stressed that the essential distinctions in (resnan and Mchombo:s typolo$y remain unchan$ed e!en if we remo!e the stipulation that incorporated pronouns must denote topics. +othin$ in the theoretical introductions of (resnan and Mchombo %4997K78;)7F2* and (resnan %2004K488)489* reDuire this to be the caseG the latter does not e!en mention topicality. "he fundamental distinction between $rammatical and anaphoric a$reement depends on whether or not the a$reement markerN incorporated pronoun can co)occur with a sub'ect +1. In summary, we can distin$uish between two kinds of a$reement , $rammatical and anaphoric , and three kinds of a$reement markers , $rammatical, anaphoric, and ambi$uous , formin$ a $rammaticalisation clineK anaphoric s ambi$uous s $rammatical. "he classification of an indi!idual a$reement marker is often difficult since it may beha!e differently in different en!ironments. 0inally, althou$h there are syntactic constructions %e.$. sentences with floatin$ topics* which may reDuire the use of an incorporated pronoun in some lan$ua$es, it is not the case in $eneral that all incorporated pronouns must ha!e topical referents. Instead, topicality is an independent dimension alon$ which incorporated pronouns can be classified.

."

Topic
hile the

e saw in L2.2.8 that when there is !ariation in a$reement, it is often the topicality of the controller referent that determines which of the a$reement choices occurs. $eneral pattern is clear, in some of the analyses discussed abo!e there is not a $reat deal of discussion of what a StopicT is. Much of the recent work on a$reement and a$reement 99

conditions has been done by researchers workin$ in the 60G framework %e.$. .alrymple and +ikolae!a 200F*, who ha!e lar$ely adopted 6ambrecht:s %4998* treatment of information structure. #onseDuently topic is considered to be an attribute of the clause or sentence, and the ways in which topics can de!elop across e3tended discourse is often i$nored. =ikhen!ald %n.d.* is an e3ception in that she e3plicitly relates the occurrence of n}inrbY$ender a$reement on the !erb to the discourse Spi!otT , in her words what the story is about. Similarly !an der (er$ characterises topic as the SthemeT or Ssub'ectT, which the messa$e is about %4999K45;*. <owe!er neither author $oes into any more detail. 2ther researchers seem to treat topicality as the constellation of a number of factors such as animacy and definiteness, but with no independent properties. Siewierska %4998K224)222* makes the important distinction between Sinherent topicalityT and Sdiscourse topicalityT, notin$ that In theory any +1 may function as the topic...of the clause but that not all +1s possess the same inherent potential for functionin$ as the topic of the clause %see also =riel 4990 and Siewierska 2008K478ff e^Yq^o[ogY{nf[q^wgYn^~YZ[\]^YX`Yn^~Yeir[qYXXYe^Yoe|[wverop[^q{{VY ^pqrq^o topicality in!ol!es definiteness, animacy, person, and the semantic role of the referent, and is independent of how the speaker mi$ht wish the hearer to construe the situation bein$ encoded. .iscourse topicality, on the other hand, in!ol!es what .ooley calls
=""-+"I2+

mana$ement , which in!ol!es %2007K40*

e3pressin$ the speaker:s current interest in, and directin$ the addressees: attention to, particular concepts as a strate$y for construin$ their comprehension in a particular way. .ooley summarises the difference between inherent and discourse topicality as follows %2007K40;*K "he abo!e factors Iincludin$ e$ocentrism, anthropocentrism, indi!iduationJ indicate referents that are easy to percei!e as topics, hence are natural candidates for discourse topics. "hey predispose addressees to consider certain kinds of referents as possible topics. +e!ertheless, the speaker controls the construal of discourse topicality and can choose to construe as topic any of a !ariety of entities... .espite her useful distinction between inherent and discourse topicality, Siewierska:s %2008* treatment of !ariation in person a$reement is an e3ample of how, in practice, researchers can downplay the si$nificance of discourse when considerin$ topicality. 99

=fter statin$ that most of the properties of the controller rele!ant to !ariation in a$reement are those related to its inherent and discourse saliency, she then states that "he factors determinin$ the inherent and discourse saliency of the controllers are those comprisin$ the familiar topicality hierarchies %2008K489*, and $i!es the followin$ fi!e dimensions , person %4 s 2 s ;*, the pronominal !s. nominal distinction, animacy %animate s inanimate*, referential status %definite s indefinite*, and focus %not in focus s in focus*20. She does not, howe!er, mention discourse topic as an independently)!aryin$ dimension rele!ant to !ariation in a$reement. .iscourse topicality will be important in 1art IE when we discuss $ender and person a$reement in #icipu. ith this in mind, this section will discuss the notion of topicality at $reater len$th. "omlin et al. %4997* and .ooley %2007* pro!ide useful re!iews of !arious treatments of topic. "omlin et al. %4997K9;* distin$uish between Sclause le!el theme or topicT and Shi$her le!el para$raph or discourse themeT , the former is discussed in L2.;.4, and the latter in L2.;.2.

#%1

0entence topic

6ambrecht:s %4998* treatment of topic, and more $enerally of information structure %includin$ definiteness, identifiability, and focus*, differs from many other approaches in that it is intended to be amenable to formalism %6ambrecht 4998K 3!, 9)4;*, and to be compatible with the theoretical assumptions common to $enerati!e theories. #onseDuently 6ambrecht:s work has been endorsed by both SformalistT and SfunctionalistT theories of synta3 %e.$. 6e3ical)0unctional Grammar , (resnan 2004, /ole and /eference Grammar , Ean Ealin 200F*. 0ollowin$ Strawson %4974*, 6ambrecht defines a sentence topic as the thin$ which the proposition e3pressed by the sentence IS =(2A" %4998K449*, and $oes on to sayK

20 +ote that althou$h Siewierska %4998* refers to these properties as makin$ up SinherentT topicality, only animacy is strictly a constant property of the referent. "he others do in fact depend on the discourse conte3t. "he difference between such factors and SdiscourseT factors is that the speaker has much less control o!er the former. Gi!en a particular speech situation, and a particular point in a discourse, the speaker has no choice but to refer to herself in the first person. Similarly if a referent is in focus pra$matically, then the speaker must formally encode this referent usin$ a referrin$ e3pression which lies within the sentence:s focal domain. "he speaker has a lot more SchoiceT concernin$ $rammatical techniDues of attention mana$ement e.$. the decision whether or not to use a passi!e construction in order to keep the topical referent in the same ar$ument position. "his choice is referred to in .ooley:s Duote abo!e. 1ayne %4997K;89* di!ides up the factors which determine topicality usin$ a different cut) off point, separatin$ %i* truly inherent properties of the referent and %ii* conte3t)imparted properties %includin$ definiteness and person*.

90

= referent is interpreted as the topic of a proposition if I+ = GIE-+ .IS#2A/S- the proposition is construed as bein$ =(2A" this referent, i.e. as e3pressin$ information which is /-6-E=+" "2 and which increases the addressee:s &+2 6-.G- of this referent %4998K427*. "he followin$ passa$e from Strawson %4974K97* helps to e3plain this idea of SaboutnessTK statin$ Ii.e. makin$ statementsJ is not a $ratuitous and random human acti!ity. e do not, e3cept in social desperation, direct isolated and unconnected pieces of information at each other, but on the contrary intend in $eneral to $i!e or add information about what is a matter of standin$ interest or concern. "here is a $reat !ariety of possible types of answer to the Duestion what the topic of a statement is, what a statement is about...and not e!ery such answer e3cludes e!ery other in a $i!en case. "opics are usually sub'ects but this is by no means always the case. 0or the most typical readin$ of %;4*, the sub'ect would indeed be the topic, but this is not true for %;2*, where in the first clause the fronted ob'ect is the matter of standin$ interest or concern. %;4* he went to S#<226 %;2* 0A+ I 62E-, but too much fun is of all thin$s the most loathsome I illiam (lakeJ. Just as topics are not always sub'ects, neither are sub'ects always topics. #onsider the followin$K %;;* % hat did the children do ne3tM* "he children went to S#<226. %;8* % ho went to schoolM* "he #<I6./-+ went to school. I6ambrecht 4998K424J 2nly in %;;* does the predicate add information about what is a matter of standin$ interest or concern, and so only this e3ample can be said to ha!e topicccomment articulation. -3ample %;8* shows rather an identi icational articulation, since the sentence identifies a missin$ referent in the open proposition went to school. Such sentences do not ha!e topic e3pressions accordin$ to 6ambrecht. It is important to note that 6ambrecht:s definition means that the topic of a sentence cannot be determined without considerin$ the surroundin$ discourse. "he proposition has to be construed as bein$ about this referent in a $i!en discourse. 6ikewise for Strawson the topic referent has to be a matter of standin$ interest or 94

concern. +e!ertheless, while acknowled$in$ the necessity of considerin$ the discourse conte3t in determinin$ sentence topics, 6ambrecht deliberately restricts his treatment to sentence rather than discourse topics %4998K447*K I will restrict my attention to S-+"-+#- "21I#S or #6=AS- "21I#S. I will ha!e little to say about the notion of .IS#2A/S- "21I#, which has more to do with discourse understandin$ and te3t cohesion than with the $rammatical form of sentences... 0or 6ambrecht, discourse topics are outside of the scope of information structure, which is limited to psycholo$ical phenomena which ha!e morphosyntactic refle3es %4998K;*K -!en thou$h information structure is concerned with such psycholo$ical phenomena as the speaker:s hypotheses about the hearer:s mental states, such phenomena are rele!ant to the lin$uist only inasmuch as they are reflected in G/=MM="I#=6 S"/A#"A/-...Information structure is not concerned with psycholo$ical phenomena which do not ha!e correlates in sentence form. e will see in chapter 9 that the notion of discourse topic is in fact reflected in #icipu $rammatical structure, and does ha!e a correlate in sentence form. Specifically, inanimate and lower animate referents may tri$$er person %rather than $ender* a$reement on person markers, but only if the referent is a discourse topic %i.e. not 'ust a sentence topic*. 0or a more comprehensi!e analysis of the distribution of #icipu a$reement markers it is therefore benefical to appeal to this notion.

#%#
#%#1

$iscourse topic
34567

Z[\]^{ Y XW Y \ef}iq Y nw^evfq~q~ Y opq Y [i|eron^wq Y e Y ~[{we}r{q Y {or}wo}rq Y [^ determinin$ how SaccessibleT referents are at any particular time, and in it he and his collaborators sou$ht to correlate two sets of !ariables %499;K4;*K 4. "he $rammatical, Spurely lin$uisticT de!ices used by the speaker to code !arious topicsNparticipants in the discourse and 2. "he e3act position of those topics in the discourse, in terms of thematic para$raph structure, distance from last appearance, the clusterin$ with potential other interferin$ topics, persistence in subseDuent discourse topics. In one way this is similar to the task I will try to perform in 1art IE , to correlate lin$uistic codin$ de!ices %le3ical +1s, $ender a$reement and person a$reement* with 92

properties of the discourse. +ote howe!er that all of the discourse properties considered jgYZ[\]^YnrqYerinfYYopqgYnrqY|re|qro[q{YeYopqYte3t itself rather than of the conceptual {or}wo}rq{YopnoYnrqY[^\ef\q~Y[^YopqY|re~}wo[e^Yn^~Ywei|rqpq^{[e^YeYopqYoqoVYZ[\]^Y[{Ye course aware of the importance of such conceptual structures %499;K42*K In particular, we ha!e attempted to assess the more concrete and readily measurable factors %a* ISlengt! o' absence 'rom t!e register TJ and %b* IS"otential inter'erence 'rom ot!er to"icsTJ. "he fact that is is not yet possible to Duantify ri$orously factors %c* ISavailability o' semantic in'ormationTJ and %d* ISavailability o' t!ematic in'ormationTJ in spite of their undeniable importance creates a de$ree of indeterminacy in the results, so that correlations between $rammatical de!ices and particular measurements appear to be less than cate$orical. It is still not yet possible to Duantify ri$orously the a!ailability of thematic information at a $i!en sta$e of a discourse, and it is difficult to ima$ine how this could e!er be done. +e!ertheless topicality cannot simply be reduced to referential density , referential density is neither a necessary nor sufficient condition for a referent to be a topic, at least for a conceptual definition of topic in!ol!in$ SaboutnessT. 2n the one hand, discourse topics may inte$rate a para$raph %in the sense that the sentences in the para$raph are construed to be related to that topic* without the topic bein$ e3plicitly mentioned, for e3ample when situations are reported from the point of !iew of one of the discourse participants %see .uchan et al. 499F*. 2n the other hand, a hi$h de$ree of referential density does not make that referent a topic, as we will see shortly.
#%## $oole-

.ooley %2007* attempts to brin$ to$ether insi$hts from se!eral different theories, includin$ #o$niti!e 6in$uistics %e.$. 6an$acker 2004, 0auconnier 4997*, !arious other functional)typolo$ical approaches %6ambrecht, #hafe, and other authors includin$ Gi!n and "omlin*, and cross)lin$uistic discourse analysis %e.$. Grimes 497F, 6on$acre 4995, .ooley and 6e!insohn 2004*. .iscourse topics in .ooleyTs framework are defined accordin$ to their function in a
.IS#2A/S- S1=#24

, which is the mental

representation the addressee tries to construct, incrementally and throu$h trial)and)error, in order to make sense of what he percei!es the speakerTs messa$e to be. .ooley assumes that, amon$st others, the followin$ elements are always contained in a
24 "he term Sdiscourse spaceT e!okes 0auconnier:s work on mental spaces %e.$. 0auconnier 4998*, but .ooley:s term is more inclusi!e.

9;

discourse space %2007K9*K /eferential entities and propositions relatin$ them =cti!ation states %as defined by #hafe, see below* =n o!erall structure that accounts for the conceptual unity of the different components of the space %called a .IS#2A/S- S#<-M=* "hemes %includin$ topics*

.ooley hypothesises that the spaces constructed by an addressee in response to a section of discourse correspond to .IS#2A/S- A+I"S, the lowest)le!el of which is the 1=/=G/=1< %to be understood independent of modality*. <e defines topic as follows %2007K74*K %;F* If a discourse unit is construed in such a way that its IdiscourseJ space is thematically inte$rated around a referent22 , that is, if the rele!ance of each of the steps in its schema is percei!ed as dependin$ on its relation to that referent and if that relation manifests as well an intrinsic interest in that referent on the part of the speaker , then the referent is called the "21I# of the discourse unit Imy italicsJ. (ut what e3actly does it mean for a discourse space to be inte$rated around a referentM hen an entity is introduced into a discourse, it brin$s with it %or, in #hafeTs
.2MI+I2+,

%4998* terms, makes semiacti!e* its

defined by 6an$acker %2000K47;)478* as

the conceptual re$ion %or the set of entities* to which Ia particular referential entityJ affords direct access. .ooley %2007K;;* notes thatK if the entity continues to be referred to throu$hout a si>eable section of a discourse, the concepts in the dominion, includin$ new ones as they are added, tend to be !iewed in relation to the referential entity. In this way the $rowin$ dominion comes to be inte$rated by the referential entityK its component elements are !iewed in relation to that entity. "omlin %4997K99* puts it like thisK Generally, in a stretch of connected discourse, one referent emer$es as central, or the one that the propositions in the discourse are about. I ha!e found .ooley:s fi$urati!e illustration a helpful one , the referent be$ins to SattractT other concepts in the discourse space somewhat as a ma$net attracts iron filin$s, so that they Spoint in its directionT %2007K74*. .ooley:s characterisation of topic is similar to 6ambrecht:s in some ways, and it is useful to compare the two theories and the predictions that they make. "he SaboutnessT relation is present in both approaches, but the scope is different, as shown in 0i$ure 20.

22 "his definition differs from #hafe:s %4998K424*, in that it is restricted to referential entities. #hafe:s discourse topics %and =sher:s e.$. 2008* correspond to what .ooley calls SthemesT.

98

!orma" C"ause

Conceptua"

#opic e pression

$e%erent

()bout(

Proposition

!orma" Paragraph

Conceptua"

&iscourse'menta" space P(graph schema ()bout( #opic e pression $e%erent

+igure 4a- 'n ormal comparison o Lambrecht's sentence topic .above2 and eooley's discourse topic .below2 =ccordin$ to 6ambrecht, a referent is a topic if the proposition encoded by the clause is construed as bein$ SaboutT that referent. 0or .ooley a referent is a topic if the conceptual schema that we ha!e to construct to process the paragraph is inte$rated by that referent %i.e. each step in the schema is construed as bein$ SaboutT that referent*. "he two approaches differ as to the status of referents with a short te3tual span. #onsider %;5* below where the $randma is only a transient participant in the discourse. %;5* I#onte3tK = Jewish $randfather has been talkin$ about the fact that his $randson is difficult to please. <e $i!es one e3ample , outmealJ =nd it:s uh $ot $ood taste, its $ood. =nd the cereal , grandma e don:t like cereal but s!e finished to the last %dish* and I en'oy , I like it too. It:s tasty? =nd I uh...<e didn:t want the cereal, doesn:t eat. I said "odd, it wouldn:t kill ya, taste it? I6ambrecht 4998K489J "he e3tract as a whole is clearly about cereal, and more $enerally the complainin$ $randson. 6ambrecht also analyses the $randma as a topic because the two propositions concernin$ Snot likin$ cerealT and Sfinishin$ it all upT are about her. <owe!er for .ooley 9F

the $randma would not be a topic, because she fails to inte$rate any te3tual span wider than a sin$le sentence. e will see in chapter 9 that sentence topics of this kind are much less likely to be inde3ed with person a$reement than true discourse topics. In .ooley:s definition of discourse topic $i!en abo!e, the additional Dualification that the speaker has an Sintrinsic interestT in the topic referent is crucial. hile talkin$ about somethin$ does increase the likelihood of that referent becomin$ a para$raph topic, it is not sufficient. #onsider the followin$ e3ample of a newspaper report on a $ame of =merican football2;K ...<offman cau$ht t!e ball. <e passed it to &in$, who punted B for fifty yards. Sed$wick cau$ht B, and on the kick)off Stacy $ot t!e ball. 0he New fork 0imes, 20th 2ctober 499F "his te3t consists of a seDuence of fi!e clauses, each of which contains a reference to the football. Moreo!er, one referential e3pression is an unstressed pronoun, and two of the references e!en in!ol!e Ssub)minimalT %i.e. null* codin$. +e!ertheless the narrati!e could not concei!ably be construed as bein$ SaboutT the ball. "he ball semantically inte$rates the te3t but does not thematically inte$rate it. =ccordin$ to se!eral theories of oe|[w Y[^wf}~[^Y Z[\]^YXW Ysee also .ooley 2007K49)22* the ball in this e3ample would Dualify as topical. If we want to e3clude such referents, then it must be reco$nised that discourse topicality in!ol!es more than 'ust referential density and semantic inte$ration. "he speaker must also ha!e an intrinsic interest in the topic referent.
#%#% 8ormal signals of discourse topic

.ooley notes that there is both conceptual and formal e!idence for discourse structure, and that "he hi$her the le!el of discourse or$ani>ation, the more the addressee can e3pect to depend on conceptual rather than formal si$nals %2007KFF*. "hus .ooley is in a$reement with the Duotation from 6ambrecht $i!en abo!e, but it is important to note that there are e3ceptions to this trend. #hafe %4998K99)99* presents e!idence for the lin$uistic rele!ance of what he calls referential importance i.e. the importance Iof referentsJ to the sub'ect matter bein$ !erbali>ed %4998K99*. =ccordin$ to #hafe, referents of primary importance are typically mentioned more freDuently and may be
2; See "omlin et al. %4997K5;* and .ooley %2007K24)22* for discussion of a similar te3t in!ol!in$ an ice) hockey $ame. My interpretation of these kinds of passa$es with respect to topicality follows .ooley rather than "omlin et al.

95

introduced into the te3t in a specific manner, especially for more structured $enres such as traditional narrati!es. I ha!e already mentioned "akelma and +a!a'o %L2.2.F.4* where special incorporated pronouns ha!e been said to mark discourse topics. Grimes: %4979* !olume contains a number of papers on Stheme oriented referential strate$iesT, which some lan$ua$es use to mana$e reference in terms of a t!ematic policy, in which one referent is distin$uished from the rest when introduced, and a special set of terms refer to it no matter how many other thin$s ha!e been mentioned more recently %Grimes 4979K !iii, emphasis in ori$inal*. .ooley and 6e!insohn %2004K449)42;* call this the SEI1T %!ery important participant* strate$y. "hey note that the strate$y can e3tend across entire te3tsK ...patterns of reference sometimes make it necessary to reco$ni>e one IparticipantJ as G62(=6 EI1. =fter bein$ introduced, the $lobal EI1 is often referred to by minimum, but !irtually constant, codin$ %2004K424*. "he (antoid lan$ua$e Mambila uses such a strate$y in folktales %1errin 4979*. "he main participant, once introduced, is referred to either by >ero %when functionin$ as sub'ect* or by the ;1S pronoun jR7otherwise, with le3ical re)mentions bein$ !ery rare. "he potential for ambi$uity is minimised because, with a few e3ceptions, 1articipants other than the main one are re)identified by a noun e!ery time they are mentioned %1errin 4979K444*. "his So!ercodin$T of non)topical participants mi$ht be thou$h surprisin$, but the same strate$y is used in at least two other =frican lan$ua$es, the Grassfields (antu lan$ua$e (abun$o and the 2motic lan$ua$e Gimira %Siewierska 2008K492)49;*. .ooley and 6e!insohn %2004K422* also analyse the thematic Spi!otT in {eiqYeYopqYrn[f[n^YrnvYfn^}nq{Yn{YnY Y{qqYUVUVaVUVUYe^Yn}inrbVY n chapter 9 I will ar$ue that the #icipu $enderNperson a$reement alternation cannot be e3plained without an appeal to discourse topicality, and that con!ersely pro$ression from $ender to person markin$ can be !iewed as a si$nal of discourse topicality.
#%#& Paragraphs

"he definition of topic in %;F* relies on the notion of Sdiscourse unitT, of which a para$raph is the lowest)le!el !ariety. .iscourse analysts are $enerally a$reed that the production and comprehension of te3ts in!ol!es the construction and reco$nition of structural units lar$er than a sin$le sentence, yet smaller than the whole te3t. =ccordin$ to "omlin et al. %4997K55*K 97

.iscourse is neither flat not linear in its or$ani>ationG it is hierarchical, with clauses formin$ hi$her)order structures, para$raphs, which in turn combine to form lar$er episodes or sections of discourse. -!idence for the e3istence of the para$raph can be either formal or conceptual. 0ormal cues of para$raph structure include hi$h pitch at the start of the initial sentence %6ehiste 4979*, lon$er pauses at para$raph boundaries %#hafe 4990*, and the pre)boundary presence of laryn$ealisation %&reiman 4992*. "he e3istence of the para$raph is also supported by at least three kinds of conceptual e!idence which con!er$e on the para$raph as a minimal unit of discourse for which the speaker can construct a coherent discourse space %see discussion in .ooley 2007K;0*. 0irstly, a para$raph is a unit of speech or writin$ that maintains a uniform orientation %<inds 4979K4;5*. #hafe %4998K429)9* identifies four orientation ~[iq^{[e^{Y{qqYnf{eYn^Yz[YXXyXYZ[\]^YXWyW`Y^VYay "he fact that consciousness cannot function without bein$ oriented in space, time, society, and on$oin$ back$round e!ents e3plains the characteristic pro!ision of what is usually called a setting as a narrati!e be$ins. %#hafe 4998K429* 2rientation dimensions are freDuently updated at the start of a new para$raph , sometimes this boundary is e3plicitly indicated with an initial constituent functionin$ as a mental Sspace)builderT %0auconnier 4997K80)84*, as in "he followin$ day,.... Secondly, e3perimental e!idence %e.$. Gernsbacher 499F* has shown that the ability of hearers to recall the Ssurface informationT %e.$. word order* of sentences is ad!ersely affected once they ha!e crossed a para$raph boundary. Gernsbacher %4998* demonstrated the same effect on sub'ects: recall of thematic information, and her e3planation is that the mental representation of the pre!ious para$raph Duickly decays %or perhaps is suppressed* as a result of the processin$ shift in!ol!ed in settin$ up the new discourse space for the new para$raph. #hafe %4998* e3presses this chan$e in terms of the
=#"IE="I2+ S"="AS

of entities. =n

=#"IE-

concept is one which is presumed to be in

addresseeTs consciousness. = S-MI=#"IE- concept is one that is in the addresseeTs current mental representation, but whose acti!ation le!el has decayed %the notion of acti!ation status is a scalar one*. 0inally, an
I+=#"IE-

concept is one that is newly introduced by an

utterance. = conseDuence of the processin$ shift identified by Gernsbacher is that pre!iously acti!e or semi)acti!e referents become less acti!e once a para$raph boundary is crossed, and therefore reDuire e3tra lin$uistic codin$ %see below*. #hafe %4998K4;9*

99

su$$ests that semiacti!e consciousness is limited to the amount of information !erbali>ed in a narrati!e schema, so that when more information is added it may ha!e to be di!ided i.e. into para$raphs. 0inally, there is also e!idence for para$raph boundaries from e3periments which test metalin$uistic 'ud$ements. It has often been shown %e.$. 6ehiste 4979, Gernsbacher 499F, Ji 2002* that sub'ects percei!e para$raph boundaries in a stretch of discourse, and that their 'ud$ements concur to a certain de$ree with each other. "hese di!isions are correlated with boundaries established by independent means %e.$. prosody in 6ehiste 4979, a!ailability of surface information in Gernsbacher 499F*.
#%#( Coding weight

Z[\]^YXWYn^~Yeopqr{ Y%e.$. =riel 4990* ha!e obser!ed that as the accessibility of a referent %and hence its a!ailability as topic* decreases, so the Scodin$ wei$htT of the rqqrr[^Yq|rq{{[e^Y[^wrqn{q{VYZ[\]^YXWyXY[\q{YopqYeffev[^Y{wnfq28K more accessible topics Ili$htJ >ero anaphora unstressedNbound pronouns %Sa$reementT* stressedNindependent pronouns full +1:s

less accessible topics Ihea!yJ

+igure 4,K7x8Ay9zG7{|}~7XPE9EBEC8V>B7VEM89C7D8CP<7GV>BD =t any $i!en point in a discourse, minimal codin$ %i.e. the minimum that will enable the hearer to identify the referent* is the norm. SAndercodin$T , the use of less than sufficient codin$ wei$ht , is, unsurprisin$ly, $enerally infelicitous, and it does not $enerally occur e3cept for stylistic effect. 2!ercodin$ can occur for at least three reasons2F. "he first relates to what .ooley %2007KF;, fn. FF* calls Ssyntactic opportunismT , the syntactic slot for a referential e3pression pro!ides an ideal opportunity to shoehorn in some additional information about the referent. In such cases the codin$ choice has a secondary discourse function which has little or nothin$ to do with the accessibility of the referent. Secondly, o!ercodin$ may occur when there are competin$ referents and therefore dan$er of confusion %as noted by Z[\]^YXWyXa. "hirdly, o!ercodin$ may occur at the start of a
28 xp[{Y[{Y}{oYopqY|pe^efe[wnfY{wnfqVYZ[\]^YXWyXY[~q^o[[q{Y{q\qrnfY~[qrq^oY{wnfq{YY{qqYnf{e =riel %4990* and Siewierska %2008K478ff*. 2F See (olin$er %4979* for others beyond those discussed here.

99

new para$raph. .ooley %2007K52G see also 6i and "hompson 4979, =nderson et al. 499;* notes thatK +ominal elements, such as participants in a narrati!e, are also $enerally updated by means of a noun phrase at the be$innin$ of a new para$raph, e!en if recent mention in the precedin$ para$raph would indicate that they are acti!e or semi) acti!e. /eally, howe!er, only the first of these should be considered to be true o!ercodin$, since in the second case while the referents concerned may both be hi$hly)accessible, neither is sufficiently more accessible than the other to allow the use of minimal codin$. "he o!ercodin$ in the third case can be e3plained if we assume, as Gernsbacher:s research su$$ests, that the acti!ation statuses of referents are reset at para$raph boundaries. "his can be looked at in two ways , either the fact that we ha!e a new para$raph %si$nalled by the other formal and conceptual cues we ha!e noted abo!e* $i!es rise to o!ercodin$, or as 0o3 %4997K459* saysK many full +1s which occur in narrati!es where one could ha!e e3pected pronouns are functionin$ to si$nal the hierarchical structure of the te3t... to demarcate new narrati!e units. It is important to be clear that, within a para$raph, referrin$ e3pressions do not always mo!e strai$ht from an +1 to minimal codin$. "here may be a codin$ Spro$ressionT within the para$raph throu$h se!eral sta$es of the codin$ wei$ht hierarchy. =ccordin$ to .ooley %2007K99*K -specially for hi$h)le!el topics, their establishment as topic commonly $oes beyond the initial introduction per se. 0or some topics, the introduction sentence is followed by another sentence with an o!ercoded reference to the topic.... <e $i!es e3amples from &oine Greek and Mby- Guarani in!ol!in$ a pro$ression from le3ical +1 s demonstrati!e pronoun s minimal codin$. Similarly <inds %4998K855* indicates a similar seDuence for topic establishment in JapaneseK e see a three step pro$ression in the identification of participants in a narrati!eK %4* the participant is introduced with the particle gaG %2* the participant is referred to with a topical noun phrase marked with waG and %;* the participant is referred to by ellipsis. Similar pro$ressions seems to e3ist in #hinese %from le3ical +1 to pronoun to ellipsis , 6i and "hompson 4979*, the "ibeto)(urman lan$ua$e Jirel %Maibaum 4979*, and the =damawa lan$ua$e 6on$uda of northeastern +i$eria %+ewman 4979*. "his idea of a 400

codin$ pro$ression will be important for the analysis of #icipu in chapter 9 when we turn to the function of $ender and person a$reement in discourse. #han$es in codin$ wei$ht as a discourse pro$resses therefore pro!ide important e!idence for para$raph structure. 2f course, para$raph structure does not ha!e a deterministic relationship with codin$ pro$ressions , as we ha!e 'ust noted, there are other reasons for o!ercodin$ and undercodin$. In particular, the use of minimal codin$ for a referent within a para$raph is not sufficient e!idence for it to be analysed as a discourse topic. =s .ooley %2007KF4* points out, minimal codin$ is also used for non) topical Srecent)reference centres of attentionT %e.$. %27* abo!e*. Some of 6ambrecht:s Ssecondary topicsT %4998K487)4F0* are simply recent)reference mentions, rather than discourse topics, and in fact this notion is necessary to pre!ent the proliferation of topics that are not matters of standin$ concern.
#%#9 Access and integration functions of topics
=##-SS

.ooley %2007K72)7;* points out that topics ha!e both

and

I+"-G/="I2+

functions,

somethin$ that is not always appreciated in the literature. "he first mention of a topic pro!ides access to the mental space reDuired for comprehendin$ what is said about itG at this point in its life history %6an$acker 2000K498* the topic has hi$her informational salience, reflected in formal si$nalsK hea!ier codin$ wei$ht, its own intonation contour, bein$ set apart linearly from its Sfocus domainT %.ooley 2007K77)79*. .ooley calls such topics Smarked topicsT25. 2nce a topic has been introduced, then its function is to integrate the discourse space, 'ust as was described abo!e. hen carryin$ out this function topics are typically SunmarkedT, and e3pressed by minimal codin$.

#%%

0ummar-

In 1art IE we will obser!e the usefulness of the notions discussed in this section for describin$ $ender and person a$reement in #icipu. "he lin$uistic codin$ of topics in #icipu discourse reflects para$raph structure in two ways. 0irst, as in many other lan$ua$es %perhaps all*, the resettin$ of the acti!ation le!el at the be$innin$ of each para$raph is reflected by increased codin$ wei$ht. Secondly, and more interestin$ly, there is an intra)para$raph codin$ pro$ression from le3ical +1, to $ender a$reement, to
25 Z[\]^YXWyYinq{YnY~[{o[^wo[e^Yjqovqq^Ydwpn[^[^[o[nfhYoe|[w{Yn^~Ydwpn[^iq~[nfhYoe|[w{VYxp[{Y[{Yn formal correlate to .ooley:s conceptual distinction between the access and inte$ratin$ functions. 6ambrecht:s %4998K4;4)4;5* use of the terms SmarkedT and SunmarkedT topic is Duite different and not to be confused with .ooley:s.

404

person a$reement.

e will see that the notion of discourse topic accounts for the

distribution of the two kinds of a$reement markers in a way that sentence topic cannot. 0urthermore it will be shown that intrinsic interest is important as opposed to mere referential density , it is not the case that a referent will become topic)marked simply by mentionin$ it enou$h. Instead the speaker must ha!e an intrinsic interest in that referent. I will finish this chapter with a Duote from .ooley %2007KF5* which bears on the nature of the implication of this study for lin$uistic theory. "he claim that discourse has hierarchical or$ani>ation which is %partially* si$nalled by lin$uistic means does not imply that this or$ani>ation is part of $rammar... "here may be and commonly are formal si$nals which help the addressee reco$ni>e discourse units, but in $eneral discourse structure is not susceptible to 'ud$ements of S$rammatical !s. un$rammaticalT in the same way that lower)le!el phenomena are. Jud$ements of discourse or$anisation are more often in such terms as Sclear !s. unclear or confusin$T, Swell)put !s. clumsily)putT, etc... If 'ud$ements of S$rammatical !s. un$rammaticalT reflect $rammatical well)formedness, then in lar$e part discourse or$ani>ation is not a matter of $rammar. So the formal si$nals we commonly obser!e in discourse must often be seen as lin$uistic correlates of what is fundamentally conceptual, hence e3tra)$rammatical, structure. "o anticipate the findin$s of 1art IE, the use of anaphoric $ender a$reement where anaphoric person a$reement would ha!e been more appropriate, or !ice !ersa, is not in $eneral un$rammatical, but perhaps should be seen as unclear or clumsily)put. "he alternation between $ender and person a$reement cannot be fully e3plained by the S$rammarT of a lan$ua$e in the way that this term is understood in $enerati!e lin$uistics, and so in this case attempts to look for, say, an e3planation in terms of 6e3ical)0unctional Grammar by appealin$ to i)structure %&in$ 4997* or $)structure %0alk 2005* would be misconcei!ed. Grammatical frameworks in the $enerati!e mould such as 60G, <1SG, Minimalism, and so on, are not intended to deal with this kind of discourse phenomenon %see +ewmeyer %4999K82)8;* for a critiDue of functionalist researchers who criticise $enerati!e approaches for failin$ to account for this kind of data*. If the alternation described in 1art IE is considered to be interestin$ and worthy of e3planation by some kind of lin$uistic theory, then we must look for that e3planation outside of theories of autonomous synta3.

.*

)hapter summary

In this chapter I set out the theoretical conte3t rele!ant for 1arts III and IE. In L2.4 I 402

discussed both the =fricanist tradition of noun class studies and more mainstream work on $rammatical $ender. Section 2.2 was concerned with a$reement, in particular !ariation in a$reement and typolo$ies of a$reement markers. 0inally L2.; introduced the notion of discourse topic and !arious associated concepts.

40;

Part II Phonological and grammatical sketch

Chapter % A phonological sketch of Cicipu


#icipu has a mid)si>ed phonemic in!entory consistin$ of 27 consonants and 5 !owels accordin$ to the analysis presented here, althou$h nasalisation is contrasti!e for all !owels, and len$th is contrasti!e for all !owels and all consonants. "he presentation of the phonemic charts is deferred until after the initial analysis of syllable and word structure in L;.4, which then informs the discussion on consonants %L;.2* and !owels %L;.;*. "he ne3t three sections deal with three important suprase$mental topics in #icipuK tone %L;.8*, !owel harmony %L;.F*, and nasalisation %L;.5*. 0inally %L;.7* I discuss some of the more important morphophonemic processes. 0or reasons of space most topics are dealt with only cursorily , as with the $rammatical sketch in chapter 8 the aim is to familiarise the reader with the phonolo$y of #icipu, as well as to pro!ide a point of departure for more detailed research in the future.

".1

0yllable and root structure

I will first consider the unambi!alent syllable structures found in noun and !erb roots %L;.4.4*, before turnin$ to the more comple3 ambi!alent cases %L;.4.2);.4.5*. Section ;.4.7 deals with prefi3es and ideophones, which allow e3tra syllable types, and in L;.4.9 I will look at the structure of nominal and !erbal roots.

%11

:nambivalent s-llables

"he only unambi!alent syllable types in #icipu noun and !erb roots are #E and E, althou$h there is a stron$ case for admittin$ #E+4 non)finally. -3amples %4)2* show #E syllables in noun rootsK %4* ka)0kulu)
+#4)!ailstone

a)0kulu)

+#2)!ailstone

hailstone %2* s0si ro mane

hailstones Ieamy00;.4;;7J i )0si ro

+#9)mane

+#;)mane

manes Ieamd020.40;;J

"he followin$ e3ample shows a #E.#E !erb root %also sita SswellT, naha Slea!eT and
4 Standard abbre!iations are used when referrin$ to syllable typesK #%onsonant*, E%owel*, and +%asal* consonant.

40F

many others*. %;* ;0ka a)


2S)tookQ/6S

you .sg.2 took Isaim004.409J E syllables may be detected root)initially because they take different sets of prefi3es, and also because the root)initial !owel coalesces with the prefi3 !owel %a special case of a more $eneral process , see L;.7.4*. "he followin$ e3amples show E.#E roots, both nominal %8)F* and !erbal %5)7*. +ote that the vi0 prefi3es in %F* and %7* could not ha!e been predicted from the correspondin$ prefi3es that occur before consonants in %2);*. %8* Iko(si )L ko0osi ) eye %F* Iv(m)L vi0m)
+#9)monkey

+#4)eye

Io(si L ) o0osi ) eyes

+#2)eye

Ieamd008.022J Ij(m)L yi0m)


+#;)monkey

monkey %5* Ima+(ja)L mi )0aya)


=GF)comeQ/6S

monkeys Ieamd020.4004J

they came Ii.e. n)0'y'y Sfish%pl.*, +#FT cameJ %7* Ivu+(wa)L vi )0uwa)
2S)!earQ/6S

Itaff004.002J

you .sg.2 heard Is!tm$004.;;9J E syllables are restricted to root)initial position, and e!en there they do not usually surface as such. If there is no prefi3 with which the root !owel can coalesce, a dummy appro3imant is added %e.$. L8.5.2.; for the imperati!e*. Eowel)initial roots are rare in #icipu %89 out of 9F0 nouns in the le3icon, 40 out of 874 !erbs*, and comparati!e e!idence su$$ests they may be deri!ed historically from consonant)initial roots, especially those be$innin$ with an appro3imant.

405

#E# syllables are only possible in restricted circumstances %see L;.4.;, L;.4.F, and L;.4.7*. (orrowed words with #E# syllables in the source lan$ua$e %in!ariably this is <ausa* are pronounced with a transitional schwa !owel as in Ika)0husXka)L SfaceT from uska, since #E# syllables are $enerally unacceptable in #icipu.

%1#

;abialisation and palatalisation


Iu)0kUaYi u)0kwari m)0'Z'Z m)0'y'y
+#8)fish

= number of consonants can be labialised or palatalisedK %9* u)0'Ua*( u)0'wa*a * u)0hZa(L * u)0hyaa

+#7)ne3tPyear

;S)passQ/6S

;S)sayQ/6S

ne*t year

ish

he passed

he said

"here is e!idence from the allomorphs of the con'unction n) SandT %L8.8.F.8* that these should be considered sin$le consonants. "he n) allomorph occurs before short consonants, while before lon$ consonants or consonant clusters we find ni ). 6abialised J Swith the dry consonants pattern with short consonants, as shown by In) kUandai seasonT. "here are only si3 such consonants in #icipu %NkU gU 'U '[ hU h[:*, and so the decision to treat them as sin$le phonemes does not $reatly increase the phoneme in!entory.

%1%

;ong consonants
+#9

6on$ consonants occur word)initially in

nouns e.$. 707a SpersonT, k0kaa SwomanT,

and word)medially in a few !erbs e.$. latta SsleepT, tanna SdescendT. "hese are StrueT $eminates %(le!ins 2008K459* and should be treated as sin$le phonetic se$ments on the timin$ tier , they ne!er ha!e an inter!enin$ epenthetic !owel %unlike the <ausa borrowin$s mentioned abo!e*, and when comple3 se$ments such as affricates are len$thened the resultin$ sound consists of a sin$le lon$ closure followed by a sin$le frication period, rather than a repetition of the short !ersion2. +e!ertheless syllables endin$ in a $eminate pattern with other hea!y syllables %i.e. #EE and #E+*, for e3ample in the habitual %L8.5.;.;* tone pattern, which depends on the number of mora in the !erb stem to which the habitual affi3 si0 attaches. If the !erb stem is #E#E %bimoraic*, then the tone pattern is 6 6 < < as in %9*.
2 =n additional test would be to check that the lon$ consonants in !erbs such as latta and tanna do not become split by the addition of the causati!e %L8.5.8.4* or pluractional %L8.5.F.8* infi3es. In the words tested so far, $eminate inte$rity is maintained %e.$. kullo SburnT, kullisso Scause to burnT*.

407

%9*

u)0si 0 ) ta'a
;S)<=()want

#E#E

66??

He wants <owe!er for #E+#E, #EE#E, and #EGE; !erbs the pattern is 6 6 <%6* 6K %40* u)0si 0 ) pn)t)
;S)<=()clap

#E+#E8

6 6 ?; ;

He claps %44* u)0si 0 ) wi *ina)


;S)<=()sell

#EE#E

6 6 ?; ;

He sells %42* u)0si 0 ) htt)


;S)<=()warmPoneselfPbyPfire

#E##E

66?;

He warms himsel by the ire 0urthermore when there is a lon$ consonant in #2F it closes off the pre!ious syllable, so that E4 is shorter than it would be in an open syllable. "herefore a word such as htt in %42* should be analysed as #E#.#E rather than #E.#E, and thus we admit the syllable pattern #EG as well as #E. GE is also reDuired to cater for the word)initial $eminates mentioned abo!e. "he internal structure of syllable)initial $eminates has been contested in recent decades %#lements and &eyser 499;, <ayes 4999, <ume et al. 4997, .a!is 4999*, the main issue bein$ whether or not they are Slon$T %bipositional* or Shea!yT %monopositional and moraic*. "he e!idence from #icipu is pu>>lin$ and reDuires further in!esti$ation , word)initial $eminates beha!e like consonant clusters in that they tri$$er epenthesisNreduplication %L8.8.F.8, LF.F.9*, yet they also beha!e like sin$le consonants with respect to phonotactic constraints %##E with two different consonants is not allowed* and minimality constraints %LF.F.7*.

%1&

;ong vowels and diphthongs

In contrast to lon$ consonants, lon$ !owels in #icipu are not sub'ect to S$eminate inte$rityT %1erlmutter 499F*. "his can be seen from the dependent imperfecti!e
; G R $eminate. 8 = few #E+#E !erbs follow the Sli$htT pattern here e.$. yinda SseeT, panda Sfor$etT, and kanda SmarkT. "his may be le3ical idiosyncrasy, althou$h the #icipu !erb system is otherwise hi$hly re$ular. =lternati!ely it may re!eal somethin$ about the phonolo$ical representation of the +# cluster %i.e. underlyin$ly N+#N or a prenasalised N#N , see L;.4.F on prenasalisation*. F #4 R first consonant in the root, E4 R first !owel in the root, and so on.

409

%L8.5.;.8* form of the !erb, where the final stem !owel chan$es to i. If the !erb stem is monosyllabic then the !owel is diphthon$ised5 and the tone is risin$K %4;* n) n) wa)yi , w0aya0i nu)utu)i n)0u0tuu0i

and ;S)come)57P.I8P> and);S)pour)57P.I8P>

he was coming, he was pouring Itapf004.00;.009J "his is e!idence that lon$ !owels are bipositional %i.e. a seDuence of !owels rather than one lon$ !owel*, since if there was only one root node linked to two timin$ slots %or two moras* we would not e3pect one half of the !owel to chan$e independently of the other. <owe!er it does not allow us to decide between a heterosyllabic #E.E or a tautosyllabic #EE analysis. <istorically, it seems likely that lon$ !owels in #icipu are deri!ed from the coalescence of two syllables with the disappearance of the intermediate consonant, as was su$$ested for !owel)initial roots in L;.4.4. Many of the lon$ !owels in "irisino ha!e co$nates in other dialects and lan$ua$es where such a consonant remains, and !ice !ersa. .iphthon$s differ from lon$ !owels %and #EyEN#EwE seDuences* in that their duration is not noticeably lon$er than short !owels, and there is no SdipT in the wa!eform. hen words are broken down into syllables by nati!e speakers then the diphthon$ is pronounced as part of one syllable. "herefore diphthon$s are considered to be a sin$le !owel with re$ard to syllable structure.

%1(

Prenasalised stops and affricates

In addition to #E and #EG %L;.4.;*, it is common to find what appear to be #E# syllables, as in Iko).don.tuL \stool]. Towever they always ha!e a nasal as the coda, and this nasal only occurs a ter nasal !owels and be ore oral stops or affricates, as in %48*. %48* Iko)dontu ko)0dotu
+#4)stool

ka) u^gu ka)0 ugu


+#4)snake

ku)m a_ ku) a_
climbQIM1

ko)ndo_L ko)do_
enterQIM1

stool

snake

climb#

enter#

1renasalisation affects all oral stops and affricates, with the e3ception of the $lottal stop and its palatalised and labialised !ariants 'y and 'w. &honetically it seems clear that two
5 In "irisino. In "ikula lon$ !owels chan$e completely to Ii +(J in monosyllabic dependent imperfecti!es.

409

distinct se$ments are in!ol!ed , the two parts can contrast in !oicin$ %e.$. Iko)dontuJ*, and the nasal component is typically much lon$er in duration than the precedin$ !owel. <owe!er the phonological status of these prenasalised consonants is problematic. "he clearest indication that the nasal component is absent from the underlyin$ representation comes from the distribution of prenasalised consonants. Since #icipu has both oral and nasal !owels, and nasal !owels do not occur directly before non)$lottal stops, it is possible to re$ard the nasal phone in words such as Iko)dontuJ as conditioned by a combination of the precedin$ !owel and the followin$ consonant. =lthou$h most instances of prenasalisation are found root)internally, the process occasionally occurs across morpheme) and e!en word)boundariesK %4F* Imu^`Ua)(nu)kUa) t%eL mu0u0gwa)anu)kwa) $e
4S)0A")seeQI//
+-G

' wouldn't know Isayb004.402J %45* kZe)L Ik)5(^ k0e) k)05


+#4)a3e

=G4)#21

it's an a*e Ieamd002.0F8J e therefore ha!e to assume the e3istence of a prenasalisation rule to account for these e3amples, Duite independently of considerations of syllable structure in roots. "he rule could then be re)used to account for root)internal prenasalisation. "here is some problematic data %e.$. the paucity of +# clusters with a !oiceless #, su$$estin$ that the # is assimilatin$ to the + in terms of IE2I#-J*, but the data lar$ely {}||ero{ Y opq Y wfn[i Y opno Y |rq^n{nf[{no[e^ Y [{ Y n Y |pe^efe[wnf Y |rewq{{ Y n^~ Y opno Y {q}q^wq{ Y nrq Y }^~qrfg[^ Y Y v[op Y opq Y ^n{nf Y {qiq^o Y {}||f[q~ Y |rq~[wonjfg7. +e!ertheless the resultant nasal se$ment %which, as mentioned abo!e, can be Duite lon$* contributes to syllable wei$ht in wei$ht)sensiti!e processes. "wo such cases are briefly discussed here. 0irstly, lon$ nasal !owels do not seem to tri$$er prenasalisation. "here are only three e3amples of lon$ nasal !owels precedin$ a consonant in the corpus, but they pattern consistently , none of them ha!e a nasal inter!enin$ between
7 2ther than in this section, the e3amples in the thesis are written with the nasal e.$. ko)0dontu. .ue to the comple3ity of the data, particular care should be taken o!er any ortho$raphic decisions relatin$ to prenasalisation.

440

the !owel and the stop. %47* ka)0hhuu$i


+#4)cloud

ka)0tii tii
+#4)foreskin

) `U * k)0`U
+#4)crow

cloud

oreskin

crow

"here does not seem to be any ob!ious reason why lon$ !owels should not tri$$er pre) nasalisation, unless the resultant nasal forms the coda of a #EE# syllable. In this case the restriction would be simply a matter of syllable wei$htK it is not unusual for lan$ua$es to ha!e special restrictions on Ssuper)hea!yT #EE# syllables. Secondly, the tone patterns on !erbs are sensiti!e to syllable wei$ht. "he habitual tone pattern has already been mentioned in WVXVW Y rqwnff Y opno Y \qrj{ Y v[op Y seDuences, as in %40*, $enerally pattern with other Shea!yT syllable patterns such as Yn^~YVY[i[fnrfgY[^YopqYwn{qYeYopqYrqnf[{Yoe^qY|nooqr^Y\qrj{Yv[opY seDuences a$ain pattern with other hea!y syllables by takin$ a fallin$ tone on the first root syllable rather than a hi$h toneK %49* u) ana) u0 ana0aTa
;S)in!ite)/6S

u)koo) u0koo0aTa
;S)die)/6S

u)kon)do) u0kondo0aTa
;S)enter)/6S

he invited

he died

he entered

In summary, prenasalisation in roots is a predictable phonolo$ical process, but the resultin$ nasal se$ment is both lon$er in duration than mi$ht be e3pected and contributes to the wei$ht of the syllable in wei$ht)sensiti!e processes. (efore lea!in$ the topic of prenasalisation, it should be noted that a small number of noun roots be$in with an +# seDuence, where # is a non)$lottal stopK %49* ma0nda
+#8)calabash

mi 0nda
+#F)calabash

calabash %20* wu0nto)


+#7)$uestPhut

calabashes vi 0 nto)
+#9)$uestPhut

guest hut

guest huts

"he +# clusters in these words should not be considered products of the prenasalisation process, at least not synchronically.

%19

Approximants

"o a!oid proliferation of syllable types, the ambi!alent !ocoids IiJ and IuJ are analysed 444

as consonants when they occur in onset position, and as !owels when they occur in nucleus position.

%1<

Prefixes and ideophones

"he precedin$ discussion was concerned with only noun and !erb roots, which may contain #E and E syllables %and, it was ar$ued, #E+ and #EG*. 2ther word classes ha!e different possibilities, in particular prefi3es %nominal or !erbal* and ideophones. +ominal prefi3es and a$reement prefi3es are all monosyllabic, and are usually of the form E, #E, or +. "he e3ceptions are the interestin$ -0 %consonant)len$thenin$* allomorphs of the
+#9

and 2S prefi3es. "he application of these prefi3es results in the

len$thenin$ of the first consonant of the root, whate!er this consonant happens to be. =ny consonant can be len$thened in this mannerG e3amples are $i!en for +#9 nouns %24* and 21S !erbs %22*. %24* 707a
/C))person

k0kaa
/C))woman

$0$')
/C))sheep

s0si ro
/C))mane

person %22* t0ta'a)


0)wantQ/6S

woman '0'p)
0)holdQ/6S

sheep l0latta)
0)sleepQ/6S

mane j0ja*nta)
0)crushQ/6S

you .sg.2 want

you .sg.2 held

you .sg.2 slept you .sg.2 crushed Ieamy0;5.004, tats00F.004.0;0J

e saw in L;.4.4 that the first !owel of a !owel)initial noun or !erb root coalesces with the precedin$ prefi3. If there is no prefi3, as in the case of imperati!es, the root is obli$atorily preceded by a w1 or y1 SdummyT consonant %L8.5.2.;*. Eowels at the be$innin$ of prefi3es beha!e sli$htly differently. Atterance)initially they are pronounced with a %non)phonemic* precedin$ $lottal stop, rather than a dummy consonant. Atterance)medially, prefi3)initial !owels usually coalesce with the precedin$ !owel, althou$h a $lottal stop may be inserted, e!en word)internally between two prefi3es %see L;.7.4 for e3amples*. "he
+#F, =GF

and 41S prefi3es are most often pronounced as syllabic nasals the followin$ consonant, but before certain consonants

homor$anic

with

%includin$ N YN, N sN, NyN and N'yN* they may surface as Ii J or IuJ, dependin$ on the followin$ !owel. "hey are assumed to be NmN underlyin$ly, since this is the !owel that surfaces before !owel)initial stems, as in %5* abo!e.

442

#ross)lin$uistically, ideophones %L8.8.4.;* are often phonolo$ically Sde!iantT %#hilds 4998K494, Eoelt> and &ilian)<at> 2004K2*. In #icipu they are characterised by #E# syllables, which as we ha!e seen do not $enerally occur in the lan$ua$e. So far ideophones ha!e been found with de!iant codas containin$ nasal consonants, the plosi!e p, and the fricati!e O. Some e3amples are $i!en in %2;*K %2;* vp splat# pass very white 8oo^ very black pm wholly

%1=

!oot structure

+oun and !erb roots in #icipu are usually disyllabic, althou$h there are a si$nificant number of mono) and tri)syllabic roots. "he monosyllabic roots almost all ha!e lon$ !owels, and many are demonstrably deri!ed from former disyllabic roots %e.$. s SdrinkT !s. so'o in * Se$$T !s. k)0k' ) in "i>oriyo*. estern &ambari, k)0k = few roots of words for birds and trees ha!e four or e!en fi!e syllables, but often these are reduplicated. In the followin$ tables <ausa loanwords are omitted because they ha!e a markedly different distribution from nati!e #icipu words, bein$ more likely to ha!e trisyllabic roots. "he distribution of nouns accordin$ to number of root syllables is $i!en in "able ;K 0able 6- eistribution o noun roots according to syllable structure Syllables in root +o. of nouns -3amples 4 2 ; 8 405 F20 98 45 ko)0luu SkneeT, ku)0t ShenT ku)0$i i.no SbackT, ku)0da.vu) SmortarT me 0 e).ri i.se) SswiftT, $i )$.$e.re) SstarT ka)0'an.ga) .la).mi ) Straditional ba$T

Eerbs follow a similar pattern to nouns, with an e!en hi$her percenta$e of disyllabic roots.

44;

0able J- eistribution o verb roots according to syllable structure Syllables in root +o. of !erbs -3amples 4 2 ; 8 49 ;44 59 2 $aa S$i!eT, yaa SdoT na.ha Slea!eT, la.sa S$reetT hee.pi.ye SaskT, mi.ri.8a StwistT gi.tu.wa.na Se3haleT, ku.si.ya.nu Ssmell %i.e. with nose*T

2f the tri) and Duadri)syllabic !erb roots, it is likely that the ma'ority were once bi) morphemic, and in some cases fossilised present)day deri!ational processes can be identified. 0or e3ample lapila SprepareT, am ala SfumbleT, and to ilo Scool liDuid by repeatedly pourin$T appear to contain a Sfro>enT pluractional infi3 ils, althou$h they cannot be used without this affi39. In other cases, while it is not possible to identify $rammatical morphemes in present)day use, patterns still emer$e. 0or e3ample, titm SthreshT, 7i7a'a Sshi!erT, and 7i7a5a StickleT all combine a reduplicated prefi3 with an inherently repetiti!e motion %this combination is also found in (antu , Schadeber$ 200;K79*. If we e3clude such cases of Seidemic resonanceT %(ickel and +icholls 2007K209* from the count, then there are no more than a do>en !erbs with more than two syllables. "he basic !erb root structure in #icipu is therefore #E#E9.

%1>

0ummar-

hile some noun and !erb roots in #icipu are analysed as be$innin$ with a !owel, noun and !erb words are almost always consonant)initial on the surface. /oots with prenasalised stops may be analysed as #E+.#E. = small number of noun roots be$in with an +# cluster %e.$. 49)20*, but with the noun prefi3 they syllabify as #E+.#E. If we may talk of underlyin$ syllabification, then #icipu allows + and E but not #E+ 40 in noun and !erb roots. 2n the surface, howe!er, #E+ occurs but not + or E. 2nly #E is possible word)finally. 1refi3es may be #E, +, or E, all of which can occur word)initially. Ideophones are freDuently #E#. +ouns and !erb roots are typically disyllabic.

9 = similar phenomenon is found in <ausa and other #hadic lan$ua$es %e.$. +ewman 2000KF49)F49*. 9 "his contrasts with (antu #E# , see L8.5.4 for discussion. 40 =ssumin$ prenasalisation is predictable %L;.4.F*.

448

". %#1

)onsonants Phonemic inventor-

"he 27 consonant phonemes of #icipu are $i!en in "able F. 0able T- 5onsonant phonemes (ilab. Stops Impl. =ffr. 0ric. +asals 6iDuid /hotic =pp3 m v s 7 n l r p 5 6ab) dent. .ental Nal!. t d 8 $ j 1ost)al!. 1alatal Eelar k g 6abio) !elar 6aryn$.44 ' 'w 'y

kw gw

h hw hy

=lthou$h the coronal consonants in the abo!e table ha!e been labelled SdentalNal!eolarT, at least NdN and NtN were found to be dental %and laminal*, rather than al!eolar. "his was only checked for one speaker, howe!er. It is not yet known whether N nN and NlN are al!eolar or dental, althou$h they mi$ht be e3pected to pattern with NdN and NtN.

%##

Allophones and general phonetic rules

"he bilabial plosi!es NpN and N N sometimes under$o lenition to IbJ and IcJ when they occur inter)!ocalically, especially in Duick speech. 0or e3ample NyapuN StwoT may surface as IjabuJ, and Njii oN Sha!e breakfastT as Id&i(coJ. =s well as the labialised and palatalised phonemes which appear in "able F there are a number of non)phonemic allomorphs which ha!e these modifications. NmN and NvN ha!e labialised allomorphs ImUJ and IvUJ before rounded !owels, while NkN and NgN ha!e palatalised allomorphs IkZJ and I`ZJ before front !owels. (efore rounded !owels N kN, N gN,N'N and NhN do not contrast with their labialised counterparts, and so the underlyin$ consonant in such seDuences cannot be determined. Similarly before front !owels N'N and NhN do not contrast with their palatalised counterparts. "he phoneme NtN is sometimes, but not always, realised as It%J before IiJ, for
44 "he three $lottal fricati!es were missin$ from the correspondin$ chart in McGill %2007*, due to an error on my part. "his chart therefore supersedes the earlier one.

44F

e3ample with the !erb tiyo S$etT. ItijoJ seems to be considered the ScorrectT form. "he fricati!e I%J is non)phonemic but does occur as an allomorph of N sN before IiJ for some speakers. "here are only two phonemic nasals in #icipu, NmN and NnN. =ll +# clusters in #icipu are homor$anic, with I^J and IdJ occurrin$ before !elar and interdental consonants respecti!ely. NrN is realised as a flapNtap42 IYJ utterance)medially. Atterance)initially, and when len$thened %L;.2.8*, it is realised as an appro3imant IeJ %or an r)coloured !owel IfJ*. Sometimes the flapNtap surfaces as the retrofle3Npost)al!eolar IgJ, especially after the !owel NaN, but unlike <ausa %+ewman 2000K;98);9F* there is no phonemic distinction between the coronal and post)al!eolar flaps. "he appro3imants NyN and NwN ha!e nasalised allophones Ih J and Iw J which occur in the nei$hbourhood of nasalised !owels. +ote that the former differs from the nasal I iJ in that it does not ha!e a closure.

%#%

$istributional restrictions and examples

"he most strikin$ distributional restriction is that NhN rarely occurs root)medially , only two of the 88 NhN:s in the le3icon are root)medial , dooho SdisappearT and naha Slea!eT. "he affricates N$N and NjN are also rare root)medially, especially in !erbs. (oth the e3amples of root)medial affricates in !erbs in!ol!e trisyllabic roots of uncertain deri!ationK Nku$'C Sshake offT and NmnjuwN S$lareT. 2ther restrictions can probably be put down to the rarity of the phonemes in Duestion. "he ne3t two tables show e3amples of all the phonemes e3cept NhwN4;. 2ther than this one e3ception, all the phonemes contrast root)initially in nati!e #icipu nouns, as demonstrated in "able 5.

42 It is not known whether this sound is a flap or tap %See 6adefo$ed and Maddieson 4995K2;0)2;4*. ) 4; NhwN occurs only in the %borrowedM* noun tu)0hwi 'i S#:6ela lan$ua$eT and in the time ad!erb hwa'ya Sday before yesterdayT.

445

0able ]- Rootcinitial consonant phonemes in nouns p 5 t d 8 k kw g gw ' 'y 'w $ j v s 7 h hy m n l r y w u)0pa$i ka)0 ara ma)05asa) ka)0tada ka)0da a u)08anga) ma)0kantu n)0kwa'a u)0galu) ma)0gwawa $i )0'adi ) m)0'y'y u)0'wi *i ka)0$a'u) ku)0jene) ka)0vara ku)0sayu a)07a $i )0havi ) ) a)0hya'a ka)0manga i )0nama) ka)0lana) ka)0rakatau ka)0yayu) ma)0waa di iculty old man mole .on skin2 palm .o hand2 bush1countryside tree kni e orphan side dea 1mute trap ish distance husk .o maiHe2 river goat hut spear people scratching arrows rope meat scar heel root dog

=part from N'yN, NkwN, NgwN and NhwN all the consonant phonemes contrast root)initially in nati!e #icipu !erbs. NkwN occurs root)medially in the !erbs dukwa S$oT and $ukwa SpraiseT, and NgwN occurs root)medially in ei$ht !erbs includin$ langwa SspoilT and hungwa SrestT. +either N'yN nor NhwN are found in !erbs at all, perhaps not surprisin$ly $i!en their o!erall rarity48.
48 N'yN also occurs in <ausa, althou$h only in a few hi$h freDuency words like 'ya'ya SchildrenT. =lthou$h ) Sday before rare in #icipu, the phoneme occurs in a !ariety of words e.$. ka)0si'ya SbreastT, hwa'ya yesterdayT, mo)0ri 'yo) SduikerT, m)0'y'y SfishT, ko)0'yuwo S$rassT, and ko)0$i 'yo)'i *i SramT. It is also found in other &ambari lan$ua$es and it seems unlikely to ha!e been borrowed from <ausa, where N'yN is a recent addition to the phonemic in!entory %+ewman 2000K;9;*.

447

0able _- Rootcinitial consonant phonemes in verbs p 5 t d 8 k g ' 'w $ j v s 7 h hy m n l r y w pasa ana 5asa tasa dasa 8asu kanda gava 'etu 'waa $a'a janta vasa sa5a 7aa hala hyaa mata naha lawa raa yaa waana cross invite slap meet castrate soak mark on wall kick dry by hanging out pass harvest crush hit embrace ind coil say give birth leave escape eat arrive twirl

%#&

;ength

hile many lan$ua$es ha!e lon$ consonants, it is rare %althou$h not unheard of* for them to occur root)initially and word)initially %(le!ins 2008K494, 6adefo$ed and Maddieson 4995K9;*. =s with #entral &ambari %#ro>ier 4998K290* any consonant can be len$thened in #icipu, includin$ the $lottal fricati!e and stops. "he difference in len$th between short and lon$ consonants can be relati!ely mild, with lon$ consonants sometimes no more than half as lon$ a$ain as their counterparts4F, and word)initially the distinction often seems to be neutralised in normal, fast, speech. 6on$ !oiceless stops may be impossible to detect utterance)initially45, but they can easily be heard utterance)
4F #ross)lin$uistically lon$ consonants tend to be lon$er than short ones by between 4.F to ; times %6adefo$ed and Maddieson 4995K92*. 45 In some lan$ua$es lon$ !oiceless plosi!es can be detected utterance)initially, perhaps from a chan$e

449

medially in careful speech. "he wa!eforms in 0i$ures 22)25 demonstrate len$th differences for NnN, NlN, and N kN. "he words in 0i$ures 22)2F all belon$ to
+#9,

which has either B1 or C1 for

allomorphs47. "he former occurs with ;0naata) Ssmall spiderT %0i$ure 22* and ;0looka)$i StimeT %0i$ure 28*, resultin$ in a short initial consonant. "he latter occurs with n0naa ScowT %0i$ure 2;* and l0lama) SnoiseT %0i$ure 2F*, resultin$ in a lon$ initial consonant.

in the amplitude of the followin$ !owel %6adefo$ed and Maddieson 4995K98*. "his has not been in!esti$ated for #icipu. 47 /ecall that C represents an underspecified consonantal wei$ht unit, which assimilates completely to the consonant to which it is attached, resultin$ in a lon$ consonant.

449

n -

a*

a -./0/112

#ime +s,

+igure 44- "ave orm o ;0naata) small spideru

n* -

a* -./02313 #ime +s,

+igure 46- "ave orm o n0naa cowu

420

" -

o*

i* -.1-5

#ime +s,

+igure 4J- "ave orm o ;0looka)$i timeu

"* -

a -.1--/

#ime +s,

+igure 4T- "ave orm o l0lama) noiseu 0i$ure 25 demonstrates the difference between an utterance)medial short and lon$ NkNK

424

%28*

ka)0dama ka,1k0kaa
+#4)word =G4R+#9)woman

the word womanu Ieamy0;2.048J

4 -

4*

a -.7/31-5

#ime +s,

+igure 4]- "ave orm o ka)0dama ka,1k0kaa Sthe word womanu Most of the lon$ consonants are formed by an e3tension of some %as in NkN* or all %as in N nN and NlN* of the short consonant. <owe!er in the case of N rN there is a Dualitati!e difference between the lon$ and short !ariants , short N rN is a tapNflap IYJ utterance) medially but lon$ NrN is realised as an appro3imant IeJ %or an r)coloured !owel IfJ*, optionally followed by a flapNtap, $i!in$ IeYJ or IfYJ as shown in 0i$ure 27K %2F* Ie)ejL, r)0rei, ) Ika)dama keeY ) ejL ka)0dama ke,1r0rei
+#4)word

+#9)town

=G4R+#9)town

towns, the word townsu Ieamy0;2.028J

422

k 0

ei 0.;95

) o townsu +igure 4_- "ave orm o ke0r0rei

"ime %s*

"he ma'ority of lon$ consonants in #icipu words are ambi)morphemic, althou$h as discussed in L;.4.; they are true lon$ consonants rather than simply seDuences of short consonants. "hey arise from the application of the C1 len$thenin$ allomorph of the +#9 prefi3 to a noun root be$innin$ with a short consonant. =ll the e3amples in the wa!eforms $i!en abo!e are of this kind. 6on$ consonants in noun roots are mostly found root)initially. 2ut of 525 nati!e #icipu noun roots, F4 start with a lon$ consonant. It is su$$ested in McGill %n.d.* that these are likely to ha!e arisen throu$h the reinterpretation of an +#9 -0 prefi3 as part of the root. 6on$ consonants in !erbs are rare. "hey do not occur root)initially, and there are only a few e3amples root)medially , 45 out of ;F9 !erbs. "he only lon$ consonants attested are NtN %ei$ht tokens*, NlN %four*, NnN %two*, NwN and N'wN %one each*. "he form and meanin$ of some of the !erbs su$$est that the lon$ consonants ha!e come about as a result of the fossilisation of affi3es %see L;.4.9*.

42;

"." %%1

1o$els Phonemic inventor0ront i e : ei :eu #entral (ack u o

0able `- 5icipu vowelcinventory #losed Mid)#losed Mid)2pen 2pen

a : ai : au

=lthou$h asymmetric !owel systems wei$hted toward the back are rare cross) lin$uistically %#rothers 4979K4;7, Schwart> et al. 4997*, there is independent e!idence for the analysis $i!en here. Israel ade %p.c.* adduced the same si3 !owels from his own independent research on "irisino, and the wordlists produced by 6o!elace %n.d.* for the nearby related lan$ua$e "su!ai also distin$uish precisely these si3 !owels. "he #icipu !owels are articulated more openly than the correspondin$ cardinal !owels, and this difference is clearly audible. "he !owels of <ausa loanwords are often raised in the tar$et form, presumably because to the #icipu ear <ausa a is more like a #icipu NeN than a #icipu NaN. -3amples are shown belowK %25* de)ge) ') k)0$$)k from daga from a'a from %aka SfromT SnoT Sba$T

%%#

Allophones and general phonetic rules


est &ain'i, where most other lan$ua$es

"he lack of a schwa phoneme is unusual in

ha!e two or e!en three %e.$. #entral &ambari* central !owels. Eowels with the phonetic !alue I9J do occur in #icipu, but only as allophones of N aN, NoN, NN, or NeN in the en!ironment P#EIhi$hJK in other words, when the ne3t !owel in the word is NiN or N uNK %27* %a* Iko)jo)^`U9li )L ko)0yo)ngoli )
+#4)ant

%b* Ido)(s9)nu*L do)oso)nu*_


swimQIM1

ant, k.o. large Ieamd022.442;J

swim# Ieamd005.089J

hen such words are pronounced carefully the underlyin$ !owel Duality is ne!er a schwa. "he distribution of the !owels NiN and NuN is problematic, 'ust as in <ausa %+ewman 428

2000K;99)800*. In some cases a nei$hbourin$ rounded !owel seems to be responsible for conditionin$ an NiN to a IuJK %29* Imu( m0uu mumpa)L mi 0 mpa) Ieab$004.0F7J 2n other occasions different !owels are found in apparently identical en!ironments, as shown by gitu and gutu in the followin$ e3ampleK %29* h0a+ya)
;1)comeQ/6S ;1)a$ainQ/6S $oPbackQrls sendQ/6S)E-+" ;1)a$ainQI//

+#F)child =GF)this

these children

e)0'esu)

gutu)

ni yu)0no)

e0'e)su)

gi t ) u)

$oPbackQI//

they again sent or them to come back again Isaim004.0F8J 0inally, for some words either NN or NaN is acceptable, as in ku)0la$i :ku)0l$i S$irlT and sa'u: s'u StouchT.

%%%

?asal vowels

0or each of the si3 !owels there is a contrastin$ oral and nasal pair, e3emplified below. "he oral !owels are considerably more common across the le3icon, with a ratio of about FK4 in the words collected so far. 0able b- ral and nasal vowel contrasts a a e e i i o o u u k0kaa * u)0taa re'e se'e ma)0jji *i u)0jii u)0ko*o u)0l kuu tuu * mo)0too k)0k woman bow .hunting2 persuade carve bird, k.o. value death saliva locust bean tree egg be older than pour

%%&

;ong vowels and diphthongs

-ach of the oral and nasal !owels has a lon$ counterpart as shown in "able 40. Short 42F

!owels are more common, by a ratio of appro3imately 7K4 in the le3icon. 0rom the rou$h measurements taken so far, lon$ !owels in citation form in elicited speech are appro3imately half as lon$ a$ain as their short counterparts. 0able ,a- ral and nasal vowel contrasts a aa a aa e ee e ee i ii i i i o oo o oo u uu u uu dapa gaata ka'a taa ke)0re'e) $i )0reene) u)pepi * u)0pee $ita $iita ka)0yi va ) ka)0hiivi tomo toono ko)'uso) * mo)0too yn tm ) ru)0k' k)0k kur kuula ku)0yu)yu* ka)0huu$i burn cut scoop out shoot tongue ireplace wind bigness sting s$uash vein navel die out come home haHe saliva pull chew pool egg grow up call sand cloud

"he !owels in monosyllabic roots are always lon$, with the e3ception of one !erb yo SbeT. "his !erb surfaces with a lon$ !owel when it occurs without a suffi3 %;0*, but with a short !owel before the perfecti!e suffi3 0nA %L8.5.;.4* as in %;2*. It thus contrasts with yoo S$oT which is consistently lon$ %;4, ;;*K

425

%;0*

u)0yo*o

;S)beQ/6S with

n)

;0mooto)
+#9)car

he has a car Ilit. Sis with carTJ I2009)08)05.004J %;4* u)0yo*o


;S)goQ/6S ItownJ

3a)kuuku) I2009)08)05.004J

he went to Iakuku %;2* 707a n0na) yo0no) n) ;0mooto)


+#9)car

+#9)person

=G9)/-6

beQ/6S)10E with

the one who has a car Ilit. Sis with carTJ I2009)08)05.004J %;;* 707a
+#9)person

n0na)
=G9)/-6

yo*o0no)

3a)kuuku) I2009)08)05.004J

goQ/6S)10E ItownJ

the one who went to Iakuku 2ther than in monosyllabic roots, lon$ !owels are !ery rare root)finally in nati!e #icipu roots. hen <ausa nouns endin$ in a lon$ !owel are borrowed into #icipu, the final !owel is usually shortened, as in ka)aka S$randparentT. Atterance)finally, the distinction between short and lon$ !owels is blurred in #icipu, 'ust as in <ausa %+ewman 2000K804*. #icipu has four diphthon$s, all of which are Sfallin$T %i.e. from an open to a close !owel Duality*K NaiN, NauN, NeiN, and NeuN. 0able ,,- eiphthongs :ai: :au: :ei: :eu: SswordT, ma0ju)'wai ) Sse3ual intercourseT ma)0gai u Scopaiba balsam treeT, ka)0sau * S$ra!eT u)0yanda SpainT, ma)0dei Scalf %body part*T u)0sei * ScottonT, 8eneu * SsmallT ke)0re)7eu

.iscountin$ loanwords, diphthon$s are only found as the last !owel of the root, and comparati!e e!idence from &ambari su$$ests they may reflect historical consonant loss from root)final syllables.

%%(

$istribution

=ll of the short oral and nasal !owels are found in both E 4 and E2 root positions. 6on$ 427

!owels are lar$ely limited to E4 %especially in !erbs*. Eowel harmony %L;.F* stron$ly constrains the !owels which are able to occur to$ether in nati!e #icipu roots. 0urthermore there is a stron$ tendency %but by no means an obli$ation* for E4 and E2 to ha!e the same Duality.

".* %&1

Tone Tone inventorku)0saa ka0ssa) lea * u)0taa


+#7)bow

"here are three contrasti!e le3ical tones in #icipuK <, 6, and <6 %fallin$*49K %;8* < !. 6K
+#9)mountain +#4)leaf

mountain %;F* < !. <6K ka)0taa shoe t0tii


+#9)container +#4)shoe

bow .hunting2 tii*


+#5Kshit

container %;5* < < !. < 6K kaaya


+#4Kroom

shit kaaya)
+#4Kbean

room $u)0kulu
+#5)tortoise

bean ka)0kulu)
+#4)hailstone

tortoise %;7* 6 6 !. 6 <6K ka0kku)ti )


+#4)shell

hailstone ka0kka)$ii*
+#4)middle

shell m0k)8)
+#8)$ossip

middle ma)0ku)da*a
+#8)sDuirrel

gossip

s$uirrel

49 /isin$ tones do occur phonetically, but they do not contrast with the other three tones in underlyin$ representations.

429

%;9*

< 6 !. 6 <6K u)0yeyu)


+#7)cold

) * ku)0yuyu
+#9)sand

cold i0 ) $i yo)
+#;)beard

sand ko0$$i )yu*u


+#4)heap

beard

heap

Statements, commands, and yesNno Duestions all ha!e their own intonational contours which are imposed on top of the le3ical tones of the sentence. Statements can be analysed as ha!in$ an utterance)final 6 tone, which means that < tones are realised as <6 at the end of utterances. 0or commands and Duestions see L;.8.5 and L8.;.5.4 respecti!ely. =s well as the statement boundary tone 'ust mentioned, there is further e!idence that <6 tones should be analysed as seDuences of < and 6 49. 0or e3ample, the realis tone pattern on disyllabic roots %L;.8.5* is < 6 if the first syllable is li$ht, but <6 6 if the first syllable is hea!y. <6 tones are mainly limited to lon$ !owels, but they may occur on phonetically)len$thened short !owels %e.$. utterance)final imperati!es L;.8.5*. In certain limited scenarios an e3tra)hi$h tone seems possible. 0or e3ample, in the "ikula dialect if the ne$ator $e occurs after a hi$h tone then it may surface as e3tra) hi$hK %;9* i )08ana
+#;)mark

$ek
+-G

not marks I"ikula, sa$b004.025J 2ther candidates for e3tra)hi$h tone are the plural morpheme aka Sassociates ofT %L8.8.2* and the topic marker gok %L9.;.2*.

%&#

$owndrift@ downstep@ and upstep

6ike many =frican lan$ua$es, tones in #icipu utterances under$o SterracedT downdrift within each intonation $roup, whereby the pitch of each < is lower than that of the one before. Successi!e 6 tones also decline in pitch, but by less, and so the distinction between < and 6 is less towards the end of an intonation $roup than at the be$innin$. .ownstep occurs in se!eral syntactic en!ironments %see L;.8.7*. In addition to downdrift and downstep, there is also e!idence for the rarer
49 "his is $enerally the case for =frican lan$ua$es %#lements 2000K4F;*.

429

phenomenon of upstep. "he rele!ant data is pro!ided by the habitual tone pattern %L;.4.;*. /ecall that the usual tone pattern for !erbs with li$ht syllables is 6 6 < <. +ormally the ne$ator $e occurs with hi$h)tone, but in %80)84* it is e3tra)hi$h. %80* a)0si 0 ) panda
;1)<=()for$et

$ek
+-G

they do not orget Isamy004.059J %84* a)0si 0 ) ta'a


;1)<=()want

$ek
+-G

they didn't want Isamy004.0;;J "his effect is not limited to e3amples in!ol!in$ the ne$ator, but also occurs with reduplicated !erbsK %82* ka0na)
=G4)=/"

ka)0si 0 ) ni JmaJ0ni maJ


=G4)<=()bite)/-.A1

some camels bite Itats008.004.049J %8;* n)0si 0 ) paJndaJ0panda)


4S)<=()for$et)/-.A1

' orget Itats00F.002.007J If we assume that upstep occurs after the habitual marker si), then e3amples %80)8;* can be accounted for without needin$ to postulate three underlyin$ tones. "he tonolo$ical representation of %8;*, for e3ample, would then be 6 6 6 6 < 6.

%&%

0preading

"one spreadin$ is !ery common in (enue)#on$o lan$ua$es, particularly of < tones. In #icipu < tones at the end of one word spread onto 6 tones at the start of the followin$ word. "one can spread onto both noun prefi3es %88* and roots %8F*K %88* u)0ka a)
;S)tookQ/6S

$e
+-G

ka0li ipi )

+#4)wron$

he didn't take the blame Icitation form ka)0li ipi J )

Isaim004.088J

4;0

%8F*

t0i nda)
41)seeQ/6S

$e
+-G

;0kaaka
+#9)$randparent

we didn't see the grandparent Icitation form ;0ka)akaJ

Ieamy0;7.429J

Gender)marked sub'ect a$reement prefi3es are also affected by <)spread from precedin$ sub'ects if the !erb is in the realis mood20 , compare %85a*, in which spread occurs after an <, with the unperturbed pattern in %85b*K %85* %a* mo)0ni
+#8)water

m0k8Qi )lR)0nu)

=G8)cutQ/6S16=#s)/-S

the water parted %b* mo)0ni


+#8)water

m0aya)
=G8)comeQ/6S

m)0k8Qi )lR)0nu)
A-4)cutQ/6S16=#s)/-S

then the water parted Isahs004.00;.044, sahs004.00;.04FJ (y contrast, personcmarked a$reement prefi3es are seemin$ly unaffected by precedin$ < tones. It seems that < tones can spread from words of any le3ical class which end in a <. -3ample %87* shows <)spread resultin$ from the imperati!e %L;.8.5* tone patternK %87* 'i )7e)'e
washQIM1

ma0tti lu_
+#8)pot

wash the pot# Icitation form ma)0tti luJ

Ieamd005.0;8J

"he ne3t two e3amples show the < from $e spreadin$ further throu$h the followin$ word until another < is reached, at which point the spreadin$ stops , a phenomenon referred to as the S1lateau 1rincipleT by &isseberth and 2dden %200;K57* %i.e. a!oid a SdipT between two <s*. %89* t0i nda)
41)seeQ/6S

$e
+-G

ti 0 7aaruma)
+#5)flea

we didn't see a lea Icitation form ti 0 ) 7a)aruma)J %89* t0i nda)


41)seeQ/6S

Ieamy0;7.45;J

$e
+-G

ka0kaana*a
+#4)crab

we didn't see a crab Icitation form ka)0ka)ana*aJ

Ieamy0;7.0F9J

20 Irrealis prefi3es are always <.

4;4

#ertain low)tone morphemes seem to be resistant to spreadin$, such as $ender)marked pronouns in the complement of E1 %L7.2* and the person)marked ob'ect enclitics %L7.;*. #ompare %F0a* where the < from han SwhereT spreads onto ma)0nnu* SbirdT with %F0b* where m0e) SitT is unaffected. %F0* %a* han
where

ma0nnu*l
/C4)bird

%b* han

where

m0e)

A-41P#6

ka'a)l
now

where is the bird

where is it now I2009)0;);4.00;J

Similarly, the habitual tone pattern is unaffected by <)spread, despite be$innin$ with a 6 tone. %F4* ma)0waa
+#8)do$

ma)0si 0 ) $i nda)
A-4)<=()wait

the dog guards Itats004.004.022J It seems there is a distinction between tones which are specified as 6 %and hence do not chan$e*, and tones which are underlyin$ly unspecified, and only surface as 6 if a < has not already spread on to them.

%&&

Polar tone

"he tone on the copula is usually polar with respect to the pre!ious syllable, 'ust as in <ausa %Ja$$ar 2004K8F7)8F9*. %F2* %a* i )0$i nto) y0i %b* ko0ggo)m o) k0e
+#4)bat =G4)C6P +#;)doorway =G;)C6P

it's a doorway %F;* %a* p0pi ya v0i )

it's a bat %b* ko)0rudu


+#4)shelter

k0e)
=G4)C6P

+#9)$uineafowl =G9)C6P

it's a guinea owl

it's a shelter

"here are e3ceptions, howe!er, and this area is not well understood yet. In particular after a fallin$ tone the copula is 6, which is une3pected if fallin$ tones are to be sub'ect to the usual <6 seDuence analysis.

%&(

;exical tone in nouns


hile noun prefi3es are affected by

+ouns in #icipu ha!e le3ically)specified tone.

<)spread %L;.8.;*, this is a phonolo$ical process and does not depend on the syntactic

4;2

confi$uration. "here are also certain syntactic en!ironments where the tone on the noun prefi3 chan$es, but these can be analysed as the result of downdrift and spreadin$ %see L;.8.7*. "he followin$ table shows the distribution of nati!e #icipu roots accordin$ to their citation tone pattern24. 0able ,4- eistribution o noun roots according to tone pattern (one "attern < <6 6 << <6 6< 6 <6 66 <<< <<6 <6< < 6 <6 <66 6<< 6<6 66< /o. o' nouns 72 2F 8 2;2 497 ;F 49 49 ;5 ; 9 4; 5 4 2F 8 7$am"les ko)0joo Sli>ardT, u)0laa SfireT * SskinT d0d* ShorseT, u)0kwaa ka0ssa) SleafT, wu0nto) S$uest hutT ka)0da a SbushT, i )08ana SlineT ku)0$i `a) ScockT, ku)0jene) Sri!erT i )0 e)ye SriceT, ma)0$$i )ji i ScalabashT k)0k)y* SsnailT, k)0pp)8* Sfro$T ko0ggo)m o) SbatT, ka0kku)ti ) SshellT i )0$i $i pu SprayersT, ka)0ppi ti tu SbubbleT ;0lwli ) SspiderT, ko0orori ) SowlT ;0'i su)moo Sstin$in$ antT, ka)0sakka)lla Stree, k.o.T ko)0$i 'yo)'i *i SramT, u)0kudi )du*u Snettle treeT ka)0$aa$u)wa) Sfine loinclothT, ke0kkeru)ke) ShornbillT ka)0$i )tuwa Se3chan$eT ti )07a)aruma) SfleaT, ;0$i )$$ere) SstarT Sprayin$ mantisT m)0ri )gi )d Sna!elT, ka)0ga)ata)5

-!ery word has at least one < , this is a $eneral constraint affectin$ all #icipu nominal and !erbal words. +ote also that <6 is restricted to root)final position, a common cross) lin$uistic pattern %Bip 2002K29*.

%&9

Grammatical tone on verbs

In contrast to nouns, !erbs are inherently toneless, and the tones with which they surface are determined entirely by $rammatical properties such as mood or aspect. "his is also
24 "he tone pattern in the plural is almost always identical to that in the sin$ular.

4;;

the case for a number of (antu lan$ua$es, and such systems are sometimes called SpredictableT %&isseberth and 2dden 200;K54*. In %F8)FF* the !erb can be found in three different moods, dependin$ on the tone pattern. %F8* %a* u)dukwa) u0dukwam;?;
;S)$oQ#;0

> @ udu)kwa) u0dukwam?;


;S)$oQI##

he1she went %FF* %a* dukwa) ;0dukwam?;


2S)$oQ#;0

he1she should go > @ du)kwa_ dukwamaT


$oQI8P

you .sg.2 went

go#

In each case, the se$mental material remains the same, but the different tone patterns superimposed on the words $i!e rise to different $rammatical cate$ories, and hence different meanin$s. In %F8a* u)dukwa) heNshe wentT has a 6<6 tone pattern indicatin$ realis mood, while in %F8b* the same se$mental seDuence appears with a <66 tone pattern, this time indicatin$ irrealis mood. -3amples %FFa* and %FFb* illustrate a similar contrast, this time between realis and imperati!e moods, and in!ol!in$ the second person. In the realis mood the basic pattern is 6<%6*, althou$h in certain circumstances this surfaces as <6. "o be precise, if the !erb has a sub'ect a$reement prefi3 capable of bearin$ tone , i.e. %a* E or #E before a consonant)initial stem, or %b* a sonorant consonant before a !owel)initial stem* , then the tone is 6<%6*. If, on the other hand, the prefi3 consonant before a !owel)initial stem is non)sonorant %c*, or if the !erb has a null prefi3 %d*, then the pattern is <%6*. "he four possibilities are illustrated belowK %F5* %a* Ika)dukUa)L ka0dukwa0;?;
=G4)$o)/6S

> @ Iw) a(ja)L w0aya0;?;


;S)come)/6S

>$@ Ika(ja)L k0aya0?;


=G4)come)/6S

>d@ IdukUa)L ;0dukwa0?;


2S)$o)/6S

it went

it came

it came

you .s.2 went

Since there is not usually a distinct boundary between appro3imants and !owels, the seDuence of a 6)tone appro3imant followed by a <)tone !owel is, in practice, realised as a risin$ contour across both se$ments. 2n monosyllabic !erbs, which always ha!e a lon$ root !owel %at least on the surface*, 6 < 6 is realised as 6 <6, with a fallin$ tone 4;8

on the !erb rootK %F7* u)$a*a u0$aa0aTa


;S)$i!e)/6S

he gave Itats004.008.087J In the irrealis mood the basic tone pattern is <%6*, with < bein$ realised on the sub'ect a$reement prefi3. 0or !owel)initial !erbs such as ewe SrefuseT the seDuence of < and 6 is realised as a fallin$ tone <6 across the first syllable. "he difference between the realis %e.$. heewe) Sthey refusedT* and irrealis forms %e.$. he*ewe) Sthey should refuseT* of !owel)initial !erbs can be difficult to detect, but nati!e speakers are !ery clear about the distinction. "he difference is clear when the pitch track of the sounds is analysed usin$ a pro$ram such as 1raat, as can be seen in 0i$ures 29 and 29.

4;F

/--

h -

e*

e -./--512

#ime +s,

+igure 4`- &itch track or heewe) they re usedu .realis2

/--

h -

e*

e -.502-52

#ime +s,

+igure 4b- &itch track or he*ewe) they should re useu .irrealis2 4;5

In the imperati!e mood there is no sub'ect a$reement prefi3. "he basic surface melody is %6*< , in other words, the final tone is <, and any pre!ious tones are 6. %F9* du)kwa_ dukwa0aT
$oQIM1

go# Isaat002.002.828J Monosyllabic imperati!es ha!e a sin$le < toneK %F9* s_ s0T

drinkQIM1

drink# Isaat002.002.;0;J If there are suffi3es after the !erb root, then the < tone is realised on the last suffi3, as in %50*. "he imperati!e tone pattern can therefore be thou$ht of as Slinin$ upT at the ri$ht) hand ed$e of the !erbal word, in contrast to the realis and irrealis patterns, which line up at the left22. %50* ya)ana_ yaa0namaT

arri!e)E-+"QIM1

arrive# Itowards deictic centreJ Ieaim00;.4802J 0inally, there is an interestin$ interaction between !owel Duality and tone in imperati!es, in that if the final !owel of the !erb is NuN or CuC then the tone is realised as fallin$ rather than as <K * %54* %a* tuu_ tuumTa
pourQIM1

%b* $i )8 o)nu*_ $i8onuma a Ta


buryQIM1

pour#

bury# Ieamd005.008, eamd005.009J

%&<

Complement tone perturbation on nouns

= sin$le comple3 perturbation affects nouns within se!eral apparently unrelated syntactic constructions. "he affected positions include the SpossessorT +1 in the associati!e construction2; %L8.8.F.4*, the complement of E1, the complement position of
22 "he 6)tone plural imperati!e suffi3 0nA) is an e3ception %n8.5.2.;* 2; See #ahill %2000K8;* on +kem, where association is marked purely by a floatin$ 6 resultin$ in

4;7

the prepositionNlocati!e proclitic A %L8.F*, after the presentati!e n)duu %L8.;.;.8*, in the +1 complement of predicate locati!e clauses %L8.;.;.2*, and optionally after the e3istential predicator o)koo:a)kwai %L8.;.;.;*28. "he followin$ e3amples show the perturbation after the locati!e A, but the same patterns are found in each of the abo!e en!ironments. "he accents in %52)59* indicate relati!e pitch within each word, and not phonolo$ical tone. %52* %a* wu0nto)
+#7)shelter

op

%b* a1wuJ0nto)
62#R+#7)shelter

q o p

shelter %5;* %a* ko0ggo)m o)


+#4)bat

in the shelter Ieamy080.004.022J q pp %b* a1koJ0ggo)m o)


62#R+#4)bat

rqpp

bat %58* %a* d0d*


+#9)horse

by the bat Ieamy080.004.04FJ op %b* a1d0ds


62#R+#9)horse

q o p

horse %5F* %a* g0gweedi e)


+#9)mainProad

on the horse Ieamy080.004.044J o op %b* e1g0gweJedi J e)


62#R+#9)mainProad

q o o p

main road %55* %a* ka0kka)$i i


+#4)middle

on the main road Ieamy080.004.009J opo %b* a1kaJ0kka)$i Ji


62#R+#4)middle

q opo

middle

in the middle Ieamy080.004.007J

In each of these e3amples the citation tone pattern is e3actly preser!ed after the locati!e, $i!in$ rise to three pitch le!els. "he data can be accounted for if we assume that downstep has occurred immediately after the locati!e morpheme, resultin$ in a downward resettin$ of the pitch re$ister from that point onward. =ll the nouns in the abo!e e3amples be$an with a < tone. 0or nouns be$innin$ with a 6 tone the root tones are preser!ed from the first < tone onward, but initial 6 tones raise to the same le!el as subseDuent <sK
subseDuent downstep. 28 See Marten %200F* for a similar phenomenon in the (antu lan$ua$e <erero, where complement case is found in the complement of E1 and of prepositions, as well as in presentational constructions.

4;9

%57*

%a* mo)0ri ngi no


+#8)spine

pqqq

%b* o1mo)0ri )ngi )no)


62#R+#8)spine

rqqqq

spine %59* %a* k0ke)eke


+#9)bicycle

on the spine Ieamy080.004.042J pq %b* e1k0ke)eke)


62#R+#9)bicycle

rqq

bicycle

on the bicycle Ieamy080.004.00FJ

"his can be accounted for by assumin$, in addition to the downstep 'ust mentioned, the < from the locati!e spreads ri$htwards , in which case the phonolo$ical representations of the %b* e3amples would be A,1mo0ri ngi no and A,1kee0ke. "he deri!ation is $i!en in %59*K %59* Anderlyin$ formK A1mo)0ri ngi no <)spreadK A1mo0ri ngi no .ownstepK q p q q q q q q q q q p p p p

A,1mo0ri ngi no

Since the same perturbations occur in all the other syntactic en!ironments mentioned abo!e, the same rules can be applied. In some of these en!ironments the tone immediately before the perturbed noun is <, as with the locati!e proclitic. In other constructions, particularly the complement of E1, the precedin$ tone is 6. If the +1 in complement position be$ins with a < tone, then this < is the same pitch as the last 6 of the !erb , this is therefore an e3ample of total downstep %#onnell and 6add 4990K2F*2F. In this case it must be admitted that it is difficult to see what the tri$$er for the necessary <)spread mi$ht be. More $enerally, the fact that this seDuence of rules has to be in!oked in apparently unrelated syntactic en!ironments su$$ests that the analysis presented here is missin$ a $eneralisation. It should therefore be re$arded as a steppin$) stone to a better solution.

".-

1o$el harmony

"here is widespread !owel harmony in #icipu. Eowels from the set o, , e, a are mutually e3clusi!e in roots, re$ardless of word class. So if a root contains NaN, its other !owels must come only from the set a, i, u, NeN only occurs with e, i ,u, and so on. "he !owel harmony system operates throu$hout the le3icon, without e3ception apart
2F =s was stated in the front matter, downstep between words is not marked in the e3amples in this thesis. So, for e3ample, ob'ects occurrin$ in the complement of E1 %the unmarked position for ob'ects* and marked as < < are actually at the same pitch as the precedin$ 6 of the !erbal word.

4;9

from some compounds %L;.F.4* and loanwords %L;.F.2*. Many prefi3es also contain harmonisin$ !owels %L;.F.;*.

%(1

$istribution of vowels in C1C1 noun roots

"able 4; below shows the distribution of !owels in #E#E noun roots. 0or simplicity the table only include roots where both !owels are short and oral, but the same restrictions hold for lon$ and nasal !owels, for both nouns and !erbs , !owel harmony is absolute in nati!e roots. "he empty cells are shaded to hi$hli$ht the pattern. 0able ,6- ^owel cococcurrence restrictions in 5^5^ noun roots where ^, and ^4 are both short oral vowels .^, down the le t, ^4 along the top2 i i e a o u
40 F 9 ; 8 ; 4 8

e
4 9

a
45

o
4;

u
48 8

29 4; 9 9

42 5 4 4;

#ompound words are not always sub'ect to harmonisation , the followin$ words both contain NeN and NaNK %70* mengeta)ari ) me0nge0t0a0ari )
+#8)child)M)+#2)man

boy %74* kwakulle) 0ku0lle) ku0a'a


+#9)day)=G9)that

then Ilit. Sthat dayTJ Eowel harmony occasionally operates beyond the word in normal, fast speech %particularly when !owel coalescence occurs at a word boundary , see L;.7.4*.

%(#

;oanwords

(orrowed noun and !erb roots !ary with respect to harmonisation, and the outcome seems to depend on how deeply entrenched the word is in the speech community. 480

.emonstrably modern borrowin$s such as roo a) SplasticT may occur with conflictin$ !owels, but in $eneral harmonisation does take place, as illustrated in %72* and %7;*. %72* ru)uka kollo kwaanu) %7;* gwede rooto) from roka from kallo from kwano from gode from reto Schattin$T SshirtT SlookT Smetal containerT SthankT Shan$in$T me)0ttegu) from taggo

+ormally it is the first !owel that chan$es, althou$h not e3clusi!ely as kwaanu) in %72* shows. "he e3amples in %7;* show that true harmonisation is in!ol!ed, rather than simply the !owel)raisin$ which typically occurs when words are borrowed from <ausa to #icipu %L;.;.4*. "here is no difference in !owel hei$ht between gwede and gode, or between rooto) and reto. Instead the !owels ha!e harmonised SsidewaysT.

%(%

Affixes and clitics

Many nominal and !erbal affi3es harmonise accordin$ to the pattern outlined abo!e. +ouns from three of the nine #icipu noun classes, classes 4, 2, and 8, are formed by addin$ to the root the harmonisin$ prefi3es kA0, A0, and mA0 respecti!ely. "he Duality of the A0 !owel is determined by the root !owels as followsK If the root contains NeN, then the prefi3 !owel will be IeJ If the root contains NoN, then the prefi3 !owel will be IoJ If the root contains NN, then the prefi3 !owel will be IJ 2therwise the prefi3 !owel will be IaJ

"hese rules do not need to be ordered, since the stron$ !owel harmony operati!e in the le3icon ensures that the four !owels in!ol!ed are mutually e3clusi!e. It should howe!er be pointed out that when speakers pronounce a word carefully, they often pronounce the prefi3 IkaJ %or IaJ, or ImaJ*, re$ardless of the root !owels which are to follow. in this thesis such !owels are written A as in kA0BG.
25 =nderson %4990a* briefly discussed prefi3)root !owel harmony in nouns for the -ast &ain'i lan$ua$e =mo. <e states that "hou$h this !owel harmony may pro!ide a phonetic Star$etT, considerable !ariation still e3ists e!en on indi!idual words %4990aK4F7*. "his statement nicely captures the situation in #icipu.

ith that

in mind, I ha!e analysed the underlyin$ !owel in these harmonisin$ prefi3es as N aN, and

484

In addition to noun prefi3es and the correspondin$ $ender a$reement prefi3es, se!eral other inflectional and deri!ational !erbal affi3es contain harmonisin$ !owelsK 0able ,J- Harmonising a i*es A''i$ A0 0wA 0wA 0wA 0nA 0nA 0nA 0kwA %78* -loss ;11 a$reement applicati!e anticausati!e separati!e perfecti!e !enti!e plural imperati!e suffi3 for borrowed !erbs %a* a0du)kwa)
3P)$oQI//

#e'erence 7$am"le L7.8 L8.5.8.2 L8.5.8.; L8.5.F.; L8.5.;.4 L8.5.F.4 L8.5.2.; L8.5.5 %78* %7F* %75* %77* %79* %79* %90* %94* %b* 0d)n)
3P)followQI//

they should go %7F* %a* ti )0yaa0wa)


41)doQ/6S)APP;

they should ollow Isaat004.005.044, saat004.005.088J %b* mi 0 do)onu)0wo)


=GF)sitQI//)APP;

we did to him %75* %a* ma)si 8u)wa) ma)0si 8u)0wa)


=G8)heatQ/6S)A/(IC

may they stay with you Isaat004.005.448, saat004.002.025J %b* golu)wo) ;0golo)0wo)
=G9)cut)A/(IC

it water spoiled Ilit. $ot hotJ %77* %a* u)0halu)0wa)


;S)coilQ/6S)S-1

he1she gets cut Itats00F.002.09;, tats008.00;.00FJ

%b* u)0'um u)0wo)


;S)closeQ/6S)S-1

it uncoiled %79* %a* a)0dukwa)0na)


;1)$oQ/6S)P>,

it opened Ieaim040.448, saat004.009.097J %b* ku)0'i ngo)0no)


=G9)$oPhomeQ/6S)P>,

they had gone %79* %a* seke)0ne)


releaseQ/6S),7/(

it had gone home Isaat004.005.009, saat004.005.099J %b* u)0yuwo)0no)


;S)fallQ/6S),7/(

he released down

he ell down I"idipo, saat002.00F.085, saat002.00F.075J 482

%90*

%a* yaa0na) _
doQIM1)P;.I8P

%b* se)ke)le)0we0ne)_
mo!eQIM1)=+"I#)P;.I8P

you.pl.2 do# %94* %a* u)0gwaanu)0kwa)


;S)understandQ/6S);2

you.pl.2 move over there# Isaat004.002.407, saat002.002.472J %b* u)0tru)0kw)


;S)pushQ/6S);2

he understands

he pushed Ieaim040.405, eab$004.428J

In e!ery other affi3 the !owel is either NiN or NuN, which are neutral with respect to !owel harmony. <armonisation often Spasses throu$hT these non)harmonic affi3es. In %92* the ;11 sub'ect prefi3 A0 harmonises with the root 7e'e SdanceT, despite the inter!enin$ habitual prefi3 si). =ll affi3es with neutral !owels in #icipu appear to be StransparentT rather than SopaDueT with respect to !owel harmony. %92* ) 'e) e)0si 0 ) 7e)'e07e
3P)<=()dance1#75DP

they dance Isahs004.002.044J 0inally, clitics show mi3ed beha!iour with respect to !owel harmony. "he locati!e proclitic A %L8.F* and the associati!e a$reement proclitics %L8.8.F.4* harmonise with the followin$ word, as shown by %9;* and %98* respecti!ely. "he latter e3ample shows the same Spassin$ o!erT of a neutral !owel as %92*. %9;* %a* e,1k0ke)eke
;6CR+#9)bicycle

%b* o,1ko0o$i )
;6CR+#9)hole

on a bicycle %98* ka)0manga ko,1ku0hoomo)wo)


+#4)rope A-1R+#9)tree@k.o.A

into a hole Itapf004.00F.044, s!tm$004.477J

rope o the kuhoomowo tree Ieaim005.48F8=J 2n the other hand, the person)marked ob'ect enclitics 8 o) %211* and re) %;11* do not harmonise with their hosts %L7.;.2*.

%(&

Cross.linguistic comparisons

"he #icipu !owel harmony system is essentially identical to that of #entral &ambari %<offmann 4972K78*, and !ery likely the other &ambari lan$ua$es as well. "he system 48;

is distinct from those usually reported in the literature %inc. .ettweiler 2000 on #:6ela*, in that the harmonisation rules cannot be stated strai$htforwardly usin$ distincti!e features. Eowels do not harmonise accordin$ to hei$ht, or roundness, or backness , instead, complete assimilation is in!ol!ed. "he most concise way to state the rule is as followsK %9F* ' there are two chigh vowels in a phonological word then they must be identical

".2

/asalisation

e ha!e already seen that #icipu has a full complement of nasal !owels, both short and lon$, as well as two nasal consonant phonemes NnN and NmN. +asalisation is not usually confined to sin$le se$ments, howe!er, and it may spread a considerable distance. "his section is concerned with the phonemes affected by the process and the direction of spread.

%91

Phonemes affected

In addition to the si3 short and lon$ nasal !owels, the appro3imants N wN and NyN ha!e nasal allomorphs Iw J and IyJK %95* ) (w aL %a* Ima`Ua ma0gwaawa
+#8)bruise

%b* Ii )h u)h u *L ) i0 ) yuyu*


+#;)fly

bruise

lies Ii.e. insectsJ

Eowels which ha!e become nasalised due to the influence of other nasal se$ments do not ha!e the same properties as underlyin$ly nasal !owels , for e3ample, the former do not tri$$er prenasalisation %L;.4.F*, as illustrated by words such as mita Imi taJ SsDuee>eT and ku)0mo$i ) Iku)mot%i )J Sold womanT. "herefore the ri$htward spread of nasalisation must occur after prenasalisation , in technical terms they are in a Scounter)feedin$T relationship. Some roots be$innin$ with a nasal such as Nmi toN Imi ntoJ Sshut mouthT do we^on[^YnYY{q}q^wqYYopq{qYnrqYn{{}iq~YoeYpn\qY}^~qrfg[^fgY^n{nfY\evqf{V /oot)internally no contrast has been found between oral and nasal !owels before a nasal consonant. <owe!er this contrast e3ists across morpheme boundaries, and since there are four !erbal suffi3es be$innin$ with NnN it is relati!ely common. #ompare the !erbs yaa SdoT and yaa Sarri!eT, shown before the !enti!e suffi3 0na in %97*K

488

%97*

* ) %a* Iu)ja(na * u)0yaa0na)


;S)doQ/6S)E-+"

%b* u)ja*(na)L u)0ya*a0na)


;S)arriveQ/6S)E-+"

he did

he arrived Ieamy0;4.2;FJ

%9#

$irection

"he spread of nasalisation from the nasals NnN and NmN is predominantly to the ri$ht, althou$h spread to the left does occur. -!ery !owel occurrin$ immediately after a nasal consonant is nasalised, whereas nasalisation to the left is more !ariableG when it can be detected, it often seems to be weaker than nasalisation to the ri$ht. If the first !owel in a ) Sni$htT and y0yuyu ) * Ih (u)h u*J SflyT. root is nasal, then the second usually is too, as in ka)0hi'i "his a$ain su$$ests spread to the ri$ht is predominant. +asal spread seems to be confined to the word. SsnakeT. =lthou$h in $eneral nasality spreads to the ri$ht, it can spread to the left from verb roots onto !erb prefi3es. "his leads to contrast between nominal %a* and !erbal %b* pairsK %99* %a* ka)0taa
/C1)shoe

ithin the word, spread to the

ri$ht is blocked by +# clusters as in u)0lenji Iu)0lend&i L SsunT and ka)0 ungu Ika)0 u^`uL

* %b* ka)0taa
A-1)shootQ/6S

shoe %99* %a* a)0taa


/C

he Ika)0 ara Sold manTJ shot I2009)02)0F.004J * %b* a)0taa


3P)shootQ/6S

)shoe

shoes

they shot I2009)02)0F.004J

".3 %<1

Morphophonemic processes Coalescence and elision

hen !owels become 'u3taposed as a result of the concatenation of two morphemes %either word)internally or across word boundaries*, the !owels often coalesce, resultin$ in a lon$ !owel27, usually bearin$ the Duality of the second. ord)internally, we ha!e already seen %L;.4.4* that !owel)initial noun and !erb roots coalesce with prefi3es to form lon$ !owels, with the second %i.e. root* !owel
27 "his is not the case in #entral &ambari , when two short !owels come to$ether the outcome is still a short !owel %<offmann 4972*.

48F

Duality dominant. =nother type of word)internal elision occurs in associati!e constructions %L8.8.F.4*, in which an a$reement clitic attaches to the noun prefi3 of the SpossessorT +1. If this prefi3 be$ins with a !owel then coalescence occurs, with the second !owel a$ain dominantK %90* ) It%i )me'i ) $i )0me'i
+#5)in

ta(hu)la) ti , 1a0hula
=G5R+#2)name

ha(vu)L ha0avu)
=G2)2S.12SS

in your .sg.2 name Ilit. Sthe inside of your nameTJ, citation form a)0hula SnameTJ Isaat004.004.044J 2ccasionally coalescence fails to happen in the associati!e construction, and a $lottal stop appears before the noun prefi3 as in %94*, 'ust as if it were utterance)initial. It is not known what, if anythin$, $o!erns the occurrence or non)occurrence of coalescence here. %94* Imo)ni mo)0ni
+#8)water

s(a)L ma'aJ ma,1a0ssa)


=G8R+#2)leaf

the colour green Ilit. Swater of lea!esT, citation form a0ssa) Slea!esTJ Ieamy004.00FJ "he con'unction n) SandNwithT and the locati!e A also coalesce with !owel)initial stems, and a$ain the second !owel in the seDuence is dominantK %92* Yi L Ina)(kant%i ni )1a)0kan$i ri
andR/C )fin$ernail

with ingernails Ieaim005.48F0J %9;* Iu(Ye)L A,1u0ree


;6CE/C7)town

in the town Ieamd00;.044J 2ne en!ironment where coalescence mi$ht be e3pected, but does not seem to occur, is pre)prefi3es %LF.;.5* , when a noun prefi3 is attached to a stem already consistin$ of a prefi3 and noun root. In this case, the !owels ne!er coalesce , as in %98*, a $lottal stop appears before the inner prefi3.

485

%98*

Ime)'i )Yi L me)0i )0ri


+#8)+#;)thin$

a small thing Icitation form i 0 ) ri Sthin$TJ

Isamoh004.422J

#oalescence also occurs across word)boundaries, in which case the Duality of either !owel may dominate, althou$h there is a preference for the second. -3amples %9F)404* show coalescence in a !ariety of syntactic en!ironments, while %97)404* additionally demonstrate the elision of NwN, NyN and N hN. %9F* )(8a)L Ia)nahi a)0naha), i 08aa
;1)lea!eQ/6S
+#;)$round

they le t the land IE-/( #2M16-M-+", samoh004.027J %95* Isu(8aJnga)L see


until

u)08anga)
+#7)tree

%ust a tree I#2+JA+#"I2+ +2A+, samoh004.059J %97* I'asuJ(na)L 'asu)


place%+#7*

wu0na)
=G7)/-6

the place I+2A+ /-6="IEIS-/, samoh004.029J %99* Ii )namelle*(na)L i )0nama) yi 0lle)


+#;)meat =G;)that

yi 0 na)
=G;)=/"

that meat I+2A+ .-M2+S"/="IE- =/"I#6-, "idipo, saat002.00;.024J %99* `o)L Ii )tumo('u i )0tumo
+#;)pre$nancy

yi )0'ugo)
=G;)throwQ/6S

there was a miscarriage ISA(J-#" SA(J-#" 1/-0I, first !owel dominates, eamd048.282J %400* Ia) aYa(vUo*(L a)0 ara
+#2)oldPman

ho0vo*o
=G2)4S.12SS

my old men I+2A+ M2.I0I-/, first !owel dominates, eamy00F.0;9J 487

%404* Ia)7i (ke)(keL a)07a ha,1i 0 ) ke)eke


+#2)person =G2R+#;)bicycle

cyclists Ilit Speople of bicyclesTJ IS12SS-SS-.T S12SS-SS2/T, tapf002.004.020J In some cases %but by no means all* neither !owel predominates and the two !owels appear to ha!e Smet in the middleT. So far, this has only been obser!ed for the !owels NaN and NiN resultin$ in IeJ as in %402)40;*, but it may be that the same process can occur with back !owels as well %this is the case for #entral &ambari , <offmann 4972*. %402* IhaYe)(hUa(Ya)L hari )
until

a)0hwaara)
;1)startQ/6S

then they started Isamoh004.025J )L %40;* I e(ta^`i aa i0 ) tangi )


+-G +#;)item

no items Isamoh004.440J Some of the most common types of relati!e clauses in!ol!e both elision and coalescenceK 'asu) wuna) I'asuJ(na)J SwhereT %lit. Sthe place thatT*, i )ri yi na) Ii )Yi (na)J SwhatT %lit. Sthe thin$ thatT*, and a)7a hana) Ia)7a(na)J Sthe ones thatT. In normal speech the elided and coalesced forms are more usual, but they are readily understood by nati!e speakers as contractions of the lon$ forms.

%<#

u.anticipation

#icipu has three homophonous 0wA suffi3es %applicati!e, anticausati!e, separati!e , see L;.F.; for e3amples*. 2n the application of any of these prefi3es the final !owel of the stem to which it is attached chan$es to IuJ, re$ardless of its underlyin$ Duality29.

29 = similar morphophonolo$ical process is found in <ausa when the feminine suffi3 caa is attached to a stem endin$ in o e.$. sabo Snew %masc.*T !s. sabuwa Snew %fem.*T %Ja$$ar 2004KF9*.

489

0able ,T- u0 anticipation be ore 0 wA su i*es Stem dama l 'p golo hala 'um o Gloss tell look or hold cut coil close =pplicati!e damuwa luw 'puw goluwo haluwa %uncoil* 'um uwo %open* =nticausati!e Separati!e

"he same process occurs with the prefi3 kwA0 found on borrowed !erbs %L8.5.5*K %408* gwaanukwa understand daamukwa koyuko dennukwe worry teach compress from <ausa gane from <ausa dama from <ausa koya from <ausa danna

"his chan$e of !owel is the result of a morphophonolo$ical rule rather than a phonolo$ical one, since it only happens on the application of a suffi3 be$innin$ with a labialised or labio!elar consonant. ithin le3ical roots other !owels are readily found before labial consonants, as illustrated belowK %40F* lawa weewe siiwa w escape re use untie steal

%<%

i.anticipation

2ne of the more unusual features of #icipu $rammar is the split in !erb and pronominal a$reement between paradi$ms that inflect for person and those that inflect for $ender %see 1art IE*. 2f interest here is the distinction between the homophonous. <owe!er consider the followin$ e3amplesK
=G9

noun class pronoun

%L5.2.4F* and the ;1S ob'ect clitic %L7.;*, both vi ). Internally, the forms are

489

%405* %a* =K han

where

Audu)l
=udu

%b* =K han

where

0vooto)
+#9)$oat

where's Audu (K mi ndi )vi ) m0i nda)1vi )


4S)seeQ/6SE30.P#6

where's the goat (K mi nda) m0i nda)


4S)seeQ/6S

vi ) v0i )
A-)1P#6

' saw him

' saw it I2007)02)0F.007J

"he person)marked clitic in %405a* tri$$ers a morphophonemic spreadin$ process which the $ender)marked pronoun in %405b* does not. "he NiN from the clitic vi ) spreads leftward, chan$in$ the last !owel of the precedin$ !erb comple329. "his happens without fail for e!ery !erb, e!en monosyllabic !erbs, and the resultin$ conte3tual neutralisation can $i!e rise to le3ical ambi$uityK * ) %407* %a* n)tiivi * n)0taa1vi )
4S)s!ootQ/6SR;S.1/2

* ) %b* n)tiivi * n)0tuu1vi )


4S)"ourQ/6SR;S.1/2

' shot him

' poured . or2 him IMB, 2007)0;J

hen combined with the ubiDuitous !owel coalescence %L;.7.4* there can be considerable displacement of !owel DualitiesK %409* It%i (vu)(Ye)L $aa1vi )
$i!eQIM1R;S.1/2

u)0ree
+#7)town

give him a town# Isaat004.00F.0F9J

29 =n alternati!e analysis would be to assi$n the ;1S ob'ect clitic the underlyin$ representation ivi, with the !owel at the end of the pre!ious !erb deletin$. Since all !erbs end in !owels and the clitics only occur immediately post)!erbally either analysis is possible. <owe!er an e3planation would be reDuired for why deletion takes place rather than coalescence %compare %405a* with %9F* which has a lon$ Ii(J*. =lso all other ob'ect clitics are #E monosyllables rather than E#E.

4F0

Chapter & A grammatical sketch of Cicipu


*.1 Introduction

=s with the phonolo$ical sketch in the last chapter, the purpose behind this chapter is twofold. 1rimarily it is intended to familiarise readers with the basic morphosyntactic structure of #icipu. <a!in$ di$ested this chapter, they will then be better eDuipped to understand the more in)depth analyses in the parts that follow. "he discussion does howe!er co!er more $round than is strictly necessary to meet this first $oal. Since the lin$uistic description of #icipu is in its infancy, a second $oal is to present a $rammatical sketch of the lan$ua$e, co!erin$ areas which, althou$h not of direct rele!ance to 1arts III and IE, are likely to be of interest to typolo$ists and =fricanists alike. +e!ertheless, limitations of time and space ha!e meant that much of the ar$umentation here does not meet the ri$orous standards that %ideally* apply to the publication of full reference $rammars. In particular, much of the analysis deri!es from inspection of the corpus rather than from metalin$uistic discussion with nati!e speakers. #onseDuently the sketch has more to say about what does occur than what does not. "here is a certain amount of una!oidable o!erlap with later chapters. (ecause $ender a$reement is so ubiDuitous in #icipu, many of the word classes introduced in this chapter will also be indi!idually discussed in chapter 5 on a$reement tar$ets. My approach is to deal with structural matters in this chapter %e.$. modifier takes a low) tone $ender a$reement prefi3*, and lea!e class)specific phenomena %e.$. the chapter 5. "he chapter is or$anised as follows. In L8.2 I $i!e a brief typolo$ical o!er!iew of the lan$ua$e. I then take a more detailed look at the structure of !erbal and non)!erbal clauses %L8.;*, nominals %includin$ ad!erbs and the structure of the +1* %L8.8*, prepositions %L8.F*, !erbs and the !erb phrase %L8.5*, ad'ecti!es %L8.7*, numerals %L8.9*, and Duantifiers %L8.9*.
=G9

prefi3

for modifier has allomorphs = and ( occurrin$ in the followin$ circumstances...* to

*.

Typological overvie$

#icipu $rammar has a lot in common with the well)known (antu sub$roup of (enue) #on$o4. "here is a robust noun class system with a$reement on a whole host of tar$ets,
4 Johnston %4949K47* wrote #uriously enou$h, there are lan$ua$es...in +i$eria...the synta3 of

4F4

as well as a producti!e system of !erbal e3tensions. In what follows, freDuent reference will be made to similar or identical constructions in (antu. 6ike most (enue)#on$o lan$ua$es, #icipu is head)initial at clause and phrase le!el, and mainly head)markin$2. =ccordin$ to the traditional terminolo$y it is hi$hly a$$lutinati!eG in (ickel and +ichols: %2007* terms most formati!es are non)fusional, non)fle3i!e, mono)e3ponential, and synthetic. #icipu is not a true polysynthetic lan$ua$e in the sense of (aker %4995*, since there is no noun incorporation and %unlike (antu* no ob'ect a$reement. 1refi3es and suffi3es occur on both nouns and !erbs, with prefi3es predominant on nouns. Suprase$mental modification %L8.5.2* and infi3ation %L8.5.4* are also found. "he former is common across =frican lan$ua$es %#reissels 2000*, not 'ust (enue)#on$o. Infi3ation is rarer, and ar$uments will be $i!en in L8.5.4 that the morphemes concerned are true infi3es, rather than 'ust word)internal affi3es. /eduplication is common and found in se!eral word classes, includin$ nouns, !erbs, ad'ecti!es, demonstrati!es, and ideophones. Stem modification other than by tonal chan$e is !ery rare in nouns %see LF.2.2 and LF.2.7 for two e3amples*, but certain !erbal cate$ories are consistently e3pressed in this way %L8.5.;.8*. Morpholo$ical processes in #icipu are in $eneral hi$hly re$ular, and suppletion is unattested. "here are formally)identifiable open classes of nouns and !erbs, and a small closed %but still formally)distinct* class of ad'ecti!es. =s for constituent order, #icipu is a confi$urational SE2 lan$ua$e, and is mainly consistent with respect to the correlations obser!ed by Greenber$ %495;*, althou$h +1)internal synta3 is unusual %L8.8.F.5*. If one only looks at the ScoreT of #icipu $rammar %e.$. !erbal morpholo$y and the noun class system* then, as mentioned abo!e, the lan$ua$e is !ery similar to many other (enue)#on$o lan$ua$es. <owe!er it should be stressed how much of the $rammar %in a broad sense of the word e.$. as concei!ed by construction $rammar or co$niti!e lin$uistics , see #roft and #ruse 2008* is influenced by the lin$ua franca <ausa, spoken fluently by !irtually all #icipu speakers. #icipu is a prime candidate for study for lin$uists interested in the influence of lan$ua$e contact on $rammarG e3amples of likely contact effects can be seen in L8.;.4.2, L8.;.;.;, L8.;.F.4, L8.;.5.4, L8.;.7, L8.8.4, L8.8.F.8, L8.F, L8.7.4, and L8.9)8.9, as well as many other places in other chapters.
construction of which freDuently recalls the (antu idiosyncrasyG but the word)roots of the !ocabulary would be found wholly dissimilar. 2 "he dependent)marked associati!eNpossessi!e construction %L8.8.F.4* is an e3ception.

4F2

Grammatical sketches of (enue)#on$o lan$ua$es spoken in northern +i$eria often fail to take into account possible influence from <ausa, and so for the benefit of readers unfamiliar with that lan$ua$e, this chapter is liberally sprinkled with references to the two main -n$lish)lan$ua$e <ausa $rammars, +ewman %2000* and Ja$$ar %2004*.

*."

)lause-level syntax

"his section be$ins with a discussion of $rammatical relations %L8.;.4*, which then informs the subseDuent account of constituent orderin$ %L8.;.2*. "he remainin$ sections co!er non)!erbal clauses %L8.;.;*, ne$ation %L8.;.8*, Duestions %L8.;.F*, and then finally clause combinations %L8.;.5*.

&%1

Grammatical relations

"here is a clear distinction in #icipu between the sub'ect and the other ar$uments of the !erb, where by Ssub'ectT I mean the sin$le ar$ument of an intransiti!e clause, or the more a$enti!e ar$ument of a transiti!e clause. "hree kinds of e!idence can be offered in support of this distinction;. 0irstly, only sub'ects can tri$$er a$reement on the !erb. Secondly, only sub'ects can occur pre)!erbally without an accompanyin$ copulaG for +1s with other $rammatical relations to occur before the !erb, they must be followed by the copula, and the !erb has to ha!e perfecti!e aspect if the clause is in realis mood %see L8.;.2 for details*. #on!ersely, only non)sub'ects can occur after the !erb. "hirdly, the sub'ect has to be omitted in #icipu infiniti!e clauses %LF.8.4*, while the ob'ect may be included8. +1s with a wide ran$e of semantic roles are found as sub'ects in intransiti!e clauses includin$ apply in #icipu. =lthou$h there is no ob'ect a$reement on the !erb, there are other tests for ob'ecthood in #icipu. If we find that the sub'ect +1 in one kind of construction consistently corresponds with the filler of a particular non)sub'ect +1 slot in a structurally)deri!ed counterpart construction, this is e!idence that the +1 picked out in the counterpart construction is an ob'ect %see #roft 4994K5)9 for discussion*. In #icipu the applicati!e %L8.5.8.2* and the anticausati!e %L8.5.8.;* both pro!ide the reDuired
; =s in (enue)#on$o as a whole, there is no case)markin$ in #icipu. 8 See =nderson %4975* and #hun$ %4999K58)59* for discussion of this test.
=G-+", 1="I-+", /-#I1I-+", "<-M-,

and

-1-/I-+#-/.

"he restriction of

sub'ect referents to topics found in some (antu lan$ua$es %Morimoto 2009* does not

4F;

correspondence. "here is not space for detailed discussion here, but the $ist of the ar$ument with respect to the applicati!e is that the sub'ect of the intransiti!e !erb in sentences such as %4a* always coincides with the immediately post)!erbal +1 in correspondin$ applicati!e constructions like %4b*, rather than with, say, an obliDue. %4* %a* ;0 i iro) vi )0tanna)
/C)1"en =G9)descendQ/6S

%b* ta)nna)0wa

descendQIM1)=116

;0 i iro)#
/C)1"en

the biro descended

lower the biro#

I2009)02)49.004J

"he cross)lin$uistic patternin$ of such constructions su$$ests that the correspondent in %4b* is a $rammatical ob'ect. "he same line of ar$ument can be applied to the anticausati!e %L8.5.8.;*. "he bindin$ of refle3i!es %L8.;.4.2* is another possible source of e!idence for ob'ecthood, but the rele!ant elicitation has not been carried out. 0inally, two other cross)lin$uistically applicable tests ser!e to distin$uish ob'ects from other types of +1s. 0irst, they are able to occur clause)internally without a preposition. Secondly, no other +1 in the clause may bear the same semantic roleF.
&%11 $ouble.ob,ect constructions

=s in many lan$ua$es, double)ob'ect constructions e3ist where one of the ob'ects bears the semantic role of
/-#I1I-+"

and the other that of

"<-M-.

"he !erbs $aa and n, both

meanin$ S$i!eT, are perhaps the clearest e3amples of this. In recipient)theme double) ob'ect constructions the primary ob'ect %i.e. the one closest to the !erb* bears the
/-#I1I-+"

role, while the secondary ob'ect encodes the "<-M-. see


unless

%2*

ni )

and

$0$a*a

2S)$i!eQ/6S I+#9)person

I707a

n0na)

=G9)/-6

wu0u0ka) a)0wa)0na)1vu)L :

;S)0A")takeQI//)=116)E-+"R2S.1/2J

Ika0

I+#4)paymentJ

i )ya)L

you have to give the person who will bring it or you 1 payment Itats004.002.022J If the primary ob'ect is pronominal and a$rees in person %rather than $ender*, then it takes the cliticised form %L8.8.;.;*, as illustrated in %;*. 1ronominal secondary ob'ects are always of the SindependentT kind %L8.8.;.4)8.8.;.2*.
F In some lan$ua$es ad'uncts form SislandsT with respect to relati!isation, which allows them to be distin$uished from the direct ar$uments sub'ect and ob'ect. <owe!er in #icipu ad'unct)internal +1s are readily relati!ised %L8.;.8*, and so this test is not applicable.

4F8

%;* kungwa

God%+#9*

u0n)18 o)

;S)$i!eQI//R21.1/2

v0i )

=G9)1/2

Iay God give you.pl.2 it Ieab$004.04;J "he primary ob'ect is not limited to the %8*, (-+-0I#I=/B %F* or SM=6-0I#I=/BT %5*. %8* n)0pi isa)nu)
4S)sprayQ/6S
/-#I1I-+"

semantic role, and may also be a

G2=6

a0aya)
/C

1bean

kwe0etu

+#9)medicine

' sprayed the beans with ertiliser %F* a)gooni v ) i) a)0goonu)1vi ) u)8angu)la) u)08angu)la)
+#7)$ather

IG2=6, 2007)04)40.04FJ

;1)helpQ/6SE30.P#6

they helped him with gathering I(-+-0I#I=/B, tapf002.002.04FJ %5* u0la)ngwa) 707a
;S)spoilQI//
/C)1"erson +#9)lun$

;0hu)uhu n0na)

=G9)/-6 =G9)beQ/6S)10E

;0yo0no)

$i 0 ) me'i )

+#5)inside

it spoils or a person the lungs that are inside ISM=6-0I#I=/BT, tats007.002.0;5J It is possible %in fact, usual* for either ob'ect to be omitted if the conte3t is clearK %7* u)0$a*a
;S)$i!eQ/6S

; m0

+#8)child

me0evi )

=G8);S.12SS

he gave his child Ii.e. to someone elseJ I1/IM=/B 2(J-#" 2MI""-., tats004.008.087J %9* * ) u)tiivi u)0tu*u1vi ) ;

;S)pourQ/6SR;S.1/2

he poured ruit or him IS-#2+.=/B 2(J-#" 2MI""-., tapf004.00;.027J "he !erbs in the e3amples abo!e admit double ob'ects without any morpholo$ical chan$e. <owe!er a much wider ran$e of !erbs ha!e the potential to take double ob'ects by means of the !alence)increasin$ applicati!e suffi3 )wA %L8.5.8.2*.
&%1# !eflexives

"here is no refle3i!e pronoun in #icipu, nor are there the refle3i!e !erbal affi3es common to (antu. Instead refle3i!es are e3pressed usin$ the associati!eNpossessi!e phrase ka)0tii* kA0tttt, literally Shead ofPPPPT. "here is a direct analo$ue in <ausa %Ja$$ar

4FF

2004K;94);9F*, but not too much should be made of this , after all, the Shead ofT construction is also cross)lin$uistically fairly common %&ni$ and Siemund 200F* and so the similarity may be coincidental rather than a contact effect. %9* u)ra*asa) u)0ra*QisRa ka)tii* ka)0tii* keevi ) kA0evi )

;S)eat#=ASsQ/6S

+#4)head

=G4);S.12SS

he served himsel Isamoh004.220J -mphatic refle3i!es are formed by addin$ the preposition n) SwithT before ka)0tii* kA0tttt. hen used ad!erbially the resultin$ phrase has an Se3clusi!eT interpretation, as in %40*. %40* huupi )si )lo) hu0u0pi )sQi )lRo) n) n) ka)tii* ka)0tii* ki ive) kA0ive)

;1)0A")break16=#sQI//

with

+#4)head

=G4);1.12SS

they will break by themselves Isamoh004.400J -mphatic pronouns are formed usin$ a construction identical to one in <ausa %Ja$$ar 2004K;9F);99*K an independent pronoun, optionally followed by the con'unction SandNwithT, and then the phrase Shead ofPPPT. %44* ka0na) k0kungwa e)vi )
=G4)/-6 +#9)God

30.P#6 wit!

n)

ka)0tii*

/C11!ead

ke0evi )

A-1130.P600

u)0$a*a0na)

;S)$i!eQ/6S)10E

k0e)

=G4)1/2

which God himsel gave Itats004.008.0;2J

&%#

Constituent order

#icipu is a confi$urational lan$ua$e with SE2 as the basic constituent order for clauses, and <ead)Modifier within the noun phrase %L8.8.F*, 'ust as e3pected for a (enue)#on$o lan$ua$e % illiamson and (lench 2000K;0);5*. 2nly two clausal word orders ha!e been obser!ed in the corpus, SE2 and 2SE, with the former much more common. =n e3ample of SE2 order is $i!en belowK %42* I$i )0kuluLs
+#5)tortoise

Iti 0 ) komo)Lv Imo0ni


=G5)co!er +#8)water

ma0na)Lo
=G8)=/"

a tortoise covered the water I"ikula, sa$b004.F29J 2SE clauses are formally distinct from their SE2 counterparts in two other ways. 0irst, the ob'ect must be followed by the copula, and second, if the mood of the clause is realis then the !erb must be marked for perfecti!e aspect. 4F5

2SE word order is also pragmatically)marked, and is only found when the ob'ect carries a much $reater Scommunicati!e loadT than the !erb. =ll three of these stipulations also apply to the more $eneral constituent)frontin$ constructions discussed below %L8.;.2.4*. -3ample %4;* illustrates the obli$atory copula, althou$h since the !erb is not realis there is no perfecti!e marker5. %4;* I#onte3tK speaker = says that when they used to $o to the market they would put on different $arments, not a loincloth %ku)0roono), +#9*. Speaker ( then picks up on the fact that the usual item of clothin$ was a loincloth...J I707aLs I;0si 0 yu*uL! (K Iku)0roono)Lo kw0i
+#9)loincloth =G9)#21 +#9)person =G9)<=()wear

it was loincloths people were wearing Isayb004.;F7J "he utterance occurred durin$ a discussion about what people wore in the past, so the idea behind the !erb yuu SwearT was already acti!ated %L2.;.2.8* , indeed the !erb itself had been used 'ust a few intonation units prior. "he referent of ku)0roono) SloinclothT can be ar$ued to be in contrasti!e focus here.
&%#1 Constituent.fronting<
E 2 2(6J.

"he basic order of elements in the clause is IS =A

Speakers may depart from

this basic order for pra$matic effect. More precisely, constituents may be fronted for two main reasons, because they are in focus %L8.;.2.4.4* or because they are marked topics %L8.;.2.4.2*.
&%#11 8ocus

If the referent of an ob'ect or an obliDue is in ar$ument focus %L2.;.4*, then it is usual to place it clause)initiallyK %48* ka)0'a77i )ki
/C11"ros"erity A-11w!at =G4)#21

ke)0yi )ni

k0e)

) vu0u0yaa0wa) 1mu)l__

2S)0A")doQI//)=116R4S.1/2

what prosperity will you bring me# Isaat002.002.;92J

5 "his was the only e3ample in the corpus of 2SE word order with two le3ical +1s %2E word order without a le3ical sub'ect is more common*. 7 "he use of the terms Sfrontin$T and Se3tractionT in this section is not intended to imply that a deri!ational process is in!ol!ed.

4F7

%4F* ;0ri si )no*o v0i )


/C)1<name=

=G9)#21

;0yo0no)

2S)beQ/6S)10E

hi na)
"21

it's a Risino you are I"ikula, sami004.424J h)Duestions like %48* ha!e focus)presupposition articulation, with the focal domain coincidin$ with the Duestioned +1, in this case a secondary ob'ect. "he conte3t for %4F* was a bit of banter concernin$ the identity of the person bein$ addressed in the e3ample, and doubts were raised about his credentials as a member of the "ikula di!ision of the =cipu. Instead, the speaker 'okin$ly says to the addressee Ii.e. you:re not a &ula but...J it's a Risino you are. -3ample %45* in!ol!es a fronted pronoun rather than an +1, but the articulation is perhaps more ob!iously focus)presupposition. "he speakers were discussin$ a time when there had been no)one willin$ to be the chief of the =cipu. = sla!e was cau$ht, and it was H'I they appointed to the chie tancy. %45* evvi )
30.C6P

a)0na8u)kwa)0na)

;1)appointQ/6S)10E

ti 0wm
+#5)chief

it was HIM they appointed to the chie tancy Isayb004.;4;J "he !erbs in %4F)45* are marked for perfecti!e aspect with the suffi3 0nA %L8.5.;.4*. If the clause is in realis mood, then focusin$ a constituent makes this suffi3 obli$atory. It is reDuired e!en when the linear order of constituents does not chan$e, as in the focusin$ of the sub'ect in %47b*. %47* %a* saa t0e)ne
or
=G5)which

$i )0kooto) ti )0pi so)


+#5)drum

=G5)breakQ/6S

every drum broke %b* t0e)ne


=G5)which /C+1drum

$i 0 ) kooto) t0i

=G5)#21

ti )0pi so)0no)l

=G5)breakQ/6S)P>,

which drum broke Ieamy005.459, eamy005.459J 2SE clauses and other kinds of SfocusT frontin$ are not strictly limited to sentences with ar$ument focus. "rue ar$ument focus %L2.;.4* is !ery rare in the corpus9, and it seems to be the case that %non)topical* constituent)frontin$ ser!es a more $eneral purpose. "he followin$ statement is an appro3imation of the conditions under which non)topical constituent)frontin$ is possibleK
9 "his seems to be the case cross)lin$uistically %e.$. 0rancis et al. 4999 on -n$lish*.

4F9

%49* or a nonctopical constituent to be ronted it must carry the ma%ority o the communicative loadu o the clause "his is !ery often the case when the ob'ect introduces a new referentK %49* wa)
/-1/"

'u*

there

MMM $i )0me'i )

+#5)inside =G5R+#7)stone

) * tu,1u0taari ku)0yuyu
/C*1sand

kw0i )

=G9)#21

a)07a*a0na)

;1)seeQ/6S)10E

it is said that there inside the stony place they saw sand Isayb004.2;0J Strictly, this sentence does not ha!e ar$ument focus since while the people seein$ the sand had already been introduced to the discourse, the e!ent of Sseein$T was not yet acti!ated. So in terms of information structure the sentence has broad %or predicate* focus, with the focal domain comprisin$ the !erb and the fronted ob'ect. <owe!er the ) * SsandT is bein$ introduced for the first time. -3ample %20* is ob'ect referent ku)0yuyu essentially the same. %20* u)0y
/C71k.o.Fclimber =G7)#21

w0i )

o)0$u'o)0no)

;1)plantQ/6S)10E

o,1$u0kuu

62#R+#5)waist

ti ,1;0gi )gi ni )ya)

=G5R+#9)delebPpalm

there was a climbing plant they had planted at the oot o a deleb palm Isaat002.002.490J -3ample %24* in!ol!es two referents introduced for the first time, but they ha!e different statuses with respect to information structure. "he utterance be$ins with a marked e3ternal topic ana) a)7a Ssome peopleT, followed by the fronted ob'ect o)modo Ssla!esT, and then finally the !erb. It is the ob'ect referent that is, to use Mithun:s %4992* term, newsworthy. %24* a0na)
;1)=/"

a)07a

+#2)person

po* Io)0modoLo h0e)


all
/C

1slave

=G2)#21

I)0'p)0n0n)Lv

;1)holdQ/6S)E-+")10E

some people they even caught slaves Isayb004.492J Gi!en that the clause is bein$ used to introduce the sla!es into the discourse by means of a particular relationship with the referent of ana) a)7a, the nature of that relationship %Scatchin$T* is fairly predictable. +ote that only o)modo is part of the focal domain of the clause , ana) a)7a is outside of the clause alto$ether.
&%#1# Topic

In addition to focused constituents, topicalised constituents can be fronted %or left) dislocated*. Most of the time these occur in what appears to be a clause)e3ternal position, offset from the clause by a pauseK 4F9

%22* ka0ta)ara) k,1w0wm ko)0ri si )no*o, ga)hi


/C11'ine A-1E/C)1c!ie' /C11<name=

a0 i )ya)
;S)payQI//

k0e),
A-11P#6

before

the fine of the chief of Korisino, be ore they could pay it, Isaim004.089J 2ther times fronted topics are phonolo$ically inte$ratedK %2;* I#onte3tK chicken, he calls eatheru...J mo)0nno'o) mo)0yoo u)0si 0 ) hyaa
/C41beer

;S)<=()say

+#8)porrid$e

beer he calls porridgeu Itats004.00;.085J %28* I#onte3tK he doesn't see things correctly, through beer %mo)yoo, +#8*,J s )da) m0i m0e) u)08aa0na) n1u)0yuu0ni ko08u*u
because
A-41P#6 =G8)#21

;S)surpassQ/6S)10E andR+#7)wear)+M6H

+#4)heart

$i )0me'i )
+#5)inside

ti ,1m0e)
=G5RA-41P#6

because it he sets his heart on it e*cessively Itats002.00;.05;J -!en when a resumpti!e pronoun is present, as in %28*, realis clauses with fronted topics are still differentiated from their non)topicalised counterparts by the perfecti!e suffi3 0nA on the !erb. "he fact that this suffi3 occurs with realis !erbs in fronted topic constructions as well as fronted focal constituents su$$ests that the occurrence of the suffi3 depends on the frontin$ of the constituent, rather than the particular pra$matic relation in!ol!ed9. -3ample %28* is also of interest because of the comple3ity of the syntactic structure from which the topic is e3tracted i.e. from the possessor of the second ob'ect of a nominalised !erb, which in turn is the ob'ect of the prepositional complement of the main !erb. In pseudo)-n$lish the construction is somethin$ like it he e*ceeds with putting heart inside o it. In comple3 constructions such as these there is rarely a $ap , instead there is a resumpti!e pronoun. "his seems to be a common state of affairs cross) lin$uistically %&roe$er 2008K499*, and also applies to #icipu relati!e clauses %L8.;.8*. "he possibility of Sisland constraintsT on topic)frontin$ has yet to be in!esti$ated. (efore lea!in$ topicality it is worth notin$ in passin$ that as well as left)dislocated topics, #icipu also allows truly Se3ternalT %&roe$er 2008K4;7* or SouterT %.ooley
9 #icipu contrasts with <ausa in this respect , in the latter only focus)frontin$ results in special !erb morpholo$y, not topic)frontin$ %Ja$$ar 2004K89;*.

450

2007K92* topics, which ha!e no anaphoric link with the followin$ clauseK %2F* ko)0ri si )no*o
/C11<name=

ka'a) m0u0gwa)anu)kwa) $e
now 4S)0A")know
+-G

ka)m

definitely

3orisino now ' really wouldn't know Iabout the settlement of &orisinoJ Isayb004.402J 0inally, we should note that there are two e3plicit topicalisers go and hi na). "hese are discussed in L9.;.2. In this subsection we ha!e seen that constituent)frontin$ does not correlate with any particular pra$matic relation. Instead the referents of fronted constituents may be topics, in focus, or 'ust SnewsworthyT. "his su$$ests that constituent)frontin$ in #icipu is similar to sentence accent in -n$lish, which can indicate topic, focus, or acti!ation %6ambrecht 4998K;22);2F*.

&%%

?on.verbal clauses

"his subsection discusses the followin$ kinds of non)!erbal clausesK predicate nominals %L8.;.;.4*, predicate locati!es %L8.;.;.2*, e3istential clauses %L8.;.;.;*, and finally presentational clauses %L8.;.;.8*.
&%%1
&%%11

Predicate nominals
?ouns
+14 +12 =G2)#21,

1redicate nominal clauses with le3ical sub'ects take the form

with the

copula a$reein$ in $ender with the second +1. "he copula is identical se$mentally to the noun class pronouns %L8.8.;.2*40. -3amples %25* and %27* show referents indicated by the second. %25* Iku)0yupu)L
+#9)crocodile +#9)creature

1/21-/ I+#6ASI2+

%1ayne 4997K448*, since the referent of the first +1 is bein$ said to belon$ to the class of I;0kwaaro)L v0i )

44

=G9)#21

the crocodile is a creature Itats002.005.004J

40 "he tone on the copula is basically polar %L;.8.8*, whereas the tone on the pronouns depends on the syntactic construction %L5.2.4F*. "his type of morpheme is sometimes called a SstabiliserT %e.$. +ewman 2000K499 for <ausa, see also elmers 497;K494 fn. 2*. 44 Kwaaro) is borrowed from the <ausa waro SinsectT, which can also mean Scrocodile, hippo, or hyenaT.

454

%27* Ika)0dama
+#4)word

ka0avu)L

=G4)2S.12SS

* Ima)0gaiL

+#8)sword

m0e)

=G8)#21

your word is a sword Ioamy004.028J "he beha!iour of the copula in a$reein$ with the predicatin$ +1 rather than the sub'ect is unusual, and is an instance of SattractionT or SbackT a$reement %#orbett 2005K5;)58* 42 %note that Duite apart from semantic considerations, the second +1 in %25)27* cannot be analysed as a sub'ect without !iolatin$ the otherwise uni!ersal restriction that #icipu sub'ects come before predicates %L8.;.4**. -3amples of
-_A="IE-

predicate nominals in!ol!in$ two le3ical +1s are non)

e3istent in the corpus. -3ample %29* was elicited, but since the initial +1 is set off from the rest of the clause by a pause, is difficult to distin$uish this from the sin$le ar$ument identificational clauses illustrated in %29);0*4;. %29* Ia)07a0mpa)L,
+#2)person)this

I)0grm)
+#2)elbow

ho)0vo*oL

=G2)4S.12SS =G2)#21

h0e)

these, they are my elbows Ieamy00F.0;7J %29* 707a


+#9)person

n0na)

=G9)/-6

doonu)0na)
sitQ/6S)10E

a,1Kangu), w0a)
62#RInameJ

;S)/-1/"

k0kg

/C)1?ausa

v0i )

A-)1C6P

and the one who settled at 3angu, they say he was a Hausa Isayb004.49FJ %;0* ma)0hu*u
+#8)truth

m0e)

=G8)#21

it's true Ilit. Sit:s truthTJ Itats004.008.099J "he position of the copula with respect to noun modifiers is !ariable, as shown in the followin$ pairs of e3amples. "he difference, if any, with respect to meanin$ or usa$e is not yet understood. %;4* ka)0nnu k0e)
+#4)bird A-11C6P =G4)bi$K/-.A1

ka)0penene*u

it's a very big hawk Isaat004.005.095J


42 "his possibility also e3ists in <ausa %+ewman 2000K452)45;, Ja$$ar 2004K8F9)8F9*, althou$h a$reement with the sub'ect is more usual. "he term SattractionT is preferable to SbackT a$reement, since the copula comes after both +1s, in both #icipu and <ausa. 4; =lon$ the same line of reasonin$, an alternati!e analysis of %25)27* mi$ht be that, rather than bein$ cases of attraction or SbackT a$reement, they consist of a left)dislocated topic followed by a sin$le ar$ument identificational clause %recall from L8.;.2.4.2 that no pause is necessary after a left) dislocated topic*.

452

%;2* m0

+#8)child

me)08ene*u m0e
=G8)small

A-41C6P

it's a small child Itats007.002.079J %;;* ka)0go)ga


+#4)rubberPba$ A-11C6P

k0e)

ka,1u0tuwa) wu,1mo0ni
=G4R+#7)fetch

=G7R+#8)water

it's the bag or etching water Isaat004.00F.059J %;8* ti 0 ) wm ti ,1ma0ga)ji ) ti ,1o0ki iso
+#5)chief =G5R+#8)priest =G5R+#2)mai$iro

t0i )

A-+1C6P

it's the priesthood Ilit. Schieftancy of the priestTJ o the Iaigiro Itats004.00;.00;J
&%%1# Personal pronouns

hen the first +1 in a predicate nominal construction is a pronoun rather than a le3ical +1, there at first si$ht seems to be a difference between the e3pression of %i* proper inclusion and %ii* eDuation. "he tone of the pronoun in the former type of construction is 6 6, and the copula appears after the second +1K %;F* I#onte3tK Speaker is $i!in$ further information about some peopleJ e)re)
3P.P#6

o)0modo h0e)

+#2)sla!e =G2)#21

they are slaves I2009)0;)07.00;J %;5* i )v)


0.P#6

;0sahwa

+#9)simpleton

v0i )

=G9)#21

you are a simpleton Is!tm$004.428J (y contrast, in the eDuati!e construction the tone on the pronoun is < 6 and there is no e3plicit copula. =dditionally, the first) and third)person pronouns differ se$mentally from the standard independent pronouns %L8.8.;.4*K %;7* I#onte3tK Speaker had pre!iously talked about some sla!es. <e then mentions another $roup of people and says...J ere)
3P.C6P
THEY

o)0modo

+#2)sla!e

ha0na)

=G2)=/"

are the slaves I2009)0;)07.00;J

45;

%;9* I#onte3tK speaker is indicatin$ the kin$ while speakin$J evvi )


HE

30.C6P

d0daa

+#9)kin$

is the king Is!$d004.002J

<owe!er this SeDuationT form of the pronoun is also found in !erbal clauses with focus) presupposition articulationK %;9* I#onte3tK Summarisin$ a discussion of the three founders of &orisinoJ ere)
3P.C6P
THEY

hwaara)0na)
startQ/6S)10E

started Ii.e. they were the ones who startedJ Isayb004.;08J

"herefore it is probably better to analyse these SeDuationT forms as focus forms of the independent pronouns. "he paradi$m is $i!en in "able 45. "hese pronouns are probably diachronically deri!ed from standard independent pronouns plus the copula %see L9.9.8*. 0able ,]- +ocused independent personal pronouns
SG 16

4 2 ;

am i ) i )v vi ) evvi )

ottu) i )8o yi ) ere)


Predicate locatives

&%%#

"hese are formed usin$ the focused form of the pronoun, as in %80*. -3ample %84* shows the same structure with a $ender)marked pronoun, and %82* with a le3ical +1. "he copula occurs immediately after the sub'ect. %80* evvi )
;S.#21

le*e

')sni )

there nearby

it was there nearby Itats008.004.0;0J %84* v0i v0e) 'u* $5)

=G9)1/2 =G9)#21

there underneath

it is there underneath Ieaim002.4;99J

458

%82* me)0i )0ri 0 mpa)


+#8)+#;)thin$)this

m0e)
=G8)#21

le*e
there

this little thing is there Isayb004.24;J 0or ne$ati!e locati!e predicati!e clauses see L8.;.F.4 below. "he fi!e #icipu demonstrati!e ad!erbs %L8.8.4.4* can also function as predicates, in which case they a$ree in $ender with the sub'ect. In response to the Duestion han GH Swhere is the MT, one possible answer is u AvM0pa*a Sthe is hereT. %8;* ka)0taari
+#4)stone

ka)0pa*a
=G4)!ere

the stone is here I2007)04)48.004J


&%%% )xistential clauses

"he simplest form of e3istential clause in #icipu consists of the predicator o)koo Sthere isNthere areT followed by an +1. "he +1 complement of the predicator is optionally sub'ect to the complement tone perturbation discussed in L;.8.7. 4)koo is borrowed from <ausa akwai, which is also used in #icipu with the same meanin$, seemin$ly in free !ariation. %88* o)koo
t!ereFis

$i 0 yi m i )
+#5)dark

there was orest Ilit SdarknessTJ Isayb004.02FJ %8F* ka'a) o)koo


now t!ereFis

ka)0ji ri gi

+#4)canoe

ka,1g0ge8u)
=G4R+#9)up

now there are aeroplanes Isayb004.;5FJ 1redication usin$ o)koo is also a common means of attributin$ Dualities to a referent. In such cases the referent nominal often occurs before the predicator, similar to the topicalised +1s discussed in L8.;.2.4.2 abo!e. %85* ke)0re'e)
+#4)ton$ue

ki 0ive)

=G4);1.12SS

o)koo

t!ereFis

u)0pa$i

+#7)difficulty

their language is hard Is!tm$004.289J =n alternati!e construction uses the preposition n) SwithT to link the two +1sK

45F

%87* v0m),

+#9)monkey

o)koo

t!ereFis ;S.1/2 with

e)vi )

n)

;05a)hanna

+#9)destruction

the monkey, it is destructive Ilit. Smonkey, it is with destructionTJ Itats002.004.078J 2n rare occasions, the complement of o)kooNa)kwai is followed by the copula. %89* o)koo
t!ereFis

mo)0yi lo

+#8)sacred

mu,1u0huna m0e)
=G8R+#7)kill

=G8)#21

there is a taboo against killing it Iku)0yupu) Scrocodile, +#9TJ 0or ne$ati!e e3istential clauses see L8.;.F.4 below.
&%%& Presentational clauses

Isaim004.090J

"he presentati!e n)duu is used in presentational clauses, similar to its <ausa eDui!alent ga %+ewman 2000K494)492, Ja$$ar 2004K859)859*. "hese can either be e3ophoric, indicatin$ referents present in the speech situation, or te3t)internal, in which case they introduce a new referent into the te3t)world. /)duu may be historically)deri!ed from the !erb inda SseeT, or perhaps e!en from n)0du*u SI stretch out s.t towardsT. %89* n)duu, mo0ni
!ereFis
+#8)water

s_

drinkQIM1

here's water, drink# Isaat002.002.;0;J %F0* n)duu,


!ereFis thief

w pa*a

here

here's the thie here Isaat002.004.449J /)duu can also operate at the clausal le!el, especially if the clause encodes a back$rounded e!ent. It seems the e!ent itself is bein$ SpresentedT for consideration, and the meanin$ can usually be con!eyed by you see in -n$lish translation. %F4* ti 0du)kwa) n)duu
41)$oQI// !ereFis 41)seeQ/6S)10E place

t0i nda)0na)

'asu) u)0dei

+#7)settlement

let's go, you see we've seen a place to stay Isayb004.459J

&%&

!elative clauses

/elati!e clauses in #icipu consist of three parts which occur in the followin$ orderK %F2* %<-=. +2A+* , /-6="IEIS-/ na) , M2.I0BI+G #6=AS-

455

"he relati!iser a$rees in $ender with the head noun, but since there is often a resumpti!e pronoun or clitic %a$reein$ in $ender or person* in the modifyin$ clause, na) is considered to be a relati!iser rather than a relati!e pronoun. If the modifyin$ clause is in realis mood then the perfecti!e aspect suffi3 )nA %L8.5.;.4* is obli$atory, 'ust as when wh)words %L8.;.5.2* or focused or topical constituents are fronted %L8.;.2.4*48. "here are few restrictions on the function of the relati!ised constituent within the modifyin$ clause, as the e3amples below illustrate. "he relati!ised constituent is indicated either by a $ap %shown as ;*, or a resumpti!e pronoun. "he $ap strate$y can apply at any point on &eenan and #omrie:s %4977* =ccessibility <ierarchy %SA(J s 2(J s
2(6

12SS-SS2/*,

as well as to time and place phrases %F9)50*. In the case of obliDues, yi 0 na) ; yi 0 ) la a)0na) yi 0 ) 'etei

the preposition is $apped as well as the +1, as illustrated by %F7*K %F;* i )0ri
/C31t!ing A-31#7; =G;)lackQ/6S)10E =G;)fine

bad things Ilit Sthin$s which lack finenessTJ ISA(J-#", G=1, oamy004.429J %F8* ma)0kuu
/C41mountain

ma0na)

A-41#7;

u)0kum a)0na)

;S)climbQ/6S)10E

; Io(J-#", G=1, "ikula, sami004.059J

the mountain which he climbed %FF* a)07a


/C

1"erson

ha0na)
A-

1#7;

ma)0gai

+#8)sword =G8)=/" =G8)<=()want

ma0na) ma)0si 0 ) ta'a

: ma0hu)na)
=G8)killQI//

the people the sword wanted 1 to kill I-M(-..-. o(J-#", G=1, samy004.04FJ %F5* o)koo
therePis

ko)0moori ka0na)
/C11task

A-11#7;

tu0u0$a)a1vu)

41)0A")$i!eQI//R2S.1/2

there's a task which we will give you IS-#2+.=/B 2(J-#", G=1, saat002.002.85;J %F7* i 0$i )nji )
/C31money A-31#7;

yi 0 na)

wu0u0ra)a

;S)0A")eatQI//

i0 ) ri

+#;)thin$

yi , 1ka0raa ; : >n)
=G;R+#4)eat

wit!

y0i @

A-31P#6

the money which he would eat ood .with it2 I2(6I_A-, G=1, tats002.00;.048J

48 "he phonolo$ical identity between the relati!iserNarticle and the perfecti!e suffi3 is presumably accidental. <owe!er the perfecti!e suffi3 0nA which occurs in relati!e clauses and clauses with fronted constituents is assumed to be a sin$le morpheme %cf. Ja$$ar %2004K452* on a similar constellation of functions assi$ned to the 0ocus "=Ms in <ausa*.

457

%F9* mo)0ni

/C41water

* po* ma0na) a)0hyaa0na)


all
A-41#7;

;1)sayQ/6S)10E therePis

o)koo

ku)0yupu)

+#9)crocodile +#5)inside

$i )me'i )

all the water said to have crocodiles inside .o it2 I12SS-SS2/, G=1, tats002.005.048J %F9* ;0looka)$i n0na)
/C)1time A-)1#7;

u)0do*oho)0no)

;S)disappearQ/6S)10E

; I"IM-, G=1, tats008.004.0;5J

when it disappeared %50* 'asu) wu0na) !ooyo) kamu)kwa)0na) ;


A-71#7;

"lace@/C7A

InameJ

wasQ/6S)10E

the place where fooyo was I16=#-, G=1, sayb004.08;J "he resumpti!e pronoun strate$y is not attested within the corpus for relati!ised possessors or sub'ects. /esumpti!e pronouns are common for relati!ised obliDues in matri3 %54)52* clauses, but less so for relati!ised ob'ects %5;*. /esumpti!e pronouns can at least be elicited for relati!ised possessors %58*, and %in subordinate clauses* for sub'ects %5F*. /elati!ised time and place constituents do not ha!e resumpti!e pronouns. %54* kw0a'a)
/C*1!ouse

ku0lle)

A-*1t!at =G9)/-6

ku0na)

u)0yo0no)

;S)beQ/6S)10E

a,1kw0i )

62#RA-*1P#6

that house which he is at I2(6I_A-, /-SAM1"IE- 1/2+2A+, tats002.00F.044J %52* i )0ri


/C31t!ing

yi 0na)

A-31#7; +#9)bi$Pspider

;0lwli )

;0yo0no)
=G9)beQ/6S)10E

n)
with

y0i
A-31P#6

a,1kw0a'a)
62#R+#9)house

kwe0evi )
=G9);S.12SS

what (ig dpider had in his house I2(6I_A-, /-SAM1"IE- 1/2+2A+, saat002.002.280J %5;* i )0ri
/C31t!ing A-31#7;

yi 0na)

ti )07a*a0na)

41)findQ/6S)10E

y0i

A-31P#6

what we ound I2(J-#", /-SAM1"IE- 1/2+2A+, "ikula, sa$b004.;09J %58* i )0ri


/C31t!ing A-31#7;

yi 0na)

)0n*0n)1tu)

;1)$i!eQ/6S)10ER41.1/2

ka)0na) ayi ) I;: ka1y0i )L


+#4)story

=G4RA-31P#6

the thing they gave us the story .o it2 I12SS-SS2/, G=1 2/ /-SAM1"IE- 1/2+2A+, "ikula, sami004.0F;J

459

%5F* ka)0 ara


/C11elder

ka0na)

ma0llu*

ma)0dama)0wa)0na)
=G8)speakQ/6S)=116)10E

wm
chief

w0a)
;S)/-1/"

A-11#7; +#8)teacher

IB: k0eL
A-11P#6

ka)0huna)
A-1)killQ/6S

ku)0yupu)
+#9)crocodile

n)
with

k)05
+#4)a3e

the elder that the teacher told the king .he2 had killed the crocodile with an a*e I-M(-..-. SA(J-#", G=1 2/ /-SAM1"IE- 1/2+2A+, 2009)08)04.002J %55* ku)0yupu) ku0na) ma0llu*
+#8)teacher

ma)0dama)0wa)0na)
=G8)speakQ/6S)=116)10E

wm w0a)
chief ;S)/-1/"

/C*1crocodile A-*1#7;

ka)0 ara ka0huna)


+#4)elder =G4)killQ/6S

<B: kw0i )L
A-*1P#6

n)
with

k)05
+#4)a3e

the crocodile that the teacher told the king the elder had killed .it2 with an a*e I-M(-..-. 2(J-#", G=1 2/ /-SAM1"IE- 1/2+2A+, 2009)08)04.002J ka0na) %57* k)05
/C11a$e A-11#7;

ma0llu*
+#8)teacher

ma)0dama)0wa)0na)
=G8)speakQ/6S)=116)10E

wm
chief

w0a)
;S)/-1/"

ka)0 ara
+#4)elder

ka0huna)
=G4)killQ/6S

ku)0yupu)
+#9)crocodile

IB :n)
wit!

k0eL
A-11P#6

the a*e which the teacher told the king that the elder had killed the crocodile .with it2 I-M(-..-. 2(6I_A-, G=1 2/ /-SAM1"IE- 1/2+2A+, 2009)08)04.002J In summary, as is typically the case cross)lin$uistically, the resumpti!e pronoun strate$y becomes more common with respect to $appin$ for relati!ised functions which are lower down the +1 =ccessibility <ierarchy. In the same way, the deeper the le!el of embeddin$ between the head noun and the $apNresumpti!e pronoun, the $reater the likelihood that a resumption pronoun will occur instead of a $ap. +o potential island constraints ha!e been tested yet. =s in many lan$ua$es %6yons 4999K54*, the #icipu relati!iser is formally identical to the article %L8.8.F.;*, but unlike the article the relati!iser is obli$atory and does not depend on the identifiabilityNuniDueness of the head noun referent , in fact it can e!en co)occur with the pre)head specific indefinite article4FK %59* o)koo
therePis 301A#(

wu0na) ka0llu)

+#4)hun$er

ka0na)

A-11#7;

* a)0yaa0na)

;1)doQ/6S)10E there

le*e

there was a certain hunger that happened there Ilit Sthey did thereTJ Isayb004.522J 0inally, it should be obser!ed that the head noun in the schema $i!en in %F2* is optional.
4F In <ausa relati!ised +1s may also be marked by both the specific indefinite determiner wani and the definiteNanaphoric linker cn %see +ewman 2000 chp. 58 for e3amples*.

459

If the referent is clear from the conte3t then the head noun may be omittedK %59* ka0na)
A-11A#(

ti )0yadda)0na)

41)a$reeQ/6S)10E

one Ika)0dama SwordT, +#4J that we agree with

Is!b$004.0;2J

&%(
&%(1

?egation
Clausal negation

"he most common means of ne$atin$ a clause is throu$h the ne$ation particle $e, which occurs post)!erbally. %70* ka)07aaki ) ka)0lapa)
+#4)lion =G4)knowQ/6S

$e
/7-

the lion didn't know Isaat004.007.040J "he tone on $e is hi$h and sometimes e3tra)hi$h %L;.8.4*. If there is a pronoun in the complement position of E1 then $e usually follows it, but nouns always come after the ne$ator %see L7.;.2 for details*. "he clause)initial ne$ation particle a*a, borrowed from <ausa, also operates at a clausal le!el45K %74*
AD ) a*a u0hyaa1vu)
/7-

;S)sayQI//R2S.1/2

he shouldn't say to you I i.e. Sit is not the case that he should say to youTJ Itats00F.004.24FJ %72* a*a mo)0$i yo)
/7=G8)$etQ/6S +#2)person

a)07a

ha0na)

=G2)=/"

it didn't get the people Isamy004.0;9J +e$ati!e locati!e predications %7;* and e3istential propositions %78)7F* can also be e3pressed usin$ a*a.

45 wa*a is sub'ect to !ariation in tone and !owel)len$th. =fter the conditional con'unction n SifT, a) usually occurs with a short low)tone !owel %e3ample %7;* is an e3ception*. wa and aa also occur, but the factors $o!ernin$ the distributions of these forms are not well understood %cf. <ausa where different forms of ba ha!e different functions , Ja$$ar 2004 chp. 5*. 47 "his kind of speech act is usually encoded in a different way, usin$ the ne$ation particle kada) %L8.5.2.2*.

470

%7;* n

if

a*a i )v)
/7-

2S.1/2

le*e

there

i you're not there Itats009.008.008J %78* a*a 707a


/7+#9)person

n0na)

=G9)/-6

wu0u0s)'u)

;S)0A")touchQI//

y0i )

=G;)1/2

nocone would touch it Ilit. Sthere does not e3ist one who would touch itTJ Isayb004.757J %7F* ma)0waa
+#8)do$

a*a ku)0$i ye)


/7-

+#9)hand

dogs don't have hands Itats009.008.024J +e$ati!e e3istential a*a clauses can also modify an +1, in which case the interpretation is that the referent of the head noun does not posses the entity marked with a*a. * %75* a)0yaa
;1)doQ/6S

u0ma)ta)

+#7)$i!ePbirth

mi 0 na)

=GF)=/"

a*a a)0'i 7a
/7-

+#2)tail

they gave birth to some Im0uu Schildren, +#FTJ without tails

Is!tm$004.2;0J

S.ouble ne$ati!esT may occur with both a*a and $e, but the meanin$ is still ne$ati!e , the two ne$ators do not cancel each other outK %77* /iger ka*i_
annoyance

a*a /igeria $e
/7/7-

Niger# not Nigeria I"ikula, sami004.8;5J


&%(# Constituent negation

+e$ation in #icipu can take scope o!er a sin$le constituent rather than a clause. "he same ne$ation morpheme $e is in!ol!ed, but this time it takes an optional prefi3 a$reein$ with the $ender of the constituent, as in %79*. If the constituent concerned is a personal pronoun then the focus form of the pronoun %L8.;.;.4.2* is used, as in %79*. %79* $i )0'i ta)ni
+#5)marria$e

u)0ko*o

+#7)death A-71/7-

wu)0$e

marriage is not death Io!fl004.004.004J

474

%79* ere)

3P.C6P

$e
/7-

o)0doonQi )sRu)0wo)0no)1vi )

;1)sitQ/6S#=ASs)=+"I#)10ER;S.1/2 there

le*e

it wasn't them that caused him to be settled there Ii.e. he came of his own accordJ Is!b$004.054J 0inally, it can be noted that #icipu has no ne$ati!e Duantifiers. "here are no words correspondin$ to nobody , instead the ne$ation particle a*a is used to assert the non) e3istence of a particular kind of referent. %90* a*a 707a
/7+#9)person

n0na)

=G9)/-6

wu0u0s)'u)

;S)0A")touchQI//

y0i )

=G;)1/2

there was nocone who would touch it Ii.e. Sthere did not e3ist one who would touch itTJ Irepeated from %78*J

&%9
&%91

Auestions
BesCno "uestions

It is likely that yesNno Duestions ha!e a special intonation pattern in all lan$ua$es %1ayne 4997K29F*, re$ardless of whate!er other formal means are a!ailable. 0or #icipu the situation is similar to <ausa and many other =frican lan$ua$es %/ialland 2009* in that the usual tonal downdrift %L;.8.2* is suspended. "he e3tra)hi$h final < characteristic of <ausa %+ewman 2000K897)899* has not been obser!ed. BesNno Duestions may also be marked with the optional sentence)final Duestion particle su. %94* w0aa
;S)/-1/"

aRd0do)ori

707a

u)0guya)
;S)canQ/6S

u)0hi ya)
+#7)$rind

kwaanu) gooma) sul


basin ten
I

62#R+#9)formerly +#9)person

it is said that in the past someone could grind ten measures Is!tm$004.049J Ou can be used across turn)transitions to Duestion the pre!ious speaker:s statement. =s the followin$ e3ample illustrates it can form a complete utterance by itself. %92* =K evi
;S.1/2

707a
+#9)person

808ene*u
=G9)small

v0vo*o
=G9)4S.12SS

he's my younger sibling (K sul


I

really I2009)0;)07.002J 472

"a$ Duestions can be formed as in <ausa %Ja$$ar 2004KF28)F2F*, by placin$ koo SorT at the end of the utterance. =s in <ausa, the tone on koo in this en!ironment is actually risin$ !i>. ko+o. "he nati!e #icipu eDui!alent saa %L8.8.F.8* followed by one of the topicalisers go or hi na) %L9.;.2* can also ha!e this function. Oaa can also be used to form indirect DuestionsK %9;* na)ha
letQIM1

n0jo)olo)0no)

4S)checkQI//)E-+"%M* or

saa

yi 0 ) ni i'wa)

=G;)soakQ/6S

let me check whether it is soaked +o word)order chan$es ha!e been obser!ed in yesNno Duestions.
&%9# wh."uestions

Isaat004.009.40;J

_uestion words are usually fronted in strai$htforward wh)Duestions, but remain incsitu for echo and rhetorical Duestions. = list of wh)words in #icipu is shown in "able 47. 0able ,_- 5icipu whc$uestion words Iuestion word yi )ni %+#;* han N ha)nu 0e)ne 8eaning Position in straig!t'orward Juestions49 what where which fronted fronted fronted or incsitu fronted or incsitu

ya)ani N ya)anu %+#9* who

-3ample %98* shows a direct reDuest for information with clause)initial yi )ni SwhatT. -3ample %9F*, on the other hand, was a $ently)mockin$ rhetorical Duestion with incsitu yi )ni , carryin$ the implication that the people bein$ indicated didn:t know anythin$. %98* yi )ni a)0hula
w!at
+#2)name

he0evi )
=G2);S.12SS

what is its name Isayb004.;80J %9F* a)07a0mpa)


+#2)person)this

a)0lapa)

yi )ni l I"ikula, sa$b004.075J

;1)knowQ/6S w!at

these ones what do they know "he word for SwhoT has two !ariants ya)ani and ya)anu. "he former is more likely to
49 "his kind of Smi3edT beha!iour %where wh)frontin$ is obli$atory in some cases but not in others* is rare in (enue)#on$o, althou$h it is reported for both <ausa and the est &ain'i lan$ua$e .uka %.ryer 200F, +ewman 2000K895*.

47;

occur before words whose first !owel is I0/2+"J, while the latter is more likely to occur otherwise, althou$h as %95* shows this is not an absolute rule %cf. L;.;.2*. %95* ya)ani ;0yo0no)
w!o
=G9)beQ/6S)10E

u0la)ngwa)1mu) n)0lavu)
+#7)spoilR4S.1/2 +#F)sleep

who's spoiling my sleep I"idipo, saat002.008.0FFJ "here are two alternati!es for the word for SwhereT. Tan occurs in fronted position, while ha)nu is found incsitu. %97* han
w!ere

ku0la$i
+#9)youn$P$irl

ku0na) 0yo0no)
=G9)/-6

u0ta)'a)l
+#7)want

2S)beQ/6S)10E

where's the girl that you want Isaat002.002.8F9J %99* i )0kama) ha)nul Isayb004.0;;J "he Duestion words so far ha!e been SpronominalT, in that they ha!e stood in for +1s. 0e)ne SwhichT is different in that it usually modifies a head noun. It takes an obli$atory prefi3 a$reein$ with the head noun either in $ender or in person. hen a$reein$ in hen $ender, the modifier can occur either before %99* or after %90* the head noun.

21)be.1S"Q/6S w!ere

where were you.pl.2

a$reein$ in person, the modifier seems to be limited to the pre)head position %94*. =n identical pattern with respect to a$reement feature and position is found with the article and the demonstrati!e , see L9.9 for more details. %99* kw0e)ne
A-*1w!ic!

ku0la$i

kw0i )

kw0aya)0na)
=G9)comeQ/6S)10E

+#9)youn$P$irl =G9)#21

which girl came Ieamd0;2.49FJ %90* aRka0kaasu)wa)


62#R+#4)market

k0e)ne
A-11w!ic!

to which market Isayb004.724J %94* w0e)ne


301w!ic!

;0'i ri )
+#9)kind

which kind Isayb004.;29J

478

!i )ni Swhat, +#;T can also function as a modifier in this way, allowin$ the Duestionin$ of particular components of a +1. In this case it too a$rees with the head noun. %92* ka)0'a77i )ki )
+#4)prosperity

ke)0yi )ni l
=G4)w!at

what prosperity Isaat002.002.;92J hen ya)anu Swho,


+#9T

is used in this way it takes on the meanin$ Show manyT, as in

%9;*. +ote the contrast between the low)tone prefi3 and short consonant in %9;* and the hi$h)tone prefi3 and lon$ consonant in %98*. "he latter two properties are characteristic of associati!e constructions with a$reement*. %9;* ko)0dontu
+#4)chair

+#9

possessors %L8.8.F.4, see also L5.8.2 on neutral k0e)l I2009)02)24.005

ka)0ya)anu

A-11!owFmany =G4)#21

How many chairs are there %98* ko)0dontu ka,1y0yaanu k0e)l


+#4)chair =G4R+#9)w!o =G4)#21

"hose chair is it I2009)02)24.005J kw0e)ne Swhich dayT can be "o Duestion the time of an e!ent the composite phrase kwaa used %9F*. Similarly the manner in which some e!ent was carried out can be Duestioned usin$ t0e)ne ShowT, which can be analysed as )e)ne SwhichT to$ether with an a$reement prefi3 %95*. %9F* va0aya)
2S)comeQ/6S
=G5

hen t0e)ne is fronted it is usually followed by the copula. kw0aa


+#9)day

kw0e)nel
A-*1w!ic!

when did you come I6e3iconJ %95* t0e)ne t0i ) ka)0 i ki ) a)0si 0 ) yaa
;1)<=()do
+#4)festi!al

ka1'a0kkwi i
=G4R+#2)deadPperson

A-+1w!ic! =G5)#21

how do they do the estival o dead people I"ikula, sa$b004.842J "o Duestion the reason for some e!ent or state, the wh)Duestion word yi )ni SwhatT can be used followed by the !erb yuu ScauseT.

47F

%97*

yi )ni yi )0yuu0na)
w!at
=G;)causeQ/6S)10E

0ti yo)0no)
2S)$etQ/6S)E-+"

$el
+-G

why didn't you.sg.2 get any Ilit. Swhat caused you didn:t $etMTJ Isaat004.008.022J 0inally, it should be noted that each wh)Duestion word has a uni!ersal Duantifier counterpart, formed by placin$ saa SwhetherNorT in front of it %see L8.9*.

&%<

Clause coordination and subordination

1erhaps nowhere else in the $rammar of #icipu is the influence of <ausa seen so stron$ly as in the methods of clause combination. +ot only are the !arious con'uncts and other particles often taken strai$ht from the lin$ua franca, but the constructions themsel!es are usually identical in format. "he discussion in Ja$$ar %2004KF92)582* could to a lar$e e3tent ha!e been reproduced here !erbatim, but for reasons of space %and scholarly inte$rity?* it is only possible to mention a few of the more common structures here.
&%<1 Coordination

"he usual +1)coordinator n) %L8.8.F.8* does not con'oin clauses in #icipu. Instead it is common to find simple 'u3taposition of clauses, either with inter!enin$ pauses %99* or without %99*. %99* u)0vondo)ro) ma)0di ya, u)0huna),
;S)slin$Q/6S
+#8)hare

u)07uwa)

;S)killQ/6S ;S)putPinsideQ/6S

u)0tpu)

a,1k0$k u)0dukwa)
62#R+#4)ba$

;S)$oQ/6S

;S)roastQ/6S

he slung it at the hare, he killed, he put it inside his bag he went and roasted I"idipo, saat002.00;.4;4J %99* w0aya)
;S)comeQ/6S ;S)openQ/6S

u)0jungo)

ru0p

+#;)$ranary

wu)0uto)0wo)0no)

;S)riseQ/6S)=116)E-+"

* ku0naa u)0hyaa
+#9)le$

;S)sayQ/6S 2&

to)

then he opened the granary he brought out the leg he said 3 Isaat002.002.540J Asually the sub'ects are identical in paratactic chains of this kind, but not always, as %400)404* show. "here is no Ssame sub'ectT marker in #icipu as there is in some (antu lan$ua$es %e.$. atters 200;K2F8* , the fact that the two !erbs in %404* ha!e different sub'ect referents can only be determined from the conte3t.

475

%400* ka0llu)
+#4)hun$er

k)0'p)1re)
=G4)holdQ/6SR;1.1/2

a)0dukwa) a,1ku0jene)
;1)$oQ/6S
62#R+#9)ri!er

hunger got hold o them and they went to the river Isaat004.005.022J %404* a)i0$a*a1re)j
;1)$i!eQ/6SR;1.1/2

a)j0dukwa)
;1)$oQ/6S

theyi gave to them% and they% went Isaat002.002.225J =s in <ausa %Ja$$ar 2004KF9;*, the particle kuma SalsoNandT may be used to pro!ide a more e3plicit si$nal of coordination. "he position of kuma !aries, but it is fairly common to find it after the sub'ect +1 %if any* and before the !erb. %402* I#onte3tK God said let the sun come out to give lightJ u)0lenji kuma wu)0uto)0no)
+#7)sun

and

=G7)riseQ/6S)E-+"

and the sun came out

Itats004.008.029J

If the sub'ect referent and the !erb are identical in both clauses then kuma occurs post) !erbally, ad'acent to the constituent bein$ contrasted. %40;* s )
because and likePthat

n)

'i ni ),

u)0laa

+#7)fire =G7)beQ/6S

wu)0yo*

n)1u)0'usu) : kuma
also

wu)0yo*

=G7)beQ/6S

kuma n)
also

and

) : wu)0yo* ti 0 ) li pai
+#5)$ood

andR+#7)power

=G7)beQ/6S

n)1u)0pa$i ) :

andR+#7)difficulty

because o that, ire has power 1 it is also good 1 it is also di icult Itats002.002.0F9J "emporally)seDuenced classes are often separated by the con'unction see

SuntilNunlessNthenT %from <ausa sai49*. If the !erb follows immediately afterwards then the tone pattern is the one indicatin$ irrealis mood %L8.5.2.2*, e!en if the clause is reportin$ an e!ent that would %in the absence of see* be e3pressed in the realis mood. %408* u)07a*a
;S)seeQ/6S

)wm,
+#9)chief

see
t!en

) u0hyaa

ak_ I"ikula, sa$b004.777J

;S)sayQI// IsurpriseJ

he saw the chie , then he said a# =s well as markin$ a temporal seDuence, see can also be found in the apodoses of conditional clauses %cf. -n$lish then*.
49 "he !owel in <ausa sai IeiJ is phonetically !ery similar to the IeJ in #icipu see. See discussion in +ewman %2000K802* and Ja$$ar %2004K44*.

477

%40F* n)
if

a*a mo)0ni
+-G +#8)water

')sni ) see i k0ka) a)


nearby t!en 2S)takeQI//

0 uuta)
+#9)waterPbottle

i there is no water close by, then you take a water bottle Itats009.00;.02FJ "he dis'uncti!e con'unction saa SorT may be used to con'oin clauses %see L8.8.F.8 for +1 dis'unction*, often followed by one of the topicalisers go20 or hi na) %L9.;.2*. %405* u)0yo*
;S)beQ/6S with

n)

mo0olo)

+#8)$uitar

saa go) u)0yo*


or
"21

;S)beQ/6S with

n)

0gu)ugel
+#9)!iolin

eid he have a guitar or did he have a violin

Isaat004.008.0;9J

Oaa has a number of other uses in #icipu that parallel those of ko SorT in <ausa %+ewman 2000K4;7, Ja$$ar 2004K;98);9F*, in particular its use in contrasti!e)focus Se!enT constructions %%407*, cf. Ja$$ar 2004KF42* and in formin$ indirect Duestions %L8.;.5.2* and uni!ersal Duantifiers %L8.9*24. %407* saa a)kwai
or therePis 2S.1/2 near

i )v

')sni ) Itats004.004.087J

even i you are close by

#ontrasti!e coordination can be e3pressed usin$ the con'unction a)maa, a$ain a borrowin$ from <ausa amma22. %409* w0wm 707a
+#9)chief +#9)person

t0to)

=G9)one

v0i ,

=G9)#21

a)maa )0wm o)0yo*


but
+#2)chief

;1)beQ/6S much

gei

chie u is one, but chie su are many


&%<# 0ubordination

Ieab$004.0F9J

Subordinate ad!erbial clauses are particularly similar in #icipu and <ausa. "he followin$ table compares the <ausa subordinators listed in Ja$$ar %2004K505)509* with their #icipu eDui!alents.

20 vo often has low)tone when it occurs after saa. "he reason why is not known. 24 Oaa may also be historically related to the counterfactual au3iliary saa %L8.5.2.8* , this latter construction does not seem to ha!e an analo$ue in <ausa. 22 In #icipu both a)maa and a)mma are heard , the former seems to be more common.

479

0able ,`- Hausa subordinators and their 5icipu e$uivalents -loss when counter actual because up until i 1when be ore even i as long as until1unless a ter like even though when instead o because until since ?ausa ~ ~n ~ei[^YY~e^ har ~n^YY[^ n[^YYni[^ ]e i}~~^ sai bayan d in kooY~ feewc^Y~YY{nn^Y~ i[ine^ {nje~ {n[Y~ o}^Y~ Cici"u n) da)a do)omi : don hari ) i )don : n ka)si : ka)mi saa muddi )n see ayan n) : ku)0$i no ku0 Sback ofT ka)ma saa da) looka)$i nna) : sa'a) nna) ma)imako vi 0 s )da) : s ) n) see n) ti nda) : tunda)

#lauses introduced by the con'unction ana) ShowNlikeT play an important cohesi!e role in #icipu discourse, in particular in tail)head linka$e %"hompson and 6on$acre 499FK209) 24;*, which is e3tremely common in #icipu narrati!e. "he dependent clause is introduced by ana) and its !erb always occurs in the perfecti!e aspect %L8.5.;.4*. %409* u)07uwa)
;S)roastQ/6S

ka)0ka'i )la*a

ana) u)07uwa)0na),

+#4)chameleon

ke0evi ) :

=G4);S.12SS

!ow ;S)roastQ/6S)P>, ;S)eatQ/6S

u)0ra*a

he roasted his chameleon 1 having roasted it, he ate I"idipo, saat002.00;.055J "his e3ample occurred as part of a lon$er chain of such back)references, dia$rammed belowK

479

0able ,b- )*ample o intracparagraph tailchead linkages 0entence no. 4 2 ; 8 ana) he had roasted)P>, ana) he had finished)P>, eatin$ ana) he had rested)P>, 5e"endent clause 8atri$ clause he roasted his chameleon he ate he rested there he said...

"he situation recapitulated in the dependent clause does not ha!e to be an e!entG it can also be a state, as illustrated by %440*. "he con'unction ana) still occurs, and so cannot be properly $lossed as SwhenT , better would be somethin$ like Sit bein$ the case that...T. %440* u)0yo* to)
2& ;S)beQ/6S with

ni )

k0kaa

+#9)woman

yapu) :
two

ana)

!ow ;S)beQ/6S)P>, with

u)0yo0no)

ni )

k0kaa

+#9)woman

yapu) :
two

he had two wives 1 3 it being the case that he had two wives 1 Isaat004.008.00FJ =s su$$ested by the phrase tail)head linka$e, this construction normally ser!es to recapitulate information that is already known to the hearer, to ser!e as a point of departure for the main e!ents to follow. = certain amount of new information may be imparted in the dependent clause, howe!er, as lon$ as it is fairly predictable and can therefore be treated as presupposed by the hearer.

*.*

/ominals

+ouns in most (antu and many other (enue)#on$o lan$ua$es ha!e two formal properties which set them apart from other word classes. 0irst, nouns tri$$er obli$atory $ender a$reement on other elements in and outside the noun phrase. Secondly, noun words themsel!es can be di!ided into a stem and a noun class affi3, usually a prefi3. "he first property, obli$atory $ender a$reement, can be re$arded as definitional for the identification of #icipu nouns, since there are no e3ceptions2;. =s for the noun class prefi3, there is a small number of nouns with irre$ular plurals %LF.2.2, LF.2.7*, which means the boundary between prefi3 and stem is not clear)cut. "he e3treme e3ample is the noun 'asu) SplaceT. .espite belon$in$ to a specific $ender accordin$ to the a$reement that it tri$$ers %sin$le class $ender 7, LF.2.9.7*, 'asu) has no o!ert morpholo$ical markin$ to indicate this.
2; i.e. there are no e3ceptions amon$st the words which we mi$ht want to be co!ered by the definition because of their morpholo$y or semantics.

490

"he structure of the noun word is simple, consistin$ of a noun class prefi3 followed by a stemK %444*
+#)S"-M

"he tone pattern on the stem does not chan$e dependin$ on the prefi3, and the tone on the prefi3 lar$ely depends on the root tones %LF.F.;*. #lass markers in #icipu are obli$atory and always prefi3es. "hus #icipu contrasts with the northwest $roup of est &ain'i lan$ua$e 1on$u where noun markers are optional %Mac.onell 2007K89)89*, and also the est &ain'i lan$ua$es which ha!e both prefi3es and suffi3es %see Smith 2007 for ut)Ma:in*. "his section co!ers ad!erbs %L8.8.4*, proper nouns %L8.8.2*, pronouns %L8.8.;*, the countNmass distinction %L8.8.8*, +1 structure and modifiers %L8.8.F*, and then finally suffi3al nominalisation %L8.8.5*. .iscussion of de!erbal nominals is deferred to LF.8 because the analysis in!ol!es indi!idual noun class prefi3es.

&&1

Adverbs
+#9

It does not seem possible to find purely formal criteria which distin$uish ad!erbs from sin$le class nouns %i.e. those without a sin$ularNplural distinction*. "he difficulty is
+#9

due to two related features of the #icipu $ender system, to be discussed in chapters F and 5 respecti!ely. 0irst, nouns can occur with a null prefi3 %LF.F.7*, makin$ them morpholo$ically indistin$uishable from ad!erbs, and indeed other word classes. Secondly, ad!erbs can tri$$er $ender a$reement in the same way as more prototypical nominals. In chapter 5 we will see that =G9 a$reement can be tri$$ered by di!erse kinds of atypical controllers %Sneutral a$reementT, L5.8*, includin$ ad!erbs. 0urthermore ad!erbs can e!en appear with
+#9

noun class prefi3es %L5.8.2*. (ecause of this noun)

like morphosyntactic beha!iour I ha!e placed the discussion of ad!erbs under the $eneral headin$ of SnominalsT. In this section I will discuss demonstrati!e ad!erbs %L8.8.4.4*, time ad!erbs %L8.8.4.2*, and ideophones %L8.8.4.;*. 0or ad!erbial Duantifiers see L8.9.
&&11 $emonstrative adverbsCpredicates

#icipu demonstrati!es show a fi!e)way contrast in dei3is. "hese distinctions are encoded in ad!erbs %which also function as predicates, L8.;.;.2* as well as modifiers %L8.8.F.2*G it is the ad!erbs which are of interest here. -ach has a < 6 tone melody 494

distributed o!er either one or two syllables, as shown belowK %442* pa*a le*e here .near speaker2 there .near hearer2

'i nde) yonder . ar rom both2 'u* very ar away or out o sight 8o*o here, our permanent place

=lthou$h #icipu has fi!e deictic distinctions, there are only three de$rees of distance in!ol!ed , %i* close to some reference point %pa*a, le*e, 8o*o*, %ii* far away from one or more reference points %'i nde)*, and %iii* very far away from one or more reference point %'u**. xa*a and le*e are not distin$uished from each other by any measure of distance, but rather by the speech)act participant functionin$ as the deictic anchor , the speaker for pa*a, the hearer for le*e. "he first three terms in %442* therefore form a speaker and addressee)anchored subsystem %6e!inson 2008*, while the meanin$ of the first four terms coincides with the <ausa system %Ja$$ar 2004K;2;);;0, 58F)587*. 6e!inson %2008K409* writes that Systems with more than four terms combine other semantic dimensions, like !isibility or !ertical distance relati!e to the speaker, or shape of the referent. "his is the case for #icipu, as the $loss for 'u* abo!e su$$ests. If the location in Duestion is out of si$ht and thus cannot be indicated by the speaker, then 'u* is used, re$ardless of distance. "he meanin$ of the fifth term 8o*o is difficult to pin down, but it in!ol!es an interestin$ combination of
S2#I=6

and

16=#-

dei3is. Its ran$e of meanin$ o!erlaps with

that of the near deictic pa*a, but there are times when only 8o*o is appropriate. 0or e3ample if two =cipu are discussin$ their !illa$e at home they could use either pa*a or 8o*o to refer to it. If they tra!elled to a nearby &ambari town, then not surprisin$ly pa*a ShereT can no lon$er be used to refer to their home !illa$e. <owe!er they can continue to use 8o*o in this way, re$ardless of their $eo$raphical location. 0or inhabitants of different !illa$es, the ran$e of locations that can be referred to usin$ this deictic is different, and so the precise meanin$ of 8o*o in any $i!en usa$e e!ent can only be determined by deictic anchora$e. =lthou$h it is clearly deictic in meanin$, 8o*o differs in character from the first four terms listed abo!e, and so it is debatable whether #icipu should be classified as ha!in$ a fi!e)term system of spatial dei3is.

492

&&1#

Time adverbs

"ime ad!erbs can occur clause)initially %44;* or clause)finally %448*, but ha!e not been obser!ed clause)internally. %44;* longo)
recently ;1)startQ/6S

a)0hwa*ara) u0yi )nda) ;0ke)eke


+#7)see

+#9)bicycle

le*e

there

recently they started getting bicycles there Isayb004.877J %448* $i )0$i pu)
+#5)#icipu

ka)m,

definitely

ti 0 ) yonno)

=G5)bePnearQ/6S

u)0doowo)

+#7)disappear

ka'a)
now

5icipu, it's nearly disappearing now I"ikula, sa$b004.707J "he set of temporal ad!erbs includes the followin$ membersK %44F* ka'a) ka'a)ka'a) longo) do)ori de)ge)le*e
&&1%
BE

now %ust now1immediately recently1in the near uture ormerly then

Ieamy0;4.044J Ieaim00;.4802J Ieamy0;4.05;J Ieamy0;4.05;J Itats00F.002.2;0J Itats004.008.074J

longo)longo) a while ago1 $uite ar into the uture

Ideophones

0orty)si3 apparently nati!e ideophones ha!e been identified, but the use of <ausa ideophones is also common, and se!eral of these are si$nificantly different from the source words phonetically, su$$estin$ true borrowin$ rather than code)switchin$. "hey usually e3press the manner of an e!ent as in %445*, but they may also describe a referent %447*2F, or function as intensifiers in stati!e clauses %449*. %445* vu*0uwa)
2S)hearQI//

* u)0taa

;S)shootQ/6S

ki ip_

soundFo'F"enetration

you would hear him shoot# Itats00F.002.099J %447* wa+0aya)


;S)comeQ/6S ;S)seeQ/6S

w0i +nda)

i0 ri

+#;)thin$

pa*a gaaa_
here !uge

then he saw something huge# Itats00F.002.47FJ


28 "his word is !ery common in narrati!e. .iachronically, at least, it is composed of de)ge) SfromT le*e SthereT %cf. <ausa daga nan SthenNfrom thereT*. 2F Anless this is describin$ the manner in which somethin$ hu$e is encountered.

49;

%449* yi 0 ) 'uyo)0no) pass


=G;)li$ht)M

snowFw!ite

it's snow white Icf. <ausa at, eamy004.040J Many #icipu ideophones are phonolo$ically unusual in some way %e.$. syllable structure L;.4.7*.

&&#
referents.

Proper nouns
hen humans are referred to by names these nouns always tri$$er
=G9

1roper nouns are assi$ned $ender in different ways accordin$ to the animacy of their a$reement, which is found with other kinds of atypical a$reement controller %L5.8*. "he names of towns or other locations, on the other hand, are usually assi$ned $ender accordin$ to their morpholo$y %or percei!ed morpholo$y , see LF.5 for SborrowedT towns*. %449* n)0ra*a
4S)eatQ/6S

m0e)

A-41P#6

' win it I3a)kuuku), a local town, +#8J %420* hari ) ko)0ri si )noo
until

Isaat004.002.099J

+#4)&orisino

ka)07amu)kwa)

A-1)becomeQ/6S

ko)0ri si )noo

+#4)&orisino

until 3orisino became 3orisino Isayb004.405J "he names of people can take part in associati!e plural constructions %.aniel and Mora!csik 200F*, which consist of the hi$h tone morpheme aka followed by the name. "he relation between the named person and their associates is usually one of kinship, but it is also commonly used to refer to a person:s a$emates or close friends. 2ccasionally the associati!e relation is broaderK %424* I#onte3tK discussion of people who would tra!el lon$ distances to $ather melukeci to make ropesJ aka me0luke$i l
16 +#8)plantPk.o.

those o the melukeci plant

Isayb004.F80J

&&%

Pronouns

"he pronominal paradi$ms of #icipu are dealt with in detail in L7.4)7.;. In this section I 'ust make some $eneral remarks and then list the paradi$ms of free pronouns and 498

pronominal clitics25. "here is no distinction between inclusi!e and e3clusi!e SweT in #icipu, unlike the +orthwest $roup of est &ain'i %e.$. Smith 2007*. "here is a distinction in number between sin$ular and plural, but there is no dual. 0or third)person pronouns there is of course an alternation between paradi$ms marked for $ender and those that are not. "his alternation is the main topic of 1art IE.
&&%1 Independent personal pronouns

0able 4a- 'ndependent personal pronouns Person 0ingular 1 3 amu i )v evi Plural otu i )8o ere

"he ;1S pronoun has an alternate form e , see L9.9.8 for discussion and possible diachronic relationship to evi .
&&%# Independent noun class pronouns

0able 4,- 'ndependent noun class pronouns in 5icipu Class 4 2 he ; yi 8 F 5 7 wi 9 !i 9 kwi Pronoun ke
&&%%

me mi ti

2b,ect clitics

"hese attach to the ri$ht)hand side of the !erbal word. 0able 44- b%ect clitics Person 0ingular 1 3
&&%&

Plural tu) 8o) re)

mu) vu) vi )

$emonstrative personal pronouns

"he sin$ular demonstrati!e pronouns are composed of the contracted form e of the independent ;1S pronoun %L8.8.;.4* plus one of the deictic locati!es %L8.8.4.4*. "hey should not be confused with the person)marked demonstrati!e modifiers %wu0mpa),
25 0or the pronominal sub'ect affi3esNa$reement markers see L7.8)7.F.

49F

wu0lle) etc... L8.8.F.2*. 0able 46- eemonstrative pronouns 5emonstrative "ronoun e0mpe) e0lle) e0''i nde) o0n8 o) this one that one yonderu one this one Corres"onding locative pa*a le*e 'i nde) 'u* 8 o*o near speaker near hearer ar rom both very ar rom both1invisible27 here, our permanent place

"here is a pu>>le about the direction of !owel harmony in these words, since in e0mpe) the !owel Duality IeJ appears to ha!e spread ri$htward from the e) to the followin$ deictic pa*a, whereas in o0n8 o) the situation is re!ersedK the IoJ seems to ha!e spread leftward from the 8o*o deictic to the e0 ;1S pronoun. "he prenasalised consonants in e0mpe) and o0n8o) are part of a more $eneral problem which also affects the demonstrati!e modifiers %L8.8.F.2*. "o e3press plurality, an alternati!e set of demonstrati!e pronouns is used, deri!ed from the word for personNpeople 707aNa)07a. "he empty cells are not attested in the corpus and no attempt has been made yet to elicit these forms. 0able 4JK7D[E9G<?><8AD7X?E9EQ9G7MD?8ADM7\?E[7YNY=7XD?GE9 0ingular 707a0mpa) 707a0lle) 707a0''i nde) 707a0''u* M Plural a)07a0mpa) a)07a0lle) a)7a0''i nde) M a)07a0n8 o) Corres"onding locative pa*a le*e 'i nde) 'u* 8 o*o near speaker near hearer ar rom both very ar rom both1invisible here, our permanent place

&&&

Count and mass nouns

=s with other (enue)#on$o lan$ua$es, count and mass nouns can be distin$uished accordin$ to the noun classes to which they belon$. 0or e3ample the word for SstoneT ka)0taari takes the
+#4

prefi3 ka)) in the sin$ular and the

+#2

prefi3 a)0 in the plural. 0or

other noun roots there is only one choice, and these tend to be either %i* abstract, %ii* nouns that denote some kind of homo$eneous mass %either liDuid or solid*, or %iii* nouns like mosDuitos or termites where it is seldom necessary to sin$le out one entity
27 "here are no occurrences of the correspondin$ demonstrati!e pronoun in the corpus.

495

* % elmers 497;K450*. -3amples of these three kinds of nouns in #icipu are su)poo Sfalsehood, +#5T, mo)0ni Swater, +#8T29, and tu0uko) Scockroach, +#5T. See LF.;.; for the deri!ational relationship that e3ists between certain count and mass nouns %e.$. lump o charcoal !s. charcoal*.

&&(

The noun phrase

"he noun phrase in #icipu is comprised of an obli$atory head followed by optional determiners, possessors, or modifiers. "he associati!e construction, demonstrati!es, and article will be discussed in this subsection. 2ther modifiers are co!ered in different parts of the chapterK for relati!e clauses see L8.;.8, and for ad'ecti!es and numerals see L8.7 and L8.9 respecti!ely. Eirtually all noun modifiers can also occur as the heads of +1s i.e. pronominally, and this property of #icipu is discussed in L8.8.F.F. "he subsection ends with a discussion of the typolo$ically)unusual orderin$ of +1 constituents %L8.8.F.5*.
&&(1 The associative DEgenitiveCpossessiveFG construction

In many (enue)#on$o lan$ua$es, one particular type of noun modification takes on a hea!y functional loadK the associati!e construction % elmers 497;K27F)279*, sometimes called the S$eniti!eT %and less often the Spossessi!eT*. In #icipu the associati!e construction is important for the understandin$ of possession, the attribution of nominal properties, possessi!e pronouns %L8.8.F.4.4*, demonstrati!es %L8.8.;.8, L8.8.F.2*, prepositions %L8.F*, ordinal numerals %L8.9.;*, and !arious de!erbal nominalisations %LF.8*, amon$st other areas of the $rammar. "he #icipu associati!e construction takes the form +14 the SpossessedT noun and +12 the SpossessorT nounK %422* Ia)hula a)0hula hak(a)(L haRk0kaa
=G2R+#9)woman

=G+14R+12,

where +14 is

+#2)name

the name o the woman +14 occurs in its citation tone pattern, the a$reement proclitic

Isaat002.002.0;0J
=G+14

is hi$h tone, while

+12 under$oes a comple3 tonal perturbation %L;.8.7*. "he most salient feature of this perturbation for the present discussion is that the pitch of the +12 !owels is consistently
29 = noun with the usual plural prefi3 correspondin$ to +#8 can in fact be elicited , n)0ni Swaters, but the same shift of meanin$ occurs as in -n$lish %cf. ''ll have two waters please*.
+#FT

497

lower than that of the a$reement clitic !owel, so that for monosyllabic roots, say, the tone pattern on the second phonolo$ical word surfaces phonetically as I< 6J. Anlike most (antu lan$ua$es % elmers 497;K275*, there is no se$mental Sassociati!eT morpheme to which the a$reement morpheme ad'oins , instead it attaches strai$ht onto the left ed$e of the second +1. "he downstep which occurs after the associati!e a$reement prefi3 may be the !esti$e of a pre!ious se$mental form29. "he a$reement morpheme harmonises with the noun root to which it is attached, as shown in %42;* below, and this is e!idence that the morpheme belon$s to the same phonolo$ical word as the followin$ noun. +ote the harmonisation Spasses throu$hT the noun class prefi3 ku0 %recall the discussion on transparent affi3es in L;.F.;*. %42;* ka)0manga koKEku0hoomo)wo)
+#4)rope A-1E+#9)treeFk.o.

rope o the kuhoomowo tree Ieaim005.48F8=J I ha!e analysed the associati!e a$reement morpheme as a proclitic rather than a prefi3, in part because it does not always attach to le3ical noun heads. If there is a pre)head demonstrati!e or article, then the associati!e a$reement morpheme will bind to it rather than the head noun, as in %428*;0K %428* m0 :
+#8)child

me0evi ) :
=G8);S.12SS

saa kuma mV,1wu0na) 707a


or more
A-4R;S)=/" +#9)person

his child, or that o another person Itats002.008.0;9J 0urthermore, the a$reement morpheme may ha!e semantic scope o!er an entire clause rather than 'ust an +1. In the e3ample below the reason is not the Sreason of the chieftancyT, but the Sreason of the chieftancy bein$ there at &adaadaT. "he clause is not nominalised, but is a non)!erbal locati!e clause %L8.;.;.2* with ti 0 ) le*e SthereT as the predicate.

29 See #ahill %2000* for lan$ua$es with purely tonal associati!e morphemes, especially +kem %2000K8;*, where the tonal associati!e morpheme also results in downstep. See also atters %4990K444* for the same phenomenon in -'a$am. ;0 "he associati!e a$reement marker m is cut off by a $lottal stop here, perhaps due to the $rammatical word boundary.

499

%42F* ma)0hwaari ma,1ti 0wm ti )0le*e


+#8)start A-4E+#5)chief

=G5)there

ti 0 ) 'i sa)nu)u0na)

=G5)standQ/6S)10EM

Ka)0daada)

+#4)InameJ

the reason why the chie tancy is there at 3adaada lit. Sthe reason of Ithe chieftancy is there it stands at &adaadaJT I"ikula, sa$b004.5;5J 2ne final ar$ument in fa!our of the clitic analysis is that the a$reement morpheme attaches only once to comple3 +1s, unlike say &ikon$o %qf} 499;K;;9*, where multiple association markers occur. =s will be clear already, the relationship between the referents of the two +1s in!ol!ed in the associati!e construction is by no means always possession. elmers %497;K275* and #ro>ier %4998K99)404* $i!e e3amples of the broad ran$e of semantic relations co!ered by the associati!e, for &iswahili and #entral &ambari respecti!ely. #icipu is no different in this matter, as the followin$ e3amples show. %425* kw0a'a)
+#9)house

ku,1n0naata)

A-*E+#9)smallPspider

the house o dmall dpider %427* k0kaa vi ,1n0naata)

I12SS-SSI2+, saat002.002.048J

+#9)woman

A-)E+#9)smallPspider

the wi e o dmall dpider %429* ko)07uvu) ka,1ku0$i ye)


+#4)di$it A-1R+#9)hand

I&I+S<I1, saat002.002.;24J

inger Ilit. Sdi$it of handTJ %429* ka)0ddi ri ka,1a0kungu


+#4)stack

I1=/"I"IE-, eamy049.004J

A-1E+#2)cornPstalk

stack o cornstalks %4;0* i )0ri yu,1u0raa


A-3E+#7)eat

I#266-#"IE-, tats002.002.0F5J

+#;)thin$

ood Ilit Sthin$s of eatin$TJ %4;4* i )0ri ) yu,1u0raa yi , 1ka0hi'i

I0A+#"I2+, tats004.00;.025J

+#;)thin$ =G;R+#7)eat

A-3R+#4)ni$ht

evening meal Ilit Sthin$s of eatin$ of ni$htTJ

I"IM- 20 AS-, eamd049.F90J

499

%4;2* ka05a)ngali ka,1ka07aaki )


+#4)slap A-1R+#4)lion

the blow o the lion %4;;* u)0dv

I/-6="I2+S<I1 (-"

--+ -E-+" =+. =G-+",

saat004.007.099J

+#7)pound

wi ,1i0da'u)

A-7R+#;)$uineacorn

the pounding o the corn I/-6="I2+S<I1 (-"

--+ -E-+" =+. 1="I-+",

2009)0;);0.004J

"here is no distinction in #icipu between between obli$atorily and optionally possessed nouns, nor between alienable and inalienable possession. =s with almost all +1 modifiers in #icipu, the ri$ht)hand side of the associati!e construction can occur without a noun head, as in %4;8* below. <ere muKEu0galu) mi$ht be translated Sthat of the sideT. %4;8* I#onte3tK one hunter was huntin$ alon$ the ri!er bank, doin$ n)0ka$i %+#F* mi K Et0tudu) lit. Shuntin$ of bankTJ * kuma muKEu0galu) ku)0jene) e0mpe) si 0 yaa
;S)this
<=()do

more

A-%E/C71side

+#9)ri!er

and that one was doing Ihuntin$J o the side o the river I"idipo, saat002.008.08;J
&&(11 Associative DEpossessiveFG pronouns

"here are si3 personal associati!e %or Spossessi!eT* pronouns, which are similar in form to the independent personal pronouns %L8.8.;.4*. "hese pronouns take re$ular a$reement prefi3es and occur after the head noun. %4;F* %a* ka)0 ara
+#4)elder

ko0vo*o

=G4)4S.12SS

my elder %b* k0kaa


+#9)woman

va0avu)

Ieamy00F.0;FJ

=G9)2S.12SS

your .sg.2 wi e "he full paradi$m is shown belowK

Isaat002.002.544J

490

0able 4T- &ossessive pronoun paradigm with gender agreement pre i*es 4 2 ; 8 F 5 7 9 9 4S ko0vo*o ho0vo*o yi0vo*o mo0vo*o mi0vo*o or n0vo*o ti0vo*o wu0vo*o vi0vo*o or v0vo*o ku0vo*o 2S ka0avu) ha0avu) ya0avu) ma0avu) ma0avu) ta0avu) wa0avu) va0avu) kwa0avu) ;S ke0evi ) he0evi ) ye0evi ) me0evi ) me0evi ) te0evi ) we0evi ) ve0evi ) kwe0evi ) 41 ko0ttu) ho0ttu) yi 0ttu) mo0ttu) mi 0 ttu) ti 0 ttu) wu0ttu) vi 0ttu) ku0ttu) 21 ko088 o) ho088 o) yi 0 88 o) mo088 o) mi 0 88 o) ti 0 88 o) wu088 o) vi 0 88 o) ku088 o) ;1 ki 0ive) hi 0ive) yi 0ive) mi 0 ive) mi 0 ive) ti 0 ive) wi 0ive) vi 0ive) kwi 0ive)

"he 2S, ;S, and ;1 forms all ha!e lon$ !owels to$ether with #) prefi3es, which su$$ests the pronouns are !owel)initial. =s with the combination of noun prefi3es and nouns %L;.4.4*, these lon$ !owels are interpreted as the product of coalescence between the prefi3 !owel and the first root !owel. =$ain, the root !owel Duality is dominant. "he #E) prefi3es %i.e. 4S, 41, and 21* tri$$ered by class 4, 2, and 8 nouns all ha!e the prefi3 !owel IoJ, which is best thou$ht of an underspecified A !owel harmonisin$ with the possessi!e pronounK %4;5* %a* ka)0 ara ko0vo*o
+#4)elder =G4)4S.12SS

my elder %b* o)08u*u


+#2)heart

ho0ttu)

Ieamy00F.0;FJ

=G2)41.12SS

our hearts %c* ma)0di ya mo088 o)


+#8)hare =G8)21.12SS

Ioamy004.4F0J

your .pl.2 hares "he other class prefi3es retain their usual !owelsK

Ieamd0;2.052J

494

) kooto) ti 0 ) vo*o %4;7* %a* $i 0 my drum %b* $0$')


+#9)sheep

+#5)drum =G5)4S.12SS

vi 0 ttu)

Ieamd0;2.400J

=G9)41.12SS

our sheep .sg.2 %c* ku)0la$i


+#9)$irl

ku088 o)

Ieamd0;2.4F4J

=G9)21.12SS

your .pl.2 girl

Ieamd0;2.475J

0inally, the tone on the a$reement prefi3 on the 41S possessi!e pronoun is unspecified, assimilatin$ to the tone immediately to the left. So in e3amples %4;5a* and %4;7a* we ha!e hi$h)tone ko0 after ka)0 ara, but low)tone ti 0 ) after $i )0kooto). =s a $eneral point, it may be obser!ed that the usual < 6 tone pattern on the associati!e pronoun matches that of the le3ical associati!e construction e3amples $i!en in the pre!ious subsection, once the effect of downstep has been taken into account. "his pro!ides us with formal e!idence for considerin$ these pronouns to be a special case of the associati!e, 'ust as elmers %497;K275* does for (antu , this ar$ument will be taken up a$ain in the discussion of neutral a$reement in L5.8.2. "he discussion abo!e is concerned with the person paradi$m. "he $ender)marked paradi$m of pronouns %L8.8.;.2* may also occur in the second part of the associati!e construction %i.e. the SpossessorT position*. "hese pronouns do not chan$e their form when they occur in this position, and so no additional paradi$m has been set up here;4.
&&(# $emonstrative modifiers

"he fi!e deictic ad!erbs discussed in L8.8.4.4 abo!e also ha!e modifier counterparts encodin$ the same distinctions. "hese occur after the noun and take a hi$h)tone #E) a$reement prefi3. "he modifier roots are shown in bold belowK

;4 "here is no under)differentiation of $ender in this en!ironment, and pronouns of all noun classes may be found %cf. &iswahili where only class 4 and 2 pronouns occur in this position , elmers 497;K275*.

492

0able 4]- eemonstrative modi iers 5emonstrative modi'ier ka0aya ka0aya ka0aya ka0mpa) ke0lle) this hut that hut Corres"onding locative pa*a le*e near speaker near hearer

ke0''i nde) that hut ko)0dondo ko0''u* that garden kwa0a'a) ku088 o* this house;;

'i nde) ar rom both * 'uCB very ar rom both1invisible 8o*o here, our permanent place

"here are also person)marked demonstrati!es wu0mpa) and wu0lle);8 which may be used to modify any noun, re$ardless of $ender. %4;9* wu0mpa) 707a
301t!is
+#9)person

ti yo)0no)

$etQ/6S)E-+" 301t!is

wu0mpa) i 0 ) ri

+#;)thin$

ha)nul_
where

this person got this thing where# %4;9* wu0lle) i )0ri


301t!at

Isaat002.002.29FJ

+#;)thin$

yi Rka0raa
=G;R+#4)eat

that ood

Itats007.002.09;J

.emonstrati!e modifiers may occur either before or after the head noun, althou$h the resultant difference in meanin$ is hard to pin down. In some lan$ua$es there is a distinction between post)head and pre)head ad'ecti!es with respect to the occurrence of a$reement %/i'khoff 2002K24*, but this is not the case for #icipuK a$reement must occur, re$ardless of position. "here is howe!er a distinction with respect to a$reement eatures , post)head demonstrati!es must a$ree in $ender, whereas pre)head demonstrati!es may be inflected either for $ender or for person alone , 'ust as we found for the Duestion word )e)ne %SwhichT* in L8.;.5.2, and 'ust as we are about to see for the article in the ne3t subsection. 6yons %4999K44F* notes that there are also pre) and post)nominal demonstrati!es in &iswahili %see elmers 497;K299 for details*, with the former construction indicatin$ e will consider this possibility for #icipu in L9.9.2 , in that the +1 referent is topical.

anticipation of that discussion, note that person a$reement in #icipu is only possible on
;2 "here are no occurrences of demonstrati!e pronouns encodin$ the Sin!isibleT 'u* deictic in the corpus. ;; cf. kwa0a'a) ku088 o) Syour %pl.* houseT from the pre!ious section. +ote the tonal difference on the modifier, cf. =nderson %4990bK8;* for a similar correspondence between deictic and 211 possessi!e pronoun in +$yembn)(amileke %Grassfields (antu*. ;8 "he other three deictic possibilities are not attested in the corpus, but they ha!e not been ruled out by elicitation either.

49;

the demonstrati!e if it occurs in the position correspondin$ to topicality in &iswahili. "he fact that lon$ consonants occur root)initially in four of the fi!e modifiers is strikin$, as is the hi$h)tone on the a$reement prefi3. In L5.8.2 on SneutralT a$reement it will be su$$ested that the demonstrati!e modifiers in #icipu are deri!ed diachronically from instances of the associati!e construction %L8.8.F.4*. "his sets #icipu apart from the (antu lan$ua$es, for which elmers states "he associati!e...is not used in constructions of noun plus modifyin$ numeral or noun plus demonstrati!e %497;K277*. =s with almost all +1 modifiers in #icipu, the $ender)marked demonstrati!es can occur as noun heads themsel!es, as in the e3ample of cataphoric reference below. "he
=G4

a$reement morpholo$y seen below is a form of SantecedentlessT a$reement %L5.;* %480* ke0lle) ma)0huu m0e)y0
+#8)truth

which occurs when the a$reement controller is a proposition.


A-11t!at =G8)#21

this is the truth-c Io!k>002.020.004J "he person)marked demonstrati!es, howe!er, do not occur as noun heads, and constructed e3amples such as %484* are re'ected. Instead the demonstrati!e pronoun e0mpe) %L8.8.;.8* would be reDuired in this e3ample;F. %484* zwu0mpa), ka)0 ara
301t!is
+#4)oldPman

ko0vo*o

=G4)4S.12SS =G4)#21

k0e

IIntended meanin$K this, it's my old manJ


&&(% Article

Ieamy00F.0;2J

Just like the demonstrati!es, the article 0na) can occur either before or after the noun head. "his time, howe!er, the difference in meanin$ is clear. 1ost)head articles indicate anaphoric reference to an +1, while pre)head articles mark an indefinite but specific +1. In both cases the article takes a hi$h)tone a$reement prefi3. =s we ha!e already seen for a number of other a$reement tar$ets, the prefi3 on the article may a$ree in either personN number, or $ender , see L9.9.4 for further details. "he difference in meanin$ is particularly clear in the con!entionalised e3pressions Sone dayT and kw0aa'a ku0na) Sthe ne3t dayT. "he more comple3 e3amples ku0na) kw0aa'a below illustrate that the pre)head article cannot occur in cases of non)specific referenceG that is, when the referent is not identifiable to either the speaker or the hearer. -3ample
;F +ote that in #icipu, as in <ausa, it is not impolite to refer to humans simply as SthisT or SthatT.

498

%482* shows this with person a$reement, %48;* with $ender a$reement. %482* do)ori
formerly ;S)<=()$et

n)

>zwu0na)@ 0ke)eke
+#9)bicycle

707a
+#9)person

yo*o
$oQ/6S

aRka)0kaasu)wa),
62#R+#4)market

when 301A#(
+-G

u)0si 0 ) $i yo $e

in the past i someone went to the market, he wouldn't get a bicycle I2009)02)04.004J %48;* do)ori
formerly ;S)<=()$et

n)
when
+-G

>zma0na)@
A-41A#(

m0
+#8)child

mo)0yo*o

aRka)0kaasu)wa),

=G8)$oQ/6S 62#R+#4)market

u)0si 0 ) $i yo $e

0ke)eke
+#9)bicycle

in the past, i a child went to the market, he wouldn't get a bicycle I2009)02)04.004J "he followin$ constructed dialo$ues further illustrate the difference. In %488* speaker ( is lookin$ for a particular stone, perhaps one that he lost earlier. "he referent of ka0na) ka0taari Sa certain stoneT is identifiable to the speaker, but not to the hearer. introducin$ such referents to a discourse, the pre)head article is !ery common;5. ) yi )ni l %488* =K ;0yuu yaa
2S)#2+" do what

hen

what are you doing (K n)0yuu )l) ka0na)


A-11A#(

ka0taari
+#4)stone

4S)#2+" lookPfor

''m looking or a IparticularJ stone ISome time later...J (K to), m0i nda) ka0taari
2& 4S)seeQ/6S
+#4)stone

z>ka0na)@
A-11A#(

3, ''ve ound the stone I2009)02)04.004J In %48F* on the other hand, speaker ( does not care what sort of stone he finds. "he referent of ka0taari Sa stoneT is identifiable neither to the speaker nor to the hearer. In such cases the pre)head article cannot occur.

;5 It is not howe!er obli$atory. If the $eneral class of referents is already a topic of discussion, then the introduction of an indi!idual referent of that class will usually be accompanied by the pre)head article. #on!ersely, if the referent has little intrinsic importance %despite bein$ specific* then the pre)head article will usually not occur. "his is similar to Ja$$ar:s %4999* findin$ that the <ausa specific indefinite determiner wani is more likely to mark the introduction of referents which persist for lon$er in the discourse %see L9.2*.

49F

) yi )ni l %48F* =K ;0yuu yaa


2S)#2+" do what

what are you doing (K n)0yuu )l) >zka0na)@ ka0taari


A-11A#( +#4)stone

4S)#2+" lookPfor

''m looking or a stone Iany old oneJ ISome time later...J (K to), m0i nda) ka0taari
2& 4S)seeQ/6S
+#4)stone

>ka0na)@
A-11A#(

3, ''ve ound the stone I2009)02)04.004J "urnin$ to speaker (:s second statement in these dialo$ues, I was told that the post)head article was obli$atory in %488* but optional in %48F*. In fact it is hard to sDuare this with the e!idence from the corpus, where +1s are freDuently used to refer anaphorically without the post)head article, althou$h the delay while searchin$ for the stone may be a factorG once the acti!ation status %L2.;.2.8* of the stone has decayed enou$h the article may be reDuired as a SreminderT to the hearer not to introduce a new referent. 1erhaps the most we can say for sure is that +1)anaphors are less likely to be marked with the post)head article if the ori$inal reference was non)specific;7. #ro>ier %4998K90)9F* ar$ues that the #entral &ambari enclitic '. is not simply a definite article but an anaphoric article %6yons 4999KF;*, a more restricti!e cate$ory which is used when referents ha!e pre!iously been mentioned in the te3t. <e contrasts its o!erwhelmin$ anaphoric use in #entral &ambari with <alliday and <asan:s %4975K7;* contention that the -n$lish definite article is primarily e3ophoric or cataphoric. 6yons %4999KF2)F;* makes a similar point re$ardin$ the <ausa definite article %see also +ewman 2000K48;, and the discussion in Ja$$ar %499FK489)4F;**. =lthou$h the #icipu article is not diachronically related to either of these, it seems to be eDui!alent in function when it occurs after the noun. -ntities which are by their nature identifiableNuniDue such as the moon or sun ne!er occur with the article. 6yons %4999K25);0*, followin$ 1ostal %4970*, ar$ues that personal pronouns are the pronominal counterpart of definite articles. "his analysis is not appropriate for #icipu, since the article may itself function pronominally %see 489)489 below*.
;7 In this respect the #icipu SdefiniteT %i.e. post)head* article is different to -n$lish the, since ' ound the stone could not, I think, be used in the settin$ $i!en for %48F*.

495

Moreo!er, personal pronouns %at least in the third person* may themsel!es be followed by the article;9K %485* I#onte3tK someone walks into a compound and asks S hich ones are the sla!esMTJ a)07a0mpa),
+#2)person)this

ere)
3P.C6P

n0na)

o)0modo

ha0na)
=G2)=/"

A-)1A#( +#2)sla!e

these ones, they are the slaves I2009)0;)07.00;J Ansurpisin$ly $i!en their contrastin$ pra$matic functions, the pre)head and post)head forms of the article ha!e not been obser!ed modifyin$ a sin$le noun. <owe!er the relati!iser %which is formally identical to the article* readily co)occurs with the pre)head articleK %487* saa 05a)a ka0na)
or 2s)ha!e
A-11A#(

ka)0na) ayi ) ka0na)


+#4)news A-11#7;

vu0u0gu)ya)
2S)0A")canQI//

u)08an0dama0wa)1tu)l
+#7)little)speak)=116R41.1/2

do you have any particular story you can tell us I"ikula, sa$b004.099J "his distributional fact supports the analytic distinction I ha!e made between article and relati!iser, as does their different internal positions within the +1 %e!en when both are post)head , see L8.8.F.5*. (oth pre)head and post)head articles can occur pronominally i.e. without a le3ical noun head. It may seem stran$e to retain the distinction between pre)head and post) head when the article itself is the head, but there is a clear difference in meanin$ between e3amples such as %489*, where a new %specific* referent is bein$ introduced into the discourse, and those such as %489* which in!ol!e anaphoric reference to pre!iously)introduced referents. %489* I#onte3tK speaker was talkin$ about a specific camel %his own* which doesn:t biteJ a)maa ka0na)
but
A-11A#(

ka)0si ) 0nu)ma)0numa)
=G4)<=()bite)/-.A1

but others Ika)0ra)kumi Scamel, +#4TJ bite

I1/-)<-=. =/"I#6-, tats008.004.049J

;9 "his is also true for <ausa %+ewman 2000K48F*.

497

%489* ma0na) ma)0'wa*a


A-41A#(

=G8)passQ/6S

m0aya)

=G8)comeQ/6S

mo)0yo*o

=G8)$oQ/6S

mo)0yuwo)

=G8)fallQ/6S

the Im0 SchildT, +#8J passed and then went and ell I12S")<-=. =/"I#6-, tapf004.008.009J Just like the person)marked forms of the demonstrati!es, wu0na) does not occur pronominally , at least, of the 44; instances in the corpus, all of them occurred with a head noun immediately afterwards. 0inally, e3ample %489* abo!e illustrated one further property of the pre)head articleK althou$h it occurs when introducin$ a new referent, the conte3t often in!ol!es contrast with a pre!ious referent. =nother e3ample is $i!en belowK %4F0* I#onte3tK speaker was talkin$ how do$s reco$nise members of the household.J a)maa i )don wu0na) 707a
but if 301A#(
+#9)person

v0i )
=G9)#21

but i it's another person Itats004.004.09FJ In this respect the article is similar to the modifier 0m ) SanotherTK %4F4* 707a
+#9)person

n0na)

=G9)/-6

dn)0n) :

followQ/6S)10E

;0si ya) vi 0m ,)
+#9)side

A-)1anot!er

+#9)thief

0 w v0i )

=G9)#21

the one who ollows the other side 1 is a thie Itats004.008.408J <owe!er 0m ) also has an SincrementalT use as in %4F2*;9, a shade of meanin$ which seems to be absent from the article. %4F2* I#onte3tK after reco!erin$ from a han$o!erJ wu0u0yo)o u0 )l) m0m )
;S)0A")$oQI//
+#7)lookPfor A-41anot!er

he'll go looking or more Imo)0yoo Sbeer, +#8TJ

Itats002.00;.0F2J

=s illustrated in %4F2*, )m ) takes a hi$h)tone $ender a$reement prefi3, occurs as the head of an +1 as well as a modifier, and can modify mass nouns as well as count nouns.
&&(& ?ominal coordination

+ominals are con'oined usin$ a sin$le proclitic meanin$ SandT. "his takes the form n) before short consonants and ni ) before lon$ consonants or a consonant cluster. If the stem is !owel)initial, then that !owel len$thens so that n)1a)07a Sand peopleT becomes
;9 cf. _uirk et al. %499FK;99* on another.

499

Ina)(7aJ. -3ample %4F;* shows ni ) before first a consonant cluster and then a !owel)initial stem, and then n) before a short consonant. -3ample %4F8* shows ni ) before se!eral words be$innin$ with a lon$ consonant. %4F;* a)0ra)kumi ni )1n0ja)ki i,
+#2)camel

andR+#F)donkey andR+#;)cow andR+#4)rest

ni 1 ) i0naa

n)1ka)0ki ngi ki 0ive)

=G4);1.12SS

camels, and donkeys, and cows, and the rest o them %4F8* n)1ka)0ra)haa7i ) ni )1k0kg
andR+#4)0ulani

Isamoh004.498J

andR+#9)<ausa andR+#9)Eai andR+#9)6elna

ni )1v0va8i

ni )1h0hwi 'i

the +ulani and the Hausa and the ^ai and the Lelna

Isaim004.029J

Sin$ular personal pronouns may not occur as the first member of a con'oined +1G instead the correspondin$ plural pronoun is used , formin$ what is sometimes called an Sinclusory pronominal constructionT %6ichtenberk 2000*80. So to e3press the meanin$ he and his wife in %4FF* the phrase ere ni )1kkaa veevi ) Sthem and his wifeT %perhaps better Sthem including his wifeT* is used. Similarly to e3press me and my wife one would say otu ni )1kkaa vvo*o Sus and my wifeT, likewise for second person. Asin$ a sin$ular personal pronoun as the first member of such a construction is considered incorrect, and will be corrected by nati!e speakers84. %4FF* $u)0kundu ere
+#5)hyena

3P.P#6 andE/C)1woman

ni )1k0kaa

ve0evi )

A-)130.P600

a,1u0dukwa
62#R+#7)$o

hyena him and his wife were going

Isaat004.009.440J

"here is no such restriction on the second member of a con'oined +1, which su$$ests that the two +1s are in a comitati!e relation rather than a co)ordinate one. hile on the sub'ect of Santicipatory pluralsT it may be noted that, as in <ausa %+ewman 2000K4;5)4;7, Ja$$ar 2004K;9;*, !erbs may take a plural a$reement prefi3 if the referents encoded in the clause are performin$ a reciprocal action. "he Duestion in the followin$ e3ample was addressed to a sin$le person, yet plural a$reement occurs on the !erb.

80 +ewman %2000K4;5* and Ja$$ar %2004K;9;* use the term Sasymmetric coordinationT for the same phenomenon in <ausa, while atters %200;K28F* uses the term Sincorporati!e bondT for Grassfields (antu. 84 =ll the e3amples in the corpus are of personal pronouns. It would be interestin$ to in!esti$ate if a similar restriction applied to the independent gender marked pronouns.

499

) si riya) %4F5* i 0

P)prepareQ/6S

n)

with

kw0i

=G9)1/2

have you .sg.2 settled with her =s well as con'oinin$ two +1s, n) can also be used to e3press the

Isaat002.002.854J
I+S"/AM-+"

semantic

role, as in %4F7*. In the case of sub'ect +1s a prepositional phrase e3pressin$


=##2M1=+IM-+"

with the sub'ect may occur after the !erb comple3 as in %4F9*. kuma, n)
more with

%4F7* wu0u0ga)va)

;S)0A")kickQI//

ku)0naa ku0n8 o)
+#9)le$ =G9)this

he would kick again, with this leg %4F9* see u0'wa)a n) ka)0ka'i )la*a ke0evi )

Isaat002.004.0;2J
=G4);S.12SS

then ;S)passQI// with

+#4)chameleon

then he went with his chameleon

I"idipo, saat002.00;.0F8J

"his con'unction n) cannot be used for con'oinin$ clauses or !erb phrases. <owe!er contrary to elmers: obser!ation %497;K;0F* that nominal con'unctions rarely ha!e any other use in =frican lan$ua$es, n) also introduces SwhenT dependent clauses %L8.5.;.8* and thus co!ers the same ran$e of meanin$s as <ausa da SandNwithNwhenT. .is'unction is e3pressed usin$ the dis'unction saa SorT. Anlike n), saa may con'oin +1s %4F9*, 11s, or clauses %see L8.;.7.4*. ) 5o$i ) %4F9* i 0
+#;)illness

saa yi 1ku0naa, saa


or
=G;R+#9)le$

or

yi 1ko0tumo, saa
=G;R+#4)belly

or

yi 1ku0$i ye),
=G;R+#9)hand

illness either o the leg, or o the stomach, or o the hand,

Ieab$004.094J

0inally, both n) and saa may optionally occur at the be$innin$ of a series of con'oined +1s, 'ust like <ausa da and ko %+ewman 2000K4;2, Ja$$ar 2004K;94, ;98*. "his was illustrated abo!e in %4F8* for n), and in %4F9* for saa.
&&(( Pronominal function of modifiers

=lmost any noun modifier in #icipu can also function as the head of an +1. "his has been obser!ed for relati!e clauses, the Sback endT of the associati!e construction, demonstrati!es, and the article, and will also turn out to be the case for ad'ecti!es %L8.7* and numerals %L8.9*. "his ability to refer independently of a head noun may of course be linked to the obli$atory a$reement prefi3 on these modifiers, which increases their

200

potential to refer unambi$uously. Gi!en this is the case, it is temptin$ to try to preser!e the left)headedness of the #icipu +1 by analysin$ the !arious pre)head modifiers we ha!e encountered as +1s in their own ri$ht, so that a more literal translation of, say, ka0mpa) ka)0taari Sthis stoneT mi$ht be this one, a stone. Anfortunately this does not seem to be possible. /ecall that each of the pre)head modifiers %0e)ne SwhichT %L8.;.5.2*, the demonstrati!es %L8.8.F.2*, and the article %L8.8.F.;** can be marked for person instead of $ender. In none of these cases can the person)marked forms function as independently)referrin$ constituentsG rather they can only occur immediately before a noun. If we ha!e to admit that wu0mpa) ka)0taari with person a$reement on the demonstrati!e has the Modifier)<ead order, then there is no reason not to assume the same for the $ender)a$reein$ ka0mpa) ka)0taari .
&&(9 ?P.internal s-ntax

/i'khoff %2002K28* made the cross)lin$uistic obser!ation that, outside of elicitation sessions, it is rare to find +1s with more than two modifyin$ constituents. "his is true for #icipu, althou$h +1s with three modifyin$ constituents do occur in the corpus %e.$. e3. 459 below*. hen multiple modifyin$ constituents do occur, they often seem to in!ol!e phrasal modifiers %see /i'khoff 2002K28 for discussion*, and nested or multiple associati!e constructions are not all that rare. hen multiple constituents do occur in an +1, they follow a re$ular %althou$h typolo$ically)unusual* orderK %450* %=/"N.-M* /6D/ ?7A5 %=/"* %.-M* %12SS* %+AMN=.J* %/-6="IE- #6=AS-* -3ceptions to this pattern can be analysed as compounds, where the modifyin$ associati!e constructionNad'ecti!e is an intrinsic part of the referrin$ e3pression, rather than merely addin$ further detail. So in %454*, for e3ample 707a 808ene*u does not mean Ssmall personT, but Syoun$er siblin$T. Similarly there is a contrast between the strai$htforward modification Schildren of the futureT in %452*, and the idiomatic compound Swater babiesT %a malicious kind of spirit* in %45;*. %454* 707a
+#9)person

808ene*u

=G9)small

ve0evi )

=G9);S.12SS

his younger sibling IAne3pected A5L before P600, sayb004.559J

204

%452* m0uu

+#F)child =GF)this

mi 0mpa) mi , 17aama)ni
=GFRfuture%+#9*

these children o the uture I578 before P600, s!tm$004.290J %45;* m0uu
+#F)child

m1mo0ni

=GFR+#8)water

mi 0 lle)

=GF)that

these water babies Ilit. Schildren of waterTJ IAne3pected P600 before 578, tats00F.002.488J =t the moment I am not aware of any morphological e!idence that the e3pressions in %454* and %45;* are compounds, howe!er cross)lin$uistically compoundin$ often has this kind of effect on orderin$ rules, and the analysis su$$ested here is supported by the non)compositional meanin$s. If such compounds are assumed to fit into the noun head slot abo!e, then almost all the e3ceptions disappear82. It should be obser!ed that nominal combinations with non)compositional meanin$s do not necessitate this kind of treatment when they occur with modifiers , they can also be consistent with the normal Skerosene lanternsT in %458* orderin$ within the +1. "he idiomatic +1 i )hi )ti la) yk below is interrupted by the demonstrati!e yi 0mpa), in !iolation of certain theories about idioms %see #roft and #ruse 2008K22Fff* but in accordance with the order set out in %450* abo!e. %458* i )0hi )ti la)
+#;)lamp =G;)this

yi 0mpa) y,10k

=G;R+#2)e$$

these kerosene lanterns Ilit. Sthese lamps of e$$sTJ I578 before P600, sayb004.;4FJ "he position of articles and demonstrati!es in)between the noun head and the numeralNad'ecti!eNrelati!e clauses is contrary to what /i'khoff %2002* predicts for confi$urational lan$ua$es, and is said to be found in SfewN!ery few lan$ua$esT by #inDue %200FK;49);20*8;. Interestin$ly, it also differs si$nificantly from the typolo$ically more freDuent order $i!en by #ro>ier %4998K78* for #entral &ambariK %45F*
/6D/ ?7A5 %+AM* %=.J* %/-6="IE- #6=AS-* %12SS* %.-M* %=/"*

"he orderin$ $i!en abo!e for #icipu is partly based on elicitation, and partly based on inspection of the hundreds of rele!ant e3amples in the corpus. More detailed elicitation
82 "here is one e3ception in the corpus which does not seem to in!ol!e a compoundK in u)0hwaara) Iwu,0tu0wmL Iwu0na)L Sthe start of chieftaincyT the order is I12SSJ I=/"J. It may be possible to e3plain this as a case of tri$$er)happy a$reement %L5.F* on the article i.e. Sstart of Ithe chieftaincyJT. 8; #inDue assumes, followin$ &ayne %4998*, that the order deri!es by mo!ement of the noun from the underlyin$ order .em +um =d' +.

202

is needed to determine the relati!e orderin$ of numerals and ad'ecti!es, and to test the e3tent to which de!iations from %450* are 'ud$ed acceptable. <ere I will 'ust list sufficient e3amples to permit the deduction of the post)head orderin$ $i!en in %450*, repeated below for con!enienceK %455*
/6D/ ?7A5 %=/"* %.-M* %12SS* %+AMN=.J* %/-6="IE- #6=AS-*

%457* ka0aya Ika0na)L Ike0''i nde)L


+#4)hut =G4)=/" =G4)that

that IaforementionedJ hut IA#( before 578, saat002.002.8F0J %459* ma)0k0ku)0naa


+#8)M)+#9)le$

Ime0lle)L Ima,1k0kaa
=G8)that

=G8R+#9)woman

ve0evi )L

=G9);S.12SS

Ima0na)
=G8)/-6

u)0$a*a0na)

;S)$i!eQ/6S)10E

;0awli )L

+#9)bi$Pspider

that little leg o his wi e that he gave (ig dpider I578 before P600, P600 before #7;, saat002.002.F95J %459* ka)0 ara
+#4)oldPman

Ika0mpa)L Ika)0yapu)L
=G4)this =G4)two

these two old men I578 before /D8, eamy00F.007J %470* m0


+#8)child =G8);S.12SS

Ime0evi )L

Imo)0to)L
=G8)one

his one child IP600 before /D8, tats004.008.079J %474* a)0va'a)


+#2)bundle =G2);1.12SS

Ihi 0 ive)L

Ihe)0pene)pene*uL
=G2)bi$K/-.A1

their large bundles IP600 before A5L, saat002.002.2;7J %472* u)0nu


+#7)work

Iwu)0to)L Iwu0na) wu)0kusu)0no)L


=G7)one =G7)/-6

=G7)remainQ/6S)10E

the one work which remains I/D8 before #7;, saat002.002.495J * %47;* $i )0kooto) Iti 0 ) 8ene*uL Iti 0na) a)0hyaa0na) 1vi )
+#5)drum =G5)small =G5)/-6

;1)sayQ/6S)10ER;S.1/2

u0va)sa)L

;S)hitQI//

the small drum which they had told him to beat IA5L before #7;, saat004.009.074J

&&9

?ominalising suffixes

Most de!erbal nominalisations in #icipu are deri!ed by addin$ one of a number of 20;

different noun class prefi3es to a !erb stem. =ccordin$ly these are deferred to LF.8, once the noun class system has been properly introduced. "here are howe!er two nominalisin$ su i*es which are independent of the noun class system, and these are mentioned below.
&&91 0tative 0ni
+#7

Suffi3ation of 0ni to a !erb stem in con'unction with the addition of the

noun prefi3

u0 %LF.;.8* $i!es rise to a nominal, usually denotin$ the state arisin$ from the acti!ity encoded by the !erb as in u)0doonuu0ni SseatedT from doonu SsitT and u)0mannanuu0ni SstuckT from mannanu SstickT. "he final !owel of the !erb is len$thened %if short* before addin$ the suffi3. "he resultant words ha!e certain nominal properties. 0or e3ample they tri$$er $ender a$reement on modifiers %478*, and can also function as sub'ects. %478* u)kooni u)0koo0ni wu!eJesu) nu)u'ungooni wu,1!eesu) n1u)0ungo0ni
A-7RInameJ

+#7)die)+M6H

andR+#7)rise)+M6H

weevi ) we0evi )

A-7);S.12SS

the death o !esus and his resurrection Itats004.008.409J <owe!er they can also function as Sad'ecti!alT predicates in a way that strai$htforward nouns cannot. #onsider the followin$ two e3amples. "he predications are similar semantically, but the constructions differ in that the de!erbal nominal u)gnjni Scur!edT can occur as a direct complement of the !erb yo SbeT in %47F*, whereas the noun u)0'wi *i Sdistance, depthT in %475* must be preceded by the preposition n). %47F* u)yo* u)0yo* u)gnjni u)0gnj0ni

;S)beQ/6S

+#7)bePcur!ed)+M6H

it is curved Ignj R Sbe cur!edTJ %475* mo)0ni ma0mpa) mo)0yo* %n*1u)0'wi *i

Ieati004.4F;;J

+#8)water =G8)this

=G8)beQ/6S

wit!E/C71distance

this water is deep Ilit Sis with depthTJ Ieati004.4F28J 0inally, 0ni can also occur after all fi!e of the demonstrati!e ad!erbs %L8.8.4.4*, in which case they ha!e < tone rather than <6, and the suffi3 has a 6 tone rather than its normal < e.$. pa*apaani ) ShereT, le*eleeni ) SthereT. 208 hether the bare ad!erbs and their

SnominalisedT counterparts differ in their synta3 or semantics is not yet known.


&&9# ;ocative 0tu

.e!erbal nominals formed by addin$ the suffi3 0tu denote the place where the action encoded by the !erb characteristically takes place, as in ka)0pi taatu Ssole, footprint, trackT from pita SstepT, i.e. Splace in!ol!ed in steppin$T, and ka)05angalaatu Supper part of the cheekT from 5angala SslapT, both basic)le!el terms for body parts. Eerbal affi3es can be included in the nominalisation, as shown by %477* for the anticausati!e %L8.5.8.;* and %479* for the pluractional %L8.5.F.8*. %477* m)'puwtu m)0'p0wa0tu

+#8)hold)=+"I#)6#EH/

handle Ilit Splace which $ets heldTJ Ieamd049.5;4J %479* ma)7aani ma)ti sa) ma)07aani ma)0ti sa)
+#8)pen

=G8)'oinQ/6S +#4)close16=#s)6#EH/

ko)'um olootu ko)0'um ilso0tu

the pen %oins the lid Ilit Splace which repeatedly closesTJ Ieaim007.485;J "he tone on the resultin$ nominal is all < e3cept a 6 on the prefi3, and as with 0ni the final !owel of the !erb is len$thened before addin$ the suffi3 88. "he noun is most often assi$ned to the 4N2 %&=N=* $ender %LF.8.F*, but 8NF and 9N; ha!e also been obser!ed.

*.-

4repositions

#icipu has prepositions but no postpositions, in line with the obser!ations made by Greenber$ %495;* for E2 lan$ua$es. "his is a small cate$ory in #icipu since the $eneral locati!e clitic A co!ers a wide ran$e of semantic relations. "he most common SprepositionT is actually a procliticK the locati!e A. "his has been analysed as a clitic rather than a free form because it harmonises with the followin$ word8F, and is therefore within the domain of !owel harmony. "he tone is always hi$h, and results in downstep85 for the rest of the intonation unit %L;.8.7*. "he followin$ e3amples demonstrate the different phonetic forms of A, as well as a sample
88 ko)0dontu Sstool, chairT from doonu SsitT is an irre$ular e3ception. 8F /ecall that A stands for a !owel which harmonises with the word to which it is attached. 85 "he (antu lan$ua$e Aks J eJ has a !ery similar locati!e prefi3 a,0, also causin$ downstep %<edin$er 4990K40*.

20F

of the wide !ariety of locati!e meanin$s it can e3press. %479* a,1$i 0ku*u
;6CR+#5)waist

tu,1u08anga)
=G5R+#7)tree

at the base o the tree I62#="IE-, tats00F.002.459J %490* a,1a07a


;6CR+#2)person

Kwangwa)
ItownJ

Ili!in$J amongst the people o 3wangwa I62#="IE-, tats00F.002.459J %494* e,1k0ke)eke on a bicycle I2+ SA/0=#-, tapf004.00F.044J %492* o,1ko0o$i ) into a hole II+SI.- #2+"=I+-/, s!tm$004.477J %49;* u)0yo*o
;S)$oQ/6S
;6CR+#9)hole ;6CR+#9)bicycle

a,1ka0da a
;6CR+#4)countryside

he went to the bush I"2 %498* see n)1)0k8)0nu) ka0tii*


+#4)head

=/.S G2=6,

saat004.002.007J ya0avu)
=G;)2S.12SS

ke0lle)
=G4)that

o1ru0'u
;6CR+#;)body

unless andR;1)cutQ/6S)/-S

you have to cut o that head from your body I= =B 0/2M S2A/#-, tats008.002.08FJ * ) : %49F* a)tiivi * a)0taa1vi ):
;1)shootQ/6SR;S.1/2

) ekri hya'a ) e,1ri 0hya'a


62#R+#;)arrow

they shot him 1 with arrows 1 II+S"/AM-+"=6, s!my004.022J %495* a,1a0naa


;6CR+#2)le$

by oot Ii.e. means of transportJ II+S"/AM-+"=6M, sayb004.;78J hile %490* has a noun denotin$ human referents as the complement of the preposition, this only seems possible with plural nouns. "o e3press a locational relation in!ol!in$ a sin$ular human such as e.$. coming towards him, the noun 'asu) SplaceT is used instead,

205

with the human as possessor e.$. uuwa)ya) 'asu) weevi ) %lit. Sin comin$ place of himT*. Gi!en that A seems to be bound to the followin$ word, the Duestion arises as to whether it should be analysed as a case marker rather than a preposition. "here are two reasons why it has been called a preposition here, despite its morpholo$ical status. 0irst, there are no unambi$uous case markers in #icipu such as an accusati!e. Secondly, the preposition is not S$o!ernedT by the !erb , it seems that no !erb reDuires one of its ar$uments to be marked with A87. "he other common preposition is n) SandNwithT %L8.8.F.8*. "i SlikeNsimilar toT and $unu) SwithoutT are only attested a few times in the corpus, but they are also candidates for prepositions. =lternati!ely, ti may 'ust be the
=G5

associati!e a$reement marker in a headless associati!e construction %L8.8.F.4, see L5.; for the connection between $ender 5 and SmannerT*. In that case structurally)faithful translations of %497* and %499* would be somethin$ like they will bring it out in the manner o a mirror, and back then it wasn't like the mannero this thing. %497* h0u0uto)0wo)
;1)0A")$oPout)=116 like

ti

;0maduu i
+#9)mirror

they will bring IitJ out like a mirror Isayb004.809J %499* ;0looka)$i vi 0 lle)
+#9)time =G9)that

ana) ti , 1i0ri 0mpa)a0ni )

how likeR+#;)thin$)this)+M6H

back then it wasn't like this Isayb004.8;0J %499* n)1a)0ka a)


andR;1)takeQ/6S

u)08anga)
+#7)tree

* $unu) a)0si 0 ) yaa

;1)<=()do wit!out

;0 i ndi ga)
+#9)$un

when they took a piece o wood they would do without guns Isamoh004.057J 2ther prepositions include de)ge) SfromT, see SuntilT, hari ) SuntilT, borrowed from <ausa daga, sai and har respecti!ely. "he first two of these are illustrated in the followin$ idiomatic e3pression %which is structurally eDui!alent in <ausa*. %490* de)ge) a)mu
'rom 4S.1/2 until

see

) ka)0maya

+#4)elderPsiblin$

ko)0vo*o

=G4)4S.12SS

%ust me and my elder brother Isayb004.00;J


87 2uwa SputT does occur with a locati!e 11 after the ob'ect but this is not obli$atory as in the case of -n$lish put. "he 11, when it does occur, could be analysed as as an ad'unct rather than a complement.

207

*.2

1erbs and the verb phrase

"his section describes the #icipu !erb phrase. =fter a preliminary section on the structure of the !erbal word %L8.5.4*, the core of the section is or$anised accordin$ to the ma'or $rammatical cate$ories distin$uished in the !erb , mood %L8.5.2*, aspect and tense %L8.5.;*, and !alence %L8.5.8*. "here follows a discussion of se!eral other miscellaneous !erbal suffi3es %L8.5.F*. "he ne3t sections discuss borrowed !erbs %L8.5.5* and semi)$rammaticalised au3iliaries %L8.5.7*. "he final two sections discuss the relati!e orderin$ %L8.5.9* and co)occurrence possibilities %L8.5.9* of the !arious affi3es. "he o!erall or$anisation of this section is functional rather than formal. So, for e3ample, the subsection on aspect and tense co!ers both morpholo$ical and syntactic e3pressions of aspect. ithin each subsection, howe!er, morpholo$ical e3ponents are treated before phrasal ones.

&91

1erbal word template

=lthou$h #icipu is spoken in the north)west of +i$eria, it is typolo$ically !ery similar to the (antu lan$ua$es of southern and eastern =frica. "his similarity manifests itself in two !ery ob!ious ways. 0irst, in the robustness and re$ularity of its noun class system, as set out in 1art III, and secondly in the structure of the !erbal word. #icipu is hi$hly a$$lutinati!eG not only is there a lar$e number of !erbal affi3es, many of them can occur simultaneously, resultin$ in !erbal words consistin$ of up to ten concurrent morphemes %includin$ the !erb root and the ob'ect enclitic*K %494* 77a 707a
+#9)person

nna) n0na)
=G9)/-6

u)to i )li )si )su)wo)wo)no)mu) u)0to i )lsi )ssi )sso0wo)0wo)0no)1mu)


;S)coolQ/6S16=#s#=ASs#=ASs)=+"I#)=116)10E R4S.1/2

shayi ) ;0shayi )
+#9)tea

the person who has caused tea to become cooled down in a orce ul and iterative ashion or me I2009)02)44.002J hile such monsters are !anishin$ly rare in e!eryday speech, it is common to find three or more se$mental affi3es on a !erb, in addition to the ubiDuitous tone pattern which e3presses the $rammatical mood. "he structure of the !erbal word in #icipu is similar to, but not identical to, the pattern found in (antu. "he #icipu !erbal word follows the followin$ templateK

209

%492*

=G/)

%#*E)

%<=(N0A")* si0:u0

root C,C,

%e3tension%s*s* E#s

%)suffi3* )#E

%R2(J.1/2* R#E

=G/

is a sin$le slot for a sub'ect a$reement prefi3, which may come from either the

$ender)marked paradi$m or the person)marked paradi$m. "hese prefi3es are discussed at len$th in chapter 7 and so no more will be said about them now. 2ther than these a$reement prefi3es, only two affi3es can come before the root, the future u0 %L8.5.;.2* and the habitual si0 %L8.5.;.;*. "he e3tension slot can be filled by either the pluractional infi3 ils %L8.5.F.8*, the causati!e infi3 iss %L8.5.8.4*, or both. "he Ssuffi3T slot can be filled by !arious di!erse !erbal affi3es, includin$ aspectual and !alence)chan$in$ de!ices. In addition to the se$mental affi3es, the tone pattern on the !erbal word e3presses the mood of the clause. More is said about orderin$ and the possible combinations of affi3es in L8.5.9)8.5.9. 0ollowin$ any suffi3es, a pronominal ob'ect clitic %L7.;* may attach to the end of the !erbal word. 2ne important difference between the #icipu and (antu !erbal word deri!es from the structure of !erb roots. (antu !erb roots ha!e a #E# shape, with a dummy Sfinal !owelT bein$ supplied accordin$ to rule. Eerbal Se3tensionsT are E#)shaped morphemes which come between the root and the final !owel. =lthou$h such e3tensions are word) internal, $i!en that the final !owel is predictable both it and the e3tension can be analysed as suffi3es rather than infi3es %e.$. Schadeber$ 200;*. 0or #icipu, howe!er, e3tensions seem to be true infi3es89. In appro3imately half %429 out of 275* of the disyllabic !erb roots, E4 and E2 differ in !owel Duality. Earious rules of increasin$ comple3ity and diminishin$ returns could be set up to deal with some of the remainder, but in many cases E2 must be le3ically)specified. -3amples %49;)498* show the !erbs pina Ssha!eT, pino SboilT, dooho SdisappearT, and goonu ShelpT to$ether with the causati!e iss and pluractional ils morphemes. "he final !owel in these e3amples cannot be determined by any phonolo$ical rules, and so must be considered as part of the root. "hus the causati!e and pluractional morphemes interrupt two portions of the root and must be analysed as true infi3es.

89 =nd other

est &ain'i lan$ua$es %e.$. #:6ela , Ste!e .ettweiler p.c.*.

209

%49;* %a* u)0pi nQi )lRa)

;S)sha!eP;ACs

he shaved Imany timesJ %b* u)0pi nQi )lRo)


;S)boilP;ACs

it boiled Imany timesJ I2009)02)42.009J %498* %a* u)0do*ohQi )sRo)


;S)disappearCAD0s

he caused s.t. to disappear %b* u)0go*onQi )sRu)


;S)helpCAD0s

he had s.o. to help I2009)02)42.009J

&9#

'ood

"he fundamental distinction in #icipu !erbs is mood. "here are a number of aspectual morphemes but none of them are obli$atory, and !ery often the !erbal word consists simply of a sub'ect a$reement prefi3 to$ether with the !erb root. Mood, howe!er, must always be specified, and this of course is tied into the way in which this cate$ory is encoded , throu$h the tonal melody of the !erbal word. +o speaker can produce a pitchless utterance, and so almost all clauses ha!e a specific !alue for moodK realis, irrealis, imperati!e89.
&9#1 !ealis

In the realis mood the basic tone pattern is 6<6%6* as in %49F*, althou$h in certain circumstances it may be <6%6* , see L;.8.5 for details. %49F* u)0dukwa) u0dukwamaTa
CO0gom{aO

he went Isaat002.002.44;J In the absence of a pastNpresent tense distinction in #icipu, clauses headed by stati!e !erbs marked with realis mood are ambi$uous between past and present interpretation e.$. u)0$uwo) can mean SheNshe was fullT or SheNshe is fullT.
89 #lauses with habitual aspect %L8.5.;.;* are an e3ception, since they ha!e another tone pattern alto$ether. .ependent imperfecti!e clauses %L8.5.;.8* also ha!e a distinct tone pattern.

240

In certain restricted conte3ts the realis can be used to encode e!ents which ha!e not yet been realised, in particular those e3pected to happen in the imminent future F0. 0or e3ample the lea!e)takin$ e3pression in %495* is used when the speaker knows they will see the addressees a$ain shortly. %495* see
until andR21)comePhomeQ#;0

n1i )0to*ono)

until you .pl.2 come home


&9## Irrealis

In the irrealis mood the basic tone pattern is <%6*, with < bein$ realised on the sub'ect a$reement prefi3. 0or further details on the tone, see L;.8.5. Irrealis mood can mark !erbs in independent clauses, as in %497*, in which case , as in <ausa %Ja$$ar 2004K49F* , it e3presses the speaker:s desire that somethin$ should or should not happen. used in the second person this is politer than usin$ the imperati!e %L8.5.2.;*. %497* u$a)a u0$aamTa
;S)$i!eQI//

hen

let him give Ioamy004.229J -3ample %499* shows the use of the irrealis with the prohibiti!e kada) %from <ausa kada*F4. %499* kada) d)n) kada) 0dnQ<6
1/2<

;1)followQI//

they shouldn't ollow Isaat004.005.088J =lternati!ely, irrealis mood can occur in the subordinate clause complements of certain !erbs %e.$. ta'a SwantT, kamaata Sbe fittin$T*. #lauses marked with irrealis mood normally encode e!ents that are yet to be realised, althou$h two classes of e3ceptions can be mentioned. "he first in!ol!es habitual e!ents, which can be encoded usin$ irrealis mood. "his is especially true for past habitual e!ents %200*, for which the habitual aspect prefi3 si0 %L8.5.;.;* is rarely used.
F0 cf. the <ausa perfecti!eNcompleti!e %Ja$$ar 2004K4F9*. F4 +e$ati!e prohibitions can also be e3pressed by usin$ a*a to ne$ate a positi!e irrealis clause, but this is much rarer in the corpus %see L8.;.F.4 for an e3ample*.

244

%499* n1a)0ku8u)wa),

whenR;1)peelPoffQ/6S ;1)rubQI## ;1)dryQI##

0h)m), e0'e)tu)

when they've trimmed the leaves, they rub them, they dry them Is!tm$004.007J ) yuu %200* ti 0
41)wearMI##

i 0yayi a :

+#;)$irdlePk.o.

) ti 0 yuu

41)wearMI##

a0da ala

+#2)beadPk.o.

) ti 0 yaa

41)doQI##

ke0ku)07e)'u) :

+#4)+#9)dance

we would wear iyayibau girdles 1 we would wear decorative beads we would dance Is!tm$004.297J = comparable constellation of functions in!ol!in$ habituals %past or present* and sub'uncti!e mood is found in <ausa %Ja$$ar 2004K494)49;*. Ja$$ar, followin$ olff %499;*F2, su$$ests there are two separate homophonous paradi$ms. If so, the parallel use of the irrealis for the same two functions in #icipu is remarkable. -ither this comin$ to$ether is less anomalous than has been supposed by <ausa scholarsF;, and is an independent property of the #icipu irrealis, or it is the result of contact with <ausa, su$$estin$ that in #icipu, at least, the irrealis tone pattern should be assi$ned a unitary function. "he second e3ception is the use of the irrealis after the sentence connecti!e see SthenT %from <ausa sai* in narrati!es to encode non)habitual past e!entsK %204* see
then

u$a)a

;S)$i!eQI//

;0awli )

+#9)lar$ePspider

then he gave (ig dpider IwaterJ Isaat002.002.080J "he conflict between the irrealis markin$ and the semantics in %204* su$$ests that the < 6 pattern on the !erb in such e3amples mi$ht be a morphotonolo$ical process caused by see, rather than an e3pression of the cate$ory of irrealis mood.
&9#% Imperative

In the imperati!e mood there is no sub'ect a$reement prefi3. "he basic melody is %6*< , in other words, the final tone of the !erbal is <, and any pre!ious tones are 6. =$ain, for further details, see L;.8.5.

F2 See also the discussion in +ewman %2000KF9;)F95*. F; In fact this combination is also found in the =ustronesian lan$ua$e &ubokota %#hambers 2009* , see 1almer %2004* for further =ustronesian and 1apuan e3amples.

242

%202* du)kwa_ dukwaQ;? $oQI8P go#

Isaat002.002.2F5J

If the !erb root is !owel)initial, then a semi)!owel is added to the start of the word, as in %20;*. "his semi)!owel surfaces as NwN unless the root)initial !owel is NiN, in which case N yN occursF8. %20;* %a* we)ewe eweQ6< re use# > @ yi )nda indaQ6<
seeQIM1

refuseQIM1

see# Ieamd005.04FJ

If an imperati!e is addressed to more than one person, then the plural imperati!e suffi3 0nA) is obli$atory. "he imperati!e tone pattern is also present, but this time the < tone falls on the penultimate syllable, and the plural imperati!e suffi3 is always 6 tone. %208* du)kwana) dukwamaT0na) $oQIM1)I8P.P; Iyou %pl.*J $o# Itats00F.004.492J 1rohibitions are not e3pressed usin$ the imperati!e mood, but rather with second person irrealis mood and the prohibiti!e kada) %L8.5.2.2*.
&9#& Counterfactual

#ounterfactual mood can be e3pressed analytically usin$ the au3iliary saa, as in %20F) 205*. "he main !erb occurs after the au3iliary and has no sub'ect prefi3. "he tone on the main !erb is often < < %especially for #E#E roots , cf. L;.4.; on the habitual tone pattern*, but is not yet fully understood. %20F* saa ko)0to*o ti )0saa
or
=G4)one

41)mig!t $et

ti yo

we might even have got every one Iif you hadn:t messed upJ Isaat004.005.0F8J

F8 a)ya Scome?T without any initial semi)!owel is an e3ception.

24;

%205* n)0saa

4S)mig!t kill

huna i )0ri

+#;)thin$ =G;)4S.12SS

yi 0vo*o

' ought to have killed my animal Isaat002.002.F48J

&9%
&9%1

Aspect and tense


Perfective

"he suffi3 )nA occurs in back$rounded clauses which encode situations !iewed as completed. It is more)or)less obli$atory in dependent clauses be$innin$ with the con'unction ana) ShowNlikeT %207* and in relati!e clauses encodin$ situations !iewed as complete %209*. %207* u)0dukwa) n0ka$i :
;S)$oQ/6S
+#F)huntin$

ana)

how

u)0dukwa)0na) n0ka$i ...


;S)$oQ/6S)P>,

+#F)huntin$

he went hunting 1 when he had gone hunting... I"idipo, saat002.00;.0;4J %209* a 707a
+-G +#9)person

n0na)

=G9)/-6

guya)0na)

canQ/6S1P>,

;0yaari )

+#9)lan$ua$e

vi 0ive)

=G9);1.12SS

there's nocone who knows their language Is!tm$004.254J hat appears to be the same suffi3 )nA also occurs whene!er a constituent is fronted throu$h focus or marked topicalisation %L8.;.4)8.;.2*. =$ain the e!ent encoded has to be !iewed as complete. %209* k0e)ne
=G4)which +#4)oldPman

ka)0 ara

k0e)

=G4)#21

ka0aya)0na)l

=G4)comeQ/6S)P>,

which old man came Ieamy00F.027J %240* ka)0 ara


+#4)oldPman =G4)this

ka0mpa) k0e)

=G4)#21

ka0aya)0na)

=G4)comeQ/6S)P>,

this old man came Ieamy00F.029J =s was mentioned in L8.;.7.2, the perfecti!e is !ery often found in tail)head linka$e constructions.
&9%# 8uture

"he future tense in #icipu is encoded by the u0 prefi3, which occurs between the sub'ect prefi3 and the !erb root. It always occurs with the irrealis %< 6* tone pattern. 248

%244* mu0u0yo)o

4S)>D()$oQI//

$i 0ni yaa
+#5)farm

' will go to the arm

Ieamd048.;48J

"he e3act $lossin$ of the future morpheme in #icipu is problematic. Speakers usually translate it with the <ausa future, and it mostly %but not always* refers to e!ents in the future accordin$ to some reference time. =n e3ception is the use of the future to indicate habitual e!ents in the past, as in %242* %cf. would in -n$lish*. In these cases it is not clear how the meanin$ differs from the basic irrealis form %cf. L8.5.2.2*. %242* a)07a
+#2)person

ha,1d0do)ori

=G2R+#9)formerly

hu0u0gu)ya)

;1)>D()takeQI//

u)0ka a) i )0tangi )
+#7)take

+#;)item

the people o be ore would be able to carry things


&9%% Habitual

Is!tm$004.0F9J

<abitual meanin$ can be e3pressed usin$ the prefi3 si1, which occurs in the same slot as the future prefi3. =dditionally there is a specifically habitual tone pattern o!erlaid on the !erbal word, distinct from each of the mood tone patterns %L8.5.2*. =s already discussed in L;.4.;, the e3act tone pattern depends on the wei$ht of the first root syllableFF. "he pattern also depends on synta3 , in particular whether a constituent has been fronted for focus or relati!ised. "he different possibilities are summarised in "able 27 and e3emplified in %24;)249*. 0able 4_- 0one on the habitual pre i* plus root, according to root structure #oot structure #E#E #EK#EN#E+#E #E#KE #EK %24;* u)0si )0ta'a 5e'ault tone "attern 6 < < %e3. 24;* 6 <6 6 %e3. 248* 6 < 6 %e3. 24F* 6 < or 6 <6 %e3. 245* >ronted tone "attern < 6 6 %e3. 247* or < < 6 < <6 6 %e3. 249* < < 6M < <6 %e3. 249*

;S)?AB1want

he wants Isamy004.029J

FF 0or mono) or disyllabic stems. "risyllabic stems show the Shea!yT pattern e.$. u)0si 0ka) a)0na) ;1S)<=() take)E-+" SheNshe brin$sT.

24F

%248* u)0si )0wi *ina)


;S)?AB1sell

he sells I2009)08)04.008J %24F* u)0si )0htt)


;S)?AB1warmFbyF'ire

he warms himsel by the ire I2009)08)04.008J ) hyaa %245* %a* u)0si 0


;S)?AB1say

%b* u)0si )0$a*aF5 he gives

;S)?AB1give

he says %247* le*e


there

I2009)08)04.008J v0i )
=G9)#21

a)0si 0ku)d)
;1)?AB1meet

it's there they meet I"ikula, s!sdt004.00FJ %249* le*e


there

v0i )
=G9)#21

a)0si 0wa*aya)
;1)?AB1come

it's there they come I"ikula, s!sdt004.00;J * %249* yi )ni i 0 ) si 0yuul


what 21)?AB1wear

what did you .pl.2 wear Isayb004.;49J Most of the time si0 occurs in utterances with a habitual or $eneric interpretationK %220* ta0a a)
+#5)tobacco =G5)?AB)kill

ti )0si 0 ) huna

a0nga

+#2)tooth

tobacco kills the teeth Itats007.002.0F9J <owe!er this is not necessarily the case, and the prefi3 may also occur in clauses denotin$ a sin$le e!ent. In these cases the prefi3 seems to indicate continuous aspect, with a meanin$ similar to that of the analytic structure described in L8.5.;.5.
&9%& $ependent imperfective

.ependent SwhenT clauses may be$in with one of two con'unctions , ana) ShowT or n) SandNwithT. =s noted in L8.5.;.4, in clauses be$innin$ with ana) the !erb is in realis
F5 Some monosyllabic !erbs pattern like hyaa, others like $aa.

245

mood and almost always carries the perfecti!e aspect suffi3 )nA. In clauses be$innin$ with n), on the other hand, the perfecti!e affi3 cannot occur. If the !erb carries the realis tone pattern with the usual final !owel, then the e!ent is interpreted as perfecti!e. If, on the other hand, the !erb has a %6*< tone pattern and the le3ically)specified final !owel is replaced by either i or u, then the interpretation is imperfecti!e, as in %224*K %224* naa)mi )ni n)1a0mina0i see a)7a*ana) see a)07a*a0na) da8a) da8a) i )ri mpa)ani ) i )0ri 0 mpa)0ni )

andR;1)skin)57P.I8P> then ;1)seeQ/6S)10E moreo!er

+#;)thin$)this)+M6H

while they were skinning, they saw something I"idipo, saat002.008.058J If the last !owel of the !erb stem is u as in %222)22;a* then it does not chan$e. =s with the final !owel in imperati!es %L;.8.5*, a < tone on an u !owel is realised as <6. -3ample %22;* demonstrates that it is the last !owel of the !erb stem that chan$es, rather than the last !owel of the root. %222* nu)ujo)olu* n)0u0joolu0i

and);S)check)57P.I8P>

while he was checking I2009)08)4;.002J %22;* %a* nu)u<e)lle)nu* n)0u0<elle0nu0i

and);S)cut)/-S)57P.I8P>

while he was skimming Ii.e. stones, lit. Scuttin$TJ %b* nu)u<e)lle)ni n)0u0<elle0na0i

and);S)cut)E-+")57P.I8P>

while he was cutting down I2009)08)4;.002J =s mentioned in L;.4.8, if the !erb stem is monosyllabic then the !owel is diphthon$ised and the tone is risin$. "here is no constraint in!ol!in$ the sub'ect of the dependent clause and the $rammatical functions in the main clause , in other words, dependent imperfecti!es are not ScontrolT structures.
&9%( Progressive

"he pro$ressi!e suffi3 )su can attach to a few !erb roots, resultin$ in a !erb appropriate

247

for describin$ actions done repeatedly %228* or to e3cess %22F*. %228* see kuma a0gi )tu)
until more ;1)$oPbackQI// ;1)soakQI//)P#6-

a08a)su)0su)

ku)0yaama)

+#9)namePofPmonth

then they urther soak the 3uyaama beer I"ikula, sami004.290J %22F* ;0'wa*a0su)
2S)passQ/6S)P#6-

ka)0tii

+#4)head

ti )0gai

+#5)len$th

your head is too long Ilit. Syou e3ceed head len$thTJ Isaat004.005.497J
&9%9 Continuous

#ontinuous aspect is e3pressed usin$ the au3iliary !erb yo SbeT followed by the locati!e A %L8.F* and then the infiniti!eN$erundial form of the !erb %LF.8.4*. %225* u)yo* u)0yo* uu )l) aRu0 l ) mo)hi'i ) mo)0hi'i

;S)beQ/6S

62#R+#7)search

+#8)blood

it was looking or blood Ilit. Sit was at lookin$ for bloodTJ I2009)02)44.004J "his combination is also found in some (antu lan$ua$es % elmers 497;K457*, and indeed similar constructions occurs freDuently in the world:s lan$ua$es %(ybee, 1erkins, and 1a$liuca 4998K4;0)4;2*. Most of the time the !owel of the au3iliary coalesces with the initial !owel of the infiniti!e to form what may be $lossed as a continuous aspect morpheme yuu, as in %227*. "he u !owel is usually lon$ but may occur shortened. +ati!e speaker consultants percei!e the full and reduced forms as identical, and when an ad!erbial phrase such as n) SreallyT is added as in %229*, the au3iliary is clearly identifiable. ma)huu %227* u)0yuu
;S)#2+" search

)l)

) mo)0hi'i

+#8)blood

it was looking or blood Isamy002.049J %229* u)yo* u)0yo* n) n) ma)huu uu )l) ma)0huu aRu0 l
/-E0truth

;S)beQ/6S with

62#R+#7)search

) mo)hi'i ) mo)0hi'i

+#8)blood

it was looking or blood Ilit. Sit was at lookin$ for bloodTJ I2009)02)44.004J It is possible, for at least some speakers, to separate the coalesced continuous morpheme yuu from the main !erb, as in %229*, which is in fact my main moti!ation for $lossin$ it 249

as a separate aspect morphemeK le*e %229* yuu 0yuu le*e uudu)kwa) sul aKEu0dukwa su
;6CE/C7)$o _

2S)C6/( there

Are you about to go I2009)08)49.002J "his last e3ample demonstrates the use of the continuous aspect in con'unction with le*e SthereT to e3press an imminent e!entG literally are you there incgoing 0inally, it should be noted that the i:u alternation %L;.;.2* applies to this morpheme. "he basic !owel seems to be u, which is not surprisin$ $i!en yuu is a contraction of yo u. <owe!er in the nei$hbourhood of an i !owel the continuous morpheme often surfaces as yi i instead.
&9%< Perfect

1ast e!ents with present rele!ance are often coded with the deictic 'u* Sthere far offT or Sthere %in!isible*T. .espite the use of the most distal deictic, the e!ent encoded is almost always recent, and the cate$ory marked is best described as a resultati!e perfect rather than an e3periential one %.ahl and Eelupillai 200F*. #onseDuently this construction is found more often in con!ersation than in other $enres. In narrati!es of past e!ents it is usually felicitously translated by the -n$lish pluperfect, as in %2;;*. %2;0* ;0'ungo) 'u*l
2S)riseQ/6S t!ere

you've risen Istandard opener in mornin$ $reetin$sJ I2009)02)02.008J %2;4* u)0dukwa) 'u*
;S)$oQ/6S t!ere

he's gone Isaid in response to a Duery about someone:s whereaboutsJ I2009)02)02.008J %2;2* n)0kt)
4S)finishQ/6S t!ere

'u*

' have inished Isaid at the end of a speechJ Isaat002.009.009J %2;;* ana)
when ;S)$oQ/6S)10E

u)0dukwa)0na), e)see,

actually

;0naata)

+#9)spider

;0ni i'wa)

=G9)meltQ/6S

'u*

t!ere

when he went, actually, dpider had already rechydrated Isaat004.009.408J

249

'u* always comes after the ob'ect enclitic %L7.;*, if there is one. If there is an ob'ect +1 in the complement of E1 then the deictic can come either before %2;8* or after %2;F* it. %2;8* anna) n)0huna)
today 4S)killQ/6S t!ere

'u*

k0kaa

/C)1wi'e A-)110.P600

v0vo*o

today ' have killed my wi e Isaat002.002.;85J %2;F* u)0ta a)ta)


;S)touchQ/6S

;0lwli )

+#9)lar$ePspider

;0ra*a

=G9)eatQ/6S

i )0nama)

/C31meat

yi 0lle)

A-31t!at

'u*

t!ere

he elt sure (ig dpider had eaten that meat Isaat002.002.528J

&9&

1alence

Anderi!ed #icipu !erbs can be classified as intransiti!e, %mono*transiti!e, or ditransiti!e. Sub'ect a$reement is obli$atory, althou$h the le3ical sub'ect may be omitted. "his state of affairs can be obser!ed across (antu %(earth 200;K422*, and the details with respect to #icipu are discussed in L7.5. "here are no ob'ect a$reement markers. 2b'ects are freDuently omittedF7, as in %2;5*, e!en when they are specific and bear an important role in the semantic representation of the !erb. "he moti!ation seems to be economy of e3pression. =s in <ausa %Ja$$ar 2004K502* ob'ect +1s with human referents are not normally omitted. %2;5* ana) u)0 d)0n)1vi ) u)0ka a)
when ;S)applyQ/6S)10EE30.P#6 ;S)takeQ/6S

mo)0to*o

+#8)sali!a

w0aya)

;S)comeQ/6S ;S)peelQ/6S

u)0ku8u)wa) ; da8a) u)08asu) ;

moreo!er

; u)0dukwa)0wa)
;S)$oQ/6S)=116

; a1ku)0jene)

62#R+#9)ri!er

u)0yo*

;S)beQ/6S

+#7)soak

when he had covered it ISpiderTs bodyJ with saliva he peeled IitJ off, he took IitJ and went with IitJ to the river and was soaking IitJ Isaat004.009.099J 2b'ect omission is also common in (antu lan$ua$es, but accordin$ to (earth %200;K42;* it ser!es a different purpose, that of limitin$ what is said to the main point. In other words, speakers omit the ob'ect not because the hearer will be able to reco!er the reference from the conte3t, but to a!oid !iolatin$ GriceTs %497F* ma3im of Duantity , they simply donTt want to specify referents if they are not rele!ant %cf. 'uve eaten already*. #icipu is closer to <ausa in this respect , in %2;5* the identity of the une3pressed ar$uments is perfectly clear %in this case it is the body of Spider*.
F7 See L8.;.4.4 for ditransiti!e !erbs.

220

Many (enue)#on$o lan$ua$es also ha!e se!eral deri!ational affi3es which affect the !alence of the !erb, and #icipu is no e3ceptionK the causati!e infi3 iss and the applicati!e suffi3 )wA increase the !alence of the !erb by one, while the anticausati!e suffi3 )wA decreases the !alence by one.
&9&1 Causative

"he causati!e iss was first mentioned in L8.5.4, alon$ with the reason for analysin$ it as an infi3. "his morpheme is doubtless deri!ed from the same source historically as the causati!e !erbal e3tension )is found across (antu %Schadeber$ 200;K7;*, althou$h in #icipu it may only be added to intransiti!e stems. It can co)occur with transiti!e roots, but only if the anticausati!e suffi3 )wA is also present so that the !alence remains at 2. "his combination differs subtly in its semantics from the bare root, as discussed in L8.5.8.; below. %2;7* hina dooho sukulu yuwo 'ungo he'we ru5a kullo ripen disappear move all rise dry sink burn hinisa doohiso yuwoso 'ungoso he'wise ru5isa kulliso cause to ripen cause to disappear cause to all raise cause to dry cause to sink cause to burn

sukulusu cause to move

0or monosyllabic !erbs the IiJ seems to be absorbed in the root !owel, as in the pairs soosooso Scrymake s.o. cryT, and raaraasa Seatser!e s.o. foodT. =s in certain (antu lan$ua$es %Schadeber$ 200;K78*, what appearsF9 to be the same infi3 has an additional functionK rather than affectin$ the !alence of the !erb, applyin$ the iss infi3 can ha!e an intensifyin$ effect insteadK %2;9* %a* )'p) )0'p)
;1)holdQ/6S

%b* )'pa)nu) )0'p)0nu)


;1)holdQ/6S)/-S

%c* )'pi )sa)nu) )0'pQi )sR)0nu)


;1)holdQ/6SCAD0s)/-S

they hold

they grab

they grab with great orce Itats00F.002.099J

F9 =s Schadeber$ notes, it is hard to see the moti!ation for the shift in meanin$.

224

) no %2;9* ko)ndi )ssi 0


enterCAD0sQIM1)E-+"

a,1ka0aya_
62#R+#4)room

come into the room# Isaid with ur$encyJ I2009)02)44.044J "he causati!e morpheme can be repeated as in e3ample %494* at the be$innin$ of this section. "he effect of this reduplication is not well understood, althou$h it may ha!e the effect of intensification in addition to causation. "here are alternati!e means of e3pressin$ causation in #icipu apart from the iss causati!eF9. 2ne way is usin$ a periphrastic construction in!ol!in$ the !erb yuu Swear, appoint, causeT. "his construction has a direct analo$ue in <ausa usin$ the !erb sa Sput, wear, appoint, causeT %Ja$$ar 2004KFF2*, and $i!en the similarity in le3ical meanin$ it seems likely #icipu has borrowed the construction. = causati!e meanin$ can also be obtained by replacin$ the final !owel of the root with NuNK e.$. kondoNkondu SenterNcause s.t. to enterT, koyoNkoyu SlearnNteachT. It is not known how producti!e this process is, nor how or if these causati!es differ in meanin$ from the iss ones.
&9&# Applicative

"he most ob!ious effect of the applicati!e suffi3 )wA is to increase the !alence of the !erb by one. If the basic !erb is intransiti!e, then the deri!ed !erb has a sin$le ob'ect. If the basic !erb is transiti!e, then the ob'ect of the basic !erb %280a* becomes the secondary ob'ect, the primary ob'ect slot bein$ filled by the Sdati!eT ob'ect as in %280b*. "he sub'ect is unaffected in all cases. I was not able to elicit applicati!e forms of any basically)ditransiti!e !erbs such as $aa S$i!eT %e.$. ' gave ood to the child or its mother*. %280* %a* u)0dama)
;S)speakQ/6S

Ii )0ri

+#;)thin$

yi 0 ) 'eteiL
=G;)fine

he spoke good things %b* u)damu)wa) u)0dama)0wa) Ia)7a)L a07a


/C

;S)speakQ/6S)APP;

1"erson

Ii )ri i )0ri

+#;)thin$ =G;)fine

yi )'eteiL yi 0 ) 'etei

he told the people good things "he semantic role of the primary ob'ect is !ariable, as was the case for the morpholo$ically)unmarked double)ob'ect constructions %L8.;.4.4*. 2ften it is the
F9 See also L8.5.8.2.

222

/-#I1I-+"

or

(-+-0I#I=/B

of the encoded e!ent, as in %284*. -Dually often the primary %28;*, %288*, and
"<-M-N1="I-+"

ob'ect encodes a participant ad!ersely affected by the action , a SM=6-0I#I=/BT, as in %282*. 2ther semantic roles include ) 1Ivu)L %284* mu0u0yaa0wa)
62#="IE#2MI"="IE-

%28F* , but not I+S"/AM-+"=6. "he primary ob'ect is often omitted, as in %288*.
4S)0A")doQI//)APP;R2S.1/2

Ika)0'a77i )ki L

+#4)prosperity

' will make you .sg.2 prosper Ilit. do you prosperityJ I(-+-0=#"IE-, saat002.002.;9;J %282* 707a
+#9)person =G9)/-6

n0na)

* si 0 ) yaa0wa) 1Ivu)L

)<=()do)APP;R2S.1/2

Iu)0 w)L
+#7)steal

the one who's been stealing rom you Ilit. doin$ you stealin$J IM=6-0=#"IE-, saat002.004.408J n) %28;* a)dukwu)wa)vi ) a)0dukwa)0wa)1Ivi )L n) wulle) i )ri wu0lle) i 0 ) ri

;S)$oQ/6S)APP;R;S.1/2 with ;S)that

+#;)thin$

they went to it Ithe spiritJ with something I62#="IE-, saat004.009.099J %288* u)dukwu)wa) u)0dukwa)0wa) IJ
;S)$oQ/6S)APP;

akuJjeJne) a,1ku0jene)
62#R+#9)ri!er

he went with Ithe corpseJ to the river I#2MI"="IE-, saat004.009.099J %28F* ta)nnu)wa ta)nna)0wa i iro) I;0 i iro)L
+#9)biro

descendQIM1)APP;

lower the biro# I"<-M-, eamy042.4829J hen the applicati!e is applied to an intransiti!e !erb with a theme or patient ar$ument as in %28F*, the effect of addin$ )wA is !ery much like a causati!e. Some intransiti!e !erbs take iss and others )wA, but I ha!e not found any which allow a choice. "here is no ob!ious semantic difference between !erbs which form causati!es with ) wA and those which form causati!es with iss , compare %28F* with 'ungo:'ungoso SriseNraiseT.
&9&% Anticausative

"he anticausati!e suffi3 )wA is formally identical to the applicati!e suffi3 'ust discussed. In terms of !alence, howe!er, they ha!e the opposite effect, and they can co) occur in the same !erbal word %as in %494* abo!e*. "herefore they are considered to be 22;

homonyms. "he function of the anticausati!e is to downplay the role of the a$entNcauser in the e!ent denoted by the !erb, so much so that it cannot be e3pressed at all, e!en in a prepositional phrase. "his inability to refer to the a$ent is why the suffi3 has been $lossed as anticausati!e rather than passi!e %.i3on and =ikhen!ald 2000K7*. +ot all !erbs can take the anticausati!e suffi3 , the ones that do seem to encode e!ents in which the patient is chan$ed in some wayK %285* %a* u)0si 8u)
;S)heatQ/6S

mo0ni

+#8)water

> @ mo)0ni

+#8)water =G8)heatQ/6S)A/(IC

ma)0si 8u)0wa)

he heated water

the water got heated I2009)02)44.044J

"he anticausati!e may combine with the causati!e iss. "he combination of a !alence) decreasin$ de!ice with a !alence)increasin$ one does not result in any net chan$e in !alence, but the doubly)deri!ed !erb differs sli$htly in its meanin$. Mu)0lenji wu)0si 8u) mo0ni Sthe sun heated the waterT sounds stran$e because it implies a$ency on the part of the sun. <owe!er u)0lenji wu)0si 8i )ssu)0wa) mo0ni Sthe sun caused the water to $et heatedT is fine, because it has no such implication. "his fits in with the established cross) lin$uistic pattern that analytical causati!es tend to imply less a$enti!ityN!olition on the part of the causer than le3ical causati!es %1ayne 4997K492*.

&9(
&9(1

2ther verbal suffixes


1entive

"he !enti!e suffi3 )nA indicates that the action encoded takes place in the direction of the speaker or some other deictic centre, for e3ample the current centre of attention in a narrati!e. It is best to !iew this as a homonym of the perfecti!e aspect suffi3 discussed in L8.5.;.4 %and indeed of the plural imperati!e suffi3, L8.5.2.;*. "he effect of the suffi3 is similar to the <ausa $rade 5 !enti!e endin$ co %Ja$$ar 2004K2F5*, and they correspond in translations $i!en by nati!e speakers. "he suffi3 is !ery producti!eK common e3amples include wuuto0no Scome outT !s. wuuto S$o outT, ka a0na Sbrin$T !s. ka a StakeT, and kondo0no Scome inT !s. kondo S$o inT, but one can also 'ust SdoT %yaa0na* somethin$ in a particular direction. hen talkin$ about the mo!ement of physical ob'ects in the speech situation %rather than, say, a narrati!e with a displaced SstoryworldT*, the deictic centre is usually the position of one of the speech participants. <owe!er for certain actions a different 228

deictic centre can become con!entionalised %e.$. the $round for fallin$ ob'ects , when droppin$ a pen from below head hei$ht, 0nA is consistently used % i iro) vi 0 ) yuwo)0no) Sthe pen fellT*, e!en thou$h the mo!ement is away from the interlocutors*. "he action encoded by !erbs with the !enti!e suffi3 does not actually ha!e to in!ol!e a direction. Instead the action may benefit the speaker %or person at the deictic centre* in some wayK %287* ni )
and 2S)$oQ/6S

;0yo*o

n0ka$i ,

+#F)huntin$

see

until

u0hu)na)0na)

;S)killQI//),7/(

i )0nama)

+#;)meat

when you .s.2 go hunting, then it kills meat Ifor youJ Itats009.008.028J "he perfecti!e and !enti!e )nA suffi3es may occur to$ether, in which case they are usually realised by )nnA, as in %289*. "he !erb yaa Sarri!eT is e3ceptional in that yaa0na0na is found rather than yaa0nna. %289* u)8anu)wa)na)mu) u)08ana)0wa)0na)1mu) ana) ana) ka)taka)dda : ka)0taka)dda
+#4)book

;S)writeQ/6S)=116)E-+"R4S.1/2

when ;S)writeQ/6S)=116)P>,1,7/(R4S.1/2

u)8anu)wa)nna)mu) u)08ana)0wa)0na)0na)1mu)

ka)taka)dda... ka)0taka)dda
+#4)book

she wrote me a letter 1 when she had written me a letter... I2009)02)42.00;J 0or the !erb dn SfollowT the !enti!e form is dnn rather than the e3pected dnn. "he perfecti!e form is dn0n. hen both the !enti!e and perfecti!e suffi3es are applied the result is dn0n0n, which su$$ests %perhaps surprisin$ly* that the perfecti!e affi3 occurs closer to the root than the !enti!e.
&9(# !esultative

If the encoded e!ent reaches some sort of natural endpoint then the resultati!e deri!ational suffi3 )nu may be used. #onsider the followin$ pairsK %289* k8 golo s vayu yuwo cut cut drink drop all k8nu golonu snu vayanu yuwonu cut o cut up drink up drop in Ii.e. into a containerJ all in

"he use of the !erbs k8, golo, and s does not imply that the action bein$ encoded 22F

was completed, or that the ob'ect referent was wholly cut or consumed. "he deri!ed !erbs k8nu, golonu, and snu do howe!er lead to this interpretation. "he !erbs vayanu and yuwonu are appropriate when the theme ends up enclosed in some way, for e3ample in a pit or a pool. )nu occurs obli$atorily with many !erbs such as $i8onu SburyT hintonu StieT, jungonu SshutT50, sttnu Sur$eT, yaddanu SabandonT %from <ausa yad da Sthrow awayT*. Most of these !erbs are SaccomplishmentsT %Eendler 49F7* with an inherent endpointG one cannot partially tie a knot, or abandon a child. "he affi3 often ser!es to indicate an action carried out with more force than normal, as in 'p'pnu Shold$rabT and ko8oko8onu Staptap forcefullyT. /elated to this, the suffi3 may also be used to indicate a certain ur$ency , when playin$ the board $ame dara players ha!e been obser!ed to call du)wa)nu_ when they are losin$ patience, rather than the usual du)wa_ Splace?T. Sometimes the effect of addin$ the affi3 is e!en more unpredictable as in naha Slea!eT !s. nahanu Slea!e s.o. an inheritanceT, dama SspeakT !s. damanu SwhisperT, and gama S'oinT !s. gamanu Smi3T. "he ar$ument structure possibilities of the !erb may also chan$e , in %2F0a* with konto Sbump intoT, the wall can only be e3pressed as the ob'ect of the !erb, whereas in %2F0b* with kontonu ScollideT it is also possible to use the locati!e A. %2F0* %a* ;0konto)
2s)bumpPintoQ/6S

ko)0pu*u

+#4)wall

you bumped into the wall %b* ;0konto)0nu)


2s)bumpPintoQ/6S)#70

%oR* ko0pu*u

%;6CR* +#4)wall

you bumped into the wall Ieaim005.4889J


&9(% 0eparative

= few !erbs take the separati!e suffi3 0wA %a$ain, presumed to be homonymous with the !alence)chan$in$ suffi3es discussed in L8.5.8*. "he only e3amples attested are halahaluwa ScoiluncoilT, 'um onu'um uwo ScloseopenT, $i8onu$i8uwo Sburyunco!er earthT, and kuntonukuntuwo Sclenchunclench fistT. In each case the !erb with the 0wA suffi3 in!ol!es separation as part of its meanin$. =s ar$ued by
50 Interestin$ly jungo means SopenT, but addin$ 0nu is not a producti!e means of opposite formation.

225

Schadeber$ %200;K79* for (antu, Sseparati!eT is a better $loss than re!ersi!e, since it predicts which member of the pair will occur with the suffi3.
&9(& Pluractional

In #icipu !erbs with the hi$hly producti!e pluractional54 infi3 ils can encode se!eral different kinds of situations, dependin$ on the source of the SpluralT component of meanin$. "his can be supplied by multiple actors, multiple patients, or multiple e!ents. "he !arious possible sources of pluractionality and different interpretations are summarised in "able 29K 0able 4`- dources o pluractional marking 8ulti"le agents 8ulti"le "atients Inter"retation Iterati!e S.isinte$rati!eT .istributi!e #ollecti!e .istributi!e /eciprocal 7$am"le %2F4* %2F2)2F;* %2F8* %2FF* %2F5* %2F7*

If there is only one a$ent and %at most* one patient, then pluractional markin$ can $i!e rise to two interpretations. "he more strai$htforward is the Siterati!eT readin$ shown in %2F4*, where the e!ent is carried out repeatedly. "he second possibility is what I ha!e called the Sdisinte$rati!eT readin$, in which the theme becomes di!ided into se!eral parts as a result of the e!ent. "his is illustrated by %2F2)2F;*. "he pluractional is also used when a sin$le a$ent acts on multiple patients, as in %2F8*. #on!ersely, it is also used when multiple a$ents act on a sin$le patient %as in the collecti!e readin$ shown in 2FF*. 0inally, the pluractional can encode multiple a$ents actin$ on multiple patients, as in the distributi!e readin$ in %2F5*. If the a$entsNpatients are actin$ on each other %e.$. the reciprocal e!ent encoded by 2F7* then pluractional markin$ can be used e!en thou$h there are only two referents in total. %2F4* wu0u0de)mpQi )lRe)
;S)0A")punchP;ACsQI//

he would pummel IhimJ Isaat002.002.F88J

54 1luractional !erbs are found in <ausa as well as (enue)#on$o. "he term was su$$ested by +ewman %e.$. +ewman 4990*.

227

%2F2* mo)0ni

+#8)water

m0aya)

=G8)comeQ/6S

m)0k8Qi )lR)0nu)

=G8)cutP;ACsQ/6S)/-S

then the water parted Ilit ScutTJ Isahs004.00;.04FJ %2F;* ka)0ji ri gi k0aya)
+#4)!essel =G4)comeQ/6S

ka)05atQi )lRa)

=G4)tearP;ACsQ/6S

then the aeroplane disintegrated I"idipo, saat002.00F.02FJ %2F8* n)0sQ)lR)


4S)drinkP;ACsQ/6S

mo0ni

+#8)water

n)

and

y0i

=G;)1/2

po*
all

' drank water rom each o them Ie.$. cupsJ I2009)02)24.008J %2FF* i )07i ya
+#;)bee

yi 0 $i tQi )lRa

=G;)stin$P;ACsQ/6S

$i )0luu

+#5)leopard

the bees stung the leopard Isaat004.005.408J %2F5* a)07a


+#2)person

ha0na)

=G2)/-6

a)0matQi )lRa)0na)18 o)

;1)fatherP;ACsQ/6S)10ER21.1/2

the people who athered you .pl.2 Isayb004.797J %2F7* see ni )1i0tasQi )lRa)0nu)
until andR21)meetP;ACsQ/6S)/-S

unless you .pl.2 meet Ii.e. meet each otherJ Itats008.004.0;FJ "here is at least one !erb whose pluractional form takes on an unpredictable meanin$K sa'asa'ila Swash hands or facewash le$sT. In some words, the infi3 seems to ha!e fossilised, and they can no lon$er occur without it e.$. havila SscratchT, pi8olo Sroll around in the dirtT, titila Sclear awayT and 8angula S$atherT. "he iterati!e component of meanin$ can clearly be seen. 0or monosyllabic roots the !owel of the pluractional infi3 is absorbed by the first !owel of the root %e.$. %2F8* abo!e, also taataala SshootT and daadaala SstretchT*. If the last consonant in the root is l then the infi3 !owel may disappear lea!in$ a double ll %cf. dndnn in L8.5.F.4*. -3amples are lapilalapilla SprepareT and daladalla Sdip food in soupT. In these cases the shortened double)l form seems to be more common than the full form.

229

&99

/orrowed verbs

Eerbs borrowed from <ausa are often con!eniently indicated by the suffi3 0kwA. %2F9* gwaanukwa understand daamukwa ha8ukwa laatukwa ma$ukwa ta5ukwa taarukwa yaahukwa koyuko dennukwe worry %oin perish close in on do s.t. once meet orgive teach compress from gane from dama from haa from lalace from matsa from taa from tare from ya e from koya from danna

Some borrowed !erbs are ne!er heard with this suffi3. 2ften this is the case for what seem to be well)established borrowin$s %e.$. gwede SthankT from gode, iya SpayT from biya*. In other cases the lack of 0kwA may be a result of code)switchin$ rather than borrowin$G speakers ha!e been obser!ed correctin$ the bare <ausa form to SproperT #icipu by addin$ the suffi3. "he historical deri!ation of this suffi3 is mysterious. It has no other synchronic function, and there is no ob!ious source candidate. Moreo!er the trisyllabic !erbs with 0kwA actually look less like nati!e #icipu !erbs than the bare stems.

&9<

Auxiliar- and aspectual auxiliar- verbs

= number of phrasal constructions in!ol!e semi)$rammaticalised au3iliary or Duasi) au3iliary !erbs. Some can be used as the only !erb in a clause, while others are limited to the constructions discussed here. "hese au3iliaries inflect for sub'ect and sometimes other !erbal cate$ories but their semantic contribution is ad!erbial, in that they do not e3press the main conceptual e!ent or state of the clause. "hree kinds can be identified, accordin$ to the way in which the %semantically* main !erb is marked. 0irst, followin$ some au3iliaries the !erb has no prefi3 at all, as with 'esu SaddT or Sdo s.t. a$ainT and 8an Sdo s.t. a littleT52, illustrated below. .esi Sdo s.t. tooT is also of this kind.

52 |an is a <ausa borrowin$G in fact, the whole construction is borrowed cf. Ja$$ar %2004K4F2*.

229

%2F9* mo)0ni

+#8)water

me)'esu)

=G8)againQ/6S

gi ta)0na)

$oPback)E-+"

the water came back again Isamoh004.429J %250* u)08an


;S)littleQ/6S push

to)su) Ieaim005.48F7#J

he pushed a little 2ther au3iliaries are followed by the


+#7

infiniti!e %LF.8.4*. -3amples are $i!en below

for 'p SholdT which can also mean Sdo s.t. repeatedlyNe3cessi!elyT, hwaara \startT with the aspectual au3iliary %(innick 4994K478* meanin$ Sstart doin$ s.t.T or Sdo s.t. firstT %<ausa ara*, and naha Slea!eT or SstopNrefrain from doin$ s.t.T. 2thers of the same kind include guya Sbe able toT, jungo SopenT or Sstart doin$ s.t.T, and the <ausa loans saa a Sbe used to doin$ s.t.T %saba* and danganu Skeep doin$ s.t.T .dinga2. %254* u)0'p)
;S)!oldM#;0

u)0pi 8a

+#7)lick

he repeatedly licked Isaat002.002.299J %252* u)0hwaara)


;S)startM#;0

u)0wuto)0no)

+#7)$oPout)E-+"

it started coming out Itats00F.002.09FJ %25;* a)0naha)


;1)leaveM#;0

u)0huna wi , 1i0nama)
+#7)kill =G7R+#;)meat

they stopped the killing o meat Isamoh004.480J "he third kind of au3iliary is followed by a main !erb with a sub'ect a$reement prefi3. "he construction is ne!ertheless distinct from a strai$htforward clause 'u3taposition, since only one e!ent is encoded. -3amples include a'a Sdo s.t. alreadyT %258* and aya ScomeT or SthenT %25F*. "he latter au3iliary is !ery common on the main e!ent line in narrati!es, especially when the situation encoded by the clause in!ol!es a temporal pro$ression with respect to that of the pre!ious clause. ) %258* ti )0 a'a
41)alreadyQ/6S

$e
+-G

ti 0 si ri )ya)

41)prepareQ/6S

we hadn't already prepared Isaat004.00F.0F9J

2;0

%25F* ka)0taari ka0mpa) k0aya)


+#4)stone =G4)this

=G4)comeQ/6S

ku0uto)

=G4)$oPoutQ/6S

then this stone went out Isayb004.22FJ

&9=

2rder of affixes

"he order of !erbal affi3es is strict, and no e3ceptions to the seDuence set out below has been found in the whole corpus, nor could any be elicited. %255*
=G/)0A"N<=()root16=#s#=AS

s)=+"I#)=116)10EN16.IM1)E-+"R2(J.1/2

"he future and habitual prefi3es do not seem to co)occur, and so they ha!e been placed in the same slot. +o e3amples ha!e been found containin$ both the perfecti!e suffi3 and the plural imperati!e suffi3, and it is probable that the two cate$ories are lo$ically incompatible. "he ob'ect clitics %L7.;* are bound to the ri$ht)hand ed$e of the !erbal word.

&9>

Co.occurrence

"he !arious affi3es and suprase$mental morphemes discussed in this section can be di!ided into three $roups, accordin$ to their beha!iour with respect to co)occurrence. "he first $roup is comprised of the tonal patterns associated with realis, irrealis, and imperati!e moods, to$ether with the habitual aspect marker si0. (ecause each of these cate$ories reDuires a particular tone pattern for their e3pression, it is lo$ically impossible for any two of them to be e3pressed at the same time. "he second cate$ory consists of the perfecti!e aspect marker 0nA and the future tense prefi3 u0. 0nA can co)occur with the realis mood tone pattern, but not seemin$ly with the others. u0, on the other hand, only occurs with the irrealis tone pattern. "hese two morphemes therefore cannot co)occur. "he third cate$ory consists of the !arious deri!ational affi3es , the resultati!e, separati!e, applicati!e, anticausati!e, and !enti!e suffi3es, to$ether with the pluractional and causati!e infi3es. -ach of these can occur with one from the first $roup to$ether with one from the second. Moreo!er it seems that in theory they could all be found in a sin$le !erbal word , recall e3ample %494* abo!e. In addition to their co)occurrence possibilities, the third cate$ory forms a natural class with respect to de!erbal nominalisations %LF.8*, which can only contain affi3es from the third cate$ory. "here is also a correlation between the abo!e three classes and 2;4

their semantics, which in turn correlates with the producti!ity of the affi3. Morphemes from the first and second cate$ories e3press inflectional "=M cate$ories and are ma3imally producti!e, whereas those from the third cate$ory e3press deri!ational cate$ories and are less producti!e. 6ittle information is a!ailable on the co)occurrence possibilities of the dependent imperfecti!e %L8.5.;.8* and the pro$ressi!e 0su %L8.5.;.F*, althou$h $i!en the former reDuires a specific tone pattern we can assi$n it to the first $roup, and the restricted producti!ity of the latter makes it a candidate for the third $roup.

*.3

Ad'ectives
+#5T*

"here are !ery few ad'ecti!al roots in #icipu, and many Sad'ecti!alT meanin$s are ) S$oodness, instead e3pressed usin$ nouns %e.$. ti 0 ) li pai or !erbs %e.$. gnj Sbe cur!edT*. =s mi$ht be e3pected, true ad'ecti!es in #icipu share properties of both nouns and !erbs. 6ike nouns they carry le3ical rather than $rammatical tone, and when they function as predicates they can be followed by the copula %e.$. %259* below, cf. L8.;.;.4*. 2n the other hand, like !erbs they do not ha!e le3ical $ender, and instead a$ree in $ender with their head noun. Some e3amples of ad'ecti!es are $i!en belowK 0able 4b- )*amples o ad%ectives * 'etei pene*u 8ene*u saavi na) kuno 'uyono) rumo)no) si lana) ine big small new old light dark red

=n ad'ecti!e can occur either as a modifier %257* or as the head of an +1 in the absence of a noun %259*, as with most other noun modifiers %L8.8.F.F*. %257* m)0'y'y ma)0si lana)
+#8)fish A-41red

red ish I"ikula, taff002.045J

2;2

* %259* a)0hyaa1mu)

;1)sayQ/6SR4S.1/2 4S)beatQI//

n0va)sa)

ti )08enene*ul_

A-+1smallN#75DP

they told me to beat the small one I$i )0kooto) Sdrum, +#5TJ# Isaat004.009.07;J hen functionin$ as the heads of +1s, ad'ecti!es can take part in predicate nominal constructions, as in %259*. "here is no formal distinction between attributi!e and predicati!e ad'ecti!es in #icipu. +ote that in addition to the le3ical tone pattern on ad'ecti!es, the %optional* presence of the copula after the complement +1 distin$uishes these constructions from !erbal clauses. %259* ka)0taka)dda ka0mpa)
+#4)book =G4)this

ko)0'uyo)no) k0e
=G4)lig!t

=G4)#21

this book is lightccoloured Ilit. Sthis book is a li$ht)coloured oneTJ Ieamy004.042J

&<1

!eduplication

"he reduplication of ad'ecti!es has different effects, dependin$ on the Duality in!ol!ed. "he meanin$s of some ad'ecti!es are intensified under reduplication, as in tei S!ery pene*uNpenene*u Sbi$N!ery bi$T, 8ene*uN8enene*u SsmallN!ery smallT and 'etei0'e fineT, whereas with colours the meanin$ is attenuated, as in si lana)0si lana) \reddish, pinkT. #icipu thus contrasts with <ausa, where ad'ecti!e reduplication always seems to $i!e rise to attenuation %cf. babbacbabba Sbi$$ishT , +ewman 2000K27)29, Ja$$ar 2004K488) 485*.

2;;

*.5 &=1
4 2 ; 8 F 5 7 9 9 40 44 42 20

/umerals Cardinal numbers


to*o yapu) ta*atu) nosi ) * tau 24 22 ;0 80 F0 50 70 90 90 400 200 4000 u)kuppa wu)yapu) n) vi )to) u)kuppa wu)yapu) n) vi )yapu) u)kuppa wu)ta*atu) u)kuppa wu)nosi ) * u)kuppa wu)tau

0able 6a- 5ardinal numbers in 5icipu

tori hi ) ti nda)ya) ku)ri llo) ku)ti tti u)kuppa u)kuppa n) vi )to) u)kuppa n) vi )yapu) u)kuppa wu)yapu)

u)kuppa wu)tori hi ) u)kuppa wu)ti nda)ya) u)kuppa wu)ku)ri llo) u)kuppa wu)ku)ti tti 8e)ri >T.@ } u)kuppa su)kuppa 8e)ri yapu) du u) >T.@

#icipu has a strai$htforward decimal numeral system with !ery few irre$ularities , it is possible to construct any number analytically usin$ only the words for 4)40. It is interestin$ to note that Mathews %4925b* $i!es hi as the #icipu for 4, with Hattu %707a t0to) Sone personT* in parentheses. hile I ne!er encountered this word in my fieldwork, est &ain'i lan$ua$es, particularly in the (asa) clear co$nates of hi e3ist in other

&amuku branch %(lench n.d. a*. "he form of tori 0hi ) Ssi3T su$$ests that to*o is a later replacement for hi5;, since the words for 5 and 7 are historically deri!ed from the phrases F and 4 and F and 2 in other for ut)Ma:in*, and more $enerally in est &ain'i lan$ua$es %e.$. Smith %2007K59*
=G9

est =frica % elmers 497;K298*. prefi3 ku0. I

Ku)ri llo) Sei$htT and ku)ti tti SnineT be$in with what looks like the Dueryin$ I was told this was not correct #icipu58.

ha!e obser!ed speakers omittin$ the ku) %e.$. m0 mo)0ri llo) Sei$ht piecesT*, but on

5; #ross)lin$uistically the word for SoneT is hardly e!er borrowed, and e!en when it is it co)e3ists with the ori$inal term %Soua$ 2007*. "o*o may be a lan$ua$e)internal inno!ation deri!ed from 707a t0to) SaloneT lit. Sone personT , cf. the eDui!alent attributi!e numeral construction in #entral &ambari, 707a t0t9), where the cardinal number is iyya %<offmann 495;K2;*. 58 In the -ast &ain'i lan$ua$e =mo the numerals 5 and abo!e take an in!ariant SconcordT prefi3 ku0 %=nderson 4990aK45;*, so that 9 and 9, for e3ample, are ku0li r and ku0ti ri respecti!ely %in their attributi!e forms*. It is not hard to see how such an in!ariant prefi3 could be reanalysed as part of the

2;8

6)0kuppa StenT can occur with or without the noun class prefi3. In $eneral, the #icipu numbers are bein$ replaced by the <ausa numbers, and youn$er speakers often ha!e difficulty with the numbers abo!e F. -!en adults may not know these numbers, and they often count in <ausa instead, e!en when the rest of the con!ersation is bein$ conducted in #icipu.

&=#

Attributive numerals

=ttributi!e %enumerati!e* numerals ha!e a lot in common with ad'ecti!es %L8.7*. "hey can occur as noun modifiers or as pronouns, they are le3ically)specified for tone, and they take the same low)tone $ender a$reement prefi3es. It may also be the case that they occur in the same syntactic slot in the +1 %see L8.8.F.5*. -3amples are $i!en belowK %270* ka)0manga ko)0to)
+#4)rope A-11one

one rope Isaat004.002.0F9J %274* w0t)


+#7)month

wu)0yapu)
A-71two

two months I2009)0;)24.002J See L5.2.2 for more details on $ender a$reement on numerals.

&=%

2rdinal numbers

If a numeral occurs in the SpossessorT slot in an associati!e construction %L8.8.F.4* then it is understood to be ordinal, 'ust as in (antu % elmers 497;K277* and many other +i$er)#on$o lan$ua$es %e.$. I$bo, elmers 497;K29;*. "here is also a clear parallel with <ausa, where the possessi!e linker occurs between the head noun and the numeral in ordinal constructions %+ewman 2000K;95, Ja$$ar 2004K;5;*. #ompare %272* with %274* abo!eK %272* w0t)
+#7)month

wu,1y0yapu)
A-7E/C)1two

the second month I2009)0;)24.002J "he ordinal forms differ from the enumerati!e forms in two ways , the tone on the prefi3 is hi$h, and the neutral
+#9

len$thenin$ allomorph -0 is inserted between the

number itself, and this may be what has happened in the #icipu forms for 9 and 9.

2;F

a$reement prefi3 and the numeric root %see L5.8.2*.

&=&
=G9

EAdverbialF numerals

"he number of times an e!ent has occurred can be e3pressed usin$ a numeral with an prefi3 but no head noun %see L5.; for further uses of =G9 a$reement morpholo$y*. %27;* n)0doonu)
4S)sitQ/6S

ku)0to)

A-*1one

' sat down once Ii.e. a sin$le timeJ Ieamy00;.4;F9J %278* saa u)0wayaa0ni
e!en
+#7)come)+M6H

ku)0yapu)
A-*1two

pa*a

here

even coming here twice Isayb004.F79J Ku0 may also occur on the wh)word ya)anu Show manyT %L8.;.5.2* in the absence of an a$reement controller, in which case the interpretation is usually Show many timesTK %27F* ku)0ya)anu kw0i )
=G9)who =G9)#21

see

) i )0si 0 yaa0wa) ko0kkwi i


+#4)corpse

ka)0 i ki )
+#4)festi!al

unless 21)<=()do)=116

how many times do you have to do a estival or a corpse I"ikula, sa$b004.84FJ

&=(

$istributive numerals

0inally, as is common in =frican lan$ua$es, the reduplication of a numeral e3presses distributi!e meanin$K %275* a)07a
+#2)person

yapu) yapu)
two two

together in twos Isamoh004.24;J

*.6

7uantifiers

-ach of the #icipu wh)Duestion words %L8.;.5.2* has a uni!ersal pro)formNDuantifier counterpart, formed by placin$ saa SwhetherNorT in front of it.

2;5

0able 6,- /niversal $uanti iers Iuestion word yi )ni %+#;* han:ha)nu 0e)ne -loss what where which Dniversal Juanti'ier -loss saa yi )ni saa ya)anu saa ha)nu saa =G30e)ne ST everything everyone everywhere every u

ya)ani :ya)anu %+#9* who

"he first three entries in the table are pro)forms and are therefore used in place of a le3ical +1. Saa Av0e)ne, on the other hand, only occurs before a noun. =s with plain )e)ne, the a$reement prefi3 in the saa %277* saa k0e)ne
or
AvM0e)ne

phrase may inflect for either $ender or person.

"he followin$ e3ample illustrates $ender a$reementK


=G4)w!ic!

ka0'i ngu)wa)
+#4)!illa$e

every village I"ikula, sami004.48FJ "his construction appears to be a calDue from the <ausa kocwa, kocme, etc... %+ewman 2000K522)525, Ja$$ar 2004K;70);7F*. 2ther Duantifiers include po* SallT, correspondin$ to duk in <ausa, the paucal ) Slittle, smallT, and the multal Duantifier gei %sometimes g0gei *. "he Duantifier $i mmai word ku)0pi lu) is also often used to e3press Duantity %cf. <ausa yawa* but since it can occur either as the sub'ect of a clause or as the head of a comple3 +1 tri$$erin$ a$reements %as in 279*, it is better analysed as an +#9 noun. %279* see
unless
=G9

* ma)0yaa

=G8)doQ/6S

ku)0pi lu)

/C*1many

o,1ko0tumo
62#R+#4)belly

until it Imo)0yoo Sbeer, stomachTJ %279* ku)0pi lu)

+#8TJ

ills the stomach Ilit. Sdoes a lot in the Itats002.00;.029J

+#9)many

* ku0mpa) ku0na) ti )0yaa0na)


A-*)this =G9)/-6

41)doQ/6S)10E here

pa*a

the many that we are here Ii.e. the lar$e numbers of people we areJ Isayb004.5;8J

2;7

Part III The noun class s-stem of Cicipu

Chapter ( The Cicipu noun class s-stem


ith respect to the main research Duestions of this thesis, the chapters so far ha!e been preliminary, e!en if they ha!e been detailed in places. #hapters 4 and 2 %1art I* introduced the #icipu lan$ua$e and set out the research conte3t to which this thesis relates. "he ne3t two chapters %1art II* consisted of a sketch phonolo$y and $rammar of #icipu, thus layin$ the foundations for the more focused study which be$ins here. "his chapter and the ne3t, which to$ether form 1art III, describe the #icipu noun class system. =s was discussed in L2.4, noun class or $ender systems can be approached from two an$les. 2n the one hand, the way in which the nouns of the lan$ua$e are di!ided up %Snoun classificationT* is of interest both to =fricanists and typolo$ists. "his topic forms the sub'ect matter for the current chapter. 2n the other hand, noun class systems, especially in (enue)#on$o, are also interestin$ because of the ubiDuitous a$reement in!ol!ed. =s we saw in L2.2 this has been the sub'ect of considerable theoretical interest in recent decades. #hapter 5 therefore describes the $ender a$reement system of #icipu, co!erin$ the e3pression of $ender on manifold a$reement tar$ets as well as !arious other topics of current theoretical rele!ance.

-.1

.vervie$

=s demonstrated in McGill %n.d.*, the #icipu noun class system is formally !ery similar to those found in the &ambari lan$ua$es, and so here I ha!e followed the %arbitrary* numberin$ system used by <offmann %495;* and #ro>ier %4998* for #entral &ambari. It should be noted that the class numbers bear no relation to the (leek)Meinhof system used for (antu lan$ua$es %e.$. elmers 497;K45F*. "he followin$ table lists the #icipu noun classes and correspondin$ prefi3es, to$ether with illustrati!e e3amples of both #) and E)initial noun stems %the !arious pairin$s i.e. $enders are $i!en in LF.2*. /ecall that - and A represent consonant and !owel wei$ht units respecti!ely, the phonetic !alues of which are determined by the stem to which they are attached. / stands for a nasal homor$anic with the followin$ consonant.

2;9

0able 64- 5icipu noun classes Class /oun "re'i$ 4 kA0 Agreement "re'i$ kA0 7$am"le ka)0 ara k)0k ko)0joo ke)0tere k05i a)0 ara )0k o)0joo e)0tere 05i i )0nama) i )08aa i )0d* y0m) ri )0pi 7i ) ) ri )0hya'a ru0usi ) ma)0di ya m)0t mo)0kooto me)0pese ma)0nnu n)0di ya n)0t n)0kooto m0pese mi 0 ) nnu ti 0 ) si'i ) tu)0mo$i ) $i 0 ) luu $u)0kulu $i 0 iya u)0pepi u)0km u)0lenji w0v ;0$i )$$ere) ;0hi i'o) $0$') d0d* 707a v0m) ku)0$i ga) ku)0$i no ku)0mi na kwe0etu elder egg liHard bone hecgoat elders eggs liHards bones hecgoats meat ground horses monkeys altar arrow rainy season hare chick kitchen hut twin bird hares chicks kitchen huts twins birds hair riendship leopard tortoise lour wind salt sun ear star insect, k.o. sheep horse person monkey cockerel back cloth medicine

A0

A0

;a

i0:y0

i0:y0

;b 8

ri0 mA0 mA0

/0, mi0

/0, mi0

ti0, tu0, $i0, $u0 ti0, tu0

u0:w0

u0:w0

;0, -0, v0

;0, -0, v0

ku0:kw0

ku0:kw0

It can be seen from the table that the number of noun classes depends on the criteria we use to distin$uish them. If the prefi3 borne by the noun is the decidin$ factor, then there 280

are ten %morpholo$ical* noun classes. <owe!er two of these classes %;a and ;b* share the same a$reement markers, and so it can also be ar$ued that there are nine %syntactic* noun classes. In (enue)#on$o studies syntactic noun classes which share the same prefi3 are $enerally assi$ned the same number, but distin$uished by a followin$ letter %e.$. ;a, ;b*. "his is the approach that has been followed with class ; here. It will be clear to anyone familiar with the (enue)#on$o or (antu literature that, superficially at least, the #icipu system is !ery different to both the su$$ested 1roto) (enue)#on$o %1(#* reconstructions %e.$. .e olf 4974* and the present)day (antu systems. "here are fewer classes, and the forms of the ori$inal 1(# prefi3es ha!e in some cases chan$ed beyond reco$nition %see McGill n.d. for a comparison*. +e!ertheless, there are also strikin$ similarities, in particular the robust and ubiDuitous alliterati!e a$reement to be discussed in chapter 5, which was no doubt the moti!ation for Johnston %4949* to include #icipu:s est &ain'i relati!es %Gurmana, &amuku, and (asa* in his Semi)(antu. Much the same could be said about the other &ain'i lan$ua$es for which we ha!e data4 , the prefi3es and class pairin$s are much chan$ed from 1(#, but the mechanics of the a$reement system ha!e been retained. "he rest of the chapter is or$anised as followsK LF.2 lists the noun class pairin$s %$enders* which occur in #icipu and characterises their semantic structure %as far as is possible*. "he ne3t two sections look at the deri!ational function of noun class prefi3es , when applied to both noun stems %LF.;* and !erb stems %LF.8*. Section F.F describes the prefi3 allomorphs for the !arious classes in more detail, and finally LF.5 in!esti$ates how loanwords fit into the system. "he basics of this chapter were presented in a somewhat different form in McGill %2007*. +one of the statistics ha!e been chan$ed, and the illustrati!e e3amples are often the same, but it has been possible to si$nificantly e3pand and impro!e the analysis.

-.

8enders

"he basic structure of the #icipu $ender system is most easily demonstrated throu$h an affi3 net, as shown in 0i$ures ;0 and ;4 below. =s described in L2.4 the lines represent double class $enders %i.e. those containin$ paired sin$ular and plural nouns*, with solid lines bein$ used for well)established $enders and broken lines for inDuorate $enders
4 0or est &ain'i, see <offmann %495;* on #entral &ambari and <offmann %4957* on #:6ela, as well as the unpublished papers listed in McGill %n.d.*. 0or -ast &ain'i, see =nderson %4990a* on =mo.

284

with only a few members %#orbett 4994K470)47F*. "he underlined prefi3es represent sin$le class $enders. 0i$ure ;0 shows the noun class prefi3es while 0i$ure ;4 shows the correspondin$ agreement prefi3es , the lines are identical in both dia$rams, with the only difference bein$ the treatment of class ; , recall from LF.4 that nouns from classes ;a and ;b tri$$er the same a$reement prefi3 yi0, althou$h they take part in different pairin$s, with ;a occurrin$ with plural nouns, and ;b with sin$ular nouns. "he Salliterati!eT nature of the a$reement is clear from the dia$rams, with the two sets of prefi3es either identical or !ery similar. /e$ardin$ the pairin$s of noun class %i.e. $enders* it can be seen that #icipu has a ScrossedT system %<eine 4992K497*K

#lass 4, ;b, 8, 5, 7 and 9 prefi3es occur with sin$ular nouns #lass 2, ;a and F prefi3es occur with plural nouns #lass 9 prefi3es can occur with either sin$ular or plural nouns

282

singular

"lural = %2* i %;a* +Nmi %F*

k= %4* ri %;b* m= %8* tiNci %5* u %7* #NN! %9* ku %9* +igure 6a- 5icipu genders according to noun pre i*es

#NN! %9*

singular

"lural h= %2* yi %;* miN+ %F*

k= %4* yi %;* m= %8* ti %5* wu %7* #NN!i %9* ku %9* +igure 6,- 5icipu genders according to agreement pre i*es

#NN!i %9*

28;

In total there are ele!en double class $enders, se!en of which are fairly well established %se!enteen or more items*. "he other four are inDuorate. =s well as the double class $enders, classes 4, ;a, ;b, and 8)9 make up nine sin$le class $enders2. 2ne noun a)0hula SnameT takes a class 2 prefi3 and tri$$ers class 2 a$reement, re$ardless of whether it is bein$ used with sin$ular or plural meanin$. "herefore it should be considered as belon$in$ to an inDuorate sin$le class $ender. "able ;; below pro!ides $i!es a rou$h idea of the distribution of #icipu nouns within the le3icon accordin$ to their $ender. "he fi$ures should be treated with caution, for two main reasons. "he database of nouns used in this study %see L4.8* was not compiled usin$ a statistical samplin$ techniDue, and so some $enders may be under) or o!er)represented with respect to the le3icon as a whole;G in particular a hi$h proportion of nouns in $ender 7N9 ha!e only been obser!ed durin$ elicitation sessions on tree names. =lso the deri!ational function of noun class prefi3es %LF.;)F.8* means that it is often difficult to 'ud$e whether the properties of a particular prefi3)stem combination are sufficient idiosyncratic to warrant treatin$ it as a separate le3ical entry. +e!ertheless, the table is included as some may find e!en these rou$h indicators useful. 1roper nouns and identifiable loanwords ha!e not been included, e!en thou$h they make up a si$nificant proportion of some $enders %especially sin$le class $enders 4 and 9*. "he se!en paired $enders which were represented with solid lines abo!e are in bold. 0able 66- eistribution o 5icipu genders .singulars down, plurals across2 #lass 4 kA0 2 A0 ;a i0 ;b ri0 F /0 5 $i0 7 u0 9 -0 9 ku0 17 +4 7% 4 2 17 %& 8 1&3 8 mA0 ; 2 A0 7 ;a i0 F /0 9 -0 Sin$le 9 4 49 7 47 F 89 49 44 4F %"otalK 74;*

2 =ccordin$ to standard (antuist terminolo$y. "here are only eight sin$le class Star$et $endersT accordin$ to #orbett:s approach, since (antuist $enders ;a and ;b would be conflated into the sin$le tar$et $ender ; %see L2.4.2.;*. "he number of double class $enders is the same under either analysis. ; 2f course, the Sle3icon as a wholeT is itself a !ery slippery notion.

288

"he remainder of this section pro!ides e3amples from each of the $enders, alon$ with some discussion of their semantic structure. In fact, the e3tent of the o!erall semantic cohesion of =frican noun class systems has been a matter of lon$)standin$ disa$reement8. <offmann set out the two e3treme !iews with respect to semantic cohesion in his paper on the #entral &ambari noun class system %495;K459*K %a* %b* +oun classes are only morpholo$ical cate$ories and are !oid of any meanin$ whatsoe!er. +oun classes ha!e an inherent meanin$ and nouns are put into the appropriate classes accordin$ to their meanin$.

ith re$ard to 1roto)(antu, /ichardson %4957K;79* was firmly in fa!our of the first positionK it is impossible to pro!e conclusi!ely by any reputable methodolo$y that nominal classification in 1roto)(antu was indeed widely based on conceptual implication. In the absence of any such proof one mi$ht eDually well assume that the assi$nment of nominals to classes was for the most part an arbitrary $rammatical de!ice. Similarly, concernin$ the much earlier 1roto)+i$er)#on$o complete in conser!ati!e +i$er)#on$o systems, while the noun classification system arose on a semantic basis in pre)+i$er)#on$o, it had already become a $rammaticalised, essentially formal system in proto) +i$er)#on$o %4999K;2*. =d!ocates of the second position include .enny and #reider %4995*, who attempted to find a sin$le abstract meanin$ behind each 1roto)(antu noun class. SubseDuently, others ha!e applied techniDues from #o$niti!e 6in$uisticsF to a number of +i$er)#on$o lan$ua$es %e.$. <endrikse 4997, Sel!ik 4997, Sa$na 2009*. "his approach seems to offer a more robust methodolo$y than .enny and #reiderTs earlier work, but it is still open to the char$e that the analyses are not falsifiable , a different lin$uist may well come up with a different analysis usin$ the same techniDues. #ontini)Mora!a %2002* applies a similar semantic network approach to &iswahili, but her analysis also takes into account the use of class markers and concords in reference)trackin$, the functional moti!ation most often proposed for noun
8 See .in$emanse %2005* for a recent re!iew. F 0ollowin$ 6akoff:s %4997* seminal analysis of the noun class system of the =ustralian lan$ua$e .yirbal.

illiamson %4999*

concluded that since the correlation between formal classes and meanin$s is ne!er

28F

classification %e.$. <eath 499;, #orbett 4994K;20);2;*. "he cate$orisation of nouns into $enders permits a parallel classification of relational elements that make reference to nouns %ad'ecti!es, demonstrati!es, numerals, !erbs etc*, and thus restricts the set of possible referents when relational elements appear in discourse. #ontini)Mora!a notes that this pra$matic function could be fulfilled whether or not there was any semantic basis to the classification5, and she su$$ests that speakers construct noun class semantic networks as a lan$ua$e)learnin$ strate$y to ease the burden of memorisation. So it is not necessary to belie!e that allspeakers construct identical networks %2002K;7*, in part because different speakers are e3posed to different !ocabulary items in different conte3ts of use %2002K9*. "he distribution of !arious nouns across different $enders, on the other hand, is shared by all speakers. "here seems little doubt that these recent approaches can be fruitful when the in!esti$ator has an in)depth knowled$e of the lan$ua$e concerned %e.$. Sa$na 2009*7. <owe!er I approached this study as a cultural outsider with no pre!ious knowled$e of #icipu, and based on only nine months of fieldwork I can claim no special insi$ht in this matter. I could draw out some #o$niti!e Grammar Ssemantic networksT, but $i!en my present limited knowled$e of #icipu semantics, such a formal treatment would merely be ScookbookT in nature, addin$ little or nothin$ of theoretical interest to e3istin$ studies. ith that in mind, "able ;8 pro!ides a summary of some of the semantic re$ularities to be found. 0or con!enience the deri!ational effects of the !arious $enders ha!e also been listed %see LF.;)F.8 for details*. Genders 4N2 and 9N2 are miscellaneous in nature, and most of the sin$le class $enders are too small for any percei!ed pattern to be worth mentionin$. "he $ender with the most ob!ious semantic coherence is 9N2, which contains only nouns denotin$ humans or spiritual bein$s9.

5 In fact <eath %497F* ar$ues that it is actually ad!anta$eous for the system not to ha!e a semantic basis, since competin$ referents !ery often share semantic properties %e.$. human, animate*. 7 See also the discussion in &atamba %200;*. 9 "he con!erse is not true , nouns for humans are found in se!eral of the other $enders. Gender 9N2 is more fle3ible in #entral &ambari, where it is not e3clusi!ely human %#ro>ier 4998K70*.

285

0able 6J- demantic content o the larger 5icipu genders -ender 0emantic content 4N2 miscellaneous 5erivational e''ect au$mentati!e %LF.;.2*, town N settlement %LF.;.F*, indi!iduati!e %LF.8.F* diminuti!e %LF.;.4*

8NF 5NF 7N9 9N2 9N; 9N2 8 5 7 9

small thin$s, possibly round thin$s and hand)held thin$s unpaired body parts, possibly unusual animals and thin$s to do with farmin$ trees, possibly places and lon$ thin$s humans and spirits only birds and animals miscellaneous liDuid nouns

a$ent nominalisation %LF.8.8*

continuous action %LF.8.;*

non)liDuid mass nouns, collecti!e nouns institution N manner of %LF.;.;*, lan$ua$e %LF.;.F* abstracts miscellaneous locati!e and de)ad'ecti!al abstracts %LF.;.8*, infiniti!e %LF.8.4* repeated action %LF.8.2*

In the list that follows, double class $enders are treated first, then sin$le class $enders.

(#1
%4*

Gender 1C#
Gender 4N2K kA0NA0 ka)0 ungu snake ke)0tere bone ko)0$i lo) abdomen k)05 a*e a)0 ungu e)0tere o)0$i lo) )05 snakes bones abdomens a*es

Gender 4N2 has the lar$est membership by far, as is the case in the &ambari cluster $enerally. <istorically there has been a lot of mo!ement into this $ender from the others %.e olf 4974K52*, and any semantic coherence it may ha!e once had has disappeared. +e!ertheless, 4N2 does ha!e a deri!ational function as an au$mentati!e %LF.;.2* and an indi!iduati!e %LF.8.F*.

(##
%2*

Gender &C(
Gender 8NFK mA0N/0 ma)0kwa'a orphan ma)0lala puppy m)0'y'y ish ma)0si'ya nipple n)0kwa'a n)0lala n)0'y'y n)0si'ya 287 orphans puppies ish .pl.2 nipples

"he $ender has a clear semantic coherence. Most ob!iously, it contains a lar$e number of nouns relatin$ to small ob'ects. In particular words for small animals and birds, and words for youn$ animals and humans occur in this $ender. 2ther e3amples are me0 e)ri ise) Sswift Ii.e. the birdJT, me)0gemi SknuckleT, and ma)0di ya ShareT. <and)held tools and round thin$s may also cluster in this class. "he word ma)0gai SswordT denotes a relati!ely lar$e ob'ect, su$$estin$ that $ender 8NF e3tends to hand) held implements re$ardless of si>e. #ompare also ka)0gi si ) SstickT %4N2* with me)0gi si ) Swalkin$)stickT %8NF*. Similarly ma)0kapa Slar$e calabashT may be classified here because its denotatum is round rather than small. Some words for disabled humans % mo)0gutu) SleperNcrippleT, ma)0gwawa SmuteT* also belon$ to this class. "he borrowed words ma)0karantaa SschoolT, ma0llu* Smaster:, ma)0 u8i SkeyT, ma0ga)ji ) SpriestT, ma)07aani SpenT, m)07ri ScatT, mo)0guugi SbrushT and ma)0'ana Smeanin$T all be$in with mac in <ausa, and these were presumably assi$ned to this $ender because of their phonolo$ical resemblance to e3istin$ 8NF words. 2ther than loanwords, it is hard to find words that do not fit semantically. 3o)0kooto SkitchenT is a candidate, althou$h a kitchen hut is usually smaller than a normal hut. Similarly me)0ttegu) SshirtT %from <ausa taggo* may ha!e ori$inally referred to an item of clothin$ shorter than the <ausa riga. I do not ha!e an e3planation for ma)0si ta Slar$e wooden supportT. = small $roup of irre$ular nouns belon$ to $ender 8NF accordin$ to their a$reement properties, althou$h they are not strai$htforwardly composed of prefi3 plus rootK %;* mengeti )kka mengeta)ari ) m young girl young boy child mi nti )kka mi nta)ari ) muu young girls young boys children

"he word for SchildT and other kinship terms often seem to be irre$ular in (enue)#on$o %e.$. =nderson 4990bK80*. "his $ender also has a deri!ational function as a diminuti!e %LF.;.4*.

(#%
%8*

Gender 9C(
Gender 5NFK $i0N/0 $i )0y yam $i )0$ii* orehead $i )0reene) ireplace $i )0k'u) chest n)0y n)0$ii* n)0reene) n)0k'u) yams oreheads ireplaces chests

"his $ender is not found in &ambari, and is a relati!ely small one in #icipu. It contains 289

* SwaistT, $i 0llu OneckT*, unpaired body parts %$i 0 ) $ii* \foreheadT, $i )0k'u) \chestT, $u)0kuu certain miscellaneous %and perhaps unusual* animals %$u)0kulu StortoiseT, $u)0kundu ShyenaT, ti 0 ) 7a)aruma) SfleaT, $i )0yuu SporcupineT, $i )0luu SleopardT, ti 0 ) ji ma) Smale wartho$T*, a few words to do with farmin$ %$i 0 ) ni yaa SfarmT, $i 0 ) y SyamT, $i 0 ) yu'u Sbi$ hoeT* plus three othersK $i 0 ) yi m i ) SdarkT, $i )'0adi ) StrapT and $i )0reene) SfireplaceT.

(#&
%F*

Gender <C=
Gender 7N9K u0CP-0, ;0, vi0 u)0lenji sun1day u)0pepi wind u)0yaa road u)08anga) tree l0lenji p0pepi y0yaa 808anga) days winds roads trees

StownT, u)0yaa SroadT, u)0re8u) Gender 7N9 contains se!eral nouns denotin$ places %u)0rei SnestT and the word for SplaceT itself 'asu)~*, some abstracts %u)0jii Se3penseT, u)0ko*o ) Swei$htT*, and some SelementalT nouns %w0t) SmoonT, u)0lenji SsunT, SdeathT, u)0ne'i u)0pepi SwindT, u)0laa SfireT*. It also contains the word for StreeT u)08anga), alon$ with many words for different species of trees and se!eral other ob'ects that are lon$ in shape * SbowT*. "here are a few e3ceptions whose %e.$. u)0sii SpestleT, u)0hiyu SstreamT, u)0taa SskinT, class membership does not appear to ha!e any semantic moti!ation %u)0kwaa wu0nto) S$uest hutT and u)0km SsaltT*. In addition to the basic place words, the class 7 prefi3 u1 also has a locati!e deri!ational function %LF.;.8*.

(#(
%5*

Gender =C#
Gender 9N2K -0, ;0, vi0NA0 k0kaa woman 707a person w0wm chie va0ari ) man a)0kaa a)07a )0wm a0ari ) women people chie s men

Gender 9N2 has the most clearly)defined semantic structure of any of the $enders, since all members denote humans or spiritual bein$s. It is the default $ender for <ausa loanwords which denote humans %see LF.5*, and is also the $ender used to form a$ent de!erbal nominalisations %LF.8.8*.

9 "he word 'asu) is also interestin$ because it does not ha!e a noun prefi3. "herefore it does not display its $ender o!ertly, but only throu$h the a$reements it tri$$ers. "his is !ery rare in #icipu, and the only other nouns which could be said to ha!e co!ert $ender are a few semi)re$ular +#5 nouns %see LF.F.5*.

289

(#9
%7*

Gender =C%
Gender 9N;K -0, ;0, vi0Ni0 d0d* horse s0si ro mane g0gaa plait v0m) monkey i0 ) d* i0 ) si ro i0 ) gaa y0m) horses manes plaits monkeys

= hi$h proportion of the nouns in this $ender denote animals or birds %;2 out of the 7F non)borrowed words*. "his is also the default $ender for loanwords referrin$ to inanimate ob'ects e.$. i ndi ga) \$unT from <ausa bindiga %see also LF.5*.

(#<
%9*

Gender >C#
Gender 9N2K ku0NA0 ku)0ti vi ) ear ku)0 eyi hoe .weeding2 ku)0$i no back ku)0t hen a)0ti vi ) e)0 eyi o)0$i no )0t ears hoes .weeding2 backs hens

SdayNmornin$T Similar to 4N2, there is no ob!ious semantic unity to this $ender. Kwa0a'a J has a !ariable irre$ular plural, either aa'wa or 'waa'wa SdaysNmornin$sT %the IkUa('a *G note the labialisation within the stem in the plural40. It e3pected form would be aa'a tri$$ers re$ular 9N2 a$reement.

(#=
(#=1

0ingle class genders


Gender 1

%9*

Gender 4K kA0 ka)0ri )ma*i ko)0ri si )no*o ka)0'a)l arka) ka)0la)ahi ya)

pleasure 3orisino hill prosperity %from <ausa albarka* wellcbeing %from <ausa la iya*

"his $ender consists almost entirely of <ausa loans, especially ones denotin$ abstract concepts. It also contains a number of proper nouns for settlements %see LF.;.F*.
(#=# Gender %a

%40* Gender ;aK i0 i )0$i $i pu i )0si pa i )0saya)tu i )05o$i )

prayers song comb illness

40 See also the irre$ular 9N9 noun kwa'a) in LF.2.9.

2F0

(#=%

Gender %b

%44* Gender ;bK ri0 ru)0we'e) ru)0'u ru)0hungwa ru0usi )


(#=& Gender &

valley body resting rainy season

%42* Gender 8K mA0 mo)0hi'i blood mo)0yoo beer * ma)0huu truth mo)07ongo re%oicing * Ssali!aT and mo)0ni SwaterT, and Se!eral $ender 8 nouns denote liDuids such as mo)0too this mA1 prefi3 can be traced back to 1roto)+i$er)#on$o % illiamson 4999*. hile some of these liDuid nouns do pluralise %takin$ class F prefi3es*, the referent chan$es its nature considerably, as is the case in other lan$ua$es %cf. ' like beer !s. ' want two beers*44. I ha!e counted all such liDuid nouns in this $ender rather than 8NF, re$ardless of whether a countable plural could be elicited or not.
(#=( Gender (

%4;* Gender FK /0 n)0haalu n)0ka$i n)0lavu) mi 0isi ) n)0sanda

%ourney hunting sleepiness ululation between the legs

"his is a small $ender with only fi!e members and could be classified as inDuorate.
(#=9 Gender 9

%48* Gender 5K $i0 $i )0$i na) ti 0 ) hi 'o ti 0 ) yeyi $u)0ku)ya*a

charcoal straw dirt earwa*

Gender 5 contains non)liDuid mass nouns. It also contains abstract nouns deri!ed from !erbs or from other nouns %LF.;.;, LF.;.F*. =t least fi!e nouns are irre$ular, or possibly !owel)initial42 rootsK
44 #orbett %2000K94* refers to this process as Srecate$orisationT. 42 ithout a correspondin$ plural, it is not clear whether tii* and ta*a are !owel)initial or not. 0or e3ample we could analyse ta*a as ti a*, in which case it would no lon$er be irre$ular, but simply the result of the normal rules for noun prefi3es before !owel)initial roots. "he same holds for tii* . "his would be a

2F4

%4F* tii* ta*a si ngi * su)poo sum o


(#=< Gender <

aeces ood hair alsehood drink made rom red sorrel sepals .Hausa >obo2

%45* Gender 7K u0 u)0yeyu) u)0ki iso u)0kwari w0v

cold the herea ter ne*t year ear

Gender 7 contains mostly abstract nouns, which may ha!e moti!ated the deri!ational use of this $ender for formin$ de)ad'ecti!al abstract nouns %LF.;.8*.
(#== Gender =

%47* Gender 9K -0, ;0, vi0 l0lama sound $0$a]a harvest ;0kungwa God ) ;0ruhu last year Most $ender 9 nouns are <ausa loans %e.$. ;0 k)i SweekT from bakwai, ;0me)nkeni ScocoyamT from mankani*. "he remainin$ nouns do not seem to ha!e any semantic coherence.
(#=> Gender >

%49* Gender 9K ku0 ) ku)0hiivi ) * ku)0yuyu ku)0suu ku)0pi lu)

breath sand smell many

=s with the double class $ender 9N2, there is no apparent semantic coherence here.

(#>

In"uorate genders

"he followin$ e3amples illustrate the $enders with less than fi!e membersK %49* Gender ;N2K ri0NA0 ) ri 0 ) hya'a arrow ri 0 ) pi 7i ) altar ru)0p granary ) a)0hya'a e)0pi 7i ) )0p arrows altars granaries

more appealin$ analysis if there were candidates for !owel)initial monosyllabic roots in other noun classes as well.

2F2

%20* Gender 8N2K mA0NA0 m)0ri )gi )d navel mo)0ri ngi no spine %24* Gender 5N2K $i0NA0 $i 0 ) kooto) drum %22* Gender 9N9K ku0N-0, ;0, vi0 kw0a'a) house %2;* Gender 2K A0 a)0hula name1names

)0ri )gi )d o)0ri ngi no o)0kooto)

navels spines drums

a''wa*a or 'wa''wa*a houses

-."

9erivational use of noun class mar,ers

=s pointed out by Mufwene %4990*, the role of noun class prefi3es in (antu is more than simply indicatin$ $rammatical $ender and number. "hey also ha!e a Sderi!ationalT use, similar to suffi3es such as cment and cness in -n$lish. "he same is true for #icipu, as will be clear from this section and the ne3t. In this section I will consider the application of !arious class prefi3es to noun stems. "he resultin$ semantic effects ha!e all been obser!ed in other =frican lan$ua$esG <eine %4992K499* commented that =u$mentati!e, .iminuti!e, <uman, and 1lace are amon$ the most producti!e deri!ati!e $enders in Snature)basedT lan$ua$es. <eine also obser!ed that e!en within a $i!en lan$ua$e $enders tend to differ considerably with re$ard to deri!ati!e producti!ity %4992K499*. "his is also true for #icipu. Oe$tion F.8 is concerned with
.-E-/(=6

nominalisations, which are formed by the

application of a number of different class prefi3es to a verb stem.

(%1

$iminutive Dgender &C(G

+ouns belon$in$ to $ender 8NF usually denote small thin$s %LF.2.2*. In addition, nouns typically occurrin$ in some other $ender may be $i!en 8NF prefi3es if the speaker wants to su$$est a smaller ob'ect than the hearer would otherwise e3pect. 0or e3ample y0yene) ScatfishT %9N;* is a $eneric term co!erin$ catfish of all si>es, but me)0yene) %8NF* denotes a small catfish, and it would be inappropriate to use this word to refer to any other kind. "his process is !ery producti!e, allowin$ speakers to readily create no!el words. 2ften the diminuti!e prefi3 occurs as a pre)prefi3 %LF.;.5*. "he semantic relation in!ol!ed in this kind of deri!ation is sometimes non) compositional and unpredictable, as the followin$ e3amples showK

2F;

%28* m)0$$)k ma)0si'ya m)0t mo)07uvu)

pocket nipple chick pinkie

k)0$$)k %4N2* %4N2* ka)0si'ya ku)0t %9N2* ko)07uvu) %4N2*

bag breast hen inger

0or e3ample m)0t Schick, 8NFT means more than 'ust an undersi>ed ku)0t Shen, 9N2T. <owe!er the 8NF noun almost always has connotations of SsmallnessT in some way, perhaps only with respect to its counterpart. It should be stressed that there many noun stems which inherently Sbelon$T to $ender 8NF and do not occur with other prefi3es, and so $ender 8NF is not a minor tar$et $ender in the sense of #orbett %4994K450*. Indeed #icipu has no minor tar$et $enders at all , the !arious deri!ational functions discussed in this section and the ne3t are all accomplished by pressin$ into ser!ice the le3ical $enders already introduced in LF.2.

(%#

Augmentative Dgender 1C#G

.espite its o!erall lack of semantic coherence, $ender 4N2 has a deri!ational function as an au$mentati!e. "he resultin$ nouns show a similar !ariety with respect to semantic compositionality as the diminuti!e 'ust discussed. %2F* ke)0yene) ka)0tti lu ka)08anga) ka)0nnu large cat ish burial pot .large2 large piece o wood hawk y0yene) %9N;* ma)0tti lu %8NF* u)08anga) %7N9* ma)0nnu %8NF* cat ish pot tree or piece o wood bird

"his does not seem to be as producti!e as the diminuti!e, and I ha!e only obser!ed a few cases, usually in!ol!in$ a stem normally occurrin$ in $ender 8NF. =s with the diminuti!es, au$mentati!es may also be formed with pre)prefi3es %LF.;.5*G in fact for au$mentati!es this may be more usual than simply replacin$ the prefi3.

(%%

Abstract "ualit- Dgender 9G

"he abstract nouns found in this $ender tend to be deri!ed either from !erbs, or from other nouns denotin$ humans. "his is in contrast to abstract nouns from $ender 7 %LF.;.8*, which do not usually ha!e counterparts in other $enders. %25* $i )0'i ta)ni $i )0pati ti 0 ) 7') ti 0 ) wm tu)0m$i ) ti 0 ) yi ri ) ti 0 ga)ji ) marriage re$uest laughter chie taincy riendship witchcra t priesthood 'ita pata 7s w0wm m0m$i ) y0yi ri ) ma0ga)ji ) marry beg1plead laugh chie %9N2* riend %9N2* witch %9N2* priest %8NF*

2F8

2ften the deri!ed $ender 5 noun can be $i!en the $loss Sinstitution ofPPPT or SmannerNbeha!iour typical ofPPPT %cf. Mufwene 4990K2F0)2F4*. "his deri!ation is relati!ely producti!e and se!eral words I tried to coin were readily accepted e.$. ti 0 ) w Sthie!eryT from 0 w SthiefT %9N2*. =bstract <ausa loanwords may also enter this class e.$. ti 0 ) ukan$i Sthe practice of nati!e medicineT from bokanci, ti 0 ) turen$i S-n$lishT from turanci4;. =ll lan$ua$e names belon$ to $ender 5 %see LF.;.F*. "he association with SmannerT also comes out in some specialised uses of $ender 5 a$reement morpholo$y %see L5.;*. 0inally, there is also a deri!ational relationship between se!eral of the Scollecti!eT48 mass nouns in $ender 5 and their indi!iduated counterparts in $ender 4N2. 0or e3ample $i )0$i na) %5* is ScharcoalT, but ka)0$i na)Na)0$i na) %4N2* means SlumpNs of charcoalT. Similarly ti 0 ) hi 'o %5* is SstrawT but ko)0hi 'oNo)0hi 'o %4N2* means SstemNs of $rassT.

(%&

;ocative and de.ad,ectival abstract nouns Dgender <G

In addition to the basic place words, the class 7 prefi3 u1 may sometimes be prefi3ed to a noun root to indicate a location, as seen in the paradi$ms below. "his pattern does not appear to be !ery producti!e %althou$h see LF.;.F on ethnic $roups and their locations*. %27* ka)0taari a)0taari ma)0taari n)0taari t0taari i0 ) taari u)0taari k0ki iso) o)0ki iso) u)0ki iso) stone %4* stones %2* pebble %8* pebbles %F* bead %9* beads %;* stony place %7* spirit %9* spirits %2* the herea ter %7*

%29*

"he

+#7

prefi3 can also be attached to a motion !erb to form a nominal with a locati!e

meanin$ %LF.8.4*. Mufwene %4990K289)289* mentions the possibility in (antu of deri!in$ nouns by addin$ a noun class prefi3 to an ad'ecti!e. =lthou$h there are not many ad'ecti!es in #icipu %L8.7*, a deri!ational relationship e3ists between at least three of them and * Sbi$nessT pene*u Sbi$T, u)0ru)mo correspondin$ nouns in class 7. "hese are u)0pee
4; "he cci suffi3 in <ausa has a similar semantic effect as the $ender 5 nominalisation bein$ discussed. 48 See #orbett %2000K447)449* on collecti!e nouns.

2FF

SdarknessT ru)mono) SdarkT, and u)0si la SrednessT si lana) SredT.

(%(

)thnic groups

0or ethnic $roups, lan$ua$es, and settlements there is a paradi$m in!ol!in$ classes 4, 2, 5, 7, and 9, shown below. "he $aps in the table are unattested. 0able 6T- )thnic group paradigm #lass 9 2 5 4 7 =cipu $0$i pu) a)0$i pu) $i 0 ) $i pu) &arishen ;0ri si )no* o)0ri si )no* ti 0 ) ri si )no* ko)0ri si )no* &adonho d0di po o)0di po ti 0 ) di po ko)0di po Maburya k0kula) a)0kula) ti 0 ) kula) u)0kula) <ausa k0kg )0kg ti 0 ) kg Gloss person .sg.2 person .pl.2 language1dialect name o settlement name o settlement

&arishen, &adonho, and Maburya are the <ausa names of three of the #icipu dialect centres %L4.2.;*G the same paradi$m applies to the other dialects as well4F. +ote that ti 0 ) kg can mean either the <ausa lan$ua$e or reli$ion %i.e. Islam*, or perhaps 'ust Sin the <ausa mannerT , similar to the suffi3 cish in -n$lish. %497;K8F2* makes a similar point for the eDui!alent class in &iswahili, kic. elmers

(%9

Pre.prefixes

Mufwene %4990K2F;* distin$uishes between primary deri!ations, such as the ones that ha!e been illustrated so far in!ol!in$ a sin$le prefi3, and secondary deri!ations, which in!ol!e the addition of a Spre)prefi3T to an already delimited noun. In #icipu pre) prefi3es seem to be restricted to au$mentati!es %29* and diminuti!es %;0*. %29* ka0k0ka)0hi ili
+#4)M)+#4)field

ko0$i )0kooto)

+#4)+#5)drum

ke0ku)07e)'u)

+#4)+#9)dance

big ield Isamoh004.488J %;0* ma0k0ka)0na) ayi )


+#8)M)+#4)story

big drum Isaat004.009.07;J ma0k0ka)0si ngi


+#8)M)+#4)hair

a lot o dancing Is!tm$004.;22J me0'0i 0 ) ri 0mpa)0ni )


+#8)M)+#;)thin$)this)+M6H

small story Is!tm$004.429J

little hair Isaat002.002.450J

this little thing Isaat002.002.;4FJ

e ha!e now seen that diminuti!es and au$mentati!es can be formed either by replacin$ the normal prefi3 %LF.;.4)F.;.2* or by addin$ a pre)prefi345. "he non)
4F 0i!e of the se!en dialect centres are formed with the u)0kum a)si. 45 2r both, in fact, for e3ample ke0k0ke)0yeneyene) denotes a catfish lar$e enou$h to swallow a human. +ote the inner prefi3 with the standard low tone, and then the pre)prefi3 with characteristic hi$h tone
+#4

prefi3. "he e3ceptions are u)0kula) and

2F5

compositionality found in cases of prefi3 replacement has not been obser!ed with pre) prefi3es , for e3ample addin$ the +#4 prefi3 to u)08anga) StreeT leads to a strai$htforward au$mentati!e ka)0u)08anga) Slar$e treeT, but replacin$ the e3istin$ prefi3 to form ka)08anga) results in the related, but unpredictable meanin$ Slar$e piece of woodT. hen it comes to a$reement, the beha!iour of nouns with pre)prefi3es is perhaps a little surprisin$ , as discussed in L5.F, a$reement may be with either the inner or outer prefi3.

-.*

9everbal nominalisation

In the pre!ious section we looked at e3amples of the deri!ational effect of noun class prefi3es when applied to noun stems. In this section we consider their effect when applied to !erbs. "here are fi!e producti!e processes in!ol!in$ different prefi3es %LF.8.4)F.8.F*, but also a few less producti!e uses of prefi3es %LF.8.5*. Jud$in$ by #ro>ier:s %4998K442)447* discussion of #entral &ambari, in #icipu there seems to be $reater meanin$ associated with the indi!idual class prefi3es.

(&1

Infinitives and locatives Dgender <G

"he most producti!e de!erbal nominalisation is the infiniti!eN$erundNparticiple formed by prefi3in$ the class 7 prefi3 u) to a !erbal stem. It seems that any !erb can be so modified, and the process is perhaps more inflectional than deri!ational. "he tone on the resultin$ noun is either 6<6 or less commonly 6<< %i.e. with 6 on the prefi3*, and is apparently le3ically)determined but otherwise arbitrary. %;4* u)0dama) u)0hullo) u)0pi 8a speaking blowing licking dama hullo pi8a speak blow lick

"hese words ha!e properties of both !erbs and nouns, which of course 'ustifies the label SparticipleT. In %;2* the participle u)0dv Spoundin$T functions as the head noun of an associati!e construction, and also tri$$ers a$reement on the relati!iser wu0na). %;2* u)0dv
/C71"ound A-7E+#;)$Pcorn

wuRi 0da'u)

wu0na) ku)0mo$i )
A-7)/-6

+#9)oldPwoman

ku)0yo0no)

=G9)beQ/6S)10E

a1u0yaa
62#R+#7)do

the pounding o guineacorn that the old woman is doing

I2009)0;);0.004J

2n the other hand, in %;;* the same participle occurs with a pluractional !erbal
and consonant)len$thenin$.

2F7

e3tension ils and a bare ob'ect +1 i )0da'u) S$uineacornT fillin$ the

1="I-+"

semantic role

of the nominalised !erb dv S$rindT. =s well as the pluractional, the applicati!e, resultati!e, and !enti!e suffi3es ha!e also all been obser!ed occurrin$ in participles. %;;* o)0yo*
;1)beQ/6S

a1u0dvi ls

62#R+#7)pound16=#s

i0 ) da'u)

+#;)$uineacorn

they are repeatedly pounding guineacorn

I2009)0;);0.004J

0urther e!idence of the participle:s !erbal nature is shown by %;8*, where it is modified by an ad!erb. %;8* ta0a a)
+#5)tobacco

ti 0 ) si 0 ) yuu

=G5)<=()put

u0ku)ngwa)

/C71age<verb=

ma7ama7a
ra"idly

tobacco causes rapid ageing

Itats007.002.094J

"hese participles in #icipu ha!e a restricted distribution compared to prototypical !erb forms , they cannot head main clauses, nor do they occur as dependent Ssentence ad!erbialT clauses. Instead they can only occur as the complements of main !erbs. 1articular subordinatin$ !erbs %L8.5.7* seem to consistently select either this form of the !erb, or a finite !erb form i.e. there is no choice. "he sub'ect +1 must be omitted, and the %lo$ical* sub'ect of the participle must be the same as that of the main clause. 0or these reasons it seems appropriate to refer to this (antu class 4F kuc infiniti!e , e.$. Eisser 4999*. Since the word forms under consideration show properties of both !erbs and nouns, it is worth askin$ whether they form an independent le3ical cate$ory. #oncernin$ the $erund in -n$lish, &roe$er %2008K8F)F0* ar$ues a$ainst a Smi3edT cate$ory analysis. Instead the $erund is ambi$uous, with indi!idual uses bein$ classifiable as either clearly nominal %i.e. a de!erbal nominalisation* or clearly !erbal %i.e. a true $erund*. "his does not seem to be the case for the #icipu participle. -3amples %;F);5* were readily accepted in elicitation sessions. In %;F* the participle takes an ob'ect like a !erb, but also functions as the head of a relati!e clause like a noun. %;F* u)0dv
/C71"ound /C31guineacorn

+#7

participle as an infiniti!e %cf. the

i0 ) da'u)

wu0na) ku)0mo$i )
A-71#7;

+#9)oldPwoman

ku)0yo0no)

=G9)beQ/6S)10E

) a1u0yaa

62#R+#7)do

the pounding guineacorn that the old woman is doing

I2009)0;);0.004J

Similarly in %;5* the participle a$ain takes an ob'ect, but itself functions as the fronted 2F9

ob'ect of the !erb yaa SdoT, as well as tri$$erin$ accompanyin$ copula. %;5* u)0dv
/C71"ound /C31guineacorn

=G7

$ender a$reement on the ) a1u0yaa

i0 ) da'u)

w0i )

A-71C6P

u)0yo0no)

;S)beQ/6S)10E

62#R+#7)do

it s pounding guineacorn she is doing

I2009)0;);0.004J

"hese elicited e3amples therefore su$$est that the participle should indeed be analysed as a sin$le mi3ed cate$ory, rather than ambi$uous between noun and !erb , certainly it appears to offer better e!idence for this than the -n$lish $erund. Anfortunately the rele!ant constructions are absent from the corpus47, so it offers no e!idence either way. -!en if the participle in the associati!e construction of %;2* is of the same cate$ory as the one in the participle ob'ect construction in %;;*, there is still a semantic difference between the two constructions. "he former seems to be more $eneral in meanin$, and there is a wide !ariety of ways in which the acti!ity encoded by the participle can be modified. In the case of a transiti!e !erb the Spossessor +1T in the associati!e construction can bear the
1="I-+"

%as in %;2**,

=G-+"

%;7*, or

I+S"/AM-+"

%;9*

role. "he participle ob'ect construction, on the other hand, is naturally limited to the same semantic roles as strai$htforward ob'ects. %;7* u)0yi nda wu0ttu)
+#7)see =G7)41.12SS

our seeing %;9* u)0yaahu)kwa)0nu)


+#7)for$i!e)/-S

Itats002.007.028J wu,1u0noo
=G7R+#7)mouth

verbal orgiveness Ii.e. not of the heartJ

Ioamy004.44FJ

2ne further difference applies in cases where the followin$ +1 corresponds to the ob'ect of the finite construction. If the +1 is definite then it is far more likely to appear in the participle ob'ect construction than the associati!e construction , in fact the latter has only been obser!ed with indefinite +1s. "he contrast can be seen in the followin$ pair of e3amplesK

47 i.e. a participle based on a transiti!e !erb, with an e3pressed ob'ect and a nominal modifier or other e!idence of =G7 a$reement morpholo$y. It is not really surprisin$ that constructions of this comple3ity are missin$ from the corpus.

2F9

%;9* n n)1u)0hwaara)
if

whenR;S)startQ/6S

u)0s

/C71drink

wu,1ta0a a)

A-7E/C+1tobacco

i he starts smoking tobacco Itats007.002.04;J %80* n n)1u)0hwaara)


if whenR;S)startQ/6S

u)0s

/C71drink /C+1tobacco

ta0a a)

ti 0 lle)

A-+1t!at

i he starts smoking that tobacco

Itats007.002.079J

Some +#7 de!erbal nominalisations ha!e Duite different meanin$s, as shown below , for * SbowT is e3ample u)0kaa Smessa$eT denotes the "<-M- of the !erb kaanu SsendT, and u)0taa the
I+S"/AM-+"

of the !erb taa SshootT. "hese should be re$arded as homophonous with death message error bow koo kaanu kure taa die send err . rom Hausa kuskure2 shoot

the correspondin$ infiniti!es rather than !a$ue. %84* u)0ko*o u)0kaa u)0kure) * u)0taa

+#7

de!erbal nominalisations are not always infiniti!es , they can also denote locations.
+#7

/ecall from LF.;.8 that locati!e nouns can be formed usin$ the

prefi3. "his prefi3


G2=6

can also be attached to a motion !erb to form a nominal denotin$ the from pasa ScrossT %note the chan$e in the final !owel*. %82* u)0pasu)
+#7)across

of the

mo!ement. -3amples include u)0kum u) SupT from kum a SclimbT and u)0pasu) SacrossT wu0mpa)
=G7)this

Ipasa R ScrossTJ

this side Iof the ri!erJ 0inally, it should be noted that


+#7

I"ikula, sami004.8;FJ de!erbal nominalisations may also take the stati!e

0ni nominalisation suffi3 %L8.8.5.4*.

(&#

!epeated action Dgender >G

"he +#9 prefi3 ku0 can be applied to a lar$e number of !erbs, perhaps the ma'ority, and the resultin$ nominal denotes a repetiti!e action. 0or some !erbs the final !owel may chan$e to IuJ. In some cases this chan$e seems to be obli$atory, as with 7e'e SdanceT in %8;*, in other cases such as kollo Slook atT it is optional. If the ori$inal !owel was nasal then the chan$e is to a nasal u.

250

%8;*

o)0kondo)

;1)enterQ/6S

ku)07e'u)

/C*1dance

I7e'e R danceJ

they started dancing

Isahs004.002.009J

In the corpus, +#9 nominalisations !ery often in!ol!e !erbs conduci!e to iteration, as in %88*. Stati!e !erbs %e.$. $uwo Sbe fullT*, on the other hand, do not seem to allow the nominalisation at all. %88* ku)0 lu) ku)0'i 7e)'u) ku)0yi ma) ku)07i 7a'u
+#7

searching washing %i.e. the action* orging shivering


+#9

l 'i7e'e yima 7i7a'a

search wash orge shiver

6ike the

infiniti!e discussed abo!e, these


=G9

nominalisations ha!e properties of

both nouns and !erbs. "hey tri$$er ku)0$i 8u) kw0i )

a$reement %8F* and can function as the head of I$i8a R pluckJ I2009)0;)48.002J I$inda 1 waitL I2009)0;)48.002J

an associati!e construction %85*, but can also occur with !erbal e3tensions %87*. %8F*
+#9)pluck =G9)#21

a lot o plucking Ie.$. of fruitJ %85* u)0yo* ku)0$i nda) ku0ttu)


+#9)wait

;S)beQ/6S

=G9)41.12SS

he's looking a ter us Ilit. She is our waitin$TJ %87* ku)08i sQi lRu driHHle I8isa R spillJ

+#9)spill16=#s

Ieamy00;.4;;5J

1erhaps the most common way of e3pressin$ repetiti!e action is throu$h a construction consistin$ of the !erb yaa SdoT, followed by kuu) and then the !erb denotin$ the repeated action, as illustrated below. %89* * a)0yaa
;1)doQ/6S

ku,1u0 am ala
=G9R+#7)fumble

they umbled Isaat004.00;.0;4J %89* u)danga)nu) u)0danga)nu) ) u)yaa u0yaa kuu Jdi v ) i) ku,1u0 d)1vi ) n) n) mo)to*o mo)0to*o

;S)keepPdoin$Q/6S

/C71do

A-*E/C71moistenR;S.1/2

with

+#8)sali!a

he kept on moistening him with saliva

Isaat004.009.095J

254

%F0*

* a)0yaa

;1)doQ/6S

ku,1u0si ri ) a)

=G9R+#7)playPfren>iedly

$i )0kooto)
+#5)drum

they played the drum renHiedly Ilit. Sthey did playin$TJ

Isaat004.009.08FJ

"he lon$ !owel and the fact that the tone pattern on I kuu) !erbJ is identical to that of the Sback endT of the associati!e construction %L8.8.F.4* su$$ests the breakdown pro!ided in the $losses to %89)F0*, that of an =G9 associati!e a$reement marker followed by the
+#7

infiniti!e, familiar from LF.8.4. "his would e3plain the presence of the bare

ob'ect +1 in %89)F0* , it is an infiniti!al ob'ect 'ust as in %80* abo!e. <owe!er it also raises the Duestion as to the identity of the missin$ Sfront endT of the associati!e construction. 2ne speculati!e candidate is ku)0pi lu) SmuchNmanyT.

(&%
+#8

Continuous action Dgender &G

de!erbal nominalisations are similar to the +#9 ones 'ust discussed, in that they both

reDuire there to be multiple occurrences of the acti!ity concerned. +onetheless they are Duite distinct, and while +#9 nominalisations highlight %or to use a term from 6an$acker 4997, SprofileT* the iterati!e nature of the e!ent,
+#8

nominalisations seem to Ssmooth

o!erT the indi!idual multiple occurrences, focusin$ instead on the homo$eneous49 state of affairs which arises , hence the term Scontinuous action nominalisationsT. 0or instance, the
+#8

nominalisation in %F4* below could not be used to refer to a sin$le

instance of kickin$, but it could denote someone:s takin$ part in a $ame of football. "o refer to a repeated kickin$ motion, the +#9 nominalisation would be more appropriate. If one did want to use this construction to refer to a simple3 e!ent, for e3ample while watchin$ a slow)motion replay of a player kickin$ a ball, then the +#7 prefi3 u0 %LF.8.4* would ha!e to be used instead. %F4* u)0yo*
;S)beQ/6S

a1ma0gava)
62#R/C4)kick

he is playing ootball Ilit. She is at kickin$TJ I2009)0;)48.002J It should be noted that not all +#8 de!erbal nominalisations are of this kind. If the initial consonant of the !erb is len$thened the resultin$ nominal denotes a diminuti!e physical ob'ect. "hese are the diminuti!es of the Sindi!iduati!eT nominalisations %LF.8.F*.

49 It is noteworthy that +#8 prefi3es also occur with liDuid mass nouns %LF.2.9.8*, which ha!e this same property of homo$eneity.

252

%F2*

m)0k0k') m)0g0gd

a ladle small lump, pimple

k' gd

scoop make into ball

(&&

Agent Dgender =C#G

=$ent nominalisations are formed by addin$ prefi3es from the human $ender 9N2 to a !erb stem, a pattern found in other (enue)#on$o lan$ua$es e.$. Aks J eJ %<edin$er 4990KF* and -simbi %Stallcup 4990K4F4*. In #icipu the tones on the resultin$ word are all < and the final !owel is nasalised, althou$h the e3act form is sub'ect to !ariation , the final !owel of all !erbs can be either %i* diphthon$ised, %ii* len$thened, or %iii* followed by an m. So for e3ample the !erb pata SreDuestNbe$T can be nominalised as , pataa , or patam , all meanin$ Sbe$$arT. 0urther e3amples of the either patau nominalisation are $i!en below. "hese nominalisations are apparently fully producti!e, althou$h they are actually rather rare in the corpus. %F;* se'em wm pa8am pi nam matam lapam kum am 8anam hunam guyam wayam dukwam rawamA~ carver thie butcher barber parent wise person rider writer killer one with ability someone coming someone going eater
+#2

se'e w pa8a pina mata lapa kum a 8ana huna guya waya dukwa raa

carve steal slaughter shave give birth know ride write kill can come go eat

"he plurals are usually re$ular

SparentsT, but there is at least plurals, e.$. a)0matau

one irre$ular pluralK tumamNa)0tamam SfarmerNfarmersT. It is not known whether any !erbal e3tensions can be included in the nominalisation, nor whether this nominalisation %or an 9N; eDui!alent* occurs with inanimate actors. In some cases there is a contrast with another morpholo$ically)related word denotin$ an a$ent. 0or e3ample, as well as the re$ular form wm SthiefT there is also a word w. "he semantic difference seems to be that the former is someone whose life is characterised by thie!in$, whereas the latter can be applied to somebody who has only committed a sin$le act of theft. = couple of other a$ent nominalisations ha!e been obser!ed, but it is not known how producti!e they areK they are smu SdrunkardT from s SdrinkT, and guyaye from
49 {aam was not accepted.

25;

guya ScanT, apparently meanin$ Shuman bein$T.

(&(

Individuative Dgender 1C#G

Gender 4N2 nominalisations usually denote a physical ob'ect, often a product arisin$ from the action. Sometimes the first consonant of the !erb stem is len$thened, 'ust like the pre)prefi3es discussed in LF.;.5, and the final !owel may chan$e to i. It does not seem possible to predict the tone on the nominal. %F8* ka0g0gi )ta) k0k0k)'i ) k)0k0k8i ka08i )si i k0ku)d ko0ku)nto)nu* a snapped o piece a scoop %i.e. a measure* a cut piece a spill a meeting a ist gita k' k8 8isa kud kuntonu snap scoop cut spill meet %i.e. $ather to$ether* make a ist

2ther semantic roles are possible includin$ ShelpT* and


62#="IE-

=G-+"

%ko0g0go)oni ShelperT from goonu

%k)0t0tpu SholeNpitT from tpu Sput insideT*. See also L8.8.5.2 for

the locati!e nominaliser 0tu, which !ery often results in a noun with 4N2 $ender.

(&9

'iscellaneous

.e!erbal nominals belon$in$ to classes 4, ;a, ;b, F, and 5 do occur, but they are less producti!e and their semantic properties are not yet understood. Earious stem chan$es may also occur, in particular a chan$e in the final !owel to i or u.

258

0able 6]- Iiscellaneous deverbal nominalisations #lass +ominalisation +#4 ka)0raa eating ke0ttepi ) talking
+#;a

Eerb raa tepe tuma sipa hatta pnt ' panda hungwa haala latta 5aya s kalla dama para taa kana dopo

eat talk cultivate sing yawn clap tire orget rest walk sleep .v.2 discipline .v.2 drink .v.2 clear speak1speak against stalk shoot cut sew

i0 ) tuma) i0 ) si pa i0 ) hattu i0 ) pntu i0 ) 'ri i0 ) pandari ru)0hungwa n)0haalu n)0lavu) $i 0 ) 5aya ti 0 ) s* $i 0 ) ka)lla $i 0 ) dami $i 0 ) pari ) $i 0 ) tai $i 0 ) kani $i 0 ) dopi )

arming song1songs yawning clapping tiredness orget ulness resting %ourney sleep .n.2 discipline .n.2 drink .n.2 clearing .i.e. location2 whisper stalking shooting cutting seam, or weaving

+#;b +#F

+#5

-.-

4refixes and allomorphs

#icipu nouns follow the prototypical (enue)#on$o pattern of noun class prefi3 plus stem %L8.8*. In this section I will discuss the phonolo$ical form of the noun prefi3es. General properties are dealt with first, followed by a discussion of the class)specific allomorphs.

((1

1owel.initial roots

Eowel)initial roots %L;.4.4* can be found in each classK 0able 6_- Noun class pre i*es occurring with vowelcinitial roots 4 2 ;a ;b 8 F 5 7 9 9 Ika(dandaL Ia(dandaL Iy(m)L Iru(maL Ima( a)L L ) Imi(si I$e(8u)L Iw(v(L Iv(m)L IkUe(tuL thorn %kA0 adanda* thorns %A0 adanda@ monkeys %i0 m)* war %ri0 uma@ anger %mA0 a a)* )@ crying %mi0 isi top o the head %$i0 e8u)* ear %u0 v monkey %vi0 m)@ medicine %ku0 etu@ 25F

"he prefi3 !owel Duality is usually subsumed by that of the root !owel, apart from class 9 where the labialised IkUJ retains a trace of the NuN prefi3 !owel, and class ;b, where some words ha!e been found with optional palatalisation , Srainy seasonT can be either Iru(si )J or IrZu(si J ).

((#

1owel harmonest &ain'i lan$ua$es %cf. <offmann 4957 for #:6ela*, where

Anlike the north)western

prefi3es often 'ust consist of a consonant followed by a short transitional s$hwa, #icipu prefi3es follow the &ambariN&amuku pattern whereby the prefi3es always ha!e phonetically full %but still short* !owels. +ouns from classes 4, 2, and 8 are formed by addin$ to the root the harmonisin$ prefi3es kA0, A0, and mA0 respecti!ely. "he Duality of the !owel in these classes is determined by the rules which were set out in the discussion on !owel harmony in L;.FK If the root contains NeN, then the prefi3 !owel will be IeJ If the root contains NoN, then the prefi3 !owel will be IoJ If the root contains NN, then the prefi3 !owel will be IJ 2therwise the prefi3 !owel will be IaJ

hile these rules predict the !owel Duality of the prefi3 in the ma'ority of cases, there is an additional factor in!ol!ed when it comes to roots which only ha!e close !owels. /oots with only IiJ or only IuJ sometimes occur with prefi3es containin$ the mid !owels IeJ and IoJ respecti!ely, rather than the e3pected IaJ. "his can be !iewed as an assimilatory process, with the underlyin$ prefi3 !owel NaN raisin$ in the en!ironment of a hi$h root !owel, but there does not seem to be any way to predict whether or not this process will occur. %FF* ke)0 i )m i *i me)0gi si ) ke)0 i ki ka)0yi vi ka08i )si i ka)0gi si ) buHHing insect, k.o. walking stick celebration . rom Hausa biki2 cold meal .tuwo2 spot stick

255

%F5* ko0$i )yu* ko)08u*u ko)0luu ka)0 ungu ka)0guutu) ka)0nuu

heap heart knee snake buttock head o corn

"here remain some nouns whose prefi3es cannot be accounted for either by harmonisation or by assimilation. "he words in %F7* should ha!e the prefi3es ma0:ka0 accordin$ to the former, and me0:ke0 accordin$ to the latter. "he roots in all such e3ceptions ha!e contained only NiN !owels, and the prefi3 !owel has always been N oN. ) %F7* mo)0hi'i mo)0sii mo)0ni ko)07i ni blood shame water ghost

Scalf %of le$*T may be an e3ception to the otherwise uni!ersal 0inally, the word ma)0dei application of !owel harmony, since it seems to ha!e a consistent IaJ !owel in the prefi3. It may be that the usual rules do not apply to diphthon$s , this is the only attested e3ample of a monosyllabic root in!ol!in$ NeiN or N euN. could 'ust be the default !alue of the prefi3. ords containin$ the diphthon$s N aiN and NauN also take prefi3es with IaJ !owels, but as discussed in L;.F this

((%

Tone
elmers %497;K479* wroteK

#oncernin$ the tone on noun class and a$reement prefi3es,

0or the most part, in a $i!en I(antuJ lan$ua$e, the noun and concord prefi3es all ha!e the same tone, far more commonly low than hi$h. Sub'ect concords may ha!e hi$h tone in some !erbal constructions, but this is a function of the !erbal construction, not of the concords themsel!es. In some lan$ua$es, howe!er, certain concords for classes 4 and 9 ha!e low tone, while all other concords ha!e hi$h tone. "he situation in #icipu is similar to elmers: first $roup of (antu lan$ua$es, in that the

tone on the noun class prefi3 is $enerally low and does not !ary accordin$ to class. Instead the tone on the noun class prefi3 is usually predictable from the structure of the noun root. It can be deri!ed as followsK

257

If the noun root is !owel)initial, then the prefi3 !owel mer$es with the first root !owel to form a lon$ !owel, as illustrated in LF.F.4. "he prefi3 tone is almost SratT is the only known e3ception*. always hi$h %ka)a'aa %F9* Ika(dandaL Ik(5i L Iko($i )L Ika)('a(L thorn %kA0 adanda@ hecgoat %kA0 5i * hole %ku0 o$i )* * rat %kA0 a)'aa

If the noun root contains only low tones then the prefi3 is hi$h. %F9* ko0ggo)m o) m0kk)8) ka0da)ngi ) wu0nto):vi 0 nto) bat rumourcmongering testicle guest hut1guest huts

2therwise the prefi3 is low. %50* ka)0 ara ko)0dondo u)0kuwo) elder garden baobab tree

+ouns with a 6 < root tone pattern are an e3ception to the abo!e rules, and these may ha!e either hi$h %54* or low %52* prefi3es. "here does not seem to be any way to predict which. %54* < ) 6 < ka0kka)$ii ka08i )si i k0ku)d k)05i )y ka)0ka)anaa ma)0$i )iji i middle spot meeting pumpkin crab calabash

%52* 6 ) 6 <

((&

Classes %a and %b

#lass ;a and class ;b nouns both tri$$er yi0 a$reement prefi3es. <owe!er they ha!e been di!ided into two morpholo$ical noun classes, accordin$ to whether the noun takes an i0 %;a* or a ri0 %;b* prefi3. "here is no difference between the two in terms of a$reement, but there is with respect to $ender pairin$s. #lass ;a nouns are either members of the sin$le class $ender %;a*, or plurals of class 9 nouns %9N;a*. #lass ;b nouns likewise ha!e their own sin$le class $ender %;b*, but they also pair up with class 2 plurals %;bN2*. "he distribution is summarised belowK

259

0able 6`- eistribution o class 6 pre i*es according to gender 1refi3 ;a %i1* ;b %ri1* Sin$ular %;bN2* 1lural %9N;a* Sin$le class $ender

"he history of these two classes can be illuminated by considerin$ comparati!e data from #entral &ambari, which has an interestin$ counterpart to class ;b. "here are at least two #entral &ambari nouns which, althou$h they tri$$er une3pected lic prefi3 %.e #icipu class ;b. .e
=G9

a$reement, ha!e an

olf 4974*. 2ne of these nouns %li 0 ) ya* SarrowT* has a co$nate

) in #icipu class ;b. 0urthermore, the li) nouns pair with class 2 plurals, 'ust like ri )0hya'a olf !iews these two #entral &ambari roots as a remnant of the 1roto)(enue)#on$o $ender lic1ac %correspondin$ to 1roto)(antu FN5*, which includes words for SarrowT, SbodyT, and SwarT, $losses which can all be found in #icipu ;bN2 %althou$h the roots themsel!es are not ob!iously co$nate*. #lass ;a, on the other hand, is most likely a refle3 of 1roto)(enue)#on$o @N %correspondin$ to 1roto)(antu 40*. "hus despite their syntactic eDui!alence, classes ;a and ;b probably ha!e different ancestor classes.

(((
"he
+#F

Class (
prefi3 has an interestin$ set of allomorphs, 'ust as in #entral &ambari n)0kantu knives mi 0 ) nnu birds )J mi 0 i si ) Imi(si crying

%<offmann 495;*K /B0 mi0 before short consonantsK before lon$ consonants and !owelsK
+#F

"here are in fact two ways of formin$ form mi1K %5;* ma)0nnu ma0llu* bird master

nouns from roots with lon$ initial

consonants. 2ne method in!ol!es addin$ an epenthetic !owel to the /0 allomorph to mi 0 ) nnu mi 0 llu* birds masters

"he other method, which is apparently not found in #entral &ambari, is to reduplicate the first syllable of the root before addin$ the /0 prefi3. methods in LF.F.9. e will come back to these two

20 /0 represents a nasal homor$anic with the followin$ consonant. Sometimes this surfaces as a nasalised !owel Ii J or IuJ , see L;.4.7 for details.

259

((9

Class 9

"he +#5 prefi3 takes a !ariety of formsK $i0, $u0, ti0, or tu0. "he !ariation between i and u is in this case lar$ely predictable %see McGill 2007K59)70 for discussion*. "he $:t alternation is more comple3. 2f the 49 paired nouns %from $enders 5NF and 5N2*, 47 be$in with I$J, and only two %ti 0 ) 7a)aruma) SfleaT and ti 0 ) ji ma) Smale wartho$T* with ItJ. "his can be treated as an idiosyncratic property to be handled in the le3icon. "he unpaired $ender 5 nouns are di!ided rou$hly eDually between I$J and ItJ, with no discernible phonolo$ical pattern to the distribution. "he reason for this !ariation is unknown, and if it were not for the phonetic similarity between the two phones24, I would ha!e no hesitation in settin$ up morpholo$ical subclasses 5a and 5b, 'ust as has been done for ;a and ;b.

((<
"he
+#9

Class =
prefi3 has a most interestin$ set of !ariants, the distribution of which is as

followsK If the root is consonant)initial, then #4 may simply be len$thened22, for e3ampleK %58* 707a k0kaa d0d* $0$') person woman horse sheep .sg.2 a)07a a)0kaa i )0d* i )0$') persons women horse sheep .pl.2 9N2 9N2 9N; 9N;

In other cases, #4 does not len$then and so there is no prefi3K %5F* ;0$i )$$ere) ;0mooto) ;0hi i'o)o star car ant, k.o. i )0$i )$ere) i )0mooto) i )0hi i'o)o stars cars ants, k.o. 9N; 9N; 9N;

If the root is !owel)initial, then vi0 is added2;. =s with all !owel)initial roots, the prefi3 !owel coalesces with E4 $i!in$ a lon$, hi$h)tone !owel with the same Duality as E4K
24 #ertain words e!en show SfreeT !ariation between the two phones, such as $iyo:tiyo S$etT, althou$h this is not the case $enerally. #oncernin$ the related lan$ua$e #entral &ambari, .a!id #ro>ier %p.c.* has remarked that the +#5 prefi3 consonant may be either I$J or ItsJ. "his time, howe!er, the alternation is principled, with I$J occurrin$ before roots with a front E4, and ItsJ otherwise. Interestin$ly this alternation does not apply more $enerally in the lan$ua$e, only to the +#5 prefi3, and N$N and NtsN are found in contrast 'ust as N$N and NtN are in #icipu. 22 I ha!e not found a lon$ #4 in any class 9 roots. 2; Since the prefi3 !owel is always replaced in the coalescence, it is not possible to be sure that the underlyin$ !alue is NiN. <owe!er the correspondin$ =G9 prefi3 is vi0, and in all other classes the prefi3 !owels are the same in their respecti!e noun and a$reement prefi3es.

270

%55*

vi0 m) vm) i0 m) ym)

monkey monkeys
+#9

9N;

"he same possibilities e3ist for the

plurals of nouns in $ender 7N9 such as

u)0lenji l0lenji SdaydaysT and w0t)v0t) SmoonmoonsT. "he distinction between the len$thenin$ prefi3 and no prefi3 at all is not always clear)cut, especially at the be$innin$ of the phonolo$ical word. #ertain nouns always seem to start with a lon$ consonant %e.$. k0kaa SwomanT*, but others are more !ariable, e!en within the speech of an indi!idual. "here seems to be a correlation between short words and the occurrence of the -0 prefi3, and a tentati!e hypothesis mi$ht be that %a* there is a trimoraic minimality constraint for noun words, and %b* syllable)initial $eminates are moraic. "hus if an
+#9

noun root is short %i.e. bimoraic*, then a

len$thened prefi3 would be necessary to satisfy the minimality constraint28. "here are three constructions in which the consonant is almost always len$thened, e!en for words such as those in %5F* which otherwise ha!e a null prefi3. "hese are after the associati!e construction proclitic %L8.8.F.4*, after the co)ordinatin$ proclitic ni ) %L8.8.F.8*, and after the locati!e proclitic A %L8.F*. It seems then that the distinction between the -1 and ;0 sets of +#9 nouns illustrated in %58* and %5F* becomes blurred in a %phonolo$ical* word)medial en!ironment. In these positions e!ery consonant has been obser!ed len$thened. Illustrati!e e3amples of nouns with the -0 allomorph are shown in "able ;9. 6on$ !oiceless plosi!es are not easily detectable utterance)initially2FG in these cases the e3amples $i!en are utterance)medial.

28 "he hi$h freDuency semi)$rammaticalised 707a SpersonT would be an e3ception. 2F 1ossibly there is a phonetic difference in the followin$ !owel which would betray its presence, such as an increase in amplitude %6adefo$ed and Maddieson 4995K98*. "his has not yet been in!esti$ated for #icipu.

274

0able 6b- )*amples o 0 i ndu) 5 $ d 8 g gw h hy hw j k kw l m n p r s t v w y 7 ' 'w 'y ni )1h0hwi 'i ) j0jeve k0kuwo) l0lenji m0ma$i ) n0naa p0pi ya r0ree s0se'u t0tii v0v)$i y) w0wm y0yumu) 707a '0'i ri ) ni )1k0kwaa
BG

N5`

consonantclengthening pre i* eamy08;.008 eamy080.002.024 saat002.008.0F0 tats004.002.040 eaim00;.4;94 tapf004.00F.0;9 eamd028.4292 saim004.029 eamy0;2.042 sayb004.080 eamy0;2.0;F eamd002.002 eamd04F.;90 saim004.0F9 eamd002.4;F eamd002.09F eamd04F.;58 tats004.004.007 eamy00;.4;08 saim004.409 eamy080.004.009 eamd002.05; taff002.0F;

tadpole python grasscutter trees sides pawpaw streams and a Lelna person an antelope baobabs and skins days riend cow guinea owl towns cocwi e container slimes chie iron person kind %from <ausa iri*

$0$ii d0di yo) 808anga) g0galu) g0gwanda) h0hoiyu

0or #entral &ambari <offmann %495;* set up two subclasses, 9a for nouns with the -0Nvi) prefi3, and 9b for those that ha!e no prefi3. Gi!en that this alternation is often not predictable in #icipu, as well as bein$ neutralised word)medially, I ha!e decided not to split the class in this description.

25 "here are no attested +#9 nouns whose stems be$in with 5, hy, 'w, or 'y. "hese are all fairly rare phonemes so this should not be surprisin$.

272

((=

)penthesis and reduplication

= number of #icipu words form their plurals by duplicatin$ the first consonant of the root to$ether with the normal method of chan$in$ the prefi3. "hese words ha!e so far only been found in $enders 4N2 and 8NF, and their roots always be$in with a lon$ consonant. "his is si$nificant because roots be$innin$ with a lon$ consonants are relati!ely rare in #icipu. Some e3amples are $i!en belowK 0able Ja- Reduplicated plurals with rootcinitial long consonants Sin$ular k)0dd ka)088ayi ka)0hhuu$i ka0kka)$ii ka)0llavu) m)0$$)k m)0ggd m)0kk') m)0pp me)0ttegu) m0kkwe)te) 1lural )0di 0 dd a)08i 088ayi ) a)0hu0hhuu$i a0ka0kka)$ii BD a)0li 0 ) llavu) n)0$i 0$$k n)0gu0ggd n)0ku0kk') m) 0pi 0 pp n)0ti 0 ttegu) n0ku)0kwe)te) Gloss edge tear i.e. a rip cloud middle dream bag, small bump ladle granary, small shirt gecko Gender 4N2 4N2 4N2 4N2 4N2 8NF 8NF 8NF 8NF 8NF 8NF

=n epenthetic short %usually hi$h)tone* IiJ or IuJ is inserted between the first %short* consonant and the ori$inal %lon$* consonant. 0or 4N2 nouns with initial lon$ consonants the reduplication is usually optional, and there does not seem to be any semantic difference between the two forms. +ouns from $ender 8NF, on the other hand, only ha!e the reduplicated possibility in the plural29. "he ma'ority of 8NF nouns form their plurals by reduplication, while 'ust a few use the prefi3 mi0K recall from LF.F.F that this is an allomorph of the
+#F

prefi3 /0 applyin$

to roots be$innin$ with a lon$ consonant. "hese two processes of reduplication and epenthesis can be seen as means to the same end , the a!oidance of consonant clusters with three timin$ units29. "he effect of reduplication is to split the /0 prefi3 from the lon$ consonant and place it to the left of the shortened reduplicant instead, resultin$ in a consonant cluster of only two timin$ units. -penthesis in n)0llu. "here is an interestin$ parallel to the /0Nmi0 alternation from outside the noun
27 Anusually the reduplicated !owel in a0ka0kka)$ii is IaJ, rather than IiJ or IuJ. 29 I ha!e come across one e3ceptionK me0 e)ri ise) SswiftT, m 0 e)ri ise) SswiftsT. 29 0or a similar phenomenon in the =ustronesian lan$ua$e 6eti see <ume et al. %4997K;7F*.
+#F

plurals a!oids the

consonant cluster alto$ether, so that we ha!e, for e3ample, mi 0 ) llu SnecksT instead of

27;

class system. "he con'unction /)0 SandNwithT %L8.8.F.8* has a !ariant ni )0, and the two allomorphs are distributed in 'ust the same way as the
+#F

prefi3esK ni )0 occurs before

lon$ consonants and !owels, and /)0 before short consonants;0. 0or completeness: sake two Duite different kinds of reduplication should be mentioned. "he first in!ol!es the plurals of nouns from $ender 9N2K %57* ku)05aa ku)0saa ku)0s ku)0ta*a marsh mountain lake debt a05a)a5a) a0sa)asa) ) ) 0ss a0ta)ata) marshes mountains lakes debt

In this construction the prefi3 is hi$h tone instead of low and there is no association with lon$ consonants. =nother difference is that with 9N2 reduplication the root syllable is duplicated, not 'ust the first consonant. It does not seem possible to predict which 9N2 nouns will under$o reduplication, althou$h all the e3amples so far ha!e been monosyllabic. 0or those that do reduplicate in the plural, the process is obli$atory. "he second kind in!ol!es the plurals of nouns from $ender 7N9, althou$h this time the reduplication is optional;4. %59* u)0ree u)0yaa u)0hoiyu town road stream r0reere)e or r0ree y0yaaya)a or y0yaa B1hoihoiyu or h0hoiyu towns roads streams

-.2

!oan$ords

6oanwords can be found in most $enders, althou$h they cluster in $enders 9, 9N;, and 4N2. 1recise distributional percenta$es are impossible to pro!ide, mainly due to the difficulty of distin$uishin$ between borrowin$ and code)switchin$. =s mentioned in chapter 4, !irtually all #icipu speakers are fluent in <ausa. "he fi$ures below $i!e a rou$h picture of the distribution, but like the numbers in LF.2, they should be treated with caution.

;0 #icipu does not $enerally allow syllables with codas, and so it has not been possible to find any further parallel situations that mi$ht $i!e rise to a consonant cluster. ;4 "his reduplication process is a potential source of the +#9 len$thenin$ prefi3. <owe!er it should be obser!ed that +#9 prefi3es are usually e3ponents of singular nouns. See McGill %n.d.* for further discussion.

278

0able J,- &roportion o loanwords in genders -ender 4N2 8NF 9N;a 9N2 7N9 9N2 5NF 4 ;a ;b 8 F 5 7 9 9 2 ;bN2 8N2 5N2 "otal Pre'i$ (otal ;oans Percentage borrowed kA:A 292 F5 20 mA:mi 44F 42 40 vi:i 449 8; ;5 ku:A 58 0 0 u:vi F4 4 2 vi:A 24 8 49 ti:mi 47 0 0 kA 4F 7 87 i 20 4 F ri 7 0 0 mA 24 ; 48 mi F 0 0 ti FF 5 44 u 49 0 0 vi 88 ;; 7F ku 4F 0 0 A 4 0 0 ri:A 8 0 0 mA:A 2 0 0 ti:A 4 0 0 975 455 49

6ea!in$ aside $enders 4 and 4N2 for the moment, $enders 9, 9N2 and 9N;a are the most o!er)subscribed. It is probably not a coincidence that the noun prefi3 for class 9 is either null or a -0 allomorphG <ausa words can Sfit inT to class 9 without lookin$ out of place;2. Some e3amples areK %59* roo a) %9* ka)ra)atuu %9* 707eni ) %9* kaataako %9N;* mi )saali ) %9N;* d0daga %9N;* plastic %roba* reading %karatu* cloth %>ane* plank %katako* e*ample %misali* belt %da$a*

=lthou$h it is presumably the ; allomorph that pro!ides the stron$est attraction to class 9, loanwords also occur with the -0 allomorph, as can be seen in %59*. 2ther e3amples include h0hu)uhu Slun$sT %huhu*, $0$aawa SnoiseT %tsawa* and e!en calDued idioms such as k0kai vi 0 ,n0haalu Son a 'ourneyT %kan ta iya, lit. Shead of a 'ourneyT*. <ausa words may also be assi$ned to other #icipu noun classes if they start with a syllable resemblin$ a noun class prefi3, as shown below. "he three e3amples referrin$ to towns illustrate that the assi$nment is purely formal and not semantic;;.
;2 See #orbett %4994K72)7;* for discussion with re$ard to loans in &iswahili. ;; "his reinterpretation seems to be producti!e, althou$h I ha!e not yet in!esti$ated $ender assi$nment

27F

%70* ma)0karantaa %8NF* ma0llu* %8NF* ma)0'anaa %8NF* ma)0 u8i %8NF* ta0a a) %5* 3a)kuuku) me %8* Kono)o ke %4* 2irindaji vi %9*

school %makaranta* master %mallami* meaning %ma:ana* key %mabui* tobacco %taba* it's Iakuku town it's 3ano town it's eirinda%i town

It seems that in #icipu borrowed words are not assi$ned $ender on the basis of their meanin$. "his contrasts with the situation found in ut)Ma:in from the +orthwest branch of est &ain'i, where loanwords are freDuently assi$ned $ender because of their semantic properties %Smith 2007K;7)55*. 2ne reason for this difference may be that the null)prefi3 class in ut)Ma:in is restricted to humans, and is therefore una!ailable as a $eneral destination for formal assi$nment. "here remains a lar$e number of loanwords in $enders 4 and 4N2, the presence of which cannot be e3plained by either formal or semantic assi$nment rules %recall that this $ender has no coherent semantic structure*. %74* ka)0la)ahi ya) %4* ka)0'a)l arka) %4* ka)0haku)ri ) %4* ka0huusi ) %4* peace %lafiya* prosperity %albarka* patience %hakuri* anger %haushi*

ka)05auna %4N2* bu alo %auna* ko)0wa)ndo %4N2* pair o trousers %wando* ka)0kaasu)wa) %4N2* market %kasuwa* ka)0huska) %4N2* ace %fuska* "here are at least two ways to account for this. 0irst, it may be that as the lar$est $ender in #icipu, 4N2 %and by e3tension 4* functions as a default, if the formal assi$nment rules e!ident in %59* and %70* fail, for whate!er reason, to come into effect;8. =n alternati!e and perhaps preferable e3planation is that the e3istence of borrowed words in 4N2 is due to a $eneral $ra!itation of nouns towards that $ender. "his trend was obser!ed for #entral &ambari by .e olf %4974*, and the same pattern is found in #icipu. It may be that these loanwords started their li!es in $ender 9N; as we would e3pect, but o!er time they ha!e shifted to $ender 4N2 alon$ with other #icipu words
for towns unknown to the =cipu. ;8 "he idea of a default $ender has been criticised by #orbett %4994K77)90*, since lin$uists may propose one e!en when nati!e assi$nment rules already account for the data. =ccordin$ to #orbett:s %4994K8;) 89* analysis of (antu)type systems i.e. $ender is assi$ned accordin$ to the noun class pairin$ of the noun, we should say that there is a de ault noun class pairing 4N2 for loanwords, which in turn means they are assi$ned 4N2 $ender in the normal way.

275

%perhaps by means of some of the deri!ational processes discussed in LF.; and LF.8*. "he nouns k)0$$)k \ba$T %from %aka*, ka)0tti lu SpotT %tulu* and ko)0ttudu) \rid$eT %tudu* may be e!idence of this shift , they all ha!e root)initial lon$ consonants which su$$ests an earlier sta$e of their e3istence in 9N;;F. If the loanwords in 4N2 ha!e mo!ed there from 9N;, then we would e3pect to find more established items in 4N2 and newer borrowin$s in 9N;, and this is indeed the case. ords for modern ob'ects such as mooto) ScarT, roo a) SrubberT;5, and ka$a*a Sbicycle chainT all belon$ to 9N;, while words denotin$ uni!ersal or more traditional concepts are more often found in 4N2.

;F See McGill %n.d.* for the su$$estion that $ender mo!ement out of 9N2 or 9N; is responsible for most root)initial lon$ consonants in nouns, whether borrowed or not. ;5 =ll -n$lish loanwords must ha!e come throu$h <ausa , as mentioned in chapter 4, hardly any =cipu speak -n$lish.

277

Chapter 9 Agreement
In the pre!ious chapter I presented the !arious noun classes and $enders %pairin$s* of #icipu, to$ether with the deri!ational functions of the noun class prefi3es. In this chapter we will be concerned with the $ender a$reement system. =lthou$h comple3 enou$h from the perspecti!e of many other lan$ua$e families, $ender a$reement in #icipu is less SmessyT than in many (antu or Grassfields (antu lan$ua$es. "here is no distinction between primary and secondary concord % elmers 497;K472)47F*, and the system is !ery close to true Salliterati!eT a$reement, where the prefi3es found on the !arious a$reement tar$ets are identical both to each other and to the correspondin$ noun prefi3. "hus there is no need here for the dauntin$ tables of $ender a$reement typical of (antu descriptions. "he chapter is arran$ed as followsK L5.4 is concerned with the form of the a$reement prefi3es, especially the
=GF

and

=G9

allomorphs. Section 5.2 pro!ides a

comprehensi!e listin$ of all $ender a$reement tar$ets, both inside and outside the +1. Section 5.; deals with two related phenomenaK antecedentless a$reement and semantic a$reementG in both cases the prefi3 on the a$reement tar$et is determined by somethin$ other than syntactic properties of the controller. Section 5.8 in!esti$ates neutral a$reement, in other words a$reement tri$$ered by atypical controllers which do not ha!e their own $ender. Section 5.F deals with comple3 +1s which offer a choice of a$reement feature !alues on the tar$et, and finally L5.5 briefly mentions $ender resolution. =s with chapter F, this chapter is based on material pre!iously published in McGill %2007*. 2nce more the basic data is lar$ely unchan$ed, but the analysis is e3panded and impro!ed.

2.1 911
=G2,

Agreement prefix allomorphs 1owel harmon+#4, +#2,

/ecall the !owel harmony shown by the and


=G8

and

+#8

prefi3es %LF.F.2*. "he

=G4,

prefi3es beha!e in an identical way, with the !alue of the prefi3 !owel

bein$ determined by the !owels of the a$reement tar$et to which it is prefi3ed. "his is demonstrated below for the =G8 prefi3 on three different demonstrati!es.

279

%4* %a*

ma)0di ya ma0mpa)
+#8)hare A-4)this

this hare Inear speakerJ %b* ma)0di ya me0lle)


+#8)hare A-4)that

that hare Inear hearerJ %c* ma)0di ya mo0''u*


+#8)hare A-4)yonder

that hare I!ery far offJ

Ieamd0;2.0F9J

91#

AG(

allomorphs

"he =GF a$reement prefi3 has two allomorphsK

mi0 before stems be$innin$ with a !owel4 or a lon$ consonant, and sometimes before stems be$innin$ with a short consonant /0 sometimes before stems be$innin$ with a short consonant
+#F

"he same allomorphs are found for the sli$htly different. In the case of the
+#F

prefi3 %LF.F.F*, but the distribution is

prefi3, mi0 and /0 are in complementary

distribution, but here there is a choice before short consonants, as can be seen from the e3tract in %2* below. "he speaker uses the n)0 allomorph as the sub'ect prefi3 on 'ugo SfallT in the first intonation unit, but switches to mi0 for the second occurrence of the !erb. %2* m0uu ana)
/C%1c!ild

mi 0 na) mi , 1808anga) m0aya)


=GF)=/" =GFR+#9)tree

=GF)comeQ/6S

n)0'ugo) :

A-%)fallQ/6S

when

mi 0 ) 'ugo)0no)...

A-%)fallQ/6S)10E

then the ruit ell 1 when it ell...

Itapf002.008.028J

Similarly, the ne3t two e3amples show the two allomorphs before the ad'ecti!e 8ene*u SsmallT. %;* m0uu
+#F)child

mi 0 ) 8ene

A-%)small

mi 0 ) 8ene*u
A-%)small

young animals Ilit. Ssmall childrenTJ

Isamoh004.094J

4 "he usual !owel coalescence occurs %see L;.4.4*. See LF.F.7 for why, before !owel)initial stems, the prefi3 !owel is assumed to be i e!en thou$h it is always subsumed by the stem !owel.

279

%8*

m0uu

+#F)child

n)08ene*u

A-%)small

small children

Itats002.008.009J

=ll tar$ets be$innin$ with short consonants seem to offer both choices, althou$h mi0 is the more common.

91%

AG=

allomorphs
=G9

"he distribution of the correspondin$


+#9

allomorphs is perhaps e!en more comple3 than that of the


+#9

allomorphs %LF.F.7*. /ecall that the

noun prefi3 has three

allomorphsK either ;0 or -0 before short consonants, and vi0 before a !owel %+#9 roots do not seem to be$in with lon$ consonants*. In the case of the a$reement prefi3, the same three allomorphs occur, but this time vi0 has a wider distribution. +ot only does it occur before !owels and all lon$ consonants, it may also be found before short consonants. "hus before short consonants all three allomorphs are possible, and the main problem here is to describe the conditions under which each of them surfaces in this en!ironment. 0irst of all, it should be admitted that it is freDuently hard to decide whether a particular instance of a prefi3 is ;0 or -0. "e3tual analysis is complicated by the fact that in fast speech, the contrast between lon$ and short consonants is often unclear. +e!ertheless speakers ha!e a metalin$uistic awareness of these two possibilities, 'ust as for the noun prefi3, and this means it is possible to ha!e fruitful elicitation sessions on the topic. "he len$thenin$ option %-0* seems to be considered to be the SbestT style of #icipu, but I could find no e!idence of a diachronic shift from -0 to ;0. =part from stylistic factors, the choice of prefi3 is influenced both by the type of a$reement tar$et, and by the semantics of the noun tri$$erin$ the a$reement. 0or most tar$ets all three possibilities vi0, ;0 and -0 can occur, as shown below for numeralsK %F* %a* mu+0uwa) %b* mu+0uwa) %c* mu+0uwa)
4S)hearQ/6S IsoundJ

'o' 'o' 'o'

n0nosi )

A-))four

4S)hearQ/6S IsoundJ

;0nosi )

A-))four

vi )0nosi )
A-))four

4S)hearQ/6S IsoundJ

' heard our 'o's

I2007)04)40.040J 290

=$reement on numerals is in fact optional, so it is not clear whether %Fb* shows a null prefi3 or a lack of a$reement alto$ether. <owe!er there is less doubt about the associati!e construction %5*G since a$reement is obli$atory for e!ery other noun class, there is no reason to analyse it as optional for e3istence of the -0:;0 noun prefi3 alternation. %5* %a* 707a
+#9)person

=G9,

especially $i!en the independent

k1ku0yi ma)

A-)E+#9)smith

blacksmith Ilit. Sperson of smithin$TJ %b* 707a


+#9)person

;1ka)0ga)ski ya)
A-)E+#4)truth

truth ul person Ilit. Sperson of truthTJ %c* l0lama


+#9)sound

vi ,1ma0waa
A-)R+#8)do$

the sound o a dog

Ieamd045.;99, tats00F.004.427, tats002.004.059J

"he articleNrelati!iser 0na) on the other hand, allows only -0 %7a* or vi0 %7b*, and cannot occur without an o!ert prefi3 %7c*K %7* %a* \ ]
IletterJ

n0na)

A-))/-6 =G9)lackQ/6S)10E

;0li a)0na)

ku)0gi ya)

+#9)hook

bu without a hook Ii.e. not 5J %b* \ ]


IletterJ

vi 0 na)

A-))/-6 =G9)lackQ/6S)10E

;0li a)0na)

ku)0gi ya)

+#9)hook

bu without a hook Ii.e. not 5J %c* z\ ]


IletterJ

;0na)

A-))/-6 =G9)lackQ/6S)10E

;0li a)0na)

ku)0gi ya)

+#9)hook

I2007)04)40.040J "he distribution of n0na) and vi 0 na) in the corpus is not random, and they are in more or less complementary distribution. /0na) almost always occurs immediately after le3ical heads %90 tokens, one e3ception*, while vi 0na) mostly %42 tokens, one e3ception* occurs as a headless relati!e, as in %9*. "his situation has a parallel in <ausa, where headless relati!es reDuire the lon$er relati!iser wanda, and headed relati!e clauses are more likely to ha!e the short form da %+ewman 2000KF;8)F;5*.

294

%9*

vi 0na) ;0si 0 7i )no)

A-))/-6 =G9)<=()chan$e

v0m),

+#9)monkey

see u0$i )yo)0no)

then ;S)collectQI//)E-+"

the one Ik0kaa Swoman, +#9TJ who changed into a monkey, she collected Isaat004.008.048J = similar distributional pattern is found with sub'ect a$reement, althou$h with a different pair of allomorphs. 0or sub'ect a$reement there is no e!idence for -0 , only vi0 %9a* or ;0 %9b* can occur. .i0 is unrestricted in distribution, but ;0 may only occur utterance)medially. %9* %a* ;0hi )ti la)
+#9)li$ht

vi )0ri 5a)

A-))sinkQ/6S

the light aded %b* ;0hi )ti la)


+#9)li$ht

;0ri 5a)

A-))sinkQ/6S

the light aded Ieaim007.4890J "able 82 summarises what is known about the distribution of the with a sin$le short consonant*. 0able J4- &ossible AG` allomorphs according to target type "ar$et +umerals =d'ecti!es +e$ati!e copula =ssociati!e construction =rticleNrelati!iser Sub'ect marker .emonstrati!e ad!erbs ni nni *i SonlyT wh)words Section L5.2.2 L5.2.; L5.2.5 L5.2.9 L5.2.40 L5.2.4; L5.2.7 L5.2.8 L5.2.F -0 ;0 vi0 M M M M M M M M
=G9

allomorphs with

respect to tar$et type %recall that a choice is only possible for tar$ets whose stems be$in

=s well as the tar$et type, the animacy of the noun tri$$erin$ the a$reement is also a factor in determinin$ the form of the prefi3. "here is a correlation between nouns denotin$ humans and -0 or ;0, and between those denotin$ animals and inanimates and vi1. "his has been in!esti$ated more fully for sub'ect prefi3es and numerals, but the correlation is belie!ed to hold for the other tar$et types which allow a choice as well. 292

+ouns for which -0 or ;0 are preferred tend to ha!e human referentsK 707a SpersonT, va0ari ) SmanT, m0ma$i ) SfriendT. +ouns for which the vi0 a$reement prefi3 is preferred are mostly animals or inanimates, such as $0$') \sheepT, v0m) SmonkeyT, and ;0ri )koda) Saudio recorderT. K0kaa SwomanT seems to be somewhere in between. "he followin$ doublet, from a te3t about do$s, consists of two consecuti!e intonation units with a strikin$ parallel between the two different SpersonNmanT tri$$ers -02. %40* u)0lapa)
;S)knowQ/6S
=G9

allomorphs used in the second associati!e construction.


=G9

In %40* the noun k0kaa SwomanT tri$$ers the vi0 ku)08oo ku,1k0kaa

allomorph, while in %44* 707a ku0lle)

+#9)!oice

=G9R/C)1woman

vi ,1kw0a'a)

A-)E+#9)house

=G9)that

it knows the voice o the woman o that house %44* u)0lapa) ku)08oo ku,1707a k,1kw0a'a)

Itats004.004.094J ku0lle)

;S)knowQ/6S

+#9)!oice

=G9R/C)1"erson

A-)E+#9)house =G9)that

it knows the voice o the man o that house


+#9

Itats004.004.092J

a$reement is a comple3 topic, and $i!en the !ariation found e!en for indi!idual

nouns, as with the noun class prefi3es it does not seem appropriate to set up separate a$reement subclasses to deal with the separate prefi3es.

2.

Agreement targets

"here is a !ery lar$e number of a$reement tar$ets within the #icipu noun phrase, and predicates also a$ree in $ender with the sub'ect. "his a$reement takes the form of prefi3es or proclitics attached to the tar$et. 2ther than for class ; %LF.F.8*, we $enerally find a one)to)one correspondence between noun class prefi3es and a$reement prefi3es, althou$h tar$ets with !owel)initial stems are under)differentiatin$ %i.e. ha!in$ fewer than e3pected distinctions* due to the phonolo$ical similarity between classes 8 and F. 2!erall, howe!er, the a$reement system is hi$hly re$ular, with only one set of a$reement markers shared by all tar$ets. 2ne $eneral complication is that the tones of the !arious tar$ets: prefi3es !ary %see below*. In addition the a$reement prefi3 !owel for class 5 may be either ti0 or tu0, as was the case for the allomorphs of the
=GF +#5

noun prefi3 %LF.F.5*. "he

and =G9 a$reement prefi3es ha!e already been mentioned %L5.4*.

2 0or more on the effect of natural $ender on !ariation in a$reement see L9.8.F.;.

29;

In the subseDuent discussion on specific tar$ets, all the prefi3es follow the basic pattern set out below. 0able J6- Agreement pre i*es ound be ore consonantcinitial and vowelcinitial stems6 Shape #E) #) (efore 4 2 h0 ; y0 8 mA0 m0 F m0 5 t0 7 w0 9 v0 9 kw0 #)initial stem kA0 hA0 yi0 E)initial stem k0 mi0:/0 ti0:tu0 wu0 vi0:-0:;0 ku0

9#1

Prefix tone

"he !arious a$reement tar$ets, summarised below, can be di!ided into two main $roups accordin$ to the tone on their a$reement prefi3es. Anlike some (antu lan$ua$es %LF.F.;*, there are no class)dependent effects. =d'ecti!es, numerals, the Duantifier 0ni ni *i SonlyT, wh)words, the ne$ati!e copula, and the demonstrati!e ad!erbs all take low)tone prefi3es. "he associati!e construction, possessi!e pronouns %with the e3ception of 41S*, demonstrati!e modifiers, the articleNrelati!iser, and the modifier 0m ) SanotherT all take hi$h)tone prefi3es. Sub'ect prefi3es can appear with either < or 6 tone, accordin$ to the !erb:s mood %see L8.5.2*. "here is no tone associated with the prefi3 on pronouns or the copula, since they are !owel)initial and take the -0 series of prefi3es.

; =nderson:s %4990a* description of the noun class system of the -ast &ain'i lan$ua$e =mo refers to SfusedT !s. Snon)fusedT prefi3es. hile the #) prefi3es in #icipu are indeed fused with the roots, the resultant forms are entirely re$ular %with the e3ception of the len$th of the fused !owel, which can be short or lon$*, and so there is no need for the abstract representations and comple3 deri!ations proposed by =nderson for =mo.

298

0able JJ- Agreement targets (arget +umerals =d'ecti!es 0ni nni *i SonlyT wh)words +e$ati!e copula .emonstrati!e ad!erbs =ssociati!e construction 1ossessi!e pronouns =rticleNrelati!iser .emonstrati!e modifiers 0m ) SanotherT Sub'ect marker #opula 1ronouns Pre'i$ can be omittedH B, freDuently + B + + + + B, but rarely + + + + + + Pre'i$ tone 6 6 6 6 6 6 < < < < < < or 6 +one %#)* +one %#)* 0ection L5.2.2 L5.2.; L5.2.8 L5.2.F L5.2.5 L5.2.7 L5.2.9 L5.2.9 L5.2.40 L5.2.44 L5.2.42 L5.2.4; L5.2.48 L5.2.4F

0et 1

0et

"hese tar$et $roupin$s are essentially the same as in #entral &ambari %#ro>ier 4998K5F) 57*, althou$h the actual tones associated with the $roups are different. In #entral &ambari the tone on the prefi3 of set 4 tar$ets is polar %as opposed to 6 in #icipu*, and for set 2 tar$ets it is 6 %< in #icipu*.

9##

?umerals

=ttributi!e numeral phrases %L8.9.2* are formed by the sin$ular noun followed by a numeral with an optional low)tone a$reement prefi3, which may occur up to but not beyond the number 98. "he followin$ e3ample shows each numeral with =G4 a$reementK %42* +o. 4 2 ; 8 F 5 7 9 9 40 ka)0 ka)0 ka)0 ka)0 ka)0 ka)0 ka)0 ka)0 ka)0 ka)0 ara ara ara ara ara ara ara ara ara ara ko)0to) ka)0yapu) ka)0ta)atu) ko)0nosi ) * ka)0tau ka)0tori )hi ) ka)0ti nda)ya) ka)0ku)ri llo) ka)0ku)ti tti kuppa one old man two old men three old men our old men ive old men si* old men seven old men eight old men nine old men ten old men

Ieamy00F, eamy005J

8 "his differs from the #entral &ambari system, where only the numbers 4)7 take a$reement prefi3es %<offmann 495;K455*. "he +orthwest $roup of lan$ua$es, like #icipu, ha!e a$reement on 4)9 %Smith 2007K74 for At)ma:in, .ettweiler n.d. for #:6ela* , no information is a!ailable on the &amuku or /eshe branches of est &ain'i.

29F

In the followin$ list the numeral is kept constant %yapu) StwoT* and the class of the head noun is !ariedK %4;* #lass 4 2 ; 8 F 5 7 9 9 ka)0 ara a)07a i )0nama) ma)0di ya n)0'y'y $i )0kooto) u)0yaa $0$') ku)0la$i ka)0yapu) ha)0yapu) yi )0yapu) ma)0yapu) mi 0 ) yapu) ti 0 ) yapu) wu)0yapu) vi )0yapu) ku)0yapu) two elders two people %plural form of noun* two meats two hares two ish %plural form of noun* two drums two roads two sheep two girls

#lasses 2 and F do not occur with sin$ular nouns and so are not normally in!ol!ed in numeral phrases. Some speakers, howe!er, optionally use the plural form of the noun in numeral phrases, as in the e3amples abo!e. "his is especially true for nouns with human referents. In this respect #icipu patterns with <ausa %Ja$$ar 2004K;F9*, althou$h accordin$ to (ickel and +ichols %2007K24;* many lan$ua$es draw the line between animate or human referents and the rest, reDuirin$ InumberJ markin$ only for nouns referrin$ to human bein$s %see also 1ayne 4997K95, #orbett 2000K70*. #icipu differs from most &ain'i lan$ua$es, for which the plural form of the noun is reDuired with numbers s 4 %e.$. #ro>ier 4998K88 for #entral &ambari, Smith 2007K74 for At)ma:in, .ettweiler n.d. for #:6ela, Mac.onell 2007KFF for 1on$u, =nderson 4990aK452 for the -ast &ain'i lan$ua$e =mo*. "his seems to be the $eneral pattern in (antuNGrassfields (antu %e.$. elmers 497;K294 for &iswahili, and se!eral papers in <yman 4990*, and it is reasonable to suppose that the #icipu reDuirement for sin$ular nouns in numeral constructions is an effect of contact with <ausa. hen countin$, the prefi3 may be omitted. =lternati!ely, if the speaker has a particular ob'ect or set of ob'ects in mind, she may use the appropriate a$reement prefi3, re$ardless of whether the a$reement controller %i.e. the noun denotin$ that ob'ect* is present or not. 0or e3ample when countin$ days %kwa0a'a), +#9*, the a$reement prefi3 on the SunitsT fi$ureK %48* u)0kuppa n)
+#7)ten

=G9

prefi3 ku0 may

be used. 0or numbers from 44)49, 24)29, ;4);9 and so on, it is obli$atory to use an vi )0to)

and

A-))one

eleven Sometimes there is no ob!ious referent, for e3ample when $i!in$ a demonstration of the 295

countin$ system for the lin$uist. In such cases the

=G9

prefi3 vi0 is used with the units

fi$ure from 44 onwards. "his is a case of SneutralT a$reement %L5.8*.

9#%

Ad,ectives
ko)07uvu) o0ggo)m o) i )0nama) ma)0di ya n)0di ya $i )0kooto) u)0yaa $0$') ku)0la$i ke)0pene*u he)0pene*u yi )0pene*u me)0pene*u m) 0pene*u ti 0 ) pene*u wu)0pene*u vi )0pene*u ku)0pene*u

=d'ecti!es take an obli$atory low tone prefi3. %4F* 4 2 ; 8 F 5 7 9 9 big inger Ii.e. thumbJ big bats big meat big hare big hares big drum big road big sheep big girl

Ieamy00F, eamy005J

9#&

ni nni *i Eonl-F

"he word ni nni *i SonlyT is another $ender a$reement tar$etF, takin$ an optional low)tone prefi3. I ha!e only collected e3amples for classes 4, 8, 5, and 9 so far, but there is no reason to think it will not be as re$ular as the other tar$ets. %45* ku)0yu)yu*
+#9)sand

ku)0ni )nni *i
A-*)only

only sand

I"idipo, saat002.008.048J

9#(

wh.words

"he wh)words yi )ni SwhatT %47*, 0e)ne SwhichT %49*, and ya)anu Show manyT %49* can all function as noun modifiers, in which case they take a low)tone a$reement prefi3. =lthou$h ya)anu means SwhoT by itself %L8.;.5.2*, when an a$reement prefi3 is attached it becomes a noun modifier meanin$ Show manyT. %47* ka)0'a77i )ki
+#4)prosperity

ke)0yi )ni

A-11what

what prosperity Isaat002.002.;92J %49* a,1ka0kaasu)wa) ke)0ene


62#R+#4)market A-1)which

to which market

Isayb004.724J

F Anless it is considered to be an ad'ecti!e. "here are no formal reasons not to do this, only semantic ones.

297

%49*

) ka)0hi'i

+#4)ni$ht

ka)0ya)anu

A-1)howPmany

how many nights "he full paradi$m is $i!en below for 0e)neK %20* 4 2 ; 8 F 5 7 9 9 ke)0ene he)0ene ye)0ene me)0ene me)0ene te)0ene we)0ene ve)0ene kwe)0ene ka0 ara o0ggo)m o) i0 nama) ma0di ya n0di ya $i 0 kooto) u0yaa $0$') ku0la$i

Isaat002.002.52;J

which elder which bat which meat which hare which hares which drum which road which sheep which girl Ieamy00F, eamy005J

9#9

?egative copula

"he ne$ati!e copula consists of the ne$ation particle $e %also found in ne$ated !erbal clauses, L8.;.F*, with a low)tone a$reement prefi3. %24* 4 2 ; 8 F5 5 7 9 9 ka)0ri )ma*i a)sa* ho)0to) i )0naa ma)0waa ti iri mpa) u)0sei B0'a)googo ku)0roono) ka)0$e he)0$e yi )0$e me)0$e ti 0 ) $e wu)0$e vi )0$e ku)0$e it's not pleasure it's not one grave it's not cows it's not a dog it's not like this thing it's not pain it's not a watch its not a loincloth Itats002.009.04FJ I"ikula, sa$b004.708J Isayb004.789J Ieamd002.0;0J Isayb004.8;0J Itats002.009.045J Isaat004.00F.099J Isayb004.;F5J

9#<

$emonstrative adverbs

"he demonstrati!e ad!erbs a$ree in $ender with the sub'ect +1 when used as predicates %L8.;.;.2*. "he prefi3es are low)tone. %22* 4 2 ; 8 F 5 7 9 9 ka)0taari a)0taari i0 ) nama) ma)0$i )iji i n)0gai $i 0 ) k'u) u)0ree ;0vooto) ku)0sayu ka)0pa*a ha)0pa*a yi )0pa*a ma)0pa*a mi 0 ) pa*a ti 0 ) pa*a wu)0pa*a vi )0pa*a ku)0pa*a the stone is here the stones are here the meat is here the calabash is here the swords are here the chest is here Ii.e. body partJ the town is here the goat is here the spear is here I2007)04)48.004J

5 3i 0 ) $e has not been obser!ed or elicited yet but there is no reason to think it is missin$ from the paradi$m.

299

-ach of the fi!e ad!erbs follows the same pattern. "his is shown below for class ;, but the pattern is re$ular across all the classes. %2;* i )0taatu) i )0taatu) i )0taatu) i )0taatu) i )0taatu) yi 0 ) pa*a yi 0 ) le*e yi 0 ) 'i nde) yi 0 ) 'u* yi 0 ) 8o*o the mat is here %near speaker* the mat is there %near hearer* the mat is over there the mat is there %far away or out of si$ht* the mat is here %permanent place* I2007)04)48.004J

9#=

Associative construction
=G+14)+12,

"he associati!e construction %L8.8.F.4* takes the form +14 the a$reement proclitic perturbation %L;.8.7*. %28* 4 2 ; 8 F 5 7 9 9 ka)0dama a)0dama i0 ) nama) ma)0di ya n)0di ya $i 0 ) kooto) u)0yaa $0$') ku)0davu) kaKEk0kaa haKEk0kaa yi KEk0kaa maKEk0kaa mi K Ek0kaa ti K Ek0kaa wuKEk0kaa vi KEk0kaa kuKEk0kaa

where +14 is the

SpossessedT noun and +12 the SpossessorT noun. +14 occurs in its citation tone pattern,
=G+14

is hi$h tone, while +12 under$oes a comple3 tonal

the word o the woman the words o the woman the meat o the woman the hare o the woman the hares o the woman the drum o the woman the road o the woman the sheep o the woman the mortar o the woman

Ieamd0;2J

hen the associati!e construction is considered as a whole, it is usually the first +1 %the SheadT in the sense that it is the semantic determinant of the whole* that tri$$ers a$reement on e3ternal tar$ets, but this is not always the case %see L5.F*.

9#>

Possessive pronouns
ka)0'i ngu)wa) ko0ttu)
+#4)!illa$e

"he possessi!e pronouns %L8.8.F.4.4* take a$reement prefi3es, usually with < tone. %2F*
A-1)41.12SS

our village "he full paradi$m is shown below, repeated from L8.8.F.4.4.

Is!b$004.0;0J

299

0able JT- &ossessive pronoun paradigm with gender agreement pre i*es 4 2 ; 8 F 5 7 9 9 4S ko0vo*o ho0vo*o yi0vo*o mo0vo*o mi0vo*o or n0vo*o ti0vo*o wu0voo* vi0vo*o or v0vo*o ku0vo*o 2S ka0avu) ha0avu) ya0avu) ma0avu) ma0avu) ta0avu) wa0avu) va0avu) kwa0avu) ;S ke0evi ) he0evi ) ye0evi ) me0evi ) me0evi ) te0evi ) we0evi ) ve0evi ) kwe0evi ) 41 ko0ttu) ho0ttu) yi 0ttu) mo0ttu) mi 0 ttu) ti 0 ttu) wu0ttu) vi 0ttu) ku0ttu) 21 ko088 o) ho088 o) yi 088 o) mo088 o) mi 0 88 o) ti 0 88 o) wu088 o) vi 088 o) ku088 o) ;1 ki 0ive) hi 0ive) yi 0ive) mi 0 ive) mi 0 ive) ti 0 ive) wi 0ive) vi 0ive) kwi 0ive)

"he a$reement markers ha!e been obser!ed omitted, but this is rare. %25* ) a0si'i
+#2)feather

ha0mpa) vo*o
=G2)this

10.P600

these my eathers

Isaat004.005.4F5J

9 # 1I

ArticleCrelativiser

=s discussed in L8.8.F.;, the same form 0na) is used for the article and the relati!iser. "he article 0na) can occur either before or after the noun head , in both cases it takes an obli$atory hi$h)tone a$reement prefi3. %27* 4 2 ; 8 F 5 7 9 9 ka)0dama o0ggo)m o) i0 ) nama) ma)0di ya n)0di ya $i 0 ) kooto) u)0yaa $0$') ku)0la$i ka0na) ha0na) yi 0na) ma0na) mi 0 na) N n0na) ti 0 na) wu0na) vi 0na) N n0na) ku0na) the elder the bats the meat the hare the hares the drum the road the sheep the man
=GF

Ieamd0;2J and =G9 prefi3es can lead to these


=G9

=s can be seen in the list abo!e, the !ariation in allomorphs.

classes bein$ undifferentiated. See L5.4.; for the distribution of the relati!iser

"he followin$ e3ample illustrates a$reement on the relati!iser, which is a$ain obli$atory. %29* i 0 ) ri
+#;)thin$

yi 0na)

A-3)/-6

yi 0 ) la a)0na)

=G;)lackQ/6S)10E

yi )0'etei

=G;)fine

bad things Ilit Sthin$s which lack finenessTJ Ioamy004.429J

290

9 # 11

$emonstrative modifier

.emonstrati!es modifiers %L8.8.F.2* take a hi$h)tone prefi3, which is $enerally obli$atory7. %29* ka0aya ka0aya ka0aya ko)0dondo kwa'a) ka0mpa) ke0lle) ke0''i nde) ko0''u* ku088o* this hut %near to speaker* that hut %near to hearer* yonder hut %far from both* that garden %!ery far away or out of !ision* this house %permanent place*

"he a$reement paradi$m is fully re$ular but is presented below for completeness. 0able J]- eemonstrative modi iers with gender agreement pre i*es 4 2 ; 8 F 5 7 9 9 this ka0mpa) ha0mpa) yi 0mpa) ma0mpa) mi 0 mpa) ti 0 mpa) wu0mpa) vi 0mpa) ku0mpa) that ke0lle) he0lle) yi 0lle) me0lle) mi 0 lle) ti 0 lle) wu0lle) vi 0lle) ku0lle) yonder ke0''i )nde) he0''i )nde) yi 0''i )nde) me0''i )nde) mi 0 ''i )nde) ti 0 ''i )nde) wu0''i )nde) vi 0''i )nde) ku0''i )nde) out of si$ht ko0''u* ho0''u* yi 0''u* mo0''u* mi 0 ''u* ti 0 ''u* wu0''u* vi 0''u* ku0''u* this %perm.* ko088 o* ho088 o* yi 088 o* mo088 o* mi 0 88 o* ti 0 88 o* wu088 o* vi 088 o* ku088 o*

9 # 1#

0m ) EanotherF
ka)0 ara a)0 ara i )0nama) ma)0di ya n)0di ya $i )0kooto) u)0yaa $0$') ku)0la$i k0m ) h0m ) yi 0m ) m0m ) mi 0 m ) ti 0 m ) wu0m ) vi 0m ) ku0m )

"he modifier 0m ) %L8.8.F.;* may be $lossed as SanotherT. It takes a hi$h)tone prefi3. %;0* 4 2 ; 8 F 5 7 9 9 another elder other elders another meat another hare other hares another drum another road another sheep another girl

Ieamy00F, eamy005J

9 # 1%

0ub,ect

Eerbs are obli$atorily prefi3ed by either a $ender or person marker. "he $ender a$reement prefi3es are of #E) shape, unless the !erb stem is !owel)initial in which case they ha!e the #) forms, with the usual prefi3Nroot !owel coalescence %L;.4.4*. "he tone depends on the mood of the !erb %L8.5.2*. Sub'ect a$reement is dealt with in more detail
7 "he =G9 Snear speakerT form is sometimes lle) rather than vi 0lle). "his could be analysed as the ;0 allomorph of the =G9 prefi3, althou$h normally that is only possible before short consonants. =lternati!ely, it could be considered as lackin$ a$reement.

294

in L7.8)7.5. %;4* ka)0 ara


+#4)oldPman A-1)passQ/6S)E-+"

ka)0'wa*a0na)

an old man passed by

Itapf002.004.0;0J

9 # 1&

Copula

"he copula consists of either an IeJ or an IiJ !owel, with a $ender a$reement prefi3 of the #) form, but without the lon$ !owel that usually results from coalescence between prefi3 and stem !owels. "he IeJ !owel is found in classes 4, 2, and 8, the prefi3es of which contain the harmonisin$ NAN !owel %L5.4.4*. It may therefore be analysed as the product of assimilation between the prefi3 !owel NAN and an underlyin$ NiN in the copula %cf. L;.7.4*. "he tone on the copula itself is usually polar, but as discussed in L;.8.8 there are e3ceptions. %;2* 4 2 ; 8 F 5 7 9 9 ko0ggo)m o) o0ggo)m o) i )0nama) ma)0di ya n)0di ya $i )0kooto) u)0yaa $0$') ku)0la$i k0e h0e y0i m0e) m0i ) t0i w0i ) v0i kw0i ) it's a bat it's bats it's meat it's a hare it's hares it's a drum it's a road it's a sheep it's a girl

Ieamy00F, eamy005J

=s noted in L8.;.;.4, in predicate nominal constructions the copula a$rees with the predicate rather than the sub'ect, an instance of SbackT a$reement or SattractionT.

9 # 1(

?oun class pronouns

+oun class pronouns take the same form as the copula 9, but this time the tone depends on the syntactic position of the tar$et rather than the phonolo$ical en!ironment. In the complement position followin$ !erbs or the locati!e proclitic A they occur with low tone as in %;;*G otherwise they take hi$h tone as in %;8*. %;;* n)07uwa)
4S)roastQ/6S

o0`i no,

+#2)$roundnut

m0i +nda)

4S)seeQ/6S

h0e)
A-

1P#6

$e
+-G

' roasted groundnuts, ' didn't see them

Isaat004.002.00;J

9 See Stassen %4997K77)9F* on Spro)copulasT.

292

%;8*

n n1a)0ka a)0na)
if

andR;1)takeQ/6S)E-+"

a)0ra)kumi a)0dukwa) n)
+#2)camel

;1)$oQ/6S with

h0e
A-

1P#6

i they took camels they went with them

Isamoh004.499J

1redicate locati!e constructions %L8.;.;.2* in!ol!e what appears to be a $ender)marked pronoun followed by the copula. %;F* k0i
A-1)1/2 A-1)#21

k0e)

pa*a

here

it Ie.$. ka)0taari Sthe stone, +#4TJ is here

I2007)04)48.004J

.espite the 1/2 $loss 'ust $i!en, the cate$ory of the first word is not totally clear, since I ha!e not found these precise forms in any other conte3t. "hey differ from strai$htforward noun class pronouns by retainin$ the i !owel throu$hout the paradi$m. "he reason for $lossin$ them as pronouns is that the structure is analo$ous to the personal pronoun predicate locati!e %e.$. i )v vi ) pa*a Syou%s$.* are hereT , L8.;.;.2*, with respect to both meanin$ and the < 6 tone pattern. %;5* 4 2 ; 8 F 5 7 9 9 k0i h0i y0i m0i m0i t0i w0i v0i kw0i k0e) h0e) y0i ) m0e) m0i ) t0i ) w0i ) v0i ) kw0i ) pa*a pa*a pa*a pa*a pa*a pa*a pa*a pa*a pa*a it is here Ie.$. batJ they are here Ie.$. batsJ it is here Ie.$. meatJ it is here Ie.$. hareJ they are here Ie.$. haresJ it is here Ie.$. drumJ it is here Ie.$. roadJ it is here Ie.$. sheepJ it is here Ie.$. $irlJ9

I2007)04)48.004J

2." Antecedentless agreement morphology and semantic agreement


In L2.2.4 I discussed the possibility of the use of a$reement morpholo$y in the absence of an a$reement controller, and noted that this does not Dualify as Sa$reementT under Steele:s definition of a$reement adopted by #orbett and others. Instead the speaker is relyin$ on the hearer makin$ an inference as to the appropriate referent, based on the use of a particular a$reement prefi3 in a particular conte3t. "he rele!ant conte3t may be te3t)e*ternal, for e3ample pointin$ to a stone while sayin$ ka0mpa) S=G4)thisT. =lternati!ely the referent may be inferred from the te3t %an Sindirect anaphorT accordin$ to Schwar>)0riesel 2007*. -3ample %;7* below is of this kind. Speaker ( assumes that,
9 "he =G9 construction is sometimes pronounced ku ku).

29;

$i!en the con!ersation is about $ra!e sites and burial, speaker = will be able to make the inference that the an Moolu. %;7* I#onte3tK Speaker =K 0he one who had the title "an Iooluu, /kula hill hasn't seen his grave yet.J (K to)
2& ;1)<=()take)/-.A1
=G4

indirect anaphor k0e denotes the corpse %ka)0kkwai,

+#4*

of

a)0si 0 ) ka a)0ka a)

o0gi )tu)

;1)$oPbackQI// with

n)

k0el

A-11P#6

3 so they are taking it I an Moolu:s corpseJ back I"ikula, sa$b004.75;J "he e3istence of the
+#4

noun ka)0kkwai ScorpseT in the le3icon was sufficient, in that

particular conte3t, to ensure coherence. It is important to distin$uish this usa$e of a$reement morpholo$y from semantic a$reement. Semantic a$reement %L2.2.;* in!ol!es the tar$et a$reein$ with semantic properties of the controller %e.$. humanness*. "he form of the a$reement morpholo$y used in %;7* was, in contrast, determined by reference to purely formal properties of another word, one that was entirely absent from the discourse. If the semantic structure of $ender 4N2 was more coherent, it mi$ht be possible to talk of the
=G4

morpholo$y

displayed by k0e in %;7* bein$ determined by the semantics of the $ender. =s thin$s are, it seems more appropriate to assume that the antecedentless a$reement is, in this case, made possible because of the prominence of the word ka)0kkwai in the parti$ular $onteMt, rather than any semantic associations of $ender 4N2. "here are times, howe!er, when semantics does seem to play a role, particularly in the case of the less incoherent $enders, and when the concept in!ol!ed is more abstract than, say,
#2/1S-.

0or e3ample

=G5

a$reement morpholo$y is freDuently used to e3press noun that can be appealed to as an e3planation for

manner, but there is no specific

+#5

this. Instead, it seems that the $ender itself has certain semantic associations that can be in!oked by the use of the correspondin$ a$reement morpholo$y. "here are also intermediate cases , in the case of
=G7

a$reement indicatin$ a place, it is not clear

whether this is possible because of the e3istence of the sin$le salient $ender 7 noun 'asu) SplaceT, or because of the fact that $ender 7N9 contains a number of words for places and therefore, to some e3tent, is semantically)inte$rated by the notion of
16=#-

%see LF.;.8

and LF.8.4 for +#7 locati!e deri!ations*. It seems likely that both these facts are rele!ant , after all, they are not independent, and a co$niti!e lin$uistics approach to $ender structure %see LF.2* mi$ht see the association with 298
16=#-

as deri!ed from a semantic

network inte$rated to a stron$ de$ree by the specific noun 'asu). "he $ender with the most semantic coherence40 is 9N2, which contains only nouns denotin$ humans or spirits. "here e3ists at least one Shybrid nounT ma0ga)ji ) Spriest, 8NFT which tri$$ers either 8NF or 9N2 a$reement. "his can be seen as e3emplifyin$ the e3treme44 end of this continuum, where the semantic structure of the $ender 9N2 is of the most importance in allowin$ the reference to be made. "he e3amples that follow are arran$ed rou$hly alon$ this continuum, startin$ with e3amples like %;7* which rely hea!ily on a sin$le word, and endin$ with a discussion of the hybrid noun ma0ga)ji , ) which relies on the coherent semantic structure of the 9N2 $ender. "he first e3ample is similar to %;7*, but instead of an antecedentless pronoun it in!ol!es an antecedentless demonstrati!e mi 0 mpa). "he topic of the discussion from which %;9* was taken was clothin$ worn in the past. "he only te3tual antecedent that would fit mi 0 mpa) semantically is the $eneric word for clothin$ i 0 ) hya$i %+#;*, but this does not fit $rammatically, and there is no semantic reason to use Instead reference by means of the antecedentless
=GF =GF

a$reement here.

demonstrati!e mi 0 mpa) is made

possible by the e3istence in the le3icon of the word me)0ttegu)Nn)0ti 0 teggu) SshirtNs, 8NFT, 'ust as in the case of ka)0kkwai ScorpseT and the antecedentless =G4 pronoun k0e. %;9* I#onte3tK discussion of clothin$ , but it wasn't these sort o clothes 1 they were something 1J mi 0 mpa)
A-%)this

mi 1o0yoyu

=GFR+#2)shirtPk.o.

mi 0mpa)
=GF)this

these In)0ti 0 teggu) Sshirts, 8NFTJ o oyoyuu these

Isayb004.729J

1erhaps more interestin$ are cases of antecedentless a$reement in!ol!in$ more abstract concepts, which conseDuently can be used in less restricted conte3ts, despite the fact that they still depend on the e3istence of indi!idual words rather than the semantic coherence of a $ender. 2ne such case is the use of
=G2

morpholo$y when names are

in!ol!ed. /ecall from LF.2.9 that the only noun in sin$le class $ender 2 is a)0hula SnameT, which of course cuts down potential referents and makes antecedentless a$reement less problematic42. "he followin$ e3tract is especially interestin$ because of
40 (arrin$, of course, inDuorate $enders containin$ only a sin$le noun. 44 ith respect to #icipu at least. Strict semantic a$reement as defined by #orbett %4994* does not occur , see the discussion at the end of this section. 42 "here are of course many other +#2 word forms, but they are all plurals, and often the conte3t will

29F

the !ariety of e3pressions %shown in bold* that are used to refer to the item of clothin$. %;9* I#onte3tK = family are tryin$ to remember the name of a kind of clothin$.J BK i 0 ) ri
/C31t!ing

yi 0 na)

a)0si 0 yu*u

=G;)/-6

a)0kamu)kwa) ;0sa'a)
;1)be.1S"Q/6S

+#9)time

vi 0lle)

=G9)that

a)0va8i

+#2)Eai

;1)<=()wear ;S)/-1/"

w0a)

ka)0ka)5a)li

+#4)leafPco!erin$

ko+ol
or

the things that they were that time the Avai are wearing they are called kakaali or "K o)'i i
yes

yes MK a)maa a)si )'mi v ) i) a)maa a)0si 0 ) 'm)1vi )


but

;1)<=()callR30.P#6

yi )ni n) yi )ni n)

what with

-i )$i pu)l -i )0$i pu)

+#5)#ipu

but what do they call it in 5icipu "K y0i


A-31P#6

y0i )

=G;)#21

$ele) ka)0ka)5a)li
that

+#4)leafPco!erin$

that's it kakaali BK h0i


A-

1P#6

h0e)

=G2)#21

le*e

there definitely

hwa

a1$i 0$i pu) ka)0ka)5a)li


62#R+#5)#ipu

+#4)leafPco!erin$

that's it de initely in 5icipu kakaali

Isayb004.;8FJ

"he e3cerpt is part of a wider discourse on clothin$, and at this moment the item of clothin$ in Duestion is the main discourse topic. "his e3plains M:s use of the person) marked ob'ect clitic vi ) in his contribution4;. ":s
=G;

pronoun y0i may be considered to

be strai$htforward %albeit cross)turn* syntactic48 a$reement with the i )0ri Sthin$T in B:s first utterance. "he most interestin$ form with respect to the present discussion is the
=G2

pronoun h0i in the final utterance. =ccordin$ to my lan$ua$e consultant, the hile the
=G;

=G2

form was used because the reference is to a name. pronoun y0i in the abo!e e3ample can be analysed as syntactic a$reement, it is also possible to !iew it as deictic reference %see L2.2.4 on the blurrin$ of the boundary between the two phenomena*. In support of the latter analysis, there are also unambi$uous cases of antecedentless =G; a$reement, which rely on the e3istence of
strai$htforwardly rule them out. 4; See L9.F.4.; for discussion of this e3ample with respect to discourse topicality. 48 In #orbett:s sense of bein$ determined by morphosyntactic properties of the controller.

295

the 9N; noun i )0ri Sthin$T4FK %80* I#onte3tK discussion of methods for trappin$ fish. there is hunting o a)0si 8a trees Ika)0si 8a Stree, k.o., 4N2TJ. 'n the past they would split split split Ii.e. the branchesJ, then they threw in the waterJ ee, to) n1a)0tu*u y0i ),
yes 2& whenR;1)pourQ/6S
A-31P#6

yes, 3 when they poured them <ere the


=G;

Itats00F.002.44FJ

pronoun y0i ) is the first e3plicit mention of the broken pieces of branches, hen
=G;

althou$h their e3istence can easily be inferred from the pre!ious sentence. Duestioned, my consultant said that were i )0ri Sthin$sT.

a$reement was used because the broken pieces

"he most common e3ploitation of antecedentless a$reement morpholo$y in the corpus is the use of
=G4

a$reement to refer to words, propositions, speech, and hile the word for SwordT,

lan$ua$es, and this will be treated at some len$th here.

ka)0dama %4N2* is no doubt stron$ly responsible for this possibility %it is the word most freDuently cited in e3planations by consultants*, the association is likely to be re) inforced by the e3istence of other 4N2 words with related meanin$s, such as ka)0na) ayi ) Snews, accountT, ko)0mi so)oni * SstoryT, and ke)0re'e) Slan$ua$eT. +e!ertheless the si>e of $ender 4N2 is so bi$ and its structure so incoherent that it is probably still better to !iew the links as links to a set of words, rather than to some abstract buildin$ up to reference to entire lan$ua$es or ways of speakin$. In %84* below the speaker uses the
=G4 2/.

concept. In what

follows I $i!e a number of e3amples startin$ from reference to a sin$le word and numeral ko)0to) with the meanin$ Sone

wordT. +owhere in the surroundin$ discourse is there a suitable antecedent %or e!en a SpostcedentT* for this indirect anaphor , howe!er the selectional restrictions of the !erb hyaa SsayT are a stron$ constraint on possible referents, and there seems little room for confusion. Ka)0dama SwordT was offered without hesitation when I asked for the reason for usin$ an =G4 prefi3. %84* u)hyii* * u)0hyaa1i ) $e $e
+-G

;S)sayQ/6SR;S.1/2

saa saa

e!en

ko)to) ko)0to)

A-11one

he didn't say one IwordJ to him


4F hile it is possible to elicit the for sin$le entities.
+#9

Isaat002.002.28FJ
+#;

Ssin$ularT r0ri , the

SpluralT i )0ri is much more common, e!en

297

=s well as actual words,

=G4

morpholo$y can be used to refer to conceptual structures,


=G4

in particular propositions. =$ain speakers consistently offer the e3istence of ka)0dama SwordT as the e3planation for the use of morpholo$y in such e3amples. In the followin$ e3ample taken from a son$, the demonstrati!e ke0lle) SthatT cataphorically links to the proposition e3pressed in the followin$ sentence. %82* ke0lle)
A-11t!at

ma)0huu m0e)K
+#8)truth

=G8)#21

Wan vi 1i0d*
6ord

=G9R+#;)horse

naha)

lea!eQ/6S

u0ree

+#7)town

* u)0yaa

;S)doQ/6S

n)0haalu

+#F)'ourney

this is the truth- "an .iid le t the town he went on a %ourney Io!k>002.020.004J "he followin$ three e3amples are similar, e3cept the reference is anaphoric and by means of a pronoun. "he collocation k0i k0e) n0na) shown in %8F* is a fairly common con!ersational response, and is unusual because of the
=G9

a$reement on the article.

"his seems to be a form of neutral a$reement %L5.8* with the non)!erbal clause k0i k0e), althou$h why this should reDuire the article is unclear , perhaps because the speaker is stressin$ that the proposition which has 'ust been !oiced is one familiar to him. %8;* I#onte3tK news o the Acipu has reached the white man's land. He ollowed the traces and thought about the land rom ar o ...J k0i
A-11P#6

k0e)

A-1)#21

w0a+ya)0na)

;S)comeQ/6S)10E here

pa*a

that's why he came here %88* kada) a0pa)nda)


1/2<

I"ikula, sa$b004.099J
+#4)head

;1)for$etQI//

-i )0$i pu) : ka)0tii AG


+#5)#ipu

ke0'e)esu)

=G4)addQI//

wu0uto)0no) :

+#7)$oPout)E-+"

k0i

A-11P#6 A-1)#21

k0e)

ka0aya)0wa)0na)1tu)

A-1)comeQ/6S)=116)10ER41.1/2

paa0ni )

here)+M6H

may 5icipu not be orgotten, let knowledge increase Ilit. Slet heads increase comin$ outTJ, that s what brought us here I"ikula, s!sdt004.408J

45 "he +#4 noun ka)0tii SheadT is unrelated to the =G4 a$reement markers in the followin$ intonation unit.

299

%8F*

=K saa hi na) a)ri *i saa hi na) a)0ra*a1i )


or
"21

;1)eatQ/6SR;S.1/2 or

saa mo)doo me) saa mo)0doo m0e)


+#8)sla!e

=G8)#21

either they de eated him or he's a slave (K k0i


A-11P#6 A-1)#21

k0e)

n0na)

A-))=/"

that's right#

I"ikula, sami004.8FFJ

"he ne3t e3ample in!ol!es the use of the =G4 relati!iser. "he speaker is emphasisin$ his family:s status within Mataari !illa$e by e3plainin$ that no decision can be made without their consent. "he <ausa translation I was $i!en for %85* was see abin da mun yadda da shi Sonly a t!ing with which we a$reeT. "he nine choices of a$reement prefi3 on the #icipu relati!iser $i!e the speaker the opportunity to be more specific about the referent, but without ha!in$ to mention the actual word ka)0dama SwordT. %85* ka'a) le*e
now there

3a)0taari see
+#8)IplaceJ

unless

ka0na)

A-11#7;

ti )0yadda)0na) :

41)a$reeQ/6S)10E

now there Iataari, only that Ii.e. a wordJ with which we agree 1 I' we speak, and the others don't agree, they won't be able to deprive us o our proper share.J Is!b$004.0;2J "he final SpropositionalT e3ample is notable because unlike the pre!ious e3amples, the immediate lin$uistic conte3t seems to be of little use for determinin$ the reference of the =G4 morpholo$y. It comes from a folktale in!ol!in$ fi!e superheroes, each of whom claimed to ha!e the best powers. =fter two of them had demonstrated their skills, the assembled thron$ inform them that they ha!e li!ed up to their wordK %87* to)
2&

i )8o

21.1/2

h0i nda)

;1)seeQ/6S

ko088 o)

A-11 P.P600

3 you .pl.2 yours IwordJ has been seen Ilit Syou %pl.* they saw yoursTJ I"idipo, saat002.00F.08;J "he point is that this sentence comes after a 90)second account of fe!erish superpower acti!ity durin$ which no SwordsT were mentioned , the crowd is referrin$ back to the heroes: ori$inal claims. "his is an ar$ument for analysin$ co!ered by the basic)le!el term ka)0dama. =s well as pointin$ to specific propositions, =G4 morpholo$y can also ha!e a more
=G4

morpholo$y as ha!in$ a

Sdefault interpretationT %in the absence of an antecedent* as referrin$ to somethin$

299

$eneral interpretation as Sknowled$e aboutT, and in the followin$ e3amples the no specific word, statement, or e!en proposition rele!ant to these e3amples. %89* I#onte3tK discussion of u)0laa \fire, +#7TJ saa ya)anu 0lapa)
or who
=G9)knowQ/6S

=G4)

marked associati!e constructions ha!e been translated usin$ about in each case. "here is

n)

with

ke1w0i )

A-1E+#7)1/2

everyone knows about it %89*

Itats002.002.044J

I#onte3tK speaker = says he doesn:t know about the foundin$ of &orisino. Speaker ( checks...J vu0u0la)pa)
2S)0A")knowQI//

$e
+-G

ke0lle)l

A-1)that

you wouldn't know about that

Isayb004.48FJ

=s usual, the e3istence of ka)0dama supplies the e3planation , and this analysis is supported by the parallel construction in %F0* which has an e3plicit antecedent and apparently identical meanin$. %F0* I#onte3tK the speaker is Sbutterin$ upT the hearer, whom he wants to start talkin$ about the traditional reli$ionJ 'i na)
some

707a

+#9)person

8a*a

surpassQ/6S some

'i na) 707a

+#9)person

u)0guya
+#7)can

ka)0dama ko1ko0ri nno)


/C11word

A-1E+#4)traditionalPreli$ion

some people know more than others about the traditional religion I"ikula, sa$b004.;0;J =ntecedentless
=G4

a$reement can also be employed for the reification of an entire

discourse durin$ its closure. -3ample %F4* came from the end of a retellin$ of the 1ear Story %L4.8.4*, and the referent of k0i seems to be the entire te3tual span of the narrati!e, or possibly the correspondin$ conceptual SstoryT. "his time consultants were less clear about the reason for the
=G4

markers , ka)0na) ayi ) SaccountT, ko)0mi so)oni *

SstoryT, and ka)0dama SwordT were all offered. It is Duite possible that the multiple possibilities reinforce rather than conflict with each other as potential referents. %F4* k0i
A-11P#6

k0e)
A-1)#21

le*e
there

k0aya)
A-1)comeQ/6S

k)0kt)
A-1)finishQ/6S

that's that it's inished Itapf004.008.02FJ

;00

Mo!in$ away from specific speech acts or e!ents, antecedentless

=G4

morpholo$y can
=G4

also be used to refer to actual lan$ua$es or ways of speakin$ %i.e. dialectsNidiolects*. Ka)0dama can mean SspeechT or SdialectT as well as SwordT and may be what the discussion of the different dialects of #icipu. %F2* "i )07o)ori yo) da) ,
+#5)Hooriyo

morpholo$y is pointin$ to in the followin$ two e3amples. "he first comes from a "i )0ddi )m* kuma ki 0ive)
+#5).odimo

da)
different

different

and

A-1);1.12SS

IaHarko dialect is di erent, 3adedan also theirs ISspeechTJ is di erent Ieadt004.2;2J "he second e3ample is $i!en mostly in -n$lish for the reader:s con!enience. "he important point to note here is the pro$ression from intonation unit. %F;* a. GrandsonK b. GrandmotherK c. d. e. f. $. h. i. '. and Cici"u%+#5* nowadays, people are not doin$ well N yes N A-+)not the same as that A-+)of)before N you %s$.* now the Cici"u%+#5* people N A-+)of)before A-+)that they would speak N if you spoke A-+)it now N children wouldn:t know N the meanin$ N they wouldn:t know the meanin$ N ka,1k0ka'a) e)re)
A-1E+#9)now

=G5

a$reement to

=G4

in the final

;1.1/2 until ;1)hearQI//

see

hu*0uwa)

k0e)

A-11P#6

dimly

i ri i ri

IlanguageJ of now they can hardly speak it

Is!tm$004.097J

"o the e3tent that we can consider the =G4 a$reement in %F;* to be SsemanticT a$reement controlled by the word -i 0 ) $i pu), the pro$ression from +1 to
=G5

to

=G4

%on both

pronouns and the associati!e construction* fits in with #orbett:s %4994K280* prediction that for a $i!en tar$et, semantic a$reement will not be found closer to the controller than syntactic a$reement47. "he e3amples up to now ha!e all been stron$ly linked to particular words, e!en if it is ar$uable that there is some semantic connection between $enders 4 and 4N2 and the idea of words or lan$ua$e. 0or the remainin$ e3amples the semantic link is stron$er. "hey in!ol!e =G5, =G7, and =G9 morpholo$y, and we will consider each of these in turn.
47 Immediately after the utterances in this e3tract, howe!er, a third speaker asks the old woman hu*uwa) t0i ) ana) ti )kg kool they speak A-+)it like <ausaMT. So #orbett:s prediction only holds within the sin$le discourse turn of the $randmother in this instance.

;04

e saw in LF.;.; and LF.;.F that there is an association between nouns from $ender 5 and SmannerT. "his association is also apparent in the use of antecedentless =G5 a$reement morpholo$y. In particular the wh)word 0e)ne %L8.;.5.2* and the relati!iser 0na) %L8.;.8* are both found with =G5 prefi3es in words meanin$ ShowTK %F8* t0e)ne t0i ) ka)0 i ki ) a)0si 0 ) yaa
;1)<=()do
+#4)festi!al

kaR'a0kkwi i
=G4R+#2)deadPperson

A-+1w!ic! =G5)#21

how do they do the estival o dead people I"ikula, sa$b004.842J %FF* ti 0 na) ;0maasii) si 0 ) suma a,1i 08aa
<=()run 62#R+#;)$round

A-+1#7; +#9)motorbike

how Ii.e. the way in whichJ a motorbike runs along the ground Itats002.005.0;9J Similarly, the SprepositionT ti %L8.F* may actually 'ust be an
=G5

associati!e a$reement

prefi3. "here is no $ender 5 word meanin$ SmannerT that has been omitted in these e3amplesG instead, it seems to be the $ender e3ponents themsel!es that are contributin$ that meanin$. /ecall from chapter F %LF.2.8, LF.;.8, LF.8.4* that $enders 7 and 7N9 are associated with places. "his association has also been obser!ed in the case of antecedentless a$reement morpholo$y. In the followin$ e3ample the use of
=G7 =G7

markin$ on the

pronoun e!okes a place , in particular, the place where #aari celebrates the festi!al. %F5* I#onte3tK discussion of the $roupin$s in which people $ather for a festi!alJ * -aari ), to) e esi ) w0i w0i ) le*e 'yau ga) a8aya ga)nga)me
InameJ 2& ;S.1/2 all
A-7)1/2 =G7)#21

there 4S.1/2 to$ether

to$ether

5aari, 3 him all it Ii.e. #aari:s placeJ is there together with me I"ikula, sami004.492J In the followin$ e3ample the sub'ect 3a)0ppaya is clearly the antecedent, and so this is perhaps better thou$ht of as semantic sub'ect a$reement49. %F7* le*e
there

3a)0ppaya wu)0yo*
+#8)InameJ A-7)beQ/6S

ana)
like

ma)0ppata
+#8)ca!e

there Iappaya is like a small cave I"ikula49, sa$b004.589J e saw in LF.8.2 that
+#9

de!erbal nominalisations denote repeated action, and 'ust as

49 =s the only e3ample of semantic =G7 a$reement obser!ed, it is su$$esti!e to find predicate a$reement $i!en #orbett:s %4994K chp. 9* =$reement <ierarchy. (ut until more data is collected nothin$ more definite can be said. 49 It is not yet known whether this phenomenon is found in "irisino.

;02

for

+#5

and

+#7,

we find a refle3 in the a$reement system. "wo different kinds ha!e


=G9

been obser!ed. 0irst, when the

prefi3 ku)0 occurs on a numeral in the absence of a

controller it usually means StimesT %e.$. he ell over three times*, as obser!ed in L8.9.820. Secondly, it seems to be possible to e!oke a reified repetition usin$ =G9 morpholo$yK %F9* I#onte3tK discussion of sla!ery. A person keeps doing this su ering, doing this su ering, doing this su ering...J to)
2&

kw0i

A-*1P#6

kw0i )

=G9)#21

ti 0modo

+#5)sla!ery

3 this is slavery

Itats002.00F.0F2J

0inally we turn to the Shybrid nounT mentioned at the start of this section. #orbett %4994K22F* describes hybrid nouns as nouns which take a$reement in more than one $ender, dependin$ on the tar$et. = well)known e3ample is the German noun Idchen S$irlT %#orbett 4994K227)229*, which may be referred to usin$ a choice of personal pronouns, either the neuter es %SsyntacticT a$reement* or the feminine sie %SsemanticT a$reement*. 2nly one such noun has been disco!ered so far in #icipu, ma0ga)ji ) SpriestT %ori$inally from <ausa maga%i SheirT*. ma0 ha!in$ been reinterpreted as an hen it was borrowed this word was presumably prefi3, as e!idenced by the
+#F

assi$ned to $ender 8NF by phonolo$ical assi$nment rules %see LF.5*, the first syllable
/-8

plural n0ga)ji ) and


=G9

SpriestsT and the +#5 deri!ation ti 0 ga)ji ) SpriesthoodT.

hile ma0ga)ji ) seems to tri$$er =G8


=G8

a$reement consistently in the associati!e construction %F9)54*, both a$reement can be found on the copula %54)52* and on !erbs %5;)58*. %F9* ma0ga)ji )
+#8)priest

ma,1ka0ngu

A-4E+#4)IplaceJ

the priest o 3angu %50* ma0ga)ji ) ma,1'yu*u0ni ), ma0na) mo)0ko*o0no)


A-4Rthere)+M6H A-4)/-6

Is!mk004.004J
A-4)dieQ/6S)10E

+#8)priest

the priest o up there, the one who died

I"ikula, sami004.022J

20 In #hadic lan$ua$es there is often a link between the word for SfootT or SfootprintT and the word for times , cf. <ausa sau SfootprintNtimesT. I do not know whether this is the case for (enue)#on$o in $eneral. In #icipu the word for SfootT is ku)0naa %+#9*, which may be reason for the =G9 morpholo$y here.

;0;

%54*

ma0ga)ji ) mu1u08a)ya)
+#8)priest

m0e
A-4)#21

A-4E+#7)IplaceJ

it's the priest o /aya %52* ma0ga)ji ) v0i )


+#8)priest

I"ikula, s!sdt004.055J 'i ni )

A-))#21

wu0u0yaa

;S)0A")doQI// likePthat

it's the priest he will do thus %5;* ma0ga)ji ) ma)0ya*a0na)


+#8)priest

Itats004.00;.029J

A-4)arri!eQ/6S)E-+"

the priest arrived %58* o0ki iso ma0ga)ji ) ;0si 0 ) hyaa


+#8)priest A-))<=()say

Ioar002.040.004J
+#2)mai$iroPspirit

the priest speaks to the kiiso spirits "he


=G8

Itats004.00;.080J
+#8

a$reement is consistent with the o!ert e3pression of

morpholo$y on the

controller, and is therefore syntactic a$reement. =G9 a$reement, on the other hand, could be ar$ued to be semantic a$reement , while there is no Ssemantic assi$nment ruleT %#orbett 4994K22F)225* assi$nin$ human nouns to $ender 9N224, we saw in LF.2.F that 9N2 is the most coherent of all the $enders, consistin$ only of humanNspirit nouns. "herefore if there are $oin$ to be hybrid nouns in #icipu at all, 9N2 is the most likely $ender for semantic a$reement. #orbett has shown how the possible a$reement forms tri$$ered by hybrid nouns is constrained by the =$reement <ierarchy %4994K chp. 9*. "he e!idence from #icipu is %currently* mea$re, but it at least does not conflict with the =$reement <ierarchy , sub'ect a$reement allows either syntactic or semantic a$reement, but noun modifiers such as the associati!e construction seem to be limited to syntactic a$reement. 0urther research would be reDuired to say anythin$ beyond this. "he phenomenon of hybrid nouns anticipates the concerns of 1art IE, which also in!ol!es an alternation between two different kinds of a$reement. "he difference is that in 1art IE we are not dealin$ with an alternation between syntactic $ender a$reement and semantic $ender a$reement restricted to a few nouns, but with one between $ender a$reement and $ender)independent person a$reement, potentially affectin$ all nouns. +e!ertheless certain of #orbett:s predictions about the differences between syntactic and
24 "here are also nouns denotin$ humans in $enders 4N2, 8NF, 9N;, and 9N2.

;08

semantic a$reement are found to hold for the $enderNperson alternation as well %L9.9*.

2.*

/eutral agreement

+eutral a$reement in #icipu %i.e. a$reement with atypical controllers , #orbett 4994* in!ol!es =G9 as the Se3ceptional case defaultT %#orbett 2007K257*, re$ardless of the type of a$reement tar$et. "his is not really surprisin$ $i!en that this class has a null prefi3, amon$st other allomorphs. #onseDuently controllers can function as a noun in this class without any structural chan$e, and yet not appear Sout of placeT either 22. In the first subsection %L5.8.4* I will demonstrate that
=G9

a$reement is used for a !ariety of


+#9

different kinds of atypical controller. In the second %L5.8.2* we will see that prefi3es may also be assi$ned to atypical controllers.

noun

9&1

AG=

neutral agreement
=G9

"he names of people tri$$er

a$reement2;. "he followin$ e3ample shows the name a$reement on the definite article, despite resemblin$ n0na)

of a person 3a)ji iji tri$$erin$ %5F* evi 3a)ji iji

=G9

an +#8 noun with a mA0 prefi3K


;S.1/2 Iname of personJ
A-))=/"

him Ia%ii%i Iwho we ha!e 'ust talked aboutJ I"ikula, sami004.0;2J "he controller in %55* below is atypical in that it is not actually a word, and therefore cannot ha!e inherent $ender. <ere the sound SoT tri$$ers =G9 a$reement on the numeral. =s usual with prefi3. %55* mu+0uwa)
4S)hearQ/6S IsoundJ
=G9

a$reement %L5.4.;*, vi0 or ;0 may be substituted for the len$thenin$ \o] n0nosi )

A-))four

' heard our o's Ii.e. the sound SoTJ

I2007)04)40.040J

"he ne3t e3ample demonstrates neutral sub'ect a$reement, a$ain with a controller denotin$ a soundK %57* \a]
IsoundJ

vi 0ita)1mu)

A-))satisfyQ/6SR4S.1/2

an ahu satis ied me Ii.e. that was the sound I wanted to hearJ I2007)02)09.002J
22 "his is also the reason why loanwords enter +#9 %LF.5*. 2; +on)human proper nouns often ha!e inherent $ender %L8.8.2*.

;0F

"he ne3t e3ample is similar, e3cept that this time it is a letter %SbT* tri$$erin$ a$reement on the relati!iser 1na) and the !erb. %59* \ ]
IletterJ

=G9

n0na)

A-))/-6

;0li a)0na)

=G9)lackQ/6S)10E +#9)hook

ku)0gi ya)

bu without a hook Ii.e. not 5J

I2007)04)40.040J

Indi!idual words without inherent $ender also tri$$er neutral a$reement. -3ample %59* comes from a metalin$uistic discussion about the !erb form haaya Sthey cameT, and shows =G9 a$reement on the numeral. %59* o)koo
therePis IwordJ

haaya vi 0 ) yapu)
A-))two

there are two haaya'su

Isamoh004.227J

1repositional phrases %70* or e!en entire clauses %74* can also tri$$er =G9 a$reement, at least on the copulaK %70* de)ge)
from

'u*

therePfarPoff

v0i )

A-))#21

it was rom ar o %74* Ika)mi i 0 ) yu*u i )0ri 0 mpa)a0ni ) o)0roono)L v0i ),

Isayb004.8F2J
+#2)loincloth A-))#21

before 21)wearQ/6S
+#;)thin$)this)+M6H

yi ,1e0 e)nte*e

=G;R+#2PloinclothPk.o.

y0aya)0na)

=G;)comeQ/6S)10E

saa
or

go)l
"21

it was be ore you wore loincloths, this thing o banteu came, or Isayb004.;50J

9&#
+#9

?eutral gender assignment

Some atypical controllers not only tri$$er =G9 a$reement, but also occur with a non)null prefi3 allomorph %LF.F.7*, as in the name of the town K0kainwa) in %72*K %72* 3aama
InameJ

vi 0 na) ;0yo0no)

=G9)/-6 =G9)beQ/6S)10E

K0kainwa)

/C)1<town=

Iaama who is at Kainwa

Isayb004.55FJ

=d!erbs too are assi$ned to $ender 9 when functionin$ as nominals. In %7;*, where do)ori SformerlyT and ka'a) SnowT both function as SpossessorT +1s, they occur with the
+#9

SmuchT %L8.9*, is always pronounced len$thenin$ prefi3. =t least one ad!erb, gei

with a lon$ g by some speakers, e!en when functionin$ ad!erbally %78*. ;05

%7;*

a)0 i ki )

+#2)festi!al =G2R/C))formerly

ha,1d0do)ori ,

n)

and

ha,1k0ka'a)
=G2R/C))now

the estivals o be ore, and o now %78* n ni ) ;0yuu s) t0i ) g0gei

Is!tm$004.297J

if when 2S)#2+" drink

=G5)1/2

/C))much

i you're smoking it a lot

Itats007.002.08FJ

=nother potential case of neutral Snoun classT assi$nment, at least historically, in!ol!es the possessi!e pronouns %L8.8.F.4.4*. "he forms of three of the si3 possessi!e pronouns %41, 21, and ;S* are e3actly what one would e3pect if they were strai$htforward associati!e constructions %L8.8.F.4*, with the ob'ect enclitics %L7.;* as possessors. -3ample %75* has been reanalysed from %8.4;7, repeated as %7F** accordin$lyK %7F* %a* $0$')
+#9)sheep

vi 0 ttu)

=G9)1P.P600

> @ ku)0la$i ku088 o)


+#9)$irl

=G9) P.P600

our sheep .sg.2 %75* %a* $0$') vi 1t1tu)

your .pl.2 girl

Irepeated from 8.4;7J

+#9)sheep =G9R/C)E1P.P#6

> @ ku)0la$i ku1818 o)


+#9)$irl

=G9R/C)E P.P#6

our sheep .sg.2

your .pl.2 girl

"he similarity between the tone patterns here and those found in the standard associati!e construction has already been obser!ed %L8.8.F.4.4*, but thinkin$ of the possessi!e pronouns in this way can also e3plain the lon$ consonants in the 41 and 21 forms. <ere they can be !iewed as the result of the -0
+#9

allomorph attachin$ to the

pronominal forms tu) and 8 o), analo$ous to the situation with re$ular nominal SpossessorsT e.$. $0$') vi , 0k0kaa Sthe sheep of the womanT. "he same analysis can also be applied to the demonstrati!e modifiers %L8.8.F.2* and ordinal numerals %L8.9.;*, which also ha!e a < 6 tone pattern and len$thened consonants. See also the diminuti!e and au$mentati!e pre)prefi3es in LF.;.528. e noted in LF.5 that #orbett assumes $ender in (antu lan$ua$es to be lar$ely morpholo$ically)assi$ned. If we adopt this !iew, then it is perhaps better to consider the e3amples 'ust mentioned as cases of neutral noun class assi$nment to +#9, with the =G9 a$reement followin$ strai$htforwardly from this.

28 Synchronically, the patterns obser!ed here can be thou$ht of as a kind of sub)morphemic Seidemic resonanceT %(ickel and +ichols 2007K209)240*.

;07

2.-

)omplex /4s and variation in agreement

"e3tbooks usually illustrate a$reement by demonstratin$ that !aryin$ the controller results in systematic chan$es on the tar$et, e.$. the boy runs, the boys run. <owe!er it is also possible for the features on the tar$et to !ary, e!en thou$h the controller remains fi3ed. "his indeterminacy can arise in two ways. "he first is for a morphosyntactically) simple noun to alternately tri$$er a$reement morpholo$y encodin$ different !alues for $ender, as we saw for the hybrid noun ma0ga)ji ) in L5.;. "he second way, which will be discussed in this section, in!ol!es morphosyntactically)comple3 nouns or +1s which offer more than one choice of controllin$ $ender, accordin$ to the features of their subparts. "wo sorts of comple3 nominal words ha!e been obser!ed. "he first is pre)prefi3ed nouns %LF.;.5*. =lthou$h we mi$ht e3pect the outer prefi3 to determine the $ender of the noun and hence the a$reements that it tri$$ers, this is not always the caseG e3ample %77* shows a$reement with the outer prefi3, and %79* with the inner. %77* ma)0ku)0naa ma,1k0kaa
/C4)+#9)le$ A-4E+#9)woman

ve0evi )

=G9);S.12SS

the little leg o his wi e Isaat002.002.558J %79* ko)0$i )0kooto) ti 0 mpa) ti 0 ) penene*u
+#4)+#5)drum A-+)this A-+)small

this big drum Isaat004.009.07;J Secondly, some compound nouns may tri$$er different a$reements. -ither the compound is treated as internally analysable as in %79a, 90a*, or as opaDue as in %79b, 90b*. %79* %a* kwakulle)
thatPday

kw0i

ku0lle) S+#9)day =G9)thatTJ it's that day Icf. full form kw0aa'a Isaat004.009.009J lle)BF v0i %b* kwaaku
thatPday
A-))#21

A-*)#21

it's that day I"ikula, sami004.29;J

lle) !aries in len$th and nasalisation , this is not thou$ht to be related to the 2F "he first !owel of kwaaku choice of $ender.

;09

%90* %a* i )0ri 0 mpa)a0ni )

+#;)thin$)this)+M6H

ye0evi )

A-3);S.12SS

yi )0to)

A-3)one

his one thingummy %b* i )0ri 0 mpa)a0ni )


+#;)thin$)this)+M6H

vi )0yapu)
A-))two

Itapf004.004.0;7J

two thingummys Itapf004.004.0;0J =s well as morpholo$ically)comple3 nouns, some +1s offer a choice of controllin$ $ender. In the case of associati!e +1s %L8.8.F.4*, a$reement can be tri$$ered either by the SpossessedT head noun or by the modifyin$ SpossessorT +1. "he phrases in %94a* and %94b* both contain the same associati!e +1 a)07a ha,0n0ka$i ShuntersT, lit. Speople of huntin$T, but in the former the copula a$rees with the %94* %a* a)07a
/C

+#2

head noun a)07a SpeopleT,

while in the latter it a$rees with the internal +1 n)0ka$i Shuntin$T.


1"erson

ha,1n0ka$i

=G2R+#F)huntin$

h0e)
A-

)#21

it was hunters %b* a)07a


+#2)person

ha,1n0ka$i

=G2R/C%1!unting A-%)#21

m0i )

Is!tm$004.474J

it was hunters

I"idipo, saat002.008.005J

Similarly in the con!ersational e3chan$e in %92* the first speaker:s copula a$rees with the
+#9

head noun of the associati!e construction, ;0looka)$i StimeT, but the second =K ;0looka)$i ;0ru0uma
/C)1time

speaker:s copula a$rees with the +#; SpossessedT noun ru0uma SwarT. %92*
=G9)+#;)war

v0i )

A-))#21

it was the time o war (K ee, ;0looka)$i ;0ru0uma


yes
+#9)time =G9)/C31war

y0i )

A-3)#21

yes, it was the time o war

Isayb004.489J

"he followin$ e3ample of an associati!e construction is presented as it was spoken. <owe!er the last two words in %9;* were actually transcribed by a nati!e speaker consultant as tina tile, a$reein$ with the inner SpossessorT noun ti 0 ) wm SchieftancyT25.
25 hen Dueried, consultants most often say that a$reement should be with the outer SpossessedT +1. hile this is more freDuent in the corpus, the e3ceptions are too common for me to be happy classin$ them all as speech errors. See #ro>ier %4998K98* for an e3ample of Sinner a$reementT in #entral &ambari.

;09

%9;* ma)0hwaari )
/C41start

ma,1ti 0 wm ma0na)
=G8R+#5)chief

A-4)=/"

me)0le*e

A-4)there

that's the start o the chie tancy Isayb004.297J "hese e3amples appear to show Stri$$er)happyT a$reement %#omrie 200;*, where conflictin$ controllers compete for a sin$le tar$et. =nother potential case in!ol!es a$reement Sout ofT prepositional phrases. #onsider the relationship between the +1 ku)0jene) Sri!erT and the a$reement tar$et $e in the followin$ e3ampleK %98* a,1ku0jene) ku)0$e
62#R+#9)ri!er A-*)+-G

it's not in the river I"ikula, taff002.04;J e mi$ht ha!e e3pected


=G9

neutral a$reement here with the 11 as an atypical

controller, but instead the ne$ator seems to seek out the more prototypical +#9 controller ku)0jene) Sri!erT, re$ardless of the fact that the scope of ne$ation is the 11, not the +1. It is not known whether neutral a$reement would be a possibility in %98* , if so, then this is another case of tri$$er)happy a$reement. "he ne3t two e3amples are similar in that the a$reement controllers are embedded within a 11. =$ain it is not known whether =G9 a$reement %i.e. with the 11 as controller* could occur in %95*, but these constructions are at least candidates for tri$$er)happy a$reement. * %9F* a,1;0'attaura vi )0hyaa
62#R/C)1(ora! A-))sayQ/6S +#9)person

707a

n0na)

=G9)/-6

po* a)0raate)0ne)
all

;1)han$Q/6S)10E

in the 0orah it says the one that they hanged Ioamy004.249J * %95* a,1m0kk)') ma)0hyaa...
62#R/C41cu" A-4)sayQ/6S

on the cup it said... I2009)08)04.00FJ

2.2

8ender resolution

SGender resolutionT is a term used by #orbett %4994K254ff* to describe what happens when two a$reement controllers are con'oined, and then to$ether tri$$er a$reement on a tar$et. =s discussed in McGill %2007K98)9F* #icipu speakers tend to find ways of a!oidin$ this structural confi$uration alto$ether. In summary, two constructions are mar$inally acceptable, illustrated by %97* showin$ a$reement with the first and more ;40

distant con'unct, and %99* showin$ syntactic resolution %the !erb a$rees with the plural form of the first con'unct*, althou$h the former may in!ol!e a comitati!e construction e.g. the sword with the spear, and the latter was not accepted by all consultants. 0urther research would be reDuired to rule out the comitati!e analysis %see <aspelmath 2008 for tests*. %97* ma)0gai
/C4)sword

n)

and

ku)0sayu)

+#9)spear

mo)0yuwo)0no)

A-4)fellQ/6S)10E

the sword and the spear ell %99* n) Mma)0gai


/C4)sword

I2007)04);4.004J

and

ku)0sayu) mi 0 ) yuwo)0no)
+#9)spear

A-%)fellQ/6S)10E

the sword and the spear ell

I2007)04);4.004J

;44

Part I1 Gender and person agreement in Cicipu

Chapter < Gender and person agreement on person markers


"he pre!ious chapter $a!e a hi$h)le!el o!er!iew of the different a$reement tar$ets found in #icipu. "his chapter will now focus in on a subset of these, specifically the different paradi$ms of person markers. I use the phrase Sperson markerT in the same way as Siewierska %2008*, as a co!er term for independent pronouns, pronominal clitics and affi3es, and $rammatical a$reement markers. <ere I consider the #icipu $ender)marked affi3es and pronouns to be person markers which inflect for the $rammatical cate$ory of
G-+.-/

%as in Siewierska:s %2008K408* analysis of &iswahili noun class pronouns*. It is


1-/S2+*

important to stress that in what follows the term Sperson markerT is used to co!er both Sperson)markedT person markers %which inflect for markers %which inflect for G-+.-/*. = distinction is made here between the pronouns, clitics, and a$reement markers 'ust mentioned, and other word classes such as the article and the demonstrati!es. =lthou$h the latter may substitute for nouns as the head of a noun phrase %and therefore can be referential*, they ha!e additional syntactic functions and are treated separately in L9.9. "his lea!es fi!e distinct paradi$ms of person markersK 0able J_- dummary o 5icipu person markers 0ub4ect "re'i$ Person person sub'ect prefi3es -ender $ender sub'ect prefi3es "here are three different
1-/S2+)marked

and S$ender)markedT person

Post1verbal ob4ect ob'ect enclitics

7lsew!ere personal pronouns

$ender pronouns paradi$ms dependin$ on the morphosyntactic


G-+.-/)marked

positionK sub'ect prefi3, post)!erbal ob'ect enclitic, and elsewhere. 0or

forms there is a two)way distinction between sub'ect prefi3 and other en!ironments. "he first si3 sections of this chapter describe the properties of these fi!e paradi$ms. 0or each of the pronounNclitic paradi$ms %L7.4)7.;* I will discuss its morpholo$ical and syntactic status accordin$ to the typolo$y of a$reement markers outlined in L2.2.F. Sections 7.8)7.F consider the phonolo$ical and morpholo$ical properties of the two paradi$ms of sub'ect a$reement markers. "he ar$uments concernin$ the syntactic status of these two paradi$ms are comple3 and are discussed in L7.5. =fter an interim summary in L7.7, I ar$ue %L7.9* that there really are separate paradi$ms of $ender and

;4;

person a$reement markers, rather than a sin$le comple3 paradi$m for each syntactic position. "o$ether with chapter 9, this chapter forms 1art IE of the thesis. "he di!ision of labour with respect to the characterisation of the $enderNperson alternation on person markers can be summarised as follows. "his chapter %especially L7.5* is concerned with the
=G/--M-+" 1/-/-_AISI"-S

%L2.2.8.4* necessary for the two different kinds of

a$reement. "here turn out to be a number of prereDuisites in!ol!in$ le3ical, phonolo$ical, and morphosyntactic factors. "hey are summarised in flowchart form in L7.5.4.;. #hapter 9, on the other hand, is concerned with the =G/--M-+" #2+.I"I2+S which bear on the choice of $ender or person a$reement in those conte3ts where the prereDuisites for both are met. "hese conditions in!ol!e the hi$her le!els of lin$uistic structureK morphosynta3, semantics, and pra$matics.

3.1

Independent personal pronouns

"he independent personal pronouns show a strai$htforward si3)way personNnumber distinction. "here is no inclusi!eNe3clusi!e 411 distinction. 0able J`- 'ndependent personal pronouns Person 1 3 0ingular amu i )v e or evi Plural otu i )8o ere

"he tone melody on these pronouns !aries accordin$ to both the phonolo$ical and the syntactic en!ironment in a way which is not yet well)understood. In isolation the tone pattern seems to either 6 < or < < for second person, and < < for the others.

<11

'orpholog-

"here are at least three reasons to classify the forms $i!en abo!e as SindependentT person markers4. 0irst, they can occur as complete utterancesK %4* a)mul_
4S.1/2

me# Isaat004.009.420J Second, they can occur in co)ordinated constructions, as in %2);*. If the first con'unct is
4 See Siewierska %2008K49* and .i3on and =ikhen!ald %2002* for the tests used here.

;48

a pronoun, then it occurs in the plural form, althou$h the meanin$ is ambi$uous between sin$ular and plural %see L8.8.F.8*. %2* e)re)
3P.P#6 and

n)

;0lwli )

+#9)lar$ePspider

o)0to*ono)

;1)$oPhomeQ/6S

he and (ig dpider went home %;* o)tu) n1i )v

Isaat004.005.20;J

1P.P#6 andR2S.1/2

me and you .sg.2 I2005)42)09.00FJ = third indication that this $roup of person markers are independent is that they can occur as the head of noun phrases, and be modified by other words. "he third)person forms %at least* can co)occur with the article, in which case they tri$$er as in %8*. "his is best seen as a form of neutral a$reement %L5.8*. %8* I#onte3tK"he speaker has been asked which ones are the sla!esJ ) ya) n) aa sha 3uusa, ere) n0na) o)0modo ha0na)
16

=G9

a$reement

Ishaya

and Musa

3P.P#6

A-)1A#(

+#2)sla!e

=G2)=/"

'shaya and Iusa, they are the slaves

I2009)0;)07.00;J

"he ;1S !ariant e is more often found in morpholo$ically)comple3 words such as the demonstrati!e pronouns e0lle) and e0''i nde) %L8.8.;.8*, but it may also occur as a free form, in apparent free !ariation with the lon$er form eviK %F* e
30.P#6

da'a)

moreo!er

w0aya)

;S)comeQ/6S ;s)pourQ/6S

u)0tuwo)

and then he poured %5* e)vi )


30.P#6

da'a)

moreo!er

u)0kondo)

Itapf002.008.087J

;S)enterQ/6S

and he entered

Isaat002.002.8;4J

<1#

0-ntax

"he independent personal pronouns are the only person)marked person forms that can occur in non)ar$ument positionsK %7* du)kwa
$oQIM1

n1a)mu_
withR4S.1/2

go with me# Isaat004.005.20;J ;4F

2ccasionally independent pronouns occur in sub'ect %i.e. pre)!erbal* position, but as with many Spro)dropT lan$ua$es, such constructions are not pra$matically)neutral. Instead they mark some kind of pra$matic relation in!ol!in$ the sub'ect referent, as in %9*, where i )v is in ar$ument focus. %9* I#onte3tK"he kin$ disco!ered who killed the crocodileJ i )v
f/

;0huna)0na) killed it

v0i )
=G9)1/2

2S.1/2 2S)killQ/6S)10E

Isaim004.0;8J

Eery rarely, the complement position of the !erb phrase may be filled by an independent personal pronoun at the e3pense of the usual ob'ect enclitic %L7.;*. I ha!e only found two e3amples of this, and in at least one of them, shown below, the sentence is pra$matically)marked, with no fewer than three left)dislocated topic +1s. "he anaphoric chains in %9* are indicated with subscripts. %9* I707a
+#9)person =G9)black

0ru)mono) n0na)J3

=G9)=/"

Ithe black oneJ3 Ievi


;S.1/2

n0na)

=G9)/-6

yo0no)

beQ/6S)10E withR+#7)power

n1u)0usu)J3

Iit which has powerJ3 Ivesi ) m)0'y'y ma0na) ma+0aya)0na)Jy


every
/C41'is! A-41#7; A-41comeM#;01P>,

see
then

IuJ30s)du)

Ie)vi J )y

;S)swallowQI// 30.P#6

Ievery fish that cameJy IitJ3 would swallow IitJy

I"ikula, taff002.020J

3.

Independent noun class pronouns

"he independent noun class pronouns are listed belowK 0able Jb- 'ndependent noun class pronouns in 5icipu #lass 4 2 he ; yi 8 F 5 ti 7 wi 9 vi 9 kwi 1ronoun ke me mi

<#1

'orpholog-

=s with the personal pronouns, the noun class pronouns2 are best analysed as independent words. "he con'unction test is not !ery useful here since there are no e3amples in the corpus of co)ordinated noun class pronouns, and neither are they
2 See L5.2.48)5.2.4F for some remarks on the phonolo$ical properties of these pronouns.

;45

attested as complete utterances. <owe!er two other tests are rele!ant. 0irstly, noun class pronouns can occur as one part of an eDuational construction %40* and in appositional noun phrases %44*. %40* w0i
A-71P#6

u)0laa
+#7)fire

it is ire Itats002.002.009J %44* kuma kw0i


and
A-*1P#6

ku)0yupu)
+#9)crocodile

ku)07amu)kwa)
=G9)beQ/6S

i0 ) ri
+#;)thin$

yi 1w0v
=G;R+#7)fear

and it the crocodile is a ear ul thing Itats002.005.042J Secondly, noun class pronouns occasionally head comple3 noun phrases, as in %42*, where the pronoun vi is followed by a modifyin$ demonstrati!eK %42* n0naa
+#9)cow

n0na)
=G9)/-6

u)0yo0no)
;S)beQ/6S)10E

v0i ),
=G9)1/2

Iv0i
A-)1P#6

vi 0 lle)L
A-)1t!at

u)0wi *ina)0na)
;S)sellQ/6S)10E

the cow that he'd had, that one he sold Isaim004.0F9J

<##

0-ntax

+oun class pronouns can occur in both ar$ument and non)ar$ument positions, althou$h it is hard to find clear e3amples of the pronouns in sub'ect position, e!en thou$h they are accepted in elicitation sessions %4;*. 0rom the corpus, %48* shows a noun class pronoun functionin$ as sub'ect in a non)!erbal clause, while %4F* is a possible e3ample of a !erbal sub'ect;. %4;* k0e
A-11P#6

ko)0yuwo)0nu)

=G4)fallQ/6S)/-S

o,1ko0o$i )

62#R+#9)hole

it Ika)0manga Sthe rope, +#4TJ ell into the hole %48* v0i u)0laa w0i )

I2007)02)0F.007J

A-)1P#6

+#7)fire

=G7)#21

it I;0gogoro S$in, +#9TJ is ire

Itats002.009.027J

; It would be better to ha!e an e3ample that ruled out inter)clausal interference.

;47

%4F*

I#onte3tK that word Ika)0daa, +#4J is not hearsay...J ) a0yaa k0e ka0wu)to)0no) a,1;0ri )kooda)
;1)doQI//
A-11P#6 =G4)$oPoutQI//)E-+"

62#R+#9)recorder

let them make it Ithe wordJ come out o the recorder I"ikula, sa$b004.749J +oun class pronouns also occur in complement position, as shown below. In %45* the pronoun occurs in the complement position of the E1, while in %47* the pronoun is in the complement of the locati!e proclitic A %L8.F, L;.8.7*. -3ample %49* shows a noun class pronoun functionin$ as the ob'ect of the preposition n) SwithT %L8.F*. %45* ma)0gai
+#8)sword

me0lle), o)0gutu)
=G4)that

;1)putPbackQ/6S

m0e)

A-41P#6

$e
+-G

a,1ku0pe'i

62#R+#9)scabbard

that sword, they didn't put it back in the scabbard Isamy004.057J %47* ka)0taari
+#4)stone

ke0lle), n)0ta'a)
=G4)that

4S)wantQ/6S 4S)takeQI//)E-+" 4S)$oQ/6S

n0ka) a)0na),

n)0yo*o

e1k0e)

62#RA-11P#6

that stone, ' wanted to bring it, ' went to it I2009)02)04.002J %49* n1a)0ka a)0na)
whenR;1)takeQ/6S)10E

a)0ra)kumi
+#2)camel

a)0dukwa) n)
;1)$oQ/6S

with

h0e
A-

1P#6

when they took camels they went with them Ithe camelsJ Isamoh004.499J

3."

.b'ect clitics

=s well as the independent personal pronouns %L7.4*, there is a correspondin$ set of si3 person)marked clitics found immediately after the !erb, as in %49)20*. "hese always e3press the ob'ect $rammatical function. %49* i g0go)onu)1mu)
2S.I//)helpQI//E10.P#6

please help me Isaat002.004.097J %20* va)yi v i )_ vayu0aT1vi )


throwPaway)IM1R30.P#6

throw him away# Isaat004.005.20;J "he full paradi$m is $i!en in "able F0K

;49

0able Ta- b%ect clitics Person 1 3 0ingular mu) vu) vi ) Plural tu) 8 o) re)

<%1

Phonolog-

"he resemblance between these #E ob'ect clitics and the E#E independent personal pronouns %amu AxO, i )v 21S, evi CxO, otu Axx, i )8o Bxx, ere Cxx* is clear, and su$$ests that the former deri!e historically from the latter, ha!in$ under$one a process of phonolo$ical reduction. Se$mentally, the clitics are !irtually identical to the second syllable of the independent pronouns, the only difference bein$ a chan$e of !owel in the 21S form vu). #ross)lin$uistically such relationships between intra)lan$ua$e person paradi$ms are common %Siewierska 2008K2F4)2FF*. "he tone on the ob'ect clitics is always 6. "his is e!en the case in the imperati!e mood, where otherwise the final < of the !erbal word spreads on to the followin$ syllable %L;.8.;*, and so the 6 tone on the clitic should be re$arded as le3ically) assi$ned. "he ;1S form vi ) is of particular interest because of its phonolo$ical similarity to the
=G9

pronoun vi. "hey are formally indistin$uishable in post)!erbal position, apart

from the fact that only the person)marked ob'ect clitic tri$$ers the leftward i)spreadin$ process described in L;.7.;. -3ample %24* is repeated from L;.7.; %;.405*. %24* %a* =K han
where

Audu)l
=udu

%b* =K han

where

0vooto)
+#9)$oat

where's Audu (K mi ndi )vi ) m0i nda)1vi )


4S)seeQ/6SR30.P#6

where's the goat (K mi nda) m0i nda)


4S)seeQ/6S

vi ) v0i )
A-)1P#6

' saw him

' saw it

<%#

'orpholog-

"he ob'ect clitics share properties with both pronouns and canonical a$reement affi3es. 6ike pronouns they may not co)occur with a le3ical +1 ob'ect, and yet like affi3es they ha!e a fi3ed position with respect to the !erb stemG they can only occur immediately to the ri$ht of the !erbal comple3. =lthou$h certain discourse markers may come between the !erb and le3ical +1 ob'ects, as in %22*, this has not been obser!ed for ob'ect clitics, ;49

despite o!er a thousand tokens of the latter in the corpus. %22* u)0dn)0n)
;S)followQ/6S)E-+" moreover

da'a)

u0yaa

+#7)road

and he ollowed the road

Isaat004.009.40FJ

"he ob'ect clitics can be distin$uished not only from le3ical +1s but also from other types of person marker, accordin$ to their position relati!e to the clausal ne$ator $e %L8.;.F.4*. =s illustrated below, le3ical +1s and independent personal pronouns always follow the ne$ator %2;)28*, ob'ect clitics always precede it %2F*, and for noun class pronouns both possibilities occur %25)27*8. %2;* a)0guya)
;1)canQ/6S

$e
/7-

$i 0 $i pu)
/C+1Ci"u

they don't know 5icipu Is!tm$004.2F;J %28* m0i +nda)


4S)seeQ/6S

$e
/7-

e)vi )
30.P#6

' didn't see H'I I2007)02)0F.007J %2F* hu0u0g)n)1re)


;1)0A")o!ercomeQI//R3P.P#6

$e
/7-

they wouldn't overcome them Isamoh004.052J %25* ma)0gai me0lle), o)0gutu)


;1)putPbackQ/6S

m0e)
A-41P#6

$e
/7-

a1ku0pe'i
62#R+#9)scabbard

+#8)sword =G4)that

that sword, they didn't put it back in the scabbard Isamy004.057J %27* ;0'i tt)
+#9)nei$hbour

ve0evi )
=G9);S.12SS

;08a)a

his neighbour doesn't have it Iu)0pepi %+#7* SspiritTJ

Av?)ha!e

$e
/7-

w0i )
A-71P#6

Itats002.008.025J

Anlike the independent personal pronouns %L7.4*, the ob'ect clitics cannot be separated from the ri$ht ed$e of the !erbal word. 0or e3ample they cannot be fronted %L8.;.2.4* , instead an independent personal pronoun must be usedK

8 -3ample %27* is the only post)ne$ation e3ample in!ol!in$ a noun class pronoun in the corpus. (oth structures can be elicited, althou$h the pre)ne$ation structure seems to be preferred.

;20

%29* %a* i )v

2S.1/2 2S.#21 4S)seeQ/6S)10E

vi )

m0i nda)0na)

it was you ' saw %b* zvu)


2S.1/2 2S.#21 4S)seeQ/6S)10E

>vi )@

m0i nda)0na)

I2009)02)04.002J 0urther e!idence comes from the !arious double)ob'ect constructions, includin$ those made possible by the applicati!e suffi3 0wA %L8.5.8.2*. =s pointed out in L8.;.4.4 the ob'ect clitic can only e3press the first ob'ect , the second ob'ect slot must be filled by an independent pronoun, as in %29*. %29* a)$i *ivi ) a)0$a*a1vi ) e)re) e)re) I2009)02)04.002J =s well as this positional restriction, the bound morpheme analysis is supported by the fact that ob'ect clitics cannot occur as complete utterances, e!en if followed by the copula. "he e!idence abo!e su$$ests that the morphemes under discussion are bound to the !erb. <owe!er they also differ from the more ob!ious !erbal affi3es discussed in L8.5, in at least three ways. 0irst, ob'ect clitics are e3cluded from the domain of certain tonolo$ical process. /ecall from L8.5.;.; that for !erbs Shea!ierT than #E#E, the habitual tone pattern is 6 6 < %6*. 0or #E#E the tone pattern is 6 6 < <. hen %undisputed* suffi3es are added to a #E#E root they contribute to the wei$ht of the !erbal word, and conseDuently the tone pattern chan$es from the li$ht pattern to the hea!y one. <owe!er when there is an ob'ect clitic after the !erb, as in %;0*, the tone pattern is not affected, and we can conclude that the clitic does not affect the wei$ht of the !erbal word. %;0* ) u)si )wosivi u0si0woso1vi ) it barks at him Itats004.004.075J "his does not actually force us to decide between affi3es, clitics and words, althou$h we do know that if the ob'ect clitics are affi3es then they are not in the first SstratumT of ;24 6 6 < < %li$ht*

;1)$i!eQ/6SR;S.1/2 ;1.1/2

they gave them to him

;S)<=()barkR;S.1/2

le3ical phonolo$y, since they must be ordered after the process which assi$ns tonal melodies to !erbal words %cf. &iparsky 4992 on -n$lish stress assi$nment and 6e!el IN6e!el II affi3es*. = second piece of e!idence sin$lin$ out the ob'ect clitics is the fact that they occur to the ri$ht of e!erythin$ else that mi$ht be considered affi3al, as illustrated below. %;4* 77a 707a nna) n0na) u)to i )li )si )su)wo)wo)no)mu) u)0to Qi )lRQi )sRQi )sRo0wo)0wo)0no)1mu) shayi ) ;0shayi )
+#9)tea

+#9)person

=G9)/-6

;S)coolQ/6S16=#s#=ASs#=ASs)1=SS)=116)10E E10.P#6

the person who has caused tea to become cooled down in a orce ul and iterative ashion or me I/epeated from %8.494*J "hirdly, there is e!idence from the domain of !owel harmony. /ecall from L;.F.; that the !owel in #icipu affi3es is either neutral i, u, as in the resultati!e suffi3 0nu, or comes from the harmonisin$ set a, e, o, like the applicati!e suffi3 0wA, in which case its phonetic !alue is determined accordin$ to the !owels of the !erb root. In contrast, the !owels in the 211 and ;11 ob'ect clitics 8 o) and re) are in!ariantF. "here is one ar$ument in fa!our of the free pronoun analysis, and this is the fact that these markers can apparently function as +1 heads, as in %;2*. =lthou$h such constructions were absent from the corpus, they were accepted without reser!ation in elicitation sessions. %;2* m0i +ndi )vi ) ni nni *i m0indaQ6<6Rvi ) ni nni *i
4S)seeQ/6SR30.P#6 alone

' saw !ust him"her I2009)02)04.002J 2n balance, there seem to be more reasons to !iew these person markers as clitics, rather than as affi3es or pronouns.

!educed ob,ect clitics


"his is an appropriate point to introduce one further set of person markers, related to the ob'ect clitics 'ust discussed, but with a more limited distribution. "hese three alternati!e forms are restricted to the sin$ular, and each of them has unusual characteristics. In contrast to the markers 'ust discussed, they are se!erely phonolo$ically)reduced, and they are most often found before the ne$ator $e. 0ollowin$ the intransiti!e e3ample in
F It should be noted that other clitics do harmonise %L;.F.;*.

;22

%;;* the three reduced ob'ect clitics are presented in an elicited paradi$m %;8);5*, and then in e3amples from the corpus %;7);9*. %;;* w0i +nda)
;S)seeQ/6S

$e
+-G

/"{A/O".

He1she didn't see I2009)02)44.00;J %;8* w0i +nda)1n


;S)seeQ/6SR10.P#6

$e
+-G

AxO

He1she didn't see me I2009)02)44.00;J %;F* w0i +nda)


;S)seeQ/6S

$1$e
0.P#6E+-G

BxO

He1she didn't see you #sg$% I2009)02)44.00;J %;5* w0i +ndi ) w0i +nda)1i );S)seeQ/6SR30.P#6

$$e $e
+-G

CxO

He1she didn't see him"her I2009)02)44.00;J %;7* vu0u07)s)0w)1n


2S)0A")lau$hQI//)=116R10.P#6

$el
+-G

won't you lirt with me Isaat002.004.049J %;9* mu0u0$a)a


4S)0A")$i!eQI//

$1$e

0.P#6E+-G

k0kaa

+#9)woman

v0vo*o

=G9)4S.12SS

' won't give you #sg$% my wi e Isaat002.002.582J %;9* n)gwaanu)kwi ) $$e ma7ama7a n)0gwaanu)kwa)1i )- $e ma7ama7a
4S)seeQ/6SR30.P#6
+-G

Duickly

' didn't see it $uickly Itats008.004.0;7J "he normal ob'ect clitics do sometimes occur before $e, as in %80*, but this is rare, and speakers ha!e stated a preference for the reduced forms in this en!ironment. %80* ) tu0u0hyaa1vu)
41)0A")sayQI//E 0.P#6

$e
+-G

we won't tell you #sg$% Isaat002.002.099J ;2;

-ach of these reduced ob'ect clitics has special phonetic properties. "he 41S form / attaches to the end of the !erbal comple3 resultin$ in a word)final consonant 5, otherwise unattested in #icipu outside of ideophones. "he ;1S form i also attaches to the end of the !erbal word, replacin$ the final !owel rather than coalescin$, since there is no increase in len$th. 2n the contrary, there is actually a decrease in len$th resultin$ from the $emination of the followin$ consonant , the $ of the ne$ator $e. "he 21S form is odder still, bein$ realised solely as the len$thenin$ of the consonant immediately followin$ the !erb. "herefore the only difference between an intransiti!e sentence such as %;;* and the correspondin$ sentence with a 21S clitic ob'ect %;F* is that the $ consonant of the ne$ator is len$thened. "he difference in len$th can be seen in the followin$ dia$ramsK

5 Anless $e is also considered to be a clitic. If so, it would be of the non)harmonisin$ kind, like the non) reduced ob'ect clitics.

;28

w 0

e 4.005

"ime %s*

+igure 64- "ave orm o e*ample .662-

wi nda) $e , ' didn't see

w 0

cc

e 4.009

"ime %s*

+igure 66- "ave orm o e*ample .6T2-

wi nda) $$e , ' didn't see you

;2F

"his kind of morpheme is cross)lin$uistically rare7, althou$h in #icipu the same SformT is also found as an allomorph of the 21S sub'ect a$reement prefi3 %L7.8.4*, and of the
+#9

and

=G9

prefi3es discussed in LF.F.7 and L5.4.;. Since the 21S reduced ob'ect clitic

is mostly found before the ne$ator $e, it usually appears as part of a lon$ consonant It(%J. <owe!er the consonant)len$thenin$ 21S sub'ect a$reement prefi3 -0 comes immediately before the !erb, and its application can therefore result in the len$thenin$ of any consonant. "herefore it seems better to also !iew the 21S reduced ob'ect clitic as an underlyin$ly)unspecified wei$ht unit -, which 'ust happens to always occur before N $N. "he ;1S form %at least* is also found as the first ob'ect in double)ob'ect constructions, either with %84* or without %82* the applicati!e suffi3 0wA %L8.5.8.2*K %84* * u)yaawi ) u0yaa0wA1i ) * $i )gaa * $i )0gaa

;S)doQ/6S)=116R30.P#6

+#5)Duarrel

he $uarrelled with her Ilit. Sdid her DuarrelTJ %82* a)$i *i a)0$a*a1i ) ma)su)uri ya) ma)0su)uri ya)
+#8)flute

Isaat004.008.024J

;1)$i!eQ/6SR30.P#6

they gave him a little lute

Isaat004.00;.089J

hile these reduced forms are most likely to occur between the !erbal word and the ne$ator $e, or in double)ob'ect constructions, they are also sometimes found before the ad!erb pa*a ShereT %8;)88*. <ere, howe!er, the reduced forms are the e3ception rather than the rule, and it is more usual to find the normal ob'ect clitics. %8;* n) mo)0ni
if
+#8)water

ma)0ka a)1m

=G8)takeQ/6SR10.P#6

pa*a ma)0ya*a
here

=G8)arri!eQ/6S

n)

with ItownJ

"ungan Ka8e Itats002.005.0;8J

i water took me rom here to 0ungan 3ae

7 Grammatical morphemes consistin$ of 'ust a consonantal wei$ht morpheme are found %e.$. the =rabic causati!e , <aspelmath %2002K22*, see also (le!ins 2008 chp. 7*, but I am not aware of any a$reement markers or incorporated pronouns e3pressed in this way. Morphemes consistin$ of a !owel wei$ht morpheme seem to be more commonG for e3ample, one of #icipu:s est &ain'i relati!es 1on$u uses a !owel wei$ht morpheme as a 21S ob'ect a$reement prefi3 %James Mac.onell p.c. 2009*. In <ausa most members of the independent series of pronouns differ from their ob'ect pronoun counterparts simply by an increase in !owel len$th %+ewman 2000K875*, althou$h this should probably not be !iewed as the result of applyin$ a !owel wei$ht morpheme.

;25

%88*

n0na)ha)

4S)lea!eQI//

p1pa*a

0.P#6Rhere

let me leave you #sg$% here Isaat004.005.098, see 0i$ure ;8J

nn 0

pp

aa 0.F249

"ime %s*

+igure 6J- "ave orm o e*ample .JJ2- nna)ha) ppa*a let me leave you .sg.2 hereu "he freDuency with which these reduced clitics are found as opposed to the normal ob'ect clitics seems to be correlated with the syntactic relationship between the marker and the word that follows, in that the more ti$htly bound the en!ironment is syntactically to the followin$ word, the more likely it is to be filled by the reduced forms. "he followin$ table may clarify mattersK 0able T,- +re$uency o reduced orms as opposed to normal ob%ect clitics >ollowing word +e$ator Secondary ob'ect =d!erb Iutterance inalJ >reJuency) o' reduced 'orms =lmost always 0reDuent 1ossible =lmost ne!er

It seems that the reduced forms occur in en!ironments where there is likely to be pressure e3erted due to the compression of speech within intonation $roups, the
9 "hese freDuencies are currently impressionistic.

;27

boundaries of which are influenced to a lar$e e3tent by syntactic structure %e.$. #ruttenden 4997K chp. 8*. It should be noted that it is possible for the reduced ;1S clitic to occur utterance)finally, as in %8F*, althou$h it is e3tremely rare. "he reduced 41S clitic has not yet been obser!ed here, and of course it would be lo$ically impossible to find the reduced 21S proclitic - utterance)finally. %8F* vi nda) v0i nda) mo)ni mo)0ni muuri )i mu0u0ra)a1vi

2S)seeQ/6S

+#8)water

=G8)0A")eatQI//R30.P#6

you see the water will overcome him I"ikula, sa$b004.45FJ

<%%

0-ntax

#ro>ier %4998K42;* makes two obser!ations about the #entral &ambari ob'ect a$reement suffi3es9K %i* they are sufficient by themsel!es to refer to the ob'ect referent, and %ii* they cannot co)occur with a le3ical ob'ect +1. "he #icipu ob'ect clitics are identical in these respects. e ha!e already seen that they are sufficient to refer to the ob'ect referent %e.$. 49)20*, and like #entral &ambari they cannot co)occur with a le3ical +1, i.e. Sclitic)doublin$T is impossible. "here is no reason to think that they are not referential, nor that they occur with anythin$ other than the ob'ect $rammatical function. +ot all ob'ects, howe!er, are e3pressed by these ob'ect clitics. =s we saw in %29)29* abo!e, focused and secondary ob'ects are e3pressed by independent pronouns instead.

3.*

4erson sub'ect prefixes

"he $enderNperson alternation is perhaps most ob!ious for the sub'ect prefi3es, since the $ender and person sub'ect prefi3es are more directly comparable than, say, the $ender) marked independent pronouns !s. the person)marked ob'ect clitics. "he person sub'ect prefi3 marks the standard si3 personNnumber distinctions, with the 21S paradi$m cell ha!in$ separate prefi3es for realis and irrealis moods.

9 #ro>ier %4998K42;* analyses the #entral &ambari eDui!alents of the #icipu ob'ect clitics as suffi3es, althou$h e3plicit reasons for this analysis are not $i!en.

;29

0able T4- &erson sub%ect pre i*es .be ore consonants1be ore vowels2 0ingular 1 /0:m0
/6S I//

Plural ti0:t0 i0:y0 A0:h0

I-0, ;0L:v0 Ii-0, ;0L:v0

u0:w0

<&1

Phonolog-

(efore !owel)initial !erb stems the usual coalescence applies %L;.7.4*K the prefi3 !owel mer$es with the root !owel, resultin$ in a lon$ !owel with the same Duality as the root !owel. /ecall from L;.8.5 that the tone on the prefi3 depends on the mood of the clause, the sonority of the prefi3 consonant, and whether or not the !erb stem is consonant) initial or !owel)initial. "he possibilities for ;1S sub'ects are set out belowK 0able T6- ^erb pre i* tone patterns with a 6&d sub%ect #)initial !erb stem dukwa S$oT /ealis u)dukwa) 6<6 <66 Irrealis udu)kwa) E)initial !erb stem aya ScomeT wa+(ja) wa*(ja) 6< 6 <6 6

1erhaps the most interestin$ thin$ to be said about the phonolo$y of the person sub'ect prefi3es concerns the 21S forms before consonant)initial stems. In the realis mood the prefi3 is either ;0 or the consonant)len$thenin$ -0 familiar from the discussion of
+#9N=G9

prefi3es %LF.F.7, L5.4.;*, as well as the reduced 21S and ;1S clitics discussed in
+#9N=G9

the pre!ious section. "he !ariation between the ;0 and -0 21S sub'ect allomorphs is comple3, as was the case for the prefi3es. =$ain the len$thened consonant is most clear word)medially, and in normal, fast speech word)initial lon$ consonants are often undifferentiated from short ones. <owe!er the difference in len$th between e3amples such as %85* and %87* comes out clearly in careful speech. %85* lelle t0ta'a)
truly 2S)wantQ/6S

truly you .sg.2 wanted Ieamy0;5.004.002J %87* i )8o


21.1/2 21)wantQ/6S

i )0ta'a)

truly you .pl.2 wanted Isayb004.;57J

;29

"he followin$ e3amples show a len$thened !oiced affricate %89*, liDuid %89*, !oiceless fricati!e %F0*, and $lottal stop %F4*. %89* j0ja*nta)
0)crushQ/6S

you (sg.) crushed Ieamy0;5.004.020J %89* l0latta)


0)sleepQ/6S

you (sg.) slept Ieamy0;5.004.04FJ %F0* s0sa*a a)


0)usedPtoQ/6S

you (sg.) are used to it Ieamy0;5.004.040J %F4* ni )1'0'p)


whenR 0)holdQ/6S

when you(sg.) hold Itats00F.004.0;0J In the irrealis mood, the 21S prefi3 can be written i-0K the root consonant is len$thened 'ust as in the realis mood, but there is also an i !owel precedin$ the $eminate. #ompare the $eminate d in the sin$ular in %F2* with the short d in the plural in %F;*K %F2* i d0da)ma)0wa)
0.I##)speakQI//)=116

you #sg$% should tell Itapf004.00;.002J %F;* i0 da)ma)0wa)


P)speakQI//)=116

you #pl$% should tell Isa$b004.2F8J "here is dialectal and e!en idiolectal !ariation here. In "ikula the irrealis 21S prefi3 is vi0 rather than the -0 len$thenin$ prefi3, and for some "irisino speakers the 21S prefi3 is vi-0, combinin$ elements of what seems to be the usual "irisino pronunciation and of the "ikula form. "he e3istence of these consonant)len$thenin$ prefi3es in both nominal and !erbal morpholo$y allows us to re)e!aluate .e olf:s %4974* hypothesis about the historical deri!ation of the &ambari lon$ consonants. "his has been done in McGill %n.d.*, where I ;;0

conclude that the len$thenin$ prefi3 -0 is deri!ed from vi0 by the loss of a !owel and the total assimilation of v to the followin$ consonant.

<&#

'orpholog-

"he morpholo$ical status of the person sub'ect prefi3es is more strai$htforward than that of the ob'ect clitics %L7.;*. "hey show none of the e3pected properties of free words, since they cannot occur independently of the !erb stem, cannot be con'oined with other words, and ne!er occur with modifiers. "hey always appear in the same positionK either prefi3ed to the !erb root, or to the future %u)* or habitual %si0* prefi3es if they are present. In addition, the ;11 prefi3 A1 harmonises with the !erb root in the same way as other affi3es.

3.-

8ender sub'ect prefixes

"he nine $ender sub'ect prefi3es are $i!en belowK 0able TJ- Gender sub%ect pre i*es #lass 4 1refi3 kA0 2 hA0 ; yi0 8 F 5 ti0 7 wu0 9 ;0:vi0 9 ku0 mA0 mi0

<(1

Phonolog-

(efore !owel)initial !erbs the usual rules of !owel coalescence apply, 'ust as for person sub'ect prefi3es. In the realis mood the prefi3 tone is usually 6, but sometimes a precedin$ < will cause it to rise. "his beha!iour contrasts with the person a$reement prefi3es which seem to be consistently 6 %L;.8.;*. Si3 of the nine $ender a$reement prefi3es are remarkably similar in form to the si3 person prefi3es 'ust discussed, particularly before !owel)initial !erbs. "he ambi$uous forms are hi$hli$hted in the tables below.

;;4

0able TT- &artial ambiguity o person pre i*es with respect to gender pre i*es, be ore consonantcinitial stems Pre'i$ kA0 hA0:A0 yi0:i0 mA0 mi0 /0 ti0 wu0:u0 vi0 ;0 ku0 Category
=G4 =G2N;1 =G;N21 =G8 =GF =GFN4S =G5N41 =G7N;S =G9 =G9N2S =G9

0able T]- 0otal ambiguity o person pre i*es with respect to gender pre i*es, be ore vowelcinitial stems Pre'i$ k.0 h.0 yi0 m.0 ti0 w.0 v.0 ku0 Category
=G4 =G2N;1 =G;N21 =G8N=GFN4S =G5N41 =G7N;S =G9N2S =G9 =G2

It is in the ambi$uity between the

and ;11 prefi3es that the real difficulty for this


+#7

study lies, since speech act participants are not normally referred to usin$ $ender a$reement prefi3es, and the semantic properties of the referents of nouns mean that they rarely tri$$er ;1S a$reement %see L9.8.;*. +#2, on the other hand, is by far the most common plural class for all kinds of nouns. In normal speech the =G2 prefi3 hA0 can be !ery hard to distin$uish from the ;11 a$reement prefi3 A0, e!en before consonant)initial stems, since word)initial NhN often elides inter!ocalically %L;.7.4*. "he problem is made worse by the fact that some of the most commonly occurrin$ !erbs ha!e !owel)initial roots, for e3ample aya ScomeT, inda SseeT, and uto S$o outT, as well as the future tense prefi3 u), and here the $ender)marked and person)marked !erbs are identical. ;;2

=mbi$uous cases of plural sub'ect a$reement are omitted from the analysis presented in chapter 9.

<(#

'orpholog-

"he $ender sub'ect prefi3es beha!e identically to the person sub'ect prefi3es in tests for morpholo$ical status, and so they too are considered to be affi3es.

3.2

0yntactic status of the sub'ect agreement prefixes

"he main aim of this section is to cate$orise the two paradi$ms of #icipu sub'ect prefi3es accordin$ to the three)way typolo$y of a$reement markers which was set out in L2.2.F. #orbett %200;bK498)492* pro!ides se!eral tests to distin$uish between $rammatical a$reement and pronominal incorporation, buildin$ on earlier work by (resnan and Mchombo %4997*, Siewierska %4999*, -!ans %4999, 2002* and others. In the first fi!e subsections of this section I will take each of #orbett:s tests in turn, applyin$ them to both of the sub'ect prefi3 paradi$ms. In L7.5.5 I will briefly consider the results of two dia$nostics offered by (resnan and Mchombo. Subsection 7.5.7 is concerned with non)o!ert a$reement, and asks whether it is better to posit null prefi3es or to assume the absence of a$reement alto$ether. 0inally L7.5.9 compares the status of the #icipu a$reement markers to their counterparts in the related lan$ua$e #entral &ambari.

<91

E'ulti.representationF

SMulti)representation40T refers to the possibility of a referent bein$ inde3ed more than once in the clause. If a !erbal marker can co)occur with a pronominal or le3ical sub'ect, this su$$ests it is a $rammatical a$reement marker rather than an incorporated pronoun. Similarly, accordin$ to (resnan and Mchombo:s %4997K7F2* locality principle, $rammatical a$reement must be local to the predicate %i.e. the a$reement relation holds between elements of the same clause*, and therefore only anaphoric a$reement can be non)local. =s will be seen in the discussion below, multi)representation is possible with both kinds of sub'ect prefi3 in #icipu, althou$h it is more common with $ender sub'ect markin$ than with person sub'ect markin$. (y this measure both $ender and person a$reement markers in #icipu come out as ambiguous a$reement markers accordin$ to
40 #orbett %200;bK49F*.

;;;

(resnan and Mchombo:s typolo$y, althou$h they are located at different points on this intermediate scale between pure $rammatical a$reement and pure anaphoric a$reement. Gender a$reement in the corpus is usually %but not always* a local relation, while person a$reement is usually %but not always* non)local. In the case of an e3plicit sub'ect +1 %i.e. $rammatical a$reement* there is usually no choice of a$reement feature, and the ne3t two subsections will make it clear which types of sub'ects can co)occur with each type of a$reement.
<911 Person sub,ect prefixes

.eterminin$ whether a particular +1 is a $rammatical sub'ect is not always strai$htforward for Spro)dropT lan$ua$es like #icipu. "he ability to tri$$er sub'ect a$reement is not pertinent since both $ender and person a$reement can be anaphoric, and so one cannot be sure that the Ssub'ectT +1 is within the clause , it mi$ht be a clause)e3ternal ad'unct, as in (aker:s %4995* analysis of polysynthetic lan$ua$es44. <owe!er, e!idence for the sub'ecthood of the +1 tri$$erin$ the !erb a$reement marker comes from the e3istence of an infiniti!e %LF.8.4* in #icipu. #onsider the followin$ comple3 sentenceK %F8* ere
;1.1/2 ;1)0A")canQI//

hu0u0gu)ya)

$e
+-G

u)07arta)

+#7)depri!ePofPshare

they won't be able to deprive us o our proper share Is!b$004.0;FJ <ere only the main clause !erb guya takes a sub'ect a$reement prefi3G the second !erb 7arta Sdepri!e s.o. of their proper shareT appears in the in!ariant infiniti!e form u)07arta), the sub'ect a$reement prefi3 ha!in$ been replaced by an
+#7

prefi3. I assume, with

(aker %4995K2F, see also &roe$er 2008K408)405*, that all infiniti!al clauses ha!e an obli$atory sub'ect , since there is no a$reement prefi3 to carry the sub'ect $rammatical function, the only other possibility is for there to be a null sub'ect %S1/2T* in the subordinate clause. "he sub'ect of the infiniti!al clause in #icipu is always understood as co)referential with that of the main clause, and so 1/2 must be controlled by a main clause sub'ect, in this case ere S;1.1/2T. 0urther e!idence that true sub'ect +1s are possible in #icipu comes from the impossibility of sub'ect a$reement bein$ tri$$ered by post)!erbal +1s, unless they are offset from the clause by a pause. If the +1 is a true sub'ect then we would e3pect it to
44 =ccordin$ to (aker these ha!e no true sub'ect +1s, only ad'uncts.

;;8

occur in the specifier position of the clause, i.e. pre!erbally, and this is precisely what we find. If, on the other hand, the Ssub'ectT +1 were an ad'unct then we mi$ht e3pect to find $reater !ariability in word order %see (aker 4995K404)402 for this ar$ument applied in re!erse to Mohawk*. "he abo!e tests are useful in demonstratin$ that the person sub'ect prefi3es ha!e the possibility of takin$ part in $rammatical a$reement, but they are of limited use when it comes to encodin$ a corpus and decidin$ whether indi!idual prefi3 tokens are instances of $rammatical or anaphoric a$reement. "herefore when encodin$ the corpus 42 I used a purely phonolo$ical dia$nostic , if there was a pause between the pre)!erbal +1 and the !erb then I assumed that it did not bear the sub'ect $rammatical functionG if there was no pause then I assumed that it did, unless the +1 was identifiable as a fronted constituent %L8.;.2.4*. In certain cases %e.$. when both ar$uments ha!e the same $ender* fronted ob'ects may be syntactically indistin$uishable from focused sub'ects, in which case one must rely on the semantics to identify the $rammatical relations. "he resultin$ counts show a stron$ tendency for person a$reement to be anaphoric rather than $rammatical. "here are ;,082 occurrences of the ;1S4; prefi3 in the corpus, and only ;24 of these %44* ha!e an o!ert sub'ect +1, accordin$ to the phonolo$ical dia$nostic 'ust described. In other words, the ;1S prefi3 is lar$ely in complementary distribution with le3ical +1s. 2nly certain kinds of +1s can be in!ol!ed in $rammatical a$reement, as can be seen from "able F7G in the ma'ority of the ;24 cases the sub'ect +1 was headed by either an independent personal pronoun %L7.4*, a demonstrati!e personal pronoun %L8.8.;.8*, or the word 707a SpersonT.

42 See L4.8 for details of the corpus. 4; "he fi$ures e3clude the "ikula te3ts %appro3. 4F of the total corpus*, which are yet to be interlinearised. =nalysis of the distribution of the ;11 prefi3 is complicated, since the distinction between it and the =G2 prefi3 %the most common $ender a$reement prefi3 for plural controllers* is often neutralised , see L7.F.4. =ccurate fi$ures ha!e not been compiled for first) and second)person prefi3es, but a brief inspection of the te3ts su$$est that co)occurrence with a sub'ect +1 is e!en rarer.

;;F

0able T_- eistribution o sub%ect N&s cococcurring with 6&d sub%ect agreement ?ead noun o' sub4ect /P ;1S personal pronoun %evi * ;1S demonstrati!e pronoun %e0mpe), e0lle), etc...* 707a SpersonT
+#9 +#9 +#8

(okens in cor"us 4F0 29 400 ;9 8 4 3 1

sub'ect with E)initial !erb sub'ect with #)initial !erb sub'ect

(otal

"he independent personal pronoun evi accounts for around half of the cases of ;1S $rammatical a$reement. /ather than bein$ set off from the clause by a pause, as we mi$ht e3pect for e3tra)clausal topicalised +1s, the pronoun actually coalesces with the person a$reement prefi3K %FF* evu)u8a*ana) evi ) u)08a*a0na) u)'usu) u)0'usu) saa ya)anu saa ya)anu
or who

;S.1/2 30)surpassQ/6S)10E

+#7)power

he is greater than everyone Ieab$004.009J "he same is true for the 41S and 411 prefi3es. %F5* a)munguya) a)mu n)0guya) me ' can't Ieaim00;.4;98=J %F7* o)ttuuto)no) o)tu) ti 0 ) uto)0no) le*e le*e 'asu) 'asu) wi ive) wi 0 ive) $e $e
+-G

4S.1/2 10)canQ/6S

41.1/2 1P)$oPoutQ/6S)E-+" there place%+#7*

=G7);1.12SS

we came rom their place there Is!b$004.02FJ "he ;1S demonstrati!e pronouns also tri$$er person sub'ect a$reementK %F9* ellu)ko*o e0lle) u)0ko*o

;S.1/2)that 30)dieQ/6S

that one died Isamoh004.2F9J In addition to the personal and demonstrati!e pronouns, the le3ical item 707a also

;;5

tri$$ers person sub'ect a$reementK48. %F9* w0aa


;S)/-1/"

a1d0do)ori

62#R+#9)formerly /C)1"erson

707a

u)0guya)

30)canQ/6S

u)0hi ya)

+#7)$rind

kwaanu) gooma sul


bowl ten
_

it's said that in the olden days one could grind ten measures Is!tm$004.049J %50* t0e)ne
=G5)which

t0i )

=G5)#21

707a

/C)1"erson

) w0u0yaa0wa) 1vu)

30)0A")doQI//)=116R2S.1/2

ka)0li ipi )l

+#4)wron$

how can anyone do you wrong Ioamy004.050J It could be ar$ued that what ha!e been classed as sub'ect +1s in %FF)F9* are actually topicalised and e3tra)clausal, despite the lack of a pause between the +1 and the !erbal word. <owe!er the sub'ect referents in %F9)50* are !ery bad candidates for topichood, since in both cases the referent is non)specific. "he fact that person sub'ect prefi3es occur here is stron$ e!idence that they can take part in $rammatical as well as anaphoric a$reement. "urnin$ back to "able F7, ;9 of the remainin$ 8; cases of $rammatical person a$reement in!ol!e the combination of an +#9 controller4F with a !owel)initial !erb stem, as in %54)52*K %54* ;0wm vu,1u0ree
+#9)chief =G9R+#7)town

w0a+ya)

30)comeQ/6S

0he chie o the town came Isaim004.022J %52* 0ku)ri )'aani ) wu0u0')p) 707a
+#9)&oran

30)0A")hold

+#9)person

the 3oran will hold a person Itats00F.004.4;8J "hese w0 prefi3es certainly look like person a$reement, but there is one other possibility that should be considered, if only to rule it out. Since !owel)initial syllables do not readily occur on their own in #icipu, could it be that the w0 prefi3 has simply been added as a kind of defaultM =s noted in L8.5.2.;, a dummy appro3imant is added when the imperati!e of !owel)initial !erbs is formed. So for e3ample the root uwa SfeelT becomes wu)uwa_ in the imperati!e. Similarly before i)initial !erb roots, a y0 is addedK
48 .espite its le3ical meanin$ of SpersonT, 707a is more often found as a $rammaticalised morpheme meanin$ somethin$ like SoneT or Sthe oneT. 4F "hese e3amples include both inherent +#9 nouns, and atypical controllers %e.$. names of people* that ha!e been assi$ned to this $ender %see L5.8*.

;;7

inda SseeT becomes yi )nda_. =ctually it is simple to show that the w0 $lossed as sub'ect a$reement in %54)52* is not there by default, since it is consistently w0, e!en before IiJ. So instead of kungwa y0i nda) in %5;* we ha!e kungwa w0i nda)K %5;* ;0kungwa w0i +nda)
+#9)God

30)seeQ/6S

God considered Itats004.008.07;J "his association between !owel)initial !erbs and person a$reement may be moti!ated by the fact that there are fewer a$reement choices before !owel)initial !erbs compared to before consonant)initial !erbs. /ecall from L5.4.; that before consonant)initial !erbs there are two
=G9

allomorphs for the sub'ect prefi3K vi0 or ;0. =s is the case across the

a$reement tar$ets in $eneral, the likelihood of vi0 occurrin$ increases as the animacy of the controller referent decreases. "his is reflected in "able F9, which shows the freDuency45 with which the vi0 and ;0
=G9

sub'ect a$reement prefi3es take part in

$rammatical a$reement before consonant)initial !erb stems. 0able T`- eistribution o the vi0 and ;0 allomorphs o the AG` sub%ect pre i* be ore 5c initial verb stems, according to animacy o re erent (y"e o' re'erent <umanNspirit =nimal Inanimate /o. o' occurrences o' vi0 2 %;* ; %47* 44 %20* /o. o' occurrences o' ;0 55 %97* 4F %9;* 88 %90*

"his distinction between vi0 and ;0 is no lon$er a!ailable when the !erb stem be$ins with a !owel , the vi0 prefi3 remains, but the null prefi3 cannot occur since #icipu nominal and !erbal words do not in $eneral be$in with a !owel. Instead it seems that nouns which would ha!e tri$$ered the ;0 =G9 allomorph before a consonant)initial !erb actually tri$$er the person a$reement prefi3 w0 before a !owel)initial !erb. So the function %that of distin$uishin$ animate47 from inanimate controller referents* of the a$reement is preser!ed re$ardless of the phonolo$ical structure of the !erb, but the feature reflected on the tar$et is not. "he situation is dia$rammed belowK

45 =$ain based on the corpus minus the "ikula te3ts. 47 It is likely that discourse topicality is in!ol!ed as well , see chapter 9 for the interaction of animacy and discourse topicality with respect to the alternation between $ender and person a$reement in $eneral.

;;9

0able Tb- dub%ect agreement pre i*es triggered by N5` controllers 6ower in animacy #)initial !erb stems vi0 %=G9* E)initial !erb stems v0 %=G9* <i$her in animacy ;0 %=G9* w0 %;1S*

In summary, person a$reement seems to ha!e been pressed into ser!ice in order to preser!e a distinction in animacyNtopicality in an en!ironment in which normally only $ender a$reement is found. "his still lea!es fi!e cases of ;1S sub'ect a$reement unaccounted for. Gi!en that there are ;,082 instances in the corpus, it seems reasonable to put them down to speech errors. <owe!er there is a certain amount of re$ularity in these e3ceptions. 0our of them in!ol!e
+#9

sub'ects with consonant)initial !erb stems , all of which ha!e uniDue

referents , kungwa SGodT, !eesu) SJesusT, d0daa Sthe kin$T, and a*ammi Spalm)wineT49. "he fifth e3ample is shown in %58* below. <ere the sub'ect is the +1 m melle) Sthat child,
+#8T,

and this une3pectedly tri$$ers person rather than $ender a$reement on the elle), e0lle), m m0 melle) wa+ya) me0lle) w0a+ya) u)$i *ivi ) u)0$a*a1vi )

au3iliary !erb aya. %58*

;S.1/2)that

+#8)child =G8)that

30)comeQ/6S ;s)$i!eQ/6SR;S.1/2

i0 ) ri 0mpa)a0ni )

+#;)thin$)this)+M6H

vi 0 ) yapu) :
=G9)two

that one, then that child gave him two thingummys 1 Itapf004.004.0;0J It seems likely that the speaker:s choice of person sub'ect a$reement here was influenced by the left)dislocated +1 elle), which is co)referential with what appears to be the true sub'ect +1 m melle), since sentences with non)+#9 $ender)marked sub'ect +1s tri$$erin$ person a$reement are re'ected in elicitation work, as %5F* illustrates. "he Ssub'ectT +1 ka)0 ara Sold man,
+#4T

cannot tri$$er person a$reement on the !erb %5Fb*,

unless there is an inter!enin$ pause %5Fc*.

49 In the sense of palm)wine in $eneral e.$. palmcwine is good or you.

;;9

%5F*

%a* ka)0 ara

+#4)oldPman

ka)0ta'a)

A-1)wantQ/6S

ka0ya)nta)

=G4)sDuee>eQI//

ka)0yi ivi

+#4)lefto!ers

the old man wants to s$ueeHe the le tovers %b* zka)0 ara
+#4)oldPman

u)0ta'a)

30)wantQ/6S

u0ya)nta)

;S)sDuee>eQI//

ka)0yi ivi

+#4)lefto!ers

IIntended meanin$K the old man wants to s$ueeHe the le toversJ %c* ka)0 ara,
+#4)oldPman

u)0ta'a)

30)wantQ/6S

u0ya)nta)

;S)sDuee>eQI//

ka)0yi ivi

+#4)lefto!ers

the old man, he wants to s$ueeHe the le tovers I2007)02)42.002J (efore lea!in$ the person a$reement prefi3es, it is worth mentionin$ a special kind of construction in!ol!in$ $rammatical person a$reement, not counted in "able F7 abo!e. It is possible to find le3ical relati!ised sub'ects tri$$erin$ irstcperson sub'ect a$reement on the relati!e clause predicate. In the relati!e clause in the appositional +1 in %55* we mi$ht ha!e e3pected ;11 a$reement, but instead the 411 sub'ect prefi3 occurs. %55* o)tu)
41.1/2

a)07a

+#2)person =G2)/-6

ha0na)

tu)0kusu)u0na)

1P)remainQ/6S)10E

us the people that %we)*remain I"ikula, sa$b004.298J It could be ar$ued that e!en thou$h the relati!iser 0na) a$rees in $ender with a)07a, the true head of the relati!ised clause is the first element of the apposition, the personal pronoun. (ut this cannot be the case in %57*, where instead of an appositional +1 we ha!e a predicate nominal construction %L8.;.;.4.2* consistin$ of two +1s and then the copula. I ha!e indicated the structure by bracketin$ the two +1s. %57* ha a)
comePon there 41.1/2

le*e

IotuL Ia)07a

+#2)person

ha0na)

=G2)/-6

ti 0 ) yo0no), h0e)

1P)beQ/6S)10E
=G2)#21

m0uu

+#F)child

mi R707a

=GFR+#9)person =G9)/-6

n0na)

ta'a)0na)

likeQ/6S)10E

a)07aL

+#2)person

5ome on# IweJ are Ipeople who %we)*are children o the one who loved peopleJ Ioamy004.499J "hese e3amples are similar in structure to the second)person a$reement tri$$ered by !ocati!e nouns in other lan$ua$es %e.$. #>ech , #orbett 2005K4;2*, but it is not known whether this construction can be used with the other first) and second)person prefi3es in #icipu, nor indeed whether it can occur with any other head noun apart from 707a

;80

SpersonT. 2!erall, then, we can see that person sub'ect markers are closer to SanaphoricT a$reement markers than to S$rammaticalT a$reement markers. It is important to stress that the difficulties in!ol!ed in determinin$ e3actly when the person sub'ect marker can co)occur with an +1 sub'ect are far from atypical cross)lin$uistically. #orbett %200;bK497* and Siewierska %2008K42F* both make this point, and we should not be too surprised by the comple3ity of the #icipu data.
<91# Gender sub,ect prefixes

6ike their person)marked counterparts, $ender sub'ect prefi3es are also ambi$uous a$reement markers in (resnan and Mchombo:s typolo$y. "he difference is that while the person prefi3es occur as anaphoric pronominals 90 of the time, $ender prefi3es are much more likely to co)occur with an e3plicit sub'ect +1. 2ut of the 98; instances of o!ert $ender sub'ect a$reement in the corpus, 895 %F9* are immediately preceded by a sub'ect +1, as in %59*49. %59* ka)0 ara
+#4)oldPman

ka)0'wa*a0na)

A-4)passQ/6S)10E

an old man passed by Itapf002.004.0;0J Just as with the person sub'ect prefi3es, there is e!idence from infiniti!al control to show that +1s like ka)0 ara in %59* are true $rammatical sub'ects. -3ample %59* below is analo$ous to the person)marked e3ample %F8* abo!e, since the sub'ect +1 ru'u yeevi ) Shis bodyT controls the sub'ect of the infiniti!e u)0he'we) Sto dryT. Just as before, this construction pro!ides e!idence that $ender sub'ect prefi3es can co)occur with true sub'ect +1s in the same clause. %59* ru0'u
+#;)body

ye0evi )

=G;);S.12SS

yi 0 ) hwaara) u)0he'we)
A-3)startQ/6S

+#7)dryPup

his body starts to dry up Itats007.004.0;7J hile $ender sub'ect a$reement prefi3es often co)occur with a sub'ect +1, they also take part in anaphoric a$reement. "hey are not licensed syntactically by any kind of
49 =$ain, the "ikula te3ts ha!e not been counted. = si$nificant proportion of these tokens come from te3ts with non)humans as topics, such as folktales and the topic)stimulation te3ts discussed in L4.8. It is likely that the incidence of anaphoric $ender sub'ect a$reement is lower in e!eryday lin$uistic practice.

;84

Sbindin$ theoryT, since they may occur arbitrarily far away from their antecedent. "he followin$ e3amples show antecedents in increasin$ly distant syntactic confi$urations, culminatin$ in %7;* where the e3pected antecedent is completely absent from the discourse. "he first e3ample shows $ender a$reement in a subordinate clause with the sub'ect of the matri3 clause. %70* ku)0longi
/C*1monitorFliQard =G9)want)/6S

ku)0ta'a)

ku0ya)nta)

A-*)sDuee>eQI//

ka)0yi vi

+#4)lefto!ers

the monitor liHard wanted to s$ueeHe the le tovers I2007)02)42.002J "he followin$ e3ample shows three 'u3taposed clauses. "he sub'ect +1 only occurs in the first clauseG in the others, the a$reement marker is anaphoric. %74* u)0laa
/C71'ire

wu)0yo*

A-7)beQ/6S

n1u)0'usu) :

andR+#7)power andR+#7)difficulty

wu)0yo*

A-7)beQ/6S

kuma n)
also

and

) : ti )0li pai

+#5)$oodness

wu)0yo*
A-7)beQ/6S

kuma n1u)0pa$i
also

ire has power 1 it is also good 1 it is also di icult Itats002.002.0F9J "he ne3t e3ample is an instantiation of a common $reetin$ pattern in!ol!in$ cross)turn anaphora , the referent kw0a'a) Shouse,
+#9T

is introduced by the Duestioner, and the


+#9T,

$ender sub'ect a$reement prefi3 ku0 is used anaphorically in the answer. "he same $reetin$ pattern can be found with nouns of any class %e.$. k0kaa Swife, Schildren, +#FT*. %72* =K t0e)ne
=G5)which /C*1!ouse

m0uu

kw0a'a)l

How's the house (K ku0yo*


=G9)beQ/6S

se)re
fine

't's ine. Ieamd00;.00;J 0inally, $ender a$reement morpholo$y can be used e!en when the ScontrollerT is completely absent from the discourse. -3ample %7;* was said by someone recountin$ what he had seen on a !ideo. "he form of the =G4 a$reement marker seems to be related to ka)0dama %+#4* SwordT %see L5.;*, but this word does not occur in the te3t itself, nor

;82

does any other potential antecedent. %7;* le*e


there

v0i )

=G9)#21

k)0kt)0n)

A-1)finishQ/6S)10E

it's there that it Ithe storyJ inished Itapf002.008.0F7J It is clear then that the $ender sub'ect prefi3es do not ha!e to be bound in any particular syntactic confi$uration in!ol!in$ their antecedent. "he antecedent may be the sub'ect +1 %in which case we ha!e $rammatical a$reement*, it may be elsewhere in the clause, e3tra)clausal, in a different sentence, a different turn, or e!en absent alto$ether20.
<91% 0ummar- of prere"uisites for sub,ect agreement

(oth $ender and person sub'ect prefi3es take part in anaphoric as well as $rammatical a$reement, and so accordin$ to the typolo$y discussed in L2.2.F they are both ambi$uous a$reement markers. <owe!er we ha!e seen that the a$reement prereDuisites differ for the two kinds of sub'ect a$reement, in!ol!in$ le3ical, phonolo$ical and morphosyntactic factors. "he flowchart in 0i$ure ;F is intended as an elucidatory aid24 to make it clear how the type of sub'ect +1 affects the kind of a$reement found on the !erb. It can be seen that the only en!ironments in which there is a choice between $ender and person a$reement are %i* when there is no sub'ect +1 at all , in other words anaphoric a$reement , and %ii* in the case of a stem.
+#9

sub'ect before a !owel)initial !erb

20 It is odd, in the li$ht of these e3amples and the freDuency data $i!en abo!e, that lan$ua$e consultants often re'ect constructed e3amples of anaphoric $ender a$reement. "his kind of elicitation seems to reDuire careful ne$otiation of the conte3t in order to supply the ri$ht felicity conditions for such utterances. 24 In particular, it is not meant to ha!e any formal status in a processin$ or te3t)$eneration model.

;8;

Start

9 p"icit sub:ect ;P<

In6pt. persona" pronoun<

&em. persona" pronoun<

JJaK (person(<

Is sub:ect ;C7<

= 4erson

Either +see chapter 7 %or agreement con6itions,

; = Vo8e"-initia" >erb<

Stop

8ender

+igure 6T- +lowchart showing which sub%ect N&s trigger gender and person agreement "he prohibition on $ender a$reement for +1s which are headed either by an independent personal pronoun, a demonstrati!e personal pronoun, or the le3ical item 707a SpersonT may seem a little arbitrary. = way round this problem is to claim that these e3pressions ha!e in common the feature
1-/S2+

but not G-+.-/. "his is easy to show for

the personal and person)marked demonstrati!e pronouns. =s will be apparent in chapter 9, person)marked independent pronouns may ha!e any kind of +1 as its antecedent, re$ardless of the $ender of the noun. Similarly the person)marked demonstrati!e pronouns can readily be used deictically to indicate any kind of ob'ect, re$ardless of which $ender the ob'ect:s basic)le!el noun belon$s to %L2.2.4*. +either show any si$ns of bein$ marked for the feature markin$ or a$reement. "o su$$est that 707a SpersonT is not marked for G-+.-/ is a more radical step. 0or a start, the word is morpholo$ically marked with the characteristic -0, and it tri$$ers
=G9 +#9 G-+.-/,

either throu$h o!ert morpholo$ical

len$thenin$ prefi3

a$reement on most of the a$reement tar$ets listed in L5.2. hile it has a SdefaultT

+e!ertheless there are two other properties that sets this word apart from other 9N2 or 9N; nouns. "he first is its e3treme fle3ibility of reference. meanin$ of SmanNpersonT %<ausa mutum*, so that in the identificational clause 707a v0i ) ;88

Sit:s a personT it can only denote a human, in other constructions %e.$. 78 below* it has the potential to refer to any concrete noun. Secondly, in addition to bein$ able to co) occur with a person a$reement prefi3, 707a is also special when it comes to $ender a$reement. =pparently uniDuely, it cannot co)occur with the vi0 allomorph of the $ender sub'ect prefi3. So only %78a* and %78b* are possible, and not %78c*22. %78* %a* 707a
+#9)person

=G9

n0na)

=G9)/-6

u)0yuwo)0no)

30)fallQ/6S)10E

the one that ell %b* 707a


+#9)person

n0na)

=G9)/-6

yuwo)0no)

fallQ/6S)10E

the one that ell %c* z707a


+#9)person

n0na)

=G9)/-6

vi 0 ) yuwo)0no)

A-))fallQ/6S)10E

I2009)08)4F.004J #ompare this with the beha!iour of other +#9 nouns such as va0ari ) SmanT %7F* and k0kaa SwomanT %75*, which allow ;0 or vi0 but not u0. %7F* %a* va0ari )
+#9)man

n0na)

=G9)/-6 A-))fallQ/6S)10E

B0yuwo)0no)

the man that ell %b* va0ari )


+#9)man

n0na)

=G9)/-6 A-))fallQ/6S)10E

vi 0 ) yuwo)0no)

the man that ell %75* %a* k0kaa n0na) B0yuwo)0no)

I2009)08)4F.004J
A-))fallQ/6S)10E

+#9)woman

=G9)/-6

the woman that ell %b* k0kaa


+#9)woman

n0na)

=G9)/-6

vi 0 ) yuwo)0no)

A-))fallQ/6S)10E

the woman that ell I2009)08)4F.004J (efore !owel)initial stems vi0 a$reement %77a* is not possible with 707a as sub'ect either, lea!in$ person a$reement as the only option %77b*, since ;0 is ruled out by phonotactic constraints %L;.4.4*.
22 /ecall from L5.4.; that the -0 !erbs in #icipu.
=G9

allomorph found on other a$reement tar$ets does not occur on

;8F

%77*

%a* z707a %b* 707a

+#9)person

n0na) n0na)

=G9)/-6

va+0aya)0na)

A-))comeQ/6S)10E

+#9)person

=G9)/-6

wa+0aya)0na)

30)comeQ/6S)10E

the one that came

I2009)08)4F.004J

+ote that this is a le3ical constraint on 707a, rather than anythin$ to do with animacy , the possibility of person a$reement and the prohibition on the vi0 allomorph hold whether 707a is bein$ used to refer to humans, animals, or e!en inanimate ob'ects. In this respect 707a patterns with the independent personal pronouns, demonstrati!e personal pronouns, and personal names, rather than other demonstrati!e pronoun e0mpe) SthisT. %79* %a* e0mpe)
;S.1/2)this 30)surpassQ/6S
+#9

nouns.

#ompare the $rammaticality 'ud$ements in %78* with those in %79* in!ol!in$ the u)08a*a e0mpe) n) ti )0li pa*i

;S.1/2)this with

+#5)$oodness

this one is better than this one Ilit. Ssurpasses with $oodnessTJ %b* e0mpe)
;S.1/2)this surpassQ/6S

8a*a

e0mpe)

;S.1/2)this with

n)

ti )0li pa*i

+#5)$oodness

this one is better than this one %c* ze0mpe)


;S.1/2)this

vi )08a*a

A-))surpassQ/6S

e0mpe)

;S.1/2)this with

n)

ti )0li pa*i

+#5)$oodness

I2007)02)4;.005J So in the same way that person sub'ect a$reement is possible with the independent and demonstrati!e personal pronouns, personal names, and the noun 707a as sub'ect +1, gender sub'ect a$reement in!ol!in$ the vi0 allomorph of the with precisely the same $roupin$ of e3pressions as sub'ect. "his can be accounted for if we assume that, despite belon$in$ to the morpholo$ical class 9N2, 707a does not in fact ha!e a specification for the morphosyntactic feature
=G9 G-+.-/. =G9

prefi3 is impossible

"his $ender)deficient noun would, by default, tri$$er

a$reement on a$reement tar$ets that necessitate some kind of $ender a$reement, in

'ust the same way as other atypical controllers such as personal names and personal pronouns %see L5.8 on neutral a$reement* , compare 707a n0na) Sthe personNthe oneT with 3a)ji iji n0na) Sthe Ma'ii'i who we:!e been talkin$ aboutT in %79*, or ere) n0na) Sthem we:!e been talkin$ aboutT in %90*. ;85

%79*

3a)ji iji n0na)


InameJ

A-))=/"

the a orementioned Ia%ii%i Ia person:s nameJ I"ikula, sami004.0;2J %90* ere)


;1.1/2

n0na)

A-))=/"

them Ii.e. those ones pre!iously mentionedJ Isayb004.84FJ "his ar$ument is stren$thened by the apparent impossibility of the plural noun a)07a SpeopleNonesT tri$$erin$ a)07a
=G2

$ender sub'ect a$reement. Instead it seems from the many a)0si 0 ) suma
3P)<=()run

e3amples in the corpus2; that only ;11 person sub'ect a$reement is found, as in %94*. %94*
+#2)person =G2)that

he0lle)

those ones were running Isamy004.024J "his e3ample also raises a Duestion, howe!erK why, if it is $ender)deficient, does a)07a tri$$er
=G2

$ender a$reement on other tar$ets such as the demonstrati!e modifier 0lle)M


AG`

e ha!e already used the neutral a$reement to account for the pronoun ere) tri$$er =G9 a$reement.

a$reement tri$$ered

by 707a, and from %90* abo!e we can see that e!en plural referents such as the ;11 In fact there is e!idence that atypical controllers with plural referents can also tri$$er
=G2

a$reement. /ecall the associati!e plural marker aa %L8.8.2*, which occurs


=G2

before names to indicate that person and their family or associates. In the identificational clause %L8.;.;.4.4* in %92*, aa 0awli )
16

a$reement is tri$$ered on the copula by

the associati!e plural of the 9N; noun28 ;0awli ) Slar$e spiderT. %92*
+#9)lar$ePspider

h0e)
A-

)#21

0here was (ig dpider and associates

Isaat004.009.00FJ

=dditional e!idence is pro!ided by the ;1S personal pronoun, which can occur with the
=G2

copula h0e) if it has a plural referent %L9.9.8*. If we accept the analysis of 707a as a $ender)deficient noun which relies on the
=G9

operation of neutral a$reement for the

morpholo$y that it does tri$$er, then the

2; +o rele!ant elicitation has been carried out, unfortunately. 28 In the conte3t of the folktale from which this e3ample is taken awli ) should probably be thou$ht of as a proper name rather a common noun , either way there is no le3ical moti!ation for =G2 a$reement.

;87

flowchart presented in 0i$ure ;F abo!e can be simplified and the analysis made more ele$ant, since there is no need to e3plicitly mention the e3ceptional classes of sub'ect +1s that allow person sub'ect a$reement. 0i$ure ;5 shows the re!ised analysis.
Start

)naphoric agreement<

Vo8e"-initia" >erb<

= Pre-re@uisites satis%ie6 %or person agreement

)ither gen6er or person +see chapter 7 %or con6itions,

Is contro""er gen6er 7? 7'2? or 7'3<

; Copy contro""er %eatures Cou"6 be gen6er or person agreement

Stop

+igure 6]- Revised lowchart modelling gender1person alternation "he fundamental rule here can be found towards the bottom)ri$ht of the chart. "he rule simply states that the controller features are to be copied onto the tar$et. So if the controller is marked for $ender, as are most le3ical +1s, then so will the tar$et be. If the controller is not marked for $ender, then only person will be marked on the tar$et. "his is what happens when the controller is a personal pronoun, a demonstrati!e personal pronoun %e.$. e0mpe) Sthis oneT, e0lle) Sthat oneT*, or, as has been ar$ued abo!e, the $ender)deficient le3ical item 707a SpersonT. "he only way that this rule can be circum!ented is if both %i* the prereDuisites for person a$reement are met %either ;89

because the a$reement is anaphoric, or because the a$reement is $rammatical with an


+#9

sub'ect and a !owel)initial !erb stem*, and %ii* the a$reement conditions to be

discussed in chapter 9 turn out in fa!our of person a$reement. = final !ersion of this flowchart will be presented in L9.7.
<91& Agreement conditions
1/-/-_AISI"-S

+ow that the a$reement with the a$reement

ha!e been established, we are eDuipped to deal

#2+.I"I2+S

which are rele!ant to the $enderNperson alternation in

#icipu. "his will be the sub'ect of chapter 9, and here I will only comment on the rele!ance of a$reement conditions in #icipu to the a$reementNpronominal affi3 distinction. #oncernin$ this topic, #orbett notes %200;bK490* that If...conditions affect the presence or absence of a marker, then this su$$ests that the marker is a pronominal affi3. "he situation for #icipu is complicated, of course, in that the two sub'ect markers are not independent. If one is not present, the other must be, and vice versa %althou$h see L7.5.7 on optional a$reement*. In L9.7 I will ar$ue that $ender a$reement should be considered the default, and person a$reement sub'ect to conditions. "his heuristic then places $ender sub'ect markers further alon$ the scale towards $rammatical a$reement than person sub'ect markers.

<9#

?umber of arguments encoded on the verb

In #icipu only one ar$ument can be encoded on the !erb, re$ardless of a$reement feature, and this outcome fa!ours neither the $rammatical a$reement nor the incorporated pronominal interpretation.

<9%

!eferentialit-

=s mentioned in L2.2, -!ans %2002* takes issue with Siewierska:s %4999* use of the term S$rammatical a$reementT to mean !erb a$reement which reDuires the presence in the clause of a le3ical +1 ar$ument. Instead he ar$ues that an a$reement marker should be called S$rammaticalT when it loses its referentiality %i.e. it can be co)inde3ed with non) referential +1s*, independently of whether it can occur in the absence of a nominal ar$ument. (oth sets of #icipu sub'ect prefi3es are able to co)occur with definite, indefinite, or non)specific2F +1s. "he followin$ e3amples show $ender sub'ect a$reement with a specific indefinite +1 %9;* and a non)specific +1 %98*.
2F See <aspelmath %4997K405)440* for the definition of non)specificity assumed here.

;89

%9;*

hari ) ka)kuya hari ) ka)0kuya


until

/C11termite

kaya) k0aya)

A-1)comeQ/6S A-1)catchQ/6SR;S.1/2

ka)ka5i )vi ) ka)0ka5a)1vi )

until a termite mound took hold o him I"idipo, saat002.008.049J %98* do)ori ,
formerly when

ni )

m0

+#8)child

mo)0yo*o

A-4)$oQ/6S

aRka0kaasu)wa) :
62#R+#4)market

u)0si 0 ) $i yo $e
;S)<=()$et
+-G

0ke)eke :

+#9)bicycle

in the past, when a child went to the market, he wouldn't get a bicycle I2009)02)04.004J -3ample %9F* shows $ender a$reement with a .)Duantified %=ustin and (resnan 4995K2;7)2;9* sub'ect +1 saa kwe)ne kula$i Se!ery $irlT. %9F* saa kw0e)ne
or
=G9)which

ku0la$i

+#9)youn$P$irl

ku0dukwa)
A-*)$oQ/6S

every girl went Ieamd0;2.498J 1erson a$reement is eDually compatible with indefinite and non)specific +1s. -3ample %95* in!ol!es an indefinite specific +1. %95* wu0na) 707a
;S)=/"
+#9)person

wa+0aya)

30)comeQ/6S with

n)

0ke)eke

+#9)bicycle

a certain person came along with a bike Itapf002.00;.04;J -3amples %97)92* show person sub'ect a$reement with !arious kinds of non)specific referents. "he first e3ample %97* occurred in the corpus as an impersonal construction %L9.8.9* with a $eneric interpretation. %97* i0 ) la8amu
+#;)li$htnin$

kuma : hu0u0yi )nda) y0i )


and 3P)0A")seeQI//
=G;)1/2

$e n)
+-G

kw0andai :
+#9)dryPseason

with

and lightning 1 it is not seen in the dry season 1 Itats002.007.02FJ "he followin$ e3amples all show person a$reement with non)specific +1s, either e0mpe) Sthis oneT %99* or 707a SpersonT %99)94*.

;F0

%99*

e0mpe)

;S.1/2)this 30)doQI//

) u0yaa

ti )0kg : e0mpe)
+#5)<ausa

;S.1/2)this 30)doQI//

) u0yaa

$e : e0mpe)
+-G

;S.1/2)this 30)drinkQI//

u0s)

mo0yoo : u0yu)wo) :
+#8)beer

e0mpe)

30)fallQI//

;S.1/2)this 30)drinkQI//

u0s)

$e
+-G

u0yu)wo) :
30)fallQI//

one would ollow 'slam 1 another wouldn't 1 one would drink beer 1 until he ell 1 another wouldn't drink until he ell 1 Isamoh004.2;4J %99* 707a 'asu)
place%+#7*
+#9)person

) wu0u0yaa wu0na)
=G7)/-6

30)0A")doQI//
"21

0kwaanika)

+#9)o!erni$htPstay

a1u0yaa :

62#R+#7)road

go u)0'*0n) :

u0hu)ngwa) :

30)tireQ/6S)10E 30)restQI//

a person would do an overnight stay on the road 1 where he got tired 1 he would rest 1 Isayb004.7FFJ %90* w0aa
;S)/-1/"

a1d0do)ori

62#R+#9)formerly /C)1"erson

707a

u)0guya)

30)canQ/6S

u)0hi ya)

+#7)$rind

kwaanu) gooma sul


bowl ten
_

it's said that in the olden days one could grind ten measures Is!tm$004.049J %94* t0e)ne
=G5)which

t0i )

=G5)#21

707a

/C)1"erson

) w0u0yaa0wa) 1vu)

30)0A")doQI//)=116R2S.1/2

ka)0li ipi )l

+#4)wron$

how can anyone do you wrong Ioamy004.050J "he final e3ample shows person a$reement with a .)Duantified sub'ectK %92* saa w0e)ne
or ;S)which

707a

+#9)person

n0na)

=G9)/-6

u)0yo0no)

30)beQ/6S)10E

ka)0rayi )
+#4)life

every person who has li e Itats004.008.055J It is clear then that both sub'ect a$reement paradi$ms can ha!e referential and non) referential uses. <owe!er e!en the free pronouns of -n$lish ha!e !arious indefinite readin$s %e.$. $eneric uses such as they always get you in the end* , the Duestion is whether the indefinite readin$s are limited to unusual syntactic and semantic conte3ts %-!ans 2002K47*. Gi!en the !ariety of e3amples presented abo!e, it seems to me that with respect to referentiality both the #icipu sub'ect prefi3es ha!e more in common with -n$lish sub'ect a$reement than -n$lish free pronouns. "hus they are non) committal with respect to referentiality, and this points towards the $rammatical a$reement analysis.

;F4

<9&

$escriptive content

/e$ardin$ descripti!e content, neither person nor $ender sub'ect affi3es seem to ha!e any semantic restrictions as to their use, and therefore we cannot say that there is any le3ical meanin$ attached to the markers themsel!es. "he most that can be said is that there is a stron$ correlation between the use of person a$reement and animacy %especially humanness*. <owe!er !irtually any kind of referent can be inde3ed with person a$reement if it is sufficiently discourse)topical %see L9.8*, and thus while the Sdescripti!e contentT criterion points towards $rammatical a$reement rather than incorporated pronouns, it does not distin$uish between the two markers.

<9(

/alance of information

#onsiderin$ Sbalance of informationT #orbett notes that for a$reement markers there are features reflected on the controller that may be absent on the tar$et, and !ice !ersa. <owe!er this is not usually the case for incorporated pronominals, and #orbett remarks %2005K40F* It is rare to find %morphosyntactic* feature distinctions marked on the noun phrase which are not also marked on the pronominal affi3es.... =s far as this criterion is concerned, we find the opposite of what we mi$ht ha!e e3pected. In common with other (enue)#on$o lan$ua$es %see L2.4.2*, the noun class %and hence the morphosyntactic features for
G-+.-/, +AM(-/

and

1-/S2+

25

* is indicated on the noun. Gender sub'ect affi3es and therefore the balance of information is

carry the same three features. 1erson sub'ect affi3es, on the other hand, are marked only
1-/S2+

and

+AM(-/,

but not

G-+.-/,

shifted towards the +1, which also marks $ender. "hus this criterion places person a$reement closer to S$rammatical a$reementT than $ender a$reement, in contrast to the multi)representational criterion discussed in L7.5.4.

<99

2ther tests

=s well as multi)representation %or the SlocalityT principle*, (resnan and Mchombo %4997K7F9)758* discuss a number of other dia$nostics for $rammatical !s. anaphoric a$reement, based on the assumed impossibility of anaphoric a$reement bein$ tri$$ered by non)topical controller referents , if anaphoric a$reement is topical a$reement, then it cannot occur in conte3ts where we know that the controller referent is not a sentence topic. So if we find an a$reement marker in such a conte3t, then it is either a
25 1-/S2+ is deri!ati!e of G-+.-/ for all +1s other than first) and second)person pronouns.

;F2

$rammatical or ambi$uous a$reement marker. #on!ersely, if an a$reement marker cannot occur in such an en!ironment, this su$$ests it is an anaphoric a$reement marker. It is important to be clear that these predictions are about different types o agreement rather than a$reement markers. It is the type of a$reement %anaphoric or $rammatical* that is associated or disassociated with topicality. Indi!idual a$reement markers are then secondarily classified as $rammatical, anaphoric, or ambi$uous. "he theory does not directly predict an association between, say, #icipu person a$reement markers and topicality, and #icipu $ender a$reement markers and non)topicality. So e!en if we find %as we did in L7.5.4* that $ender a$reement is more common when there is an e3plicit sub'ect +1 and that person a$reement is more commonly anaphoric, we should not e3pect (resnan and Mchombo:s subseDuent dia$nostics to pick out any asymmetry between the two paradi$ms. +e!ertheless there is some merit in discussin$ a couple of their predictions, since e!en if they do not distin$uish between the two paradi$ms, they at least ser!e to confirm that #icipu $ender and person a$reement sub'ect prefi3es are ambi$uous a$reement markers, rather than pure anaphoric a$reement markers.
<991 Auestioned sub,ects

(resnan and MchomboTs first prediction concerns wh)Duestions. =ccordin$ to their theory, if a sub'ect a$reement marker is an incorporated pronoun %and hence inde3ed to a sentence)topical referent*, then it should not be able to co)occur with Duestioned sub'ect +1s, assumin$ that %i* a sentence)topic +1 cannot also be in focus, and %ii* Duestioned constituents are in focus. =ssume for a moment that $ender sub'ect a$reement prefi3es were pure anaphoric a$reement markers, and so in sentences like %9;*, repeated from %59*, the apparent sub'ect was actually a topicalised ad'unct. %9;* ka)0 ara
+#4)oldPman

ka)0'wa*a0na)

=G4)passQ/6S)10E

an old man passed by If that were the case, then by the lo$ic outlined in the pre!ious para$raph we would e3pect $ender sub'ect a$reement to be impossible with Duestioned sub'ect +1s. <owe!er this is not the case, and both yi )ni Swhat, +#;T and ya)anu Swho, +#9T beha!e 'ust like strai$htforward le3ical nouns, tri$$erin$ $ender a$reement in %98* with yi )ni Swhat, ;F;

+#;T and

%9Fa* with ya)ani N ya)anu Swho, +#9T. yi )ni *ihwaaru)wa)na) y0i )ni yi )0hwaara)0wa)0na) what started 3orisino Isayb004.79;J ko)0ri si )no*ol ko)0ri si )no*o

%98*

/C31w!at A-3)startQ/6S)=116)10E

+#4)&arishen

%9F*

%a* ya)ani ;08ava)0na)


w!o

A-))wea!eQ/6S)10E

i0 ) taatu

+#;)mat

yi 0mpa)l
=G;)this

who wove this mat %b* zya)ani u)08ava)0na)


w!o 30)wea!eQ/6S)10E

i )0taatu yi 0 mpa)l
+#;)mat =G;)this

I2007)04)22.008J In #icipu there is nothin$ wron$ with the co)occurrence of a focused wh)word and a co) inde3ed $ender sub'ect a$reement prefi3, and this adds to the e!idence that $ender sub'ect prefi3es are ambi$uous a$reement markers. "he test also shows that wh)words beha!e in the same way as le3ical nouns with respect to person a$reement. -3ample %9Fb* abo!e was 'ud$ed un$rammatical, 'ust like the analo$ous %5Fb* abo!e with an ordinary, non)focused sub'ect. _uestioned sub'ects can co)occur with person a$reement, but only under the same circumstances as ordinary non)Duestioned +1s, i.e. +#9 nouns before !owel)initial !erb stems. In this en!ironment person a$reement is possible with ya)ani N ya)anu Swho, +#9T as the wh)word, but not with yi )ni Swhat, con!enience. %95* ya)anu wu0u0'i )ngo)
who 30)0A")$oPhomeQI// withR4S.1/2
+#;T.

#ompare %95* with %52* abo!e, repeated as %97* below for n1a)mul

who will go home with me Isaat004.009.050J %97* 0ku)ri )'aani ) wu0u0')p) 707a
+#9)&oran

30)0A")hold

+#9)person

the 3oran will hold a person "he fact that person a$reement is possible in %95* su$$ests that, like $ender a$reement, person a$reement prefi3es are ambi$uous a$reement markers rather than pure anaphoric a$reement markers.

;F8

<99#

!elativised constituents

=nother of (resnan and Mchombo:s predictions concerns relati!ised constituents, which they take to be uni!ersally topics27. In #icipu the same situation obtains for relati!ised sub'ects as for Duestioned sub'ects. Mostly we find $ender a$reementK %99* ka)0 ara
+#4)oldPman

ka0mpa) ka0na)
=G4)this

=G4)/-6

ka)0'wa*a0na)

A-1)passQ/6S)10E

pa*a

here

this old man who passed away here Isayb004.877J If the relati!ised sub'ect is not +#9, then person a$reement is impossibleK %99* %a* m0
+#8)child

ma0na) mo)0dooho)0no)
=G8)/-6

A-4)disappearQ/6S)10E

the child who disappeared %b* zm0


+#8)child

ma0na) u)0dooho)0no)
=G8)/-6

30)disappearQ/6S)10E

I2007)02)07.002J 1erson a$reement is howe!er possible, but apparently only in the same restricted en!ironments as before, e.$. with the head noun 707a SpersonTK %400* %a* 707a
+#9)person

n0na)

=G9)/-6

dooho)0no)

disappearQ/6S)10E

the one who disappeared %b* 707a


+#9)person

n0na)

=G9)/-6

u)0dooho)0no)

30)disappearQ/6S)10E

the one who disappeared I2007)02)07.002J If we accept (resnan and Mchombo:s assumption that relati!ised sub'ects are sentence topics, then the data further confirms that both $ender and person a$reement markers in #icipu are ambi$uous rather than pure anaphoric, but a$ain it does not ser!e to distin$uish the two kinds of a$reement. ^f[qYopqY~[qrq^wqY[^Yp[wpqnYjqovqq^ ambi$uous sub'ect a$reement markers and anaphoric ob'ect a$reement markers, the asymmetry between #icipu $ender and person sub'ect a$reement cannot be unco!ered by considerin$ sentence topics. In chapter 9 we will see that topicality is an important factor in the distribution of $ender and person a$reement across all person markers, not 'ust sub'ect prefi3es. <owe!er it is the notion of discourse topicality %L2.;.2* that turns out to be crucial, rather than sentence topicality.
27 =$ain sentence topic is meant rather than discourse topic %L2.;*.

;FF

<9<

Are sub,ect prefixes obligator-L

(efore attemptin$ to tackle the semantic and discourse factors $o!ernin$ the alternation between $ender and person a$reement, we should consider a third possibilityK the absence of a$reement alto$ether. e saw in L7.5.4 that o!ert sub'ect +1s are not obli$atory with either the $ender or the person sub'ect prefi3es. (ut what about the prefi3es themsel!esM =re they obli$atory, or can the sub'ect +1 sometimes be followed directly by a bare !erb stemM It has already been mentioned %L7.5.4* that o!ert sub'ect prefi3es are obli$atory before !owel)initial !erb stems. ith consonant)initial !erb o!ert sub'ect prefi3es may be absent with either +#9 or 21S sub'ectsG recall that the +#9 noun prefi3 %LF.F.7*, the =G9 a$reement prefi3 %L5.4.;*, and the 21S a$reement prefi3 %L7.8.4* were all analysed as ha!in$ ;0 as one of their allomorphs, in order to fill the $aps in the paradi$ms of otherwise o!ert prefi3es29. -3ample %404* shows an a$reement. %404* ana)
when
+#9

sub'ect with no o!ert sub'ect

k0kaa

+#9)woman =G9)that A-))$i!ePbirthQ/6S

vi 0lle)

;0mata)

when that woman gave birth Isayb004.459J =s well as nouns belon$in$ to


+#9,

members of another $roup of nouns can occur as

sub'ect without tri$$erin$ an o!ert sub'ect prefi3. "his is the same $roup of nouns for which person a$reement is possible before consonant)initial nouns %L7.5.4.4*K the independent personal pronouns, the demonstrati!e personal pronouns, and the word 707a SpersonT. e saw in %78* and %79* abo!e %the latter is repeated below for con!enience* that these nouns either tri$$er the u)0 ;1S sub'ect prefi3, or they occur without any o!ert prefi3. %402* %a* e0mpe) u)08a*a e0mpe) n) ti )0li pa*i

;S.1/2)this 30)surpassQ/6S

;S.1/2)this with

+#5)$oodness

this one is better than this one Ilit. Ssurpasses with $oodnessTJ %b* e0mpe)
;S.1/2)this surpassQ/6S

8a*a

e0mpe)

;S.1/2)this with

n)

ti )0li pa*i

+#5)$oodness

this one is better than this one %c* ze0mpe)


;S.1/2)this

vi )08a*a

A-))surpassQ/6S

e0mpe)

;S.1/2)this with

n)

ti )0li pa*i

+#5)$oodness

29 See (aker %4995K24* and Siewierska %2008K22)28* for 'ustification of this approach.

;F5

=nticipatin$ the analysis in the rest of this subsection, the o!erall situation can be summarised as followsK 0able ]a- Agreement possibilities be ore 5cinitial verb stems (y"e o' sub4ect /P +on)+#9 noun
+#9

Possible "re'i$es be'ore a C1initial stem Gender a$reement only vi0 %=G9* or ;0 %=G9* u)0 %;1S* or lack of a$reement

noun

Independent personal pronoun, person)marked demonstrati!e pronoun, 707a SpersonT

"he problem here is how to handle the lack of o!ert a$reement shown by the last two cate$ories of nouns in "able 50. e saw in L5.4.; that many #icipu a$reement tar$ets
+#9

offer either vi0, -0 or ;0 a$reement prefi3es with

nouns as controllers. (earin$ in


+#9

mind that many +#9 nouns also ha!e ;0 noun prefi3es %LF.F.7*, the most strai$htforward analysis seems to be that the ;0 prefi3es before !erbs and other tar$ets with controllers are true null =G9 a$reement prefi3es, as indicated in the table. 0or the person)marked +1s in the third row it would be con!enient if the ;0 prefi3 could be $lossed as a null allomorph of the ;1S person prefi3 u0, since this would allow us to maintain that sub'ect a$reement is always obli$atory. <owe!er there is no independent e!idence for this from the other tar$ets that allow person a$reement29. 0urthermore, plural personal pronouns %40;)408* and other plural +1s lackin$ a $ender specification %40F* may also co)occur with !erbs lackin$ o!ert morpholo$y, and so to maintain this ar$ument there would ha!e to be a proliferation of null allomorphs. %40;* e)re)
3P.P#6 for$etQ/6S

panda)

they orgot Ieamd008.0J %408* o)tu)


1P.P#6 understandQ/6S

gaanu)kwa)

$e
+-G

we don't understand I"ikula, sa$b004.;;7J

29 i.e. the demonstrati!es and the article. "he copula and the wh)word 0e)ne ha!e !owel)initial stems, and so some kind of a$reement is obli$atory. See L9.9 for details.

;F7

%40F* ti 0na)

=G5)/-6

aa Agri )$
P;

agriculture sayQ/6S)10E heatin$

* hyaa0na)

heating Isamoh004.097J

how the &griculture people say heating 2!erall it seems best to analyse the lack of o!ert person a$reement as the lack of a$reement alto$ether. It has not been possible to determine when $enderless)+1s such as e0mpe) and 707a will tri$$er person a$reement and when they will not, and so no more will be said about this alternation with respect to $rammatical a$reement. In clauses lacking a sub'ect +1, there is almost always an o!ert sub'ect a$reement marker on the !erb. So, for e3ample in %405* below, the only possible allomorph of the
+#9

$ender a$reement prefi3 is the %405* z>vi )0@7amu)kwa) i 0 ) ri


A-))beQ/6S

=G9

allomorph vi);0. "he correspondin$ sentence

without o!ert a$reement was 'ud$ed un$rammatical.


+#;)thin$

yi R;05ahanna

=G;R+#9)destruction

Iv0m) Smonkey, +#9TJ it is a thing o destruction

Itats002.004.028J

"here are a handful of instances of this un$rammatical structure in the corpus, but they are each either sentence fra$ments repeatin$ part of a pre!ious clause, as in %407*, or hesitations with an immediately precedin$ sub'ect +1, as in %409*. %407* ;0ku)ri )'aani ) vi )0si ) 0p)l)0pl)
+#9)&oran =G9)<=()knockPdown)/-.A1

707a.

+#9)person

;0p)l)0pl)

A-))knockPdown)/-.A1

707a.

+#9)person

0he 3oran knocks down people. 3nocks down people. Itats00F.004.0;2J %409* ;0gogoro >.@ B0he'wQi )sRe)0he'wi )se)
+#9)$in A-))dryPupQ/6S#=ASs)/-.A1

707a

+#9)person

gin ..2 dries people up Itats009.002.00;J If the non)o!ert a$reement in these e3amples reflects, as has been assumed here, a true null prefi3 rather than the absence of a$reement, then this =G9 ;0 allomorph seems to be a true $rammatical a$reement prefi3, since it cannot occur without an e3plicit sub'ect +1.

;0 "he ;1S person sub'ect prefi3 u0 would also be a possibility, as lon$ as the monkey was a discourse topic %see L9.8*.

;F9

<9=

Comparison with Central +ambari

(efore lea!in$ this section, it is worth takin$ a moment to look a$ain at the $enderNperson alternation in the related lan$ua$e, #entral &ambari %L2.2.8.2.;*. "here are two ma'or differences between the alternation in #icipu and #entral &ambari. 0irst, in #entral &ambari $ender sub'ect a$reement is $rammatical rather than ambi$uous, and is not found without an e3plicit +1 %#ro>ier 4998K249*;4. Secondly, in #entral &ambari person sub'ect a$reement is more readily found after le3ical sub'ect +1s than in #icipu. It seems that in #icipu neither kind of a$reement has pro$ressed as far alon$ the $rammaticalisation cline as their #entral &ambari eDui!alents. Gender a$reement in #icipu is sometimes anaphoric %althou$h usually it is not* and so it can be classified as ambi$uous a$reement, and placed further to the left of #entral &ambari $ender a$reement on Siewierska:s %4999* scale. 6ikewise person a$reement in #icipu is usually anaphoric, and should be placed close to the far left of the cline. "he situation in the two lan$ua$es can be !isualised as in 0i$ure ;7K =naphoric #entral &ambari #icipu s Person Person -ender =mbi$uous s Grammatical -ender

+igure 6_- Relative positions o person and gender agreement in 5entral 3ambari and 5icipu on diewierska's grammaticalisation scale So while in both lan$ua$es there is an en!ironment in which the alternation between the two a$reement patterns is found, it is a different en!ironment in each lan$ua$e. In #icipu the alternation is mainly found when there is no sub'ect +1, in other words when there is anaphoric a$reement. In #entral &ambari, on the other hand, the contrast is after le3ical +1s, as in <offmann:s %495;* e3amples Duoted in L2.2.8.2.; %2.2;* and repeated belowK;2

;4 Siewierska %4999K2;9)2;9* notes that this is actually !ery rare outside Indo)-uropean. ;2 I discussed the #icipu eDui!alents of <offmann:s e3amples with one of my lan$ua$e consultants and he 'ud$ed them unacceptable, unless there was a pause inserted between the +1 and the !erb. In that case, the +1 is most likely e3tra)clausal rather than a true sub'ect.

;F9

%409* %a* ma0nun


+#8)bird

u0kuwete
30)die

the bird died %b* ma0nune me0kuwete


+#8)bird A-4)die

a bird died

3.3

4erson mar,er summary

e ha!e now e3amined the fi!e different person marker paradi$ms in #icipu. I concluded that there are two paradi$ms of free pronouns , one person)marked, and one $ender)marked. "he independent personal pronouns contrast with a paradi$m of person) marked ob'ect clitics, which attach to the ri$ht)hand ed$e of the !erbal word. S#litic) doublin$T is not possible. "he sin$ular members of the ob'ect clitic paradi$m ha!e reduced counterparts with particularly interestin$ phonolo$ical properties. "here are also two paradi$ms of sub'ect a$reement prefi3es, one $ender)marked and one person) marked. (oth sets of prefi3es are ambi$uous between anaphoric a$reement and $rammatical a$reement accordin$ to (resnan and Mchombo:s %4997* typolo$y, but the $ender)marked prefi3es are closer to the S$rammatical a$reementT end of the $rammaticalisation cline. "able 54 summarises the morpholo$ical and syntactic status of the fi!e person marker paradi$ms. 0able ],- dummary o properties o 5icipu person markers Paradigm Independent 1-/S2+ Independent G-+.-/ 2b'ect clitics Sub'ect 1-/S2+ 8or"!. -rammatical >eatures -rammatical Ana"!oric status 'unction agreementH agreementH ord ord #litic 1refi3 =ny =ny 2b'ect Sub'ect 1erson Gender 1erson 1erson +o +o +o Sometimes, depends on sub'ect +1 Asually, depends on sub'ect +1 Bes Bes Bes Bes

Sub'ect G-+.-/

1refi3

Sub'ect

Gender

Bes

I also determined the a$reement

1/-/-_AISI"-S

for the two kinds of sub'ect a$reement


+#9

%0i$ure ;5 abo!e*. "here are two distinct en!ironments where prefi3es from either paradi$m can occurK %i* before !owel)initial !erb stems with controllers, and %ii* in the absence of a sub'ect +1 %i.e. anaphoric a$reement*. 2therwise the features encoded ;50

on the tar$et are a strai$htforward match for the features encoded on the sub'ect. "he syntactic status tests discussed in L7.5 were inconclusi!e and sometimes pointed in conflictin$ directions, as summarised in the followin$ tableK 0able ]4- dummary o tests or syntactic status o the sub%ect pre i*es 0ection (est L7.5.4 L7.5.2 L7.5.; L7.5.8 L7.5.F Multi)representation 6utcome
G-+.-/ 1-/S2+

closer to $rammatical a$reement than

+umber of ar$ument roles Inconclusi!e inde3ed on !erb /eferentiality .escripti!e content (alance of information 1oints towards $rammatical a$reement for both 1oints towards $rammatical a$reement for both
1-/S2+ G-+.-/

closer to $rammatical a$reement than

L7.5.5.4 _uestioned sub'ects L7.5.5.2 /elati!ised sub'ects

+either G-+.-/ nor 1-/S2+ is pure anaphoric +either G-+.-/ nor 1-/S2+ is pure anaphoric

2ther than multi)representation, the tests su$$est that, if anythin$, person a$reement markers are closer to true a$reement markers than incorporated pronominals. <owe!er as we saw in L7.5.4.4, when we look at the distributional data in the corpus, person prefi3es rarely take part in $rammatical a$reement. 1erhaps the only thin$ we can say for sure is that neither of the sub'ect prefi3es is a pure anaphoric a$reement marker. "he analysis in the followin$ chapter will pro!e to be more fruitful when it comes to characterisin$ the differences between these two paradi$ms. "here I will consider the factors that influence the $enderNperson alternation in the en!ironments which offer a choice of feature, in other words the a$reement
#2+.I"I2+S.

(efore doin$ this, howe!er,

it is important to be clear that we really are dealin$ with separate paradi$ms of $ender and person markers, rather than one comple3 paradi$m.

3.5

+eatures involved
hat does seem to be rather rarer is for a

It is not uncommon for the !arious a$reement tar$ets of a lan$ua$e to differ with respect to their o!ert features %#orbett 2005K79*. single tar$et to !ary in its a$reement features. In the case of #icipu, it is important to make sure that what we ha!e so far been callin$ S$enderT and SpersonT a$reement really do contrast with each other in re$ard to the a$reement features
G-+.-/

and 1-/S2+. =s an

alternati!e analysis we mi$ht posit a sin$le comple3 paradi$m in!ol!in$ $ender and ;54

person, with $ender as a subdi!ision of third person. "his is assumed by Siewierska %2008K408* in her representation of &iswahili person markers, and also by %497;K475* with respect to (antu as a whole. because of the implicit contrast with the state of affairs in #icipu. In a number of (antu lan$ua$e $rammars, there are references to personal pronouns in connection with !erbs or with e3pressions of possession %which represent only one of many uses of the associati!e construction*. "he implication seems to be that morphemes or sets of morphemes meanin$ SheNshe, himNher, hisNherT and Sthey, them, their %personal*T belon$ with first and second person pronoun forms, and that referents to non)personal nouns ha!e some kind of a secondary status. Actually, o course, the re erents or personal nouns, singular and plural, are nothing more than the concords or nouns o class , and 4. ' there is to be any dichotomy, it should be between irst and second person morphemes on the one hand and all class concords, including those or classes , and 4, on the other. =ctually, first and second person morphemes can also be treated with the noun class and concord system. "here are no first and second person noun prefi3es, to be sure, and first and second person concepts are hardly e3pected with attributi!es or demonstrati!es. (ut there are sub'ect and ob'ect forms, and forms used after the associati!e morpheme %indicatin$ possession*. 0urther, there are independent referents for first and second person sin$ular and plural, and also for nouns of all classes, which must be treated in a uniform way...In the li$ht of these similarities, althou$h the parallelism is not perfect, it would seem more ele$ant to include first and second person forms in the concord system Imy italics , S.M.J. "he contrastin$ distributions of the $ender and person sub'ect prefi3es %L7.5* ar$ues a$ainst this approach for #icipu, as does the fact that <)tone spreadin$ distin$uishes between them, affectin$ only the $ender prefi3es %L;.8.;*. +e!ertheless, if we were to i$nore this e!idence and try to construct a composite paradi$m for #icipu, then we mi$ht start as in "able 5;, which presents all the possible realis sub'ect;; a$reement forms that occur before a consonant)initial !erb, e.$. dukwa S$oT. 0able ]6- dub%ect agreement pre i*es or verbs in the realis mood .single lat paradigm2 4S n0 2S ;S 41 ti0 21 i0 ;1 a0
=G4 =G2 =G; =G8 =GF =G5 =G7 =G9 =G9

elmers

elmers is worth Duotin$ at len$th

-0:;0 u0

ka0

ha0

yi0

ma0 mi0 ti0

wu0 vi0:;0 ku0

"heoretically it would be possible to treat these tar$et forms as showin$ a$reement in a sin$le feature with fifteen !alues. <owe!er, as in most lan$ua$es, there are certain relationships between #icipu pronouns of a $i!en person that su$$est we need at least the
+AM(-/

and

1-/S2+

features, in particular the anticipatory plurals discussed in

;; =ny person marker could ha!e been chosen here, since they all ha!e an identical paradi$m structure. Similarly the irrealis mood makes the same distinctions as the realis.

;52

L8.8.F.8. 2nce we add


+AM(-/

and

1-/S2+

as separate features, then the paradi$m becomes

more comple3 if we still try to include the cells for the $ender)marked prefi3es. "his is because the class markers differ with respect to numberK some are sin$ular, some are plural, and some can occur with both sin$ular and plural nounsK 0able ]J- dingle paradigm with &)RdN and 5LAdd as a subset o
1-/S2+ +AM(-/ SG 16 &)RdN

;
#6=SS .-0=A6"

4 ka0

; yi0

8 ma0

F mi0

5 ti0

9 vi0

9 ku0

/0 -0:;0 ti0 i0

u0 a0

wu0 vi0

ha0 yi0

+ote that the u)Na) cells ha!e been included alon$ with the noun classes, and are bein$ considered as the morpholo$ical realisations of two Sminority tar$et $endersT in #orbett:s %4994K450* terminolo$y. "hey Dualify as such because there are no nouns that must tri$$er these a$reement forms, but e!ery noun may tri$$er them. In fact, the paradi$m in "able 58 is still not complete, since we ha!e not taken into account sin$ular classes, which typically contain abstract or mass nouns. "hese may show the same morpholo$y and a$reement as the plural classes 2, ;, F, and 9, and yet still tri$$er u0 a$reement under special circumstances %see L9.8.;*. "hus a further re!ision of the paradi$m is reDuired %"able 5F*, and it should be clear that the interaction between person, $ender and number is too comple3 for all the a$reement forms to be insi$htfully represented in the same paradi$m. 0able ]T- Revised single paradigm with person and number eatures but not gender
1-/S2+ +AM(-/ SG 16 M=SS

;
#6=SS

.efault 4 u0 a0 u0 ka0 ka0

2 ha0 ha0

; yi0 yi0 yi0

8 ma0

F mi0

5 ti0

9 vi0

9 ku0 ku0

/0 -0:;0 ti0 i0

wu0 vi0 wu0 vi0

ma0 mi0 ti0

Instead it is much simpler to construct two paradi$ms, one in!ol!in$ the features I1-/S2+J and I+AM(-/J, and the other in!ol!in$ the feature I#6=SSJ, correspondin$ to #orbett:s tar$et $enderK

;5;

0able ]]- &erson1number paradigm


0 P

1 3

/0 u0

ti0 a0

-0:;0 i0

0able ]_- Noun class paradigm 4 ka0 2 ha0 ; yi0 8 ma0 F mi0 5 ti0 7 wu0 9 9 vi0:;0 ku0

In summary, we ha!e two distinct paradi$ms of a$reement markers affectin$ the same a$reement tar$et. 1erson a$reement in!ol!es 1-/S2+ and +AM(-/K
1-/S2+K +AM(-/K

4, 2, ; sin$ular, plural

Gender a$reement in!ol!es G-+.-/ %or #6=SS*. "ar$et $enderK 4, 2, ;, 8, F, 5, 7, 9, 9

3.6

)hapter summary

"he bulk of this chapter %L7.4)7.5* $a!e an account of the phonolo$ical, morpholo$ical, and syntactic properties of the fi!e paradi$ms of #icipu person markers, and was summarised in L7.7. In particular, L7.5 set out the prereDuisites necessary for the two different kinds of sub'ect a$reement, which turned out to in!ol!e le3ical, phonolo$ical, and morphosyntactic factors. "hese were summarised in flowchart form in 0i$ure ;5. In the followin$ chapter I turn to the conditions which bear on the choice of $ender or person a$reement, in those conte3ts where the prereDuisites for both are met.

;58

Chapter = Gender and person agreement in Cicipu discourse


5.1 Introduction

"he main aim of this chapter is to show that both the semantic notion of animacy and the discourse)pra$matic notion of topicality are necessary in order to fully describe the alternation between $ender and person a$reement introduced in the pre!ious chapter. /ecall that chapter 7 identified the a$reement prere$uisites %le3ical, phonolo$ical, and morphosyntactic* rele!ant to this alternation on sub'ect prefi3es. <ere I will show that animacy and discourse topicality are a$reement conditions %L2.2.8.4*, in other words factors which help determine the relati!e likelihood of the two competin$ a$reement patterns occurrin$, $i!en that the prereDuisites for them both ha!e been met. "he a$reement conditions discussed in this chapter apply to all three of the morphosyntactic positions discussed in chapter 7K sub'ect prefi3es, post)!erbal ob'ect, and elsewhere. In the case of the sub'ect prefi3es, we will in this chapter only be concerned with those that meet the prereDuisites for both kinds of a$reementK %i* anaphoric a$reement prefi3es, and %ii* $rammatical a$reement prefi3es with
+#9

controllers before !owel)initial !erb

stems %recall 0i$ure ;5 from chapter 7*. If the prereDuisites for either $ender or person a$reement are not met, then there can be no competition in a$reement, and hence no a$reement conditions. =s a preliminary obser!ation, it should be stressed that re$ardless of potential conditions on a$reement such as animacy or discourse topicality, the initial referrin$ e3pression has an important role to play in determinin$ the features inde3ed on the subseDuent anaphors. If the initial referrin$ e3pression is marked for $ender, then this increases the likelihood of subseDuent person markers a$reein$ in $ender. #on!ersely if the initial referrin$ e3pression is unmarked for $ender %e.$. a personal pronoun*, then subseDuent anaphors are far less likely to be marked for $ender. If a speaker wishes to refer e3ophorically to some ob'ect, a tape)recorder, say, then they ha!e two choices of Snear)speakerT demonstrati!eK %i* e0mpe), which is the ;1S form appropriate for any %sin$ular* ob'ect, re$ardless of animacy or of the $ender of the noun most commonly used to refer to it, or %ii* vi 0mpa), the =G9 demonstrati!e which a$rees in $ender with the noun ;0ri )kooda) SrecorderT. <owe!er once the recorder has been introduced usin$ one or

;5F

the other of these terms, subseDuent mentions are likely to match the features and feature !alues of the ori$inal form. So while either of the SconsistentT con!ersations %4a* or %4b* would be natural, a Smi3T of differently)marked antecedent and anaphor would be less likely4. %4* %a* =K e0mpe)
30.P#61t!is

hi na)l
"21

(K u)08a*a

30)surpassQ/6S ;S.1/2)this with

e0mpe)

n)

* ti )0li pai

+#5)$oodness

"hat about this one %b* =K vi 0 mpa)


A-)1t!is

It's better than this one. (K vi 0 ) 8a*a


A-))surpassQ/6S =G9)this

hi na)l
"21

vi 0 mpa) n)

with

* ti )0li pai

+#5)$oodness

"hat about this one

It's better than this one.

I2007)02)4;.005J

In such cases the sub'ect markin$ is nothin$ to do with animacy or discourse topicality, but it is influenced to a stron$ de$ree by the form of the initial referrin$ e3pression. Gender)markin$ could occur in %4a*, in the sense that it would be $rammatical, but would probably be re$arded as Sclumsily)putT %see L2.;.;*. +e!ertheless, howe!er influential the initial referrin$ e3pression may be in $uidin$ the form of future anaphora, it is not the only factor. =lthou$h in the scenario 'ust presented a chan$e of feature would be unlikely, there are other circumstances where a switch from $ender)markin$ to person)markin$ is more probable. "o anticipate the later analysis, we will find that animacy and discourse topicality are both positi!ely correlated with person a$reement. +either of the two notions is sufficient on its own, and moreo!er the two conditions cannot be SrankedT in the manner of traditional 2ptimality "heory, since neither SconstraintT consistently wins out2. "he most we can say is that sub'ect a$reement markin$ is conditioned on both inherent topicality and discourse topicality %see L2.;*. /eferents that are $ood inherent topics are almost %but not Duite* always coded usin$ person a$reement morpholo$y. /eferents that are bad inherent topics almost %but not Duite* ne!er achie!e the status of bein$ person)marked. "he situation can be !isualised as in 0i$ure ;9, where li$hter shadin$ indicates a hi$her probability that the !erb will a$ree in person rather than in $ender.

4 "he e3ample $i!en in!ol!es a turn)transition, but the effect is 'ust as stron$ in monolo$ues. 2 =lthou$h see =issen %200;* for a discussion of !ariation in a$reement and Sweak bidirectional 2"T.

;55

Person

&iscourse topica"ity Gen6er Inherent topica"ity


+igure 6`- ^ariation in agreement according to inherent and discourse topicality /eferents situated at the bottom)left and top)ri$ht corners do not admit a choice , those that are neither inherently) nor discourse)topical are always coded with $ender a$reement morpholo$y, while those with both kinds of topicality always tri$$er person a$reement. /eferents with only one kind of topicality, on the other hand, may be found with either kind of a$reement. In addition to animacy and discourse topicality, there is one further a$reement condition on the $enderNperson alternation, this time in!ol!in$ the noun class of the controller. If the controller:s $ender is 9N2 %the default $ender for humans*, 9N; %the default $ender for animals and inanimates*, or the sin$le class $ender 9, then person a$reement is more likely than if the controller belon$s to any other $ender. 0inally, there is also an effect of SnaturalT $ender. =lthou$h #icipu, as is typical for +i$er)#on$o lan$ua$es, does not distin$uish male and female referents within its $ender system, this distinction does play a role in determinin$ the likelihood of person !s. $ender a$reement. 0emale referents appear to fa!our $ender a$reement to a $reater e3tent than males. It is tentati!ely su$$ested here that this difference should be incorporated into the #icipu animacy hierarchy, rather than treated as an independent parameter. "he rest of this chapter is or$anised as follows. In L9.2 I e3plain the methodolo$y used for codin$ and analysin$ the corpus. Section 9.; pro!ides a preliminary o!er!iew of participant reference in #icipu, includin$ a brief discussion on marked and

;57

contrasti!e topics. "he key parts of the chapter are L9.8, in which indi!idual e3amples from the corpus are chosen to demonstrate the combined effect of animacy and discourse topicality, and then L9.F, where it is shown that there is a codin$ pro$ression within para$raphs from le3ical +1s s $ender a$reement s person a$reement. Section 9.5 considers two possible alternati!e e3planations for the distribution of $ender and person a$reement. "he final two sections take a step back from the data , L9.7 asks which of the two kinds of a$reement is SmarkedT and which is SunmarkedT, while L9.9 considers how the two a$reement paradi$ms can be understood in terms of codin$ wei$ht and anaphoric !s. deictic reference. 0inally L9.9 turns to the remainin$ a$reement tar$ets that alternate between $ender and person a$reement, and considers to what e3tent the analysis de!eloped for person markers also applies to them.

5.

9ata analysis

"he methodolo$y in!ol!ed in collectin$ the data used for this study was set out in L4.8. In summary, the corpus consists of appro3imately si3 hours of transcribed and interlinearised audio and !ideo recordin$s, supplemented by elicitation sessions with nati!e speaker consultants. In addition to more traditional $enres such as folktales and historical narrati!es, I also recorded people talkin$ about a number of set topics of !aryin$ animacy, in order to stimulate te3ts with a hi$her density of a$reement markers with non)human controller referents. "his section describes how the corpus was analysed and coded. "he corpus is small accordin$ to the standards of corpus lin$uistics, with only about twel!e thousand clauses;. +e!ertheless, an e3hausti!e codin$ of e!en a small proportion of this corpus, co!erin$ the identification of discourse units and para$raph topics, and the enumeration and trackin$ of all referents and their a$reement markers was beyond the scope of this thesis. "he crucial e3amples which most clearly demonstrate the intra)para$raph pro$ressions from $ender to person a$reement are spread thinly throu$h the corpus, and such an enterprise, while it would ha!e increased the methodolo$ical ri$our of this study, would ha!e consisted mainly of the encodin$ of thousands of the less)interestin$ discourse)topical human referents, which always tri$$er person a$reement and therefore admit no choice. 6ikewise, the ma'ority of non) human referents are non)topical and do not persist in the discourse. 2nly rarely are they
; =lthou$h see 6@pke %2005* on Ssmall is beautifulT.

;59

encoded as le3ical +1 sub'ects, and when they do they always tri$$er $rammatical $ender a$reement, a$ain offerin$ no choice. It was considered to be a better use of the time a!ailable to take a more Dualitati!e approach by concentratin$ on the more interestin$ e3amples, and these are discussed at len$th in L9.8 and L9.F. "hus my approach to the data was to consider all cases of anaphoric a$reement, apart from the ma'ority cases where the referent was human and tri$$ered only person a$reement, in which case it was i$nored. In particular I was interested in the way in which a referent was coded between le3ical mentions. "he followin$ properties were coded for each person markerK 0able ]`- &roperties coded or person markers Pro"erty =nimacy 6in$uistic form .iscourse topical Categories human, spirit, anthropomorphised animal, animal, inanimate sub'ect prefi3, ob'ect clitic, free pronoun, possessi!e pronoun yes, no

=$reement features $ender, person, null

"he first three properties are simple to determine8. "he last property, discourse topicality, is much harder to code independently of the $ender and person markers that %it is to be ar$ued* pro!ide part of the lin$uistic e!idence for it. Identifyin$ the statuses of referents in the minds of interlocutors can be difficult e!en when it is the lin$uist:s nati!e lan$ua$e %see +ewmeyer 4999K4;; for a pessimistic !iew*, and an undeniable disad!anta$e of the methodolo$y used here is the fact that the analysis has been carried out by a non)nati!e speaker, whose intuitions with respect to conceptual structure stem from his understandin$ of the -n$lish translations that he himself has prepared. .espite such difficulties, as "omlin et al. %4997K404* ha!e obser!ed, "he most con!entional method of lin$uistic analysis , the introspecti!e e3amination of discourse data , remains a central and important strate$y in discourse studiesF, and some of the ma'or problems associated with introspection as a research techniDue are eliminated by the reliance in this study on relati!ely SauthenticT discourse data, rather than e3amples constructed in elicitation sessions. Moreo!er, there are not too many !iable alternati!e research methodolo$ies when
8 "he absence of o!ert a$reement was interpreted accordin$ to the analysis in L7.5.7. F SIntrospecti!eT in the sense that the lin$uist reconstructs the discourse structure of the te3t throu$h introspection, rather than the data itself bein$ created throu$h introspection.

;59

it comes to the analysis of topics. "e3t)countin$ strate$ies %"omlin et al. 4997K404)402, Myhill 2004K45FX`Y{}wpYn{ Yope{qYqi|fegq~YjgYZ[\]^YXWYvqrqYiq^o[e^q~ Y [^ L2.;.2.4. "he verYeYZ[\]^Yn^~Yp[{Yweffnjernoer{Ypn{Yjqq^Y\qrgY[^f}q^o[nf and it is undeniable that this kind of methodolo$y can lead to a better understandin$ of the pra$matic function of referential e3pressions. <opper and "hompson %4998K744* talk about the SmanipulabilityT %or SdeployabilityT* of forms, notin$ that the codin$ of referents often depends on how important they are $oin$ to be in the discourse subseDuent to their introduction %i.e. their discourse referentiality as opposed to the semantic notion of ob'ecti!e referentiality , 1ayne 4997K258)255*. = strai$htforward measure of discourse importance is the number of times that referents $o on to be mentioned followin$ their initial introduction. 0or e3ample, Ja$$arTs %499;, 4999* research on <ausa narrati!es shows that if the initial +1 is marked with the specific indefinite determiner wani, then the referent is likely to be mentioned more often in the subseDuent discourse5, a correlation which holds re$ardless of either the animacy of the referent or the $rammatical function of the +1. "he use of Duantitati!e methodolo$y thus pro!ides hard e!idence allowin$ Ja$$ar to impro!e upon the earlier analyses of wani described in Ja$$ar %4999K85*. So why not take a similar approach to the present problem and test the correlation between the use of $enderNperson a$reement and the number of subseDuent mentionsM A priori we can predict that such an analysis would indeed show that person a$reement is associated with a hi$her number of subseDuent mentions than $ender a$reement, since there is of course o!erlap between Gi!]^ Y n^~ Y zeefqgh{ Y ~[qrq^o Y n||renwpq{ Y oe discourse topic %L2.;*. It is usually the case that what we e3plicitly refer to in te3ts are those thin$s which hold intrinsic interest for us. "he hi$h de$ree to which discourse topics %in .ooleyTs sense* inte$rate te3ts conceptually makes them prime candidates for continued lin$uistic e3pression, and con!ersely the more the speaker mentions a referent, the more likely it is that the hearer will be encoura$ed to construe the unfoldin$ discourse as rele!ant to that particular referent %i.e. it will become for him a discourse topic*. +e!ertheless this is by no means always the case, and while a te3t) countin$ approach would almost certainly re!eal $eneralities at the corpus le!el, it is not fine)$rained enou$h to account for the indi!idual e3amples to be discussed in L9.8. =s was stressed in L2.;, discourse topicality in!ol!es more than 'ust referential density
5 =s noted in L8.8.F.; the pre)nominal article in #icipu may well function in a similar way.

;70

, the interlocutors must also ha!e an intrinsic interest in the referent, and it is impossible to determine this solely by the inspection of formal properties of the te3t. "he other ma'or methodolo$ical strate$y mentioned by "omlin et al. is e3perimental studies. I ha!e already discussed the problems I e3perienced in this area in L4.8. +ow that the alternation between $ender and person a$reement is better understood, the possibility is open for more tar$eted e3perimental work in the future %see L9.;*. In addition to the properties coded for indi!idual markers, I also recorded the nature of the series of a$reement markers %i.e. anaphoric chain* that they took part in. "he followin$ a priori possibilities e3istK

#onsistent $ender)markin$ #onsistent person)markin$ Gender)markin$ followed by person)markin$ 1erson)markin$ followed by $ender)markin$ More comple3 mi3es of $ender) and person)markin$

e will see in L9.F that the last two possibilities are stron$ly disfa!oured in #icipu. In other words, series of person markers are either consistent with respect to the a$reement features they encode, or they show a pro$ression from $ender)markin$ to person) markin$.

5." =%1

4articipant reference and mar,ed topics in )icipu Participant reference

=s in many lan$ua$es %.ooley and 6e!insohn 2004K449)42;*, the way in which participants are introduced into a te3t in #icipu depends on their referential importance %#hafe 4998K99*, or to put it another way, the de$ree to which they will $o on to inte$rate the te3t, in the sense that the remainder of the te3t is construed to be about them %see L2.;.2.2*. "he first)mentioned discourse participant in a narrati!e %which often turns out to be the most important* may be introduced with a special presentational construction, particularly in folktales %e3amples ;)5 below*, which ha!e a lar$ely formulaic be$innin$ in #icipu. -3ample %2* is taken from a #icipu 1ear Story narrati!e %see L4.8*.

;74

%2*

wu0na) 707a
301A#(

/C)1"erson

v0i ),

A-)1C6P

ana)

when ;S)climbQ/6S

u)0kum a)

u08anga)
+#7)tree

there was a man, when he climbed a tree %;* aa ;0lwli )


P;

Itapf002.008.009J ;0naata)
+#9)smallPspider

/C)1largeFs"ider

h0e),
A-

1C6P ;1.1/2

e)re)

n)

and

'ig (pider and associates, him and dmall dpider %8* wu0na) 707a
301A#(

Isaat004.009.00FJ

/C)1"erson

go), elle) u)0yo*


(6P

that

;S)beQ/6S and

n)

$i 0 ) ni yaa
+#5)farm

a,1u0noo

62#R+#7)mouth

wu,1ku0jene)
=G7R+#9)ri!er

once there was a man, he had a arm by the side o a river %F* ;0naata) go
(6P

Isaat004.00;.00FJ

/C)1s"ider

ere

;0naata),

+#9)spider

* u)0yaa

;1.1/2

n)

and

ku)0longi ,

+#9)monitorPli>ard

;S)doQ/6S

tu)0mo$i )

+#5)friendship

(pider him and LiHard, dpider, he was riends %5* wu0na) ma)0gai
301A#(

Isaat004.005.00FJ

/C41sword

m0e)

A-41C6P

go
(6P

once there was a sword

Isamy002.00FJ

"he initial construction in each of these e3amples is a !erbless e3istential clause, consistin$ of an +1, often prefaced with the person)marked indefinite article %L9.9.4* as in %2*, %8* and %5*, and always followed by either the copula %2);*, the topicaliser go %8) F*, or both the copula and the topicaliser %5*. 2ther important participants may be introduced in a marked construction such as an e3tra)clausal left)dislocated +1, but more commonly they occur in the focal domain of a clause as the ob'ect of a !erb or instrumental preposition %e.$. they saw a..., he was with a...*. Mar$inal participants or props are mainly introduced as the ob'ects of !erbs or prepositions. +ew participants may also be introduced usin$ sub'ect +1s, contrary to #hafe:s %4998K92)92* Sli$ht sub'ect constraintT hypothesis. #hafe found that sub'ects in con!ersational -n$lish always e3press either referents that are acti!e or semi)acti!e, or that are tri!ial in importance. Important referents are ne!er introduced by +1s

;72

functionin$ as sub'ect. <owe!er in the nine 1ear Story narrati!es that I recorded in #icipu, in e!ery sin$le one the main character7 was introduced by the sub'ect of an intransiti!e motion !erb. In more traditional #icipu narrati!es the use of sub'ect +1s to introduce main characters is less freDuent, but it does happen. #learly the li$ht sub'ect constraint does not apply as stron$ly to #icipu as it does to -n$lish. =s has been pointed out %e.$. &horoun'aia and "olchinsky 2008*, #hafe:s hypothesis does not distin$uish between transiti!e and intransiti!e sub'ects, and .u (ois: %4997* 1referred =r$ument Structure is a better fit with the #icipu data, since it places no restriction on intransiti!e sub'ects %S*, only on transiti!e sub'ects %=*, which speakers prefer to reser!e for S$i!enT referents. 2nce a participant has been introduced it tends, unsurprisin$ly, to be referred to usin$ minimal codin$ %L2.;.2.F*, especially if it is human. 0or sub'ects the minimal codin$ is an a$reement prefi3, for ob'ects it is a clitic, and for non)ar$uments it is a free pronoun. =naphors may a$ree with their antecedent in either $ender or person, as discussed in chapter 7. Introduction with a full +1 and then minimal sub'ect codin$ is illustrated in %7*. +ote how $ender sub'ect a$reement occurs at first, followed by person sub'ect a$reement in the last clause of the e3tract. "his foreshadows the pro$ression from $ender to person a$reement that will be discussed in L9.F. %7* ana)
when ;S)climbQ/6S)10E then certain

u)0kum a)0na) : see 'i na) m0


/C41c!ild

707a

/C)1"erson

ya*a0na) :

arri!eQ/6S)E-+"

'i na)

certain

ma)0ya*a0na)

=G8)arri!eQ/6S)E-+"

e,1k0ke)eke :

62#R+#9)bicycle

when he had climbed 1 then a certain person arrived 1 a certain child arrived on a bike 1 ana)
when

ma)0ya*a0na)0na) :

A-4)arri!eQ/6S)E-+")10E

m0aya)

A-4)comeQ/6S

m)0plu)

A-4)putPdownQ/6S

0ke)eke

+#9)bicycle

u)0ka a)0na) :

30)takeQ/6S)E-+"

mo)0kulo)otu*
+#8)basket

mo)0to) :
=G8)one

when he arrived 1 then he came and he put down the bike and he took 1 one basket 1 Itapf002.008.048J Most discourse topics are human or at least animate. It is in our nature to talk about, and to be interested in, thin$s that are most like oursel!es. +e!ertheless, sometimes we do ha!e an interest in inanimate entities, in which case they may become discourse topics.
7 "he boy on the bicycle.

;7;

+e!ertheless the kind of topic chain we can see in %7* abo!e is rarely seen for inanimate entities, re$ardless of how interested in them the interlocutors mi$ht be. In part this is because inanimates cannot $enerally be a$enti!e and hence are rarely found as sub'ects in any case. (ut e!en in places where an inanimate referent is repeatedly encoded as sub'ect, len$thy unbroken anaphor chains are not the norm, and instead the le3ical +1 is repeated more often. It is as if, despite the discourse topicality of the referent, its lack of inherent topicality reDuires it to be more freDuently)mentioned than would be the case for a human or animate referent. "his is illustrated in %9* below9. /eferences to the &oran are in bold. +ote how freDuently the le3ical +1 occurs, despite the ob!ious discourse topicality of the referent and the lack of any competin$ referents. %9* 2& t!e 3oran, A-))has a !erse for e3ample of %.* of this thin$ %.* of SpermissionsT %.* of God %.* t!e 3oran A-))has in a permission of God %.* if you do, if you hold t!e 3oran %.* 2& fore!er %.* 2& I know for sure, t!e 3oran A-))knocks down people %.* knocks down people %.* t!e 3oran %.* A-))kills people %.* t!is big 3oran %.* when they place A-)1P#6, and then you touch A-)1P#6 %.* 2& A-))kills people Itats00F.004.028J "his property of inanimate referents should be borne in mind throu$hout this chapter, especially in L9.F on a$reement pro$ressions. 6on$ a$reement pro$ressions startin$ with a le3ical +1 are rare for inanimate referents, in lar$e part because the le3ical +1 is often repeated.

=%#

'arked and contrastive topics

In L2.;.2.5 we noted the difference between SunmarkedT and SmarkedT topics. =ccordin$ to .ooley:s %2007* use of the terms, unmarked topics, which are e3pressed usin$ minimal codin$, carry out the integration function of topics, whereas marked topics, which ha!e more than minimal codin$, often carry out the access function. Marked topics may also be used to set up a contrast between two or more different topics. #icipu has two o!ert topicalisers which may be used to mark topics, the particles go9 and hi na). It is not easy to pinpoint the difference between the two. (oth can be found after +1s, in which case they are commonly left)dislocated topics, althou$h they
9 0or reasons of readability and space the -n$lish translation is $i!en here and for other e3amples in this chapter. See the front matter for the key to the $rammatical markup. 9 "he tone on go is !ariable and not well)understood. <i$h)tone seems to predominate, howe!er, and in some cases the tone e!en seems to be e3tra)hi$h.

;78

may also occur at clause)le!el. Ti na), howe!er, is more likely to be found after +1s, and it is most often used for contrasti!e topics as in the metalin$uistic statement in %9*. %9* v0aari )
+#9)man =G9)one

vi )0to),

a0ari

+#2)man

hi na)
(6P

o)0yo*

;1)beQ/6S

gei

many

manu is one, as or menu they are many

Ieab$004.089J

vo often occurs when introducin$ ma'or discourse participants, as was noted in L9.;.4. +aturally such participants are also discourse topics for lar$e portions of the te3t. vo also re$ularly occurs in the protasis of conditionals %40*, or after sentence ad!erbials functionin$ as Sspace)buildersT %0auconnier 4998* or Spoints of departureT %.ooley and 6e!insohn 2004K59)59*, as in %44* and %42*. %40* n go) v0u0n)1mu),
if
(6P

2S)0A")$i!eQI//R4S.1/2

to)

2& 2S)sitQI//

i d0do)onu) I"idipo, saat002.00;.0F7J

i you'll give me meat, then sit down %44* anna) go h0u0yi )nda)
(6P

today

;1)0A")seeQI//

707a

+#9)person

n0na)

=G9)/-6

w0u0ra)a

;S)0A")eatQI//

today they'll see the person who'll win %42* ki ila) go saa ko)0to*o ti 0 ) saa
(6P

I"idipo, saat002.00F.082J

perhaps

or

=G4)one

41)mi$ht

ti yo
$et

perhaps we might have got every one

Isaat004.005.0F8J

"he topic)like nature of conditionals and other space)builders has been obser!ed by <aiman %4979, see also .ancy$ier and Sweetser 200FK42F*, and the fact that go in #icipu marks both discourse topics and space)builders pro!ides further e!idence for this connection. =lthou$h the particles go and hi na) are the most e3plicit indicators of topics in their access function, the rest of this chapter will not be particularly concerned with them. Most topics are not in fact marked with either of these particles, and in any case, this chapter focuses on the inte$ration function of anaphoric a$reement, since that is principally where the $enderNperson alternation is found. (efore closin$ the discussion on contrasti!e topics, it should be noted that they may also be e3pressed throu$h a contrast in dei3is. "he near)hearer demonstrati!e pronoun e0lle) %L8.8.;.8* seems to imply a continued use of a current topic, whereas the ;7F

yonder demonstrati!e e0''i nde) reacti!ates an earlier topic in contrast to the current one. "his can be seen in the followin$ e3ampleK %4;* I#onte3tK "his e3cerpt is taken from a discourse on tobacco and alcohol. =fter talkin$ about the dan$ers of alcohol, the speaker then mo!ed on to tobacco. "he discussion was summarised as followsKJ v0i nda) e0lle) u)0ko*o, ;0sana)di ya) vi ,1ta0a a), e0''i nde) kuma,
2S)seeQ/6S 30.P#61t!at ;S)dieQ/6S +#9)cause
=G9R+#5)tobacco

30.P#61yonder and

w0aya)

;S)comeQ/6S ;S)dieQ/6S

u)0ko*o,

s )

because

;0gogoro
+#9)$in

you see, this one died, because o tobacco, and that other one, he died, because o gin Isamoh004.2F9J

5.* Animacy and discourse topicality as agreement conditions


"he main aim of this section is to show that both animacy and discourse topicality are conditions on a$reement for #icipu person markers. 0or didactic purposes it seems best to hold the conte3t)independent property of animacy constant while !aryin$ the conte3t) dependent property of discourse topicality. So for each cate$ory of animacy startin$ from the inanimates and buildin$ up to human referents 40, I will consider the e3pression of first non)topical referents and then topical referents. If the hypothesis put forward at the be$innin$ of this chapter %which was dia$rammed in 0i$ure ;9* is accurate, then we e3pect that for each le!el of animacy topical referents will be more likely to be person) marked than non)topical referents. =nd as we ascend the animacy hierarchy, we also e3pect that the Scut)off pointT between $ender)markin$ and person)markin$ should become lower , in other words the more animate the referent, the less topical it should ha!e to be in order to achie!e person)markin$. In addition to animacy and topicality, the noun class of the controllin$ referent is also a condition for a$reement. -!erythin$ else bein$ eDual, nouns from class 9 %$ender 9N2 or 9N;, or the sin$le class $ender 944* are more likely to tri$$er person a$reement than nouns from other classes. "his condition is treated first %L9.8.4* since it is rele!ant for parts of the later discussion. #lear e3amples of the alternation where all rele!ant factors are held constant e3cept discourse topicality are !ery rare, only occurrin$ in the corpus SaccidentallyT
40 2nly third)person referents are rele!ant here , there is no $ender a$reement for first) and second) person referents. 44 I ha!e no data with respect to the plurals of $ender 7N9.

;75

rather than as a result of the topic)stimulation sessions. +e!ertheless, when such contrasts do occur they are strikin$, and pro!ide the clearest e!idence for the account bein$ put forward in this chapter. "herefore I will be$in the discussion of animacy and discourse topicality by considerin$ a Sminimal pairT of this kind %L9.8.2*. "he rest of the e!idence for the rele!ance of discourse topicality comes from $eneralisations across te3ts, and the remainin$ obser!ations in this subsection %L9.8.;* are of this kind42.

=&1
+#9

?oun class =
=G9

nouns are a special case, in that althou$h there are perfectly $ood

a$reement

forms for the person markers, in comparison to the other noun classes they are less often used. Instead we often find person a$reement where we mi$ht ha!e predicted $ender a$reement. "his is especially ob!ious in te3ts where the same referent is referred to usin$ more than one noun. In one topic)stimulation te3t about the crocodile, for e3ample, the crocodile is denoted both by the 9N2 noun ku)0yupu) ScrocodileT, and by its 9N; hypernym kwaaro) ScreatureT %from <ausa waro*, as can be seen in %48*, taken from the start of the te3t. In the part of the e3cerpt followin$ the 9N; noun %b)k*, the speaker pro$resses to person a$reement much more readily than in the part followin$ the 9N2 noun %l)t*.

42 =lthou$h see L9.F.2 for a te3t)le!el pro$ression in discourse topicality in!ol!in$ a sin$le referent.

;77

%48* a. t!e crocodile @ku)0yupu), /C*A N b. it:s a creature @kwaaro), /C)A N c which A-))is..., 30.P#6 N d. 30)is in the water all the time N e. li!in$ place)30.P600 is there, in the water N f. when 30)comes to the bank N $. then 301wants 301take some air N h. 301will be able to come outside, 301lies down and 301takes air N i. '. l. m. after a little time, 301$oes back to the water N whene!er not in the water N and A-*1P#6 t!e crocodile @ku)0yupu), /C*A N
A-*1is

k. 2&, 30.P#6, 301doesn:t understand N a fearful thin$ for people N

n. any water where there is said to be a crocodile @ku)0yupu), /C*A inside N o. if a crocodile @ku)0yupu), /C*A is sensed to be there inside N p. 2&, people won:t..., people will be afraid to enter that water N D. since, t!e crocodile @ku)0yupu), /C*A N r. s. t.
A-*)$rabs A-*)can A-*)will

people N

$rab a person N kill N

u. in the water N !. because a person in the water doesn:t ha!e wei$ht N w. when someone enters N 3. 30.P#6 when 30)comes and 30)takes someone and 30)pulls Itats002.005.004J = similar contrast is found across two separate topic)stimulation sessions in!ol!in$ the $ender 9 noun gogoro S$inT %a$ain borrowed from <ausa* and the $ender 8 noun mo)0yoo SbeerT respecti!ely. "he anaphors of gogoro pro$ress much more Duickly to person a$reement than those of mo)0yoo, e!en thou$h there is little or no difference between the two referents with respect to animacy. "his condition applies to both human nouns from $ender 9N2 as well as lower animates and inanimates from 9N;.

=&#

A Eminimal pairF showing contrast in topicalit-

#onsider the followin$ two e3amples, which come from separate te3ts. "he first

;79

e3ample %4F* is from a topic)stimulation session about monkeys, the second %45* from one about do$s. #ommon to both te3ts is that they include a section about a do$ chasin$ another animal , the monkey in the first instance, and a hare in the second. "he crucial difference is that the do$ is not a discourse topic in the first e3cerpt, but it is in the second. "he SchaseesT also differ with respect to topicality , in %4F* the monkey is a discourse topic and the central participant of the te3t. (y contrast, in %45* the hare is incidental to the discourse and is only mentioned in order to demonstrate a particular characteristic of the do$ , any other suitable animal could ha!e taken its place. 2bser!e the contrastin$ means by which the do$ is referred to in each of the e3amples %bold references are to the dog, underlined references are to the monkey and the hare*K %4F* n)
and ;S)seeQ/6S

w0i +nda)

): ma)0waa mo)0woso
/C41dog

=G8)barkQ/6S

hali )

e!en

m0u+uwa)

A-4)feelQ/6S

ku)0suu

+#9)smell

kweevi ) kw0eevi )

=G9);S.12SS

mo)'ungo)no) mo)0'ungo)0no)

A-4)$etPupQ/6S)E-+"

m)dni )vi ) m)0dn)1vi )

A-4)followQ/6SR;S.1/2

ni )ila8i : n1i 0 ) la8i :

andR+#;)speed

i it sees the dog bark 1 it eels its smell it gets up and it ollows it ast to), n1u)0suma) :
2& andR;S)runQ/6S ;S)lea!eQ/6S

u)0naha)

u)0gi ta)0na)

+#7)return)E-+"

ma7ama7a
hurriedly

3, when it runs 1 it won't come back in a hurry %45* to), looka)$i wu0na) ma)0waa, n)
2& time 301A#(
/C41dog

Itats002.004.0F;J ma)0di ya :
+#8)hare

and

m)0dn)

A-4)followQ/6S

n1u)0huna)

andR30)killQ/6S

ma)0di ya a,1ka0da a :
+#8)hare 62#R+#4)bush

3, some dogs lit. a certain dogu, when it ollows a hare 1 when it kills a hare in the bush 1 u)0si 0 ) ra)a $e 'u* : u)0si 0 ) ra)a $e :
30)<=()eat
+-G

there

30)<=()eat

+-G

see

unless 30)comePbackPhomeQI//

u0to)ono)

n)

with

m0e :

=G8)1/2

it doesn't eat it there 1 it doesn't eat it 1 then it comes back home with it 1 Itats004.004.0;9J In %4F*, where the monkey is discourse)topical and the do$ is not, the do$ is consistently inde3ed by sub'ect $ender a$reement %mA0*, while the monkey is consistently inde3ed with person a$reement %in a !ariety of morphosyntactic positions*. In %45*, where the do$ is topical, it is at first marked with $ender a$reement, but crucially, it $raduates to person a$reement and doesn:t re!ert to $ender a$reement, at least until after the ne3t ;79

le3ical mention followin$ this e3cerpt. "he hare, on the other hand, is either marked with $ender a$reement or not at all4;. =ll references to the do$ across both the e3amples are sub'ects, and in %4F* the referent of a sub'ect inde3ed for $ender acts on the referent of an ob'ect marked for person %m)0dni )1vi * ) , illustratin$ that it is not the semantic role of the referent that determines the tar$et a$reement features. It is also important to note that the propositions encoded in se!eral of the indi!idual clauses in %4F* are about the do$, and so the do$ is a sentence topic accordin$ to 6ambrecht:s approach %see discussion of the S$randmaT e3ample %2.;5* in L2.;.2.2*. It is only by considerin$ the role of the participants in the wider discourse %i.e. discourse topicality* that the differences in morphosyntactic codin$ can be accounted for. "he third intonation unit in %4F* %be$innin$ n1u)0suma)* pro!ides an interestin$ parallel to the second intonation unit in %45* %be$innin$ n1u)0huna)*. In both cases the pre!ious clause encoded the do$ as sub'ect by means of anaphoric $ender a$reement. In %45* the sub'ect remains the same $oin$ into the second intonation unit, seamlessly transitionin$ from $ender a$reement to person a$reement. <owe!er in %4F*, the sub'ect of the !erb u)0suma) is now the monkey, no lon$er the do$. +ote that apart from the association of person a$reement with topic that I am ar$uin$ for here, e!erythin$ else points to topic continuity. "he sub'ect is e3pressed usin$ minimal codin$ and there is no discontinuity of actionK the do$ is followin$ the monkey with speed in one clause, and then somethin$ is runnin$ off in the ne3t and ne$lectin$ to come back. <ad the speaker been makin$ a point that do$s often disappeared off on their own while chasin$ monkeys, then no doubt this sentence could ha!e been used with the do$ remainin$ as the sub'ect referent. <owe!er because the hearer understands that in %4F* the monkey and not the do$ is topical, the speaker can use the ;1S marker u0 without fear of ambi$uity. =nimal referents can only be encoded usin$ person a$reement if they are discourse)topical, and so u) can only refer to the monkey in this te3t. "his pair of e3amples also brin$s home the fact that discourse topicality is about more than 'ust referential density %recall the SfootballT e3ample from L2.;.2.2*. In %4F* four references to the do$ are made in the space of two intonation units, but it nowhere tri$$ers person a$reement. Simply talkin$ about a referent is not enou$h to make it a discourse topic. Instead, it is the speaker:s intrinsic interest that ele!ates freDuently) mentioned referents to the status of discourse topic, and, as far as non)human referents
4; See L8.5.8 for ob'ect omission.

;90

in #icipu are concerned, allows them to $raduate from anaphoric $ender a$reement to anaphoric person a$reement. e now mo!e on to look more methodically at the anaphors used to refer to non) topical and topical referents, mo!in$ up the animacy hierarchy from inanimates throu$h animals to humans. e will find that in principle, most referents can be found with either $ender or person a$reement pro!ided they are discourse)topical enou$h48. +e!ertheless human referents are far more likely to be found tri$$erin$ person a$reement than animal referents, which in turn are more likely to tri$$er person a$reement than inanimate referents. "his is true for each of the morphosyntactic positions co!ered in chapter 7K sub'ect prefi3, post)!erbal ob'ect, and elsewhere.

=&%
=&%1

Inanimates
Inanimates@ non.topical

/eferents which are both inanimate and non)topical rarely function as the sub'ects of clauses, and in clauses where they mi$ht be e3pected to occur as ob'ects %i.e. because they are a central participant in the !erb:s semantic structure* they are often omitted alto$ether %L8.5.8*. 2n occasion, howe!er, anaphoric chains of inanimate non)topical referents do occur. =s predicted by the dia$ram in 0i$ure ;9 such referents are without e3ception coded usin$ $ender a$reement. "wo e3amples are $i!en below. "he first describes how a monkey escapes from bein$ tied up. "he discourse topics are the monkey, and also the owner of the monkey. "he rope is only of interest because when it slips off the monkey can escapeG by itself it holds no intrinsic interest for the interlocutors. Similarly in the second e3ample, the hoe is only the means by which one of the main discourse participants is made to wake up. It is of no interest by itself. %47* ka)0manga ko)0si ro)0no)
/C11ro"e =G4)slipPoffQ/6S)E-+"

k)0p')

ka)0naha)

A-1)lea!eQ/6S

v0m)

+#9)monkey

'u*

A-1)remo!eQ/6S

ko)0yuwo)0no),

A-1)fallQ/6S)E-+"

wayPo!erPthere

the rope slipped down it came o it ell, it le t the monkey up there Isaat002.002.F;5J

48 = set of e3amples from the corpus can be inspected in =ppendi3 #, co!erin$ the 49 combinations of a$reement feature %32 , person, $ender*, animacy le!el %3; , human, animal, inanimate*, and morphosyntactic position %3; , sub'ect prefi3, post)!erbal ob'ect, and SelsewhereT*.

;94

%49* see

until

ku)0 eyi ku)0sa'u)0na)


/C*1!oe

ana) ku)sa'u)ni )vi ) ana) ku)0sa'u)0na)1vi )


when

=G9)touchQ/6S)E-+"

707a

+#9)person

'0'i nde) :

=G9)yonder

A-*)touchQ/6S)10ER;S.1/2

ku)gaata)ni )vi ) ku)0gaata)0na)1vi )

A-*)cutQ/6S)E-+"R;S.1/2

) : $i mmai ) : $i mmai
little

then the hoe touched that one 1 when it touched him, it cut him slightly 1 I"idipo, saat002.008.0F;J
=&%# Inanimates@ topical

+1s referrin$ to inanimate referents rarely tri$$er person a$reement, e!en when they are topical. In the whole corpus there are only se!enteen inanimate referents cross) referenced by person a$reement markers, ten of which surfaced in topic)stimulation te3ts deliberately desi$ned to stimulate this kind of infreDuent data. +e!ertheless, despite the rarity of the combination, a fairly wide !ariety of inanimate referents ha!e been obser!ed inde3ed by person a$reement markers. -3amples from topic)stimulation te3ts include beer, $in, tobacco, the (ible, and the &oran. -3amples from other kinds of te3ts include palm)wine, fish trappin$, trees, a lake, a fence, and clothin$. In each of the se!enteen cases the referent is a discourse topic, and in the ma'ority of cases there is no reason to think personification or metonymy mi$ht be in!ol!ed. =s obser!ed in the pre!ious subsection, person a$reement ne!er occurs with non)topical referents. -3amples %49* and %20* show person a$reement with +1s denotin$ the &oran and tobacco respecti!ely, taken from topic)stimulation sessions. "he topicality of the referents should be self)e!ident in both e3amples. %49* a. c e. f. t!e 3oran @ku)ri )'aani . ) /C)A =G9)forbids all kinds of thin$s N 30)forbids beer N you will do eatin$, you will eat you will eat you will eat you are full to burstin$? N 2& 30)forbids, 30)forbids that thin$ N

b. the thin$s 30)forbids, are two places, you see 30)forbids beer N d. 30)forbids, e!il eatin$ N

$. yes, like that 333 it is forbidden like that N h. but 30.P#6 t!e 3oran, 30)stands between us and God Itats00F.004.499J

;92

%20* a. c e. f.

2& 30.P#6 tobacco@ta0a a). /C+A when you are smokin$ A-+1P#6 N you will find your heart blackenin$, heart blackenin$ N 2& heart, near to the heart N 301will block the holes of the lun$s Itats007.002.0;;J

b. if you are healthy N d. until 301does, 301spoils for a person the lun$s that are inside N

"he information $i!en in %24* below was unsolicited, and comes from a passa$e about !arious methods for trappin$ fish. "he speaker lists se!eral different techniDues and describes each one. "he different kinds of trappin$ therefore form a series of contrasti!e discourse topics. "he
=G4

a$reement markers deri!e from a $eneral word for trappin$,

ka)0yi ma %+#4*. +ote how in the second line the person)marked article wu0na) %L9.9.4* is used to introduce a new kind of trappin$, in contrast to the $ender)marked ka0na) in the first line. /eference to this ma0ruwo trappin$ continues with person a$reement markers in the third line. %24* a)kwai
therePis

ka0na)
A-11A#(

) ka,1ma0gi ta : a)0si 0 hyaa


;1)<=()say
=G4R+#8)trappin$Pk.o.

ee, a0ku)m a) ma0ruwo :

gi ihwa :

yes ;1)climbQI// side


+#8)trappin$Pk.o.

ee, o)koo, ee, evi ,

) ki ,1i0ri 0 wu0na) a)0si 0 hyaa, mpa) :


;1)<=()say
=G4R+#;)thin$)this

yes therePis 301A#(

u)0yo*

n)1i )0ri 0mpa)a0ni ),

a)0ggaji )
+#2)shrubPk.o.

yes 30.P#6 30)beQ/6S andR+#;)thin$)this)+M6H

there is another Itrappin$ techniDueJ called o magitau 1 yes, they climb the side 1 yes, there is, another Itrappin$ techniDueJ called, o this thing 1 maruwou 1 yes, it, it is with this thing, agga%iu 1 Itats00F.002.40;J So there are indisputable occurrences of person a$reement with inanimate referents, as lon$ as they are discourse topics. +e!ertheless bein$ topical is not always sufficient to ensure that anaphors are person)marked, and in fact topical inanimate referents are more often encoded by $ender a$reement, as demonstrated by the remainin$ e3amples in this subsection. -3ample %22* is a con!ersational aside recorded durin$ a session on the history of the =kula di!ision of the =cipu. "he speaker had 'ust noted a sacrificial pot which had been knocked o!er. "he e3ample shows a nice contrast between the inanimate pot, which is encoded with $ender a$reement e!en thou$h it is a discourse topic %and clearly marked as such by its introduction usin$ a presentational construction*, and the human, ;9;

who is marked with person a$reement despite bein$ non)topical and unknown to the speaker. %22* v0i *nda)
2S)seeQI//

ma0ti llu me0lle), 'i na) 707a


/C41"ot A-41t!at =/" +#7)stand)+M6H

+#9)person

'ungQo)sRo)0no) m0e)
A-41P#6

rise#=ASsQ/6S)10E

m0e) :

A-41P#6

ma)0kama)

A-4)be.1S"Q/6S

u)0'i sanuu0ni , u)05ongu)lo)0wo)

;S)fallPdownQ/6S)=116

5ongu)lo)0wo)

fallPdownQ/6S)=116

you see that pot, someone caused it to rise up 1 it was standing, he1she pushed it over I"ikula, s!sdt004.070J "he ne3t e3ample comes from the summin$ up of a topic)stimulation te3t concernin$ SfireT. 6ike other non)indi!iduated nouns such as water or li$htnin$, fire pro!es to be hi$hly resistant to person)markin$. %2;* s )
because and likePthat

n)

'i ni ),

u)0laa

/C71'ire =G7)beQ/6S

wu)0yo*

n)1u)0'usu) : kuma
also

wu)0yo*

A-7)beQ/6S

kuma n)
also

and

) : wu)0yo* ti 0 ) li pai
+#5)$ood

andR+#7)power

A-7)beQ/6S

n)1u)0pa$i ) :

andR+#7)difficulty

because o that, fire has power 1 it is also good 1 it is also di icult Itats002.002.0F9J "he final two e3amples show relati!ely lon$ anaphoric chains. In the second e3ample %2F* the referent that meat of that day is introduced usin$ a marked topic construction with a resumpti!e pronoun, but ne!ertheless does not achie!e person a$reement. %28* a. =K there Mappaya it is like a small ca!e, t!e water @mo)0ni . stays clean N b. (K hmm N c =K clean? N d. e. only $oodness N they are scramblin$ to $et A-41P#6 N
A-41P#6, A-41would /C4A =G8)

f. (K and when you fetched A-41wouldn:t finishM $. =K h. i. '.


A-41finished

finish all at once or

but A-41didn:t finish Duickly

they did di$$in$


A-41spurted

sweetness? that of &okko:o in sweetness I"ikula, sa$b004.589J

A-41surpassed

;98

%2F* a.

you t!at meat @i )0nama). A-31P#6 yetM he said aha?

/C3A

o' t!at day e!en now you ha!en:t eaten

b. he said no? I ate B lon$ a$o c e. f. $. i. '. d. I knew thin$s would turn out like this really A-3)lasted a lon$ time
A-3)lasted A-3)lasted

a lon$ time

h. I know now for sure you:!e eaten A-31P#6 he said 2& since you:!e eaten A-31P#6 Isaat002.002.529J It is not clear why a speaker will sometimes choose person a$reement to refer to a topical inanimate referent and other times $ender a$reement, and it may not be possible to $i!e a deterministic account. =s mentioned in L2.2.8.4 some lin$uists ha!e treated topicality as more of an a$reement SprereDuisiteT than a SconditionT, such that the presence or absence of a$reement can be precisely determined accordin$ to the topicality of the controllin$ referent. #oncernin$ >ero anaphors in #hinese, for e3ample, 6i and "hompson %4979K;42* maintain that their distribution is in principle predictable, but that the principle contains !ariables dependent on the speaker:s perception of the pra$matic situation. "his position does not seem tenable for the #icipu alternation, since in the e3amples abo!e the referents concerned are indisputably discourse topics, and yet they are still not marked with person a$reement. In #icipu topicality is necessary %but not sufficient* for person a$reement to occur with an inanimate referent.

k. you:!e $ot to $i!e me one wife

=&&

Animals

In the te3ts I ha!e collected, and of course more $enerally in discourse, animal referents are rarely topics. =lthou$h they ha!e $reater potential than inanimates to take part in anaphoric chains, and are more often the sub'ects of clauses, they are rarely of intrinsic interest to the interlocutors. Instead the situations in which they take part re!eal somethin$ about or of interest to some other topical discourse participant, usually one of the speech participants or another human. =nimal referents fall in)between inanimates and humans in the !arious !ersions of ;9F

the cross)lin$uistic animacy hierarchy %e.$. Sil!erstein 4975, 0rawley 4992K9F*. Similarly in #icipu animals can be distin$uished from inanimates on the one hand and humans on the other with respect to the $enderNperson alternation. =s we saw abo!e, inanimate referents are rarely encoded usin$ person a$reement markers, e!en when topical. <umans, on the other hand, are rarely encoded usin$ $ender a$reement markers, e!en when non)topical. "he alternation has a more balanced character as far as animal referents are concerned, and thus animals reflect the importance of discourse topicality more clearly than other kinds of referents.
=&&1 Animals@ non.topical

+on)topical animal referents are without fail encoded usin$ $ender a$reement. "he do$ in %4F* abo!e was one such referent. =nother e3ample is $i!en in %25* below, taken from a folktale about a hunter stuck to a rock. "he hare is a minor character in this te3t, only bein$ introduced ri$ht at the end, in unit 420 of 4;2. "he sole purpose of the hare is to pro!ide a means of freein$ the hunter, who is the main character and a discourse topic. 2bser!e that the hare is able to speak in this story, and so has been anthropomorphised, and is therefore hi$her up the animacy hierarchy than normal animals. +e!ertheless it is not encoded usin$ person markin$, despite ample opportunity for this. "his is because it is not a discourse topic. %25* a. b. then A-4)came t!e !are @ma)0di ya. /C4A =G8)came A-4)said 2& $et ready, A-41came t!e !are@/C4A =G8)came A-4)went yonder far off N
A-4)came A-4)ran

towards A-4)came A-4)collided with him crash? N I"idipo, saat002.00;.427J

=&&#

Animals@ topical

In contrast to non)topical animals, topical animals are !ery often inde3ed usin$ person a$reement, and in fact the corpus does not contain any e3amples of topical animal referents which fail to pro$ress from $ender)markin$ to person)markin$. -3ample %27* comes from a topic)stimulation te3t about the tortoise, which is easily seen to be a discourse topic in the e3cerpt presented.

;95

%27* a. c e. f.

t!e tortoise @$i )0kulu. /C+A N when God created 30.6BL N because whene!er N 30)tra!els N

b. God created 30.6BL well N d. he beat 30.6BL to$ether with an iron bowl N

$. the sun doesn:t bother 30.6BL N h. if 30)feels the sun botherin$ 30.6BL N i. '. then 301enters self)30.P600 into house)30.P600 N then 30)hides... Itats008.002.005J "he ne3t e3ample comes from a folk history of a$riculture. "he donkeys are initially marked with $ender a$reement in %a* but as the discourse pro$resses they are referred to usin$ person a$reement markers. %29* a. you see t!e donkey @ma)0ja)ki i. 4C%A if you do $uineacorn, it:s that A-%1will carry the $uineacorn for you N when they took camels%+#2* they went with =G2)them N
A-%1t!em

b. 301will take for you to the house, camels%+#2* formerly it is said N c d. 3334F)were destroyin$ the farm it was said it would be better to return the camels and come back to donkeys@/C%A N e. f. they %people* would do for 3P.6BL %donkeys* load standin$ N after they finished loadin$, you see they %people* took 3P.6BL %donkeys* N Isamoh004.495J 2ne further e3ample comes from an inter!iew of an old man conducted by his son. "he main topic of the passa$e below is a particular horse, introduced by the <ausa loanword ma0ha)uka)$i Slunatic,
+#8T.

"he ne3t anaphoric reference is a $ender)marked pronoun

%b*, but thereafter the horse is referred to usin$ only person a$reement morpholo$y. "he use of the word ma0ha)uka)$i seems to indicate a certain anthropomorphism, and it mi$ht be thou$ht that this is the reason for the person)markin$ rather than topicality. <owe!er note the plural reference in %$*, which refers to the %non)anthropomorphised* horses in $eneral.
4F Anfortunately it is not clear from the recordin$ whether the camels are denoted here usin$ person % a)* or $ender %ha)* a$reement.

;97

%29* a. SonK

so you knew !orse @d0d*. /C)A wellM N


/C4A

b. 0atherK a? !orse now back then one lunatic @ma0ha)uka)$i . !orse we had, if &ooki wasn:t there N c. d. SonK f. $. h. i. '. k. there was no)one who could catch A-41P#6 N 2& N by Ma$u'i there, by "i>ebi N where they would find 3P.6BL in the marsh N $ra>in$ N with arri!in$ N the one who knew unless he called 30.6BL, now there N you would find 3s1would come N

o' a

e. 0atherK 2& $oin$ there, by Molloci N

Isayb004.F00J 0inally e3ample %;0* is in the form of a riddle. /iddle)tellin$ of course de!iates from the normal co)operation that is assumed to occur between interlocutors %Grice 497F*, in that the speaker does not want to be %too* easily understood. If the hearers do not know the riddle, then they do not ha!e the same referents and potential topics a!ailable to them as the speaker. =t one le!el the main discourse topic of a riddle is the answer, in this case Sa broomT. <owe!er on the surface le!el the broom is not mentioned at all, and it is the animate cows of our father that are the discourse topic, occurrin$ left) dislocated in a marked topic construction, with a person)marked resumpti!e pronominal clitic. %;0* i 0 ) naa
/C31cow

y1;0s

) a)0si 0 va)'a1re)

A-3E/C)1'at!er

vi 0 ttu)

A-)11P.P600

patti ) ka)0manga ko)0to)


all
+#4)rope

=G4)one

k0e

=G4)#21

;1)<=()tieR3P.P#6

all the cattle of our father it's with one rope they tie them

Isaat004.002.0F9J

=&(

Humans

<umans are of course the archetypal topics %e.$ &uno 4975, 1ayne 4997K4F4*, but it does not follow that all human referents encoded in a te3t are discourse topics. <umans can be of tri!ial importance in a te3t, despite their inherent topicality. If a referent fails to persist and therefore does not inte$rate the discourse schema for a te3t, or if the interlocutors ha!e no intrinsic interest in that referent, then it cannot be said to be a

;99

discourse topic. So we mi$ht e3pect to find some e3amples of non)topical human referents in the corpus, and thus it is a non)tri!ial Duestion as to whether they will be marked with $ender or person a$reement. In fact, the possibility of $ender)markin$ for human referents is restricted by a le3ical property of the word for SpersonT 707a, discussed in L7.5.4.4. "he ma'ority of male45 referents in the corpus are introduced into the discourse either by name, or usin$ an +1 headed by 707a, and neither of these cases allows the possibility of $ender a$reement. Moreo!er, most of the remainin$ +1s used to introduce human referents are headed by
+#9

nouns , recall from L9.8.4 that these also fa!our person rather than
+#9

$ender a$reement. /ulin$ out referents introduced by 707a and other

nouns lea!es

only ele!en corpus e3amples of nouns used to introduce adult humans 47, ei$ht of which ha!e discourse topical referents. "hose with non)topical referents will be discussed first.
=&(1 Humans@ non.topical

"wo of the three non)topical cases in!ol!e $ender sub'ect markin$ which persists not only beyond the initial !erb, but also into the ne3t intonation unit. "his can be seen in the followin$ two e3amples. -3ample %;4* is taken from a sermon about for$i!eness. "he point of the passa$e from which the e3cerpt is taken is to demonstrate certain facets of the character of Jesus, who is a discourse topic of this section %note the left)dislocated personal pronoun in the first intonation unit*. "he officials are of no intrinsic interest, and are only mentioned in order to reflect the nature of the discourse topic. "hey are introduced in this e3cerpt and they are not referred to a$ain. =lthou$h they do actually pro$ress to person markin$ at the end of the third intonation unit, $ender a$reement persists for lon$er than mi$ht ha!e been e3pected for human referents, i.e. beyond the intonation unit in which the +1 occurs.

45 0or female referents the word for SwomanT k0kaa is the Sbasic)le!elT term, rather than Q07a %at least for male speech*. See L9.8.F.; on Snatural $enderT. 47 #hildren in the corpus are usually introduced usin$ the word m Schild, +#8T %e.$. e3. 7 abo!e*. 2n the basis of the limited data a!ailable they seem to come in between animals and adult humans in the animacy hierachy.

;99

%;4* e)vi e)vi

;S.1/2

i )ri gaati i )0ri gaati

/C31o''icialFk.o.

yi )lwi )vi ) : yi )0lw)1vi ) :

=G;)dri!ePawayQ/6SR;S.1/2

yeewi )vi ) : ye0ewe)1vi ) :

A-3)hateQ/6SR;S.1/2

a)maa !eesu) ta'a)re) : a)maa !eesu) ta'a)1re) :


but Jesus

lo!eQ/6SR3P.P#6

him the senior officials drove him away 1 they hated him 1 but !esus loved them 1 Ioamy004.244J In e3ample %;2*, taken from a discussion about family history, the referent of their elder holds no intrinsic interest for the Duestioner. <e is tryin$ to disco!er when the inter!iewee first mo!ed to his present !illa$e, and is usin$ the childhood of the referent of their elder as a temporal standard for comparison. "he referent has no other role and plays no further part in this discussion, and so cannot be said to be a discourse topic. =s with the pre!ious e3ample, this may e3plain the persistence of the $ender a$reement markin$ across se!eral intonation units. %;2* a. c. they said t!eir elder @ka)0 ara. /C1A, A-1)$rew up here in this house N
A-1)$rew

b. 2& when A-11$rew up here, were you here N up here or, were you across there N Isayb004.098J
=&(# Humans@ topical

2f the ei$ht topical adult human referents introduced usin$ non)+#9 nouns, fi!e are encoded e3clusi!ely usin$ person a$reement. "he other three do admit anaphoric $ender a$reement, but only within the same intonation unit, in contrast to the non)topical referents discussed in %;4* and %;2* abo!e. hen the reference is taken up a$ain in the ne3t intonation unit, person a$reement takes o!er, as can be seen in %;;* and %;8*, taken from the same folktale. "he teacher and the prostitute are two of the main characters in the story, and are discourse topics of the respecti!e e3cerpts. %;;* ma0llu* w0aya)
/C41teac!er

m0aya) evi

=G8)comeQ/6S A-4)writeQ/6S

ma)0ru u)ta)

ma0taka)dda :
+#8)book =G7)30.P600

30)comeQ/6S 30.P#6 ;S)0A")showQI//

w0u0nu)una)

u)0matuwaa0ni we0evi )
+#7)ma$ic)+M6H

le*e

there

then the teacher wrote a little piece o paper, he was going to demonstrate his magic there Isahs004.00;.008J

;90

%;8* de)ge)lee ka)0kaaru)wa)


then

* ka)0yaa

/C11"rostitute

k0aya)

=G4)comeQ/6S

* ka)0yaa

A-1)doQ/6S

;0klli ya mo0ni

+#9)makeup

A-1)doQ/6S

;0klli ya : w0aya)
+#9)makeup

30)comeQ/6S

u)0ri va)

30)lookQ/6S

+#8)water

then the prostitute put on makeup she put on makeup 1 then she looked at the water Isahs004.00;.040J
=&(% E?aturalF gender

In addition to the three)way di!ision between humans, animals, and inanimates, there is also an ar$ument for settin$ up two sub)cate$ories for men and women, as dia$rammed belowK <umans Men omen <i$h probability of $ender a$reement <i$h probability of person a$reement =nimals Inanimates

+igure 6b- How animacy a ects the 5icipu gender1person agreement alternation "his subdi!ision between men and women must be re$arded as tentati!e. It is based partly on my own inspection of the corpus, and partly on discussion with lan$ua$e consultants. "he main e!idence from the corpus is the fact that althou$h there are relati!ely few mentions of women, most of the e3amples of $ender)marked human anaphors ha!e female referents. "his is true with respect to both pronounsNclitics and affi3es. In $eneral, post)!erbal ob'ects with human referents are almost always e3pressed by person)marked ob'ect clitics, rather than $ender)marked pronouns. "he few e3amples of $ender)marked pronouns in this position all in!ol!e female referents49. In %;F* the a$reement markers deri!e from ku)0la$i S$irl, 9N2T, the sin$ular form of )0l$i . %;F* ka'a) $i )0me'i )
now
+#5)inside =G5R/C

=G9

t,10l$i

1girl

ha0na) ho)0yo0no)
=G2)/-6

=G2)beQ/6S)10E

pa*a,
here

v0i nda)

2S)seeQ/6S

ku0na) 'ii,

;S)sayQ/6S yes ;S)seeQ/6S

* u)0hyaa

A-*1#7;

;0yo0no)

2S)beQ/6S)10E

a,1u0ta'a

62#R+#7)want

sul :
_

w0i nda)

=G9)1/2

kw0i ) :

now among the girls that are here, do you see the one that you want 1 he said yes, he saw her 1 Isaat002.002.42;J
49 Gender)marked pronouns with male referents can be elicited howe!er %2007)02)2;.004*.

;94

-3ample %;5* below shows a direct contrast between female and male referents. "he pronoun referrin$ to the woman is marked with $ender a$reement, while in the parallel construction afterwards, her husband is inde3ed with a person)marked ob'ect clitic49. %;5* n)0yo
4S)beQ/6S and more

n)

): kuma wu0na) ka)0maya


;S)=/"

/C11elderFsibling

' also have another elder sibling 1 ka0na)


=G4)/-6

ko)0yo0no)

=G4)beQ/6S)10E

u)0'i ta)

+#7)marry

k0e) : pa*a : a)0si 0 ) hyaa


here ;1)<=()say

A-11P#6

who is married here 1 they call her 1 m) 0panda) :


4S)for$etQ/6S

Oaahi )ya) :
InameJ

' orget 1 dahiya 1 n) n) vaari ) v0aari ) veevi ) ve0evi ) * ) a)si )hyiivi * a)0si 0 ) hyaa1vi ) 3amma) : 3amma) : Isaat002.009.004J

and

/C)1!usband A-)130.P600

;1)<=()sayR30.P#6 InameJ

and her husband they call him Iamma 1

In addition to pronounsNclitics, sub'ect a$reement markers also seem more likely to be inflected for $ender when they ha!e female referents. "he followin$ three e3amples come from the same folktale, and all in!ol!e topical female referents tri$$erin$ anaphoric $ender sub'ect a$reement. %;7* vi 0 lle)
=G9)that

t0to)

=G9)one

kuma, vi 0 ) si 0 ) 7i )no)07i no)


more

A-))<=()turnPinto)/-.A1

v0m)

+#9)monkey

and that one Ik0kaa Swoman, +#9TJ, she habitually turned into a monkey Isaat004.008.009J %;9* n)
when

vi 0 ''i nde) ;0yo*o


=G9)that

=G9)$oQ/6S

vi 0 ) $i 8a)0na)

A-))pluckQ/6S)10E

when that one Ik0kaa Swoman, +#9TJ went picking %;9* k0kaa vi 0 lle) va0avu), vi 0 ) si 0 ) 7i no)07i no

Isaat004.008.04;J v0m)
+#9)monkey

+#9)woman

=G9)that =G9)2S.12SS

A-))<=()turnPinto)/-.A1

that wi e o yours, she habitually turns into a monkey Isaat004.008.029J It should be stressed that $ender sub'ect a$reement is still not the norm for the women
49 It should be mentioned that there is an alternati!e e3planation for this e3ample, since it could be ar$ued that the difference in feature is due to the class of the controllin$ noun , recall from L9.8.4 that +#9 nouns are more likely to tri$$er person a$reement than nouns of other classes.

;92

in this story , these are the only three occurrences. <owe!er it does seem that for

+#9

nouns, female referents ha!e a $reater potential than male referents to tri$$er $ender sub'ect a$reement. .espite the far $reater number of nouns with male referents in the corpus, there is only a sin$le e3ample of $ender sub'ect a$reement in!ol!in$ an noun with a male referent, taken from a son$20K %80* d0daa
+#9)kin$ A-))sayQ/6S)10E

+#9

* vi 0 ) hyaa0na)

) n0yaa0wa) 18 o)

4S)doQI//)=116R21.1/2

the king told me to tell you .pl.2 Ilit. Sdo for youTJ Isa$'004.040.004J =lthou$h these e3amples may not by themsel!es be enou$h to con!ince the sceptic, more than one lan$ua$e consultant has independently su$$ested that $ender a$reement is more likely to occur with female controllers than with male24. 2n one occasion, when I was in!esti$atin$ the phrases in %84*, I was told that usin$ the $ender)marked pronoun as in %84b* implied that it was a woman that was told, not a man. %84* %a* n)damu)wi )vi ) n)0dama)0wa)1vi )
4S)tellQ/6S)=116R30.P#6

%b* n)damu)wa) n)0dama)0wa)


4S)tellQ/6S)=116

vi ) v0i )
A-)1P#6

' told him1her1it

' told her1it I2009)0;)29.004J

=&9

0entence and discourse topic

e ha!e now seen the effect of topicality on the inde3in$ of referents at all le!els of animacy. "he effect is most ob!ious for inanimates and animals, which are ne!er inde3ed by person a$reement unless they are topics. "he rele!ant data for humans is limited to a small number of e3amples in!ol!in$ non) +#9 nouns, but there is still a discernible trend for $ender a$reement to persist lon$er with non)topical humans at the e3pense of person a$reement, as in %;4);2* abo!e. Such e3amples recall the discussion of 6ambrecht:s S$randmaT e3ample %2.;5* in L2.;.2.2, where transient discourse participants were classed as topical by 6ambrecht, but non)topical accordin$ to .ooley:s theory of discourse topic. 2ther e3amples such as %4F*, %47)49* and %25* are similar in that either %i* the referent is only transient in the discourse, or %ii* e!en if it
20 "he son$ is of the Scall)and)responseT type common in sub)Saharan =frica. "he female caller consistently used $ender a$reement for this line, whereas the male responder consistently used person a$reement. 24 "hese consultants were all men, and the Duestion remains as to whether this pattern is repeated or re!ersed in women:s speech.

;9;

semantically inte$rates a para$raph it does not thematically inte$rate it, since it holds no intrinsic interest for the speaker. hen one considers indi!idual clauses and propositions in isolation, then the referents encoded by $ender a$reement in the abo!e) mentioned e3amples Dualify as sentence topics, 'ust like the discourse)topical referents encoded by person a$reement. "hus the notion of sentence topicality does not help us to distin$uish between the wholly $ender)marked referents discussed in the Snon) topicalT sections abo!e, and the person)marked referents in the StopicalT sections. If we are to look to topicality to help account for the distinction between $ender and person a$reement markin$, then it is discourse topicality we reDuire rather than sentence topicality.

=&<

Conflict with speakersF intuition

.espite the e3istence of te3tual e3amples of unambi$uous person a$reement occurrin$ with non)human or inanimate sub'ect referents, nati!e speakers, when Duestioned, maintain that person a$reement can only occur with human referents, 'ust as they also often maintain %a$ain, contrary to the te3tual e!idence* that $ender a$reement cannot occur anaphorically. It is interestin$ that neither <offmann nor #ro>ier mention such difficulties with respect to #entral &ambari, and it seems likely that person a$reement is more common with non)human referents in #entral &ambari than in #icipu. "his partial loss of a semantic restriction %Sbleachin$T* fits in with the su$$estion made in L7.5.9 that the #entral &ambari markers are further down the $rammaticalisation cline than the #icipu ones. It was mentioned in L7.5.4.2 %fn. 20* that with careful ne$otiation of the conte3t in elicitation sessions, nati!e speakers will accept instances of anaphoric $ender a$reement as $rammatical. It pro!ed much harder to achie!e this for constructed e3amples of person a$reement with non)human controller referents. 0or e3ample, for the followin$ e3chan$e my consultant adamantly refused to accept person a$reement %o0* in the answerK

;98

%82*

=K han

where

a)0mangal
+#2)rope

where are the ropes (K %a* ho)0yuwo)0nu)


A-

ko0o$i )
+#9)hole

)fallQ/6S)/-S

they ell down the hole %b* o)0yuwo)0nu)


3P)fallQ/6S)/-S

ko0o$i )
+#9)hole

I2007)02)0F.007J "his refusal to admit person a$reement was presumably due to the difficulty, in an artificial en!ironment, of persuadin$ speakers to !iew referents low in inherent topicality as intrinsically interestin$. "he e3ample a$ain illustrates the importance of reco$nisin$ the notion of discourse topic %rather than 'ust sentence topic*. =ccordin$ to 6ambrecht:s theory of information structure %L2.;.4*, the rope in %82* is the topic of the answer sentences. <owe!er non)human sentence topics ne!er achie!e the le!el of accessibility reDuired for minimal codin$ unless they are also discourse topics. "o e3plain the reluctance of nati!e speakers to accept sentences such as %82b*, we therefore need to appeal to the notion of discourse topic. In contrast to my e3perience with constructed e3amples in elicitation sessions, when recorded e3amples of person a$reement with non)human controllers were taken from less artificial te3ts and played back to speakers in conte3t, they were ne!er re'ected as wron$.

=&=

:nexplained data

"here is one ma'or class of e3ceptions that remains unaccounted for, and this consists of the !arious kinds of constructions that in!ol!e non)referential sub'ects, includin$ e3pleti!es, weather !erbs, and impersonal constructions. Sub'ect a$reement is obli$atory in #icipuG it cannot simply be omitted. Instead third)person a$reement prefi3es are found in these constructions. "he problem for the current account of #icipu a$reement is that these kind of sub'ects are neither animate nor topical. -3pleti!es make use of ;1S a$reement, as in %8;)88*, while in impersonal constructions we find ;11 %8F)85*, a common techniDue in +i$er)#on$o and in other lan$ua$es that lack true passi!es %Siewierska 200F*. ;9F

%8;* u)0kama)ata)

30)bePsupposedPtoQ/6S

$e
+-G

ti 0s)

41)drinkQI//

t0aa a)

+#5)tobacco

it is not proper or us to smoke tobacco %88* u)0'wa*a ;0kaa'i )da)

Itats007.002.077J

30)passQ/6S

+#9)principle

it's beyond reason %8F* i 0 ) la8amu kuma : hu0u0yi )nda) y0i )


and 3P)0A")seeQI//

Itats007.002.005J
=G;)1/2

+#;)li$htnin$

$e
+-G

n)

with

kw0andai :

+#9)dryPseason

and lightning 1 it is not seen in the dry season 1 Ilit. St!ey will not see itTJ Itats002.007.02FJ %85* I#onte3tK there is only a sin$le prota$onist in this narrati!eJ ) ka)0yapu) see ha0asa) ka)0hi'i
then 3P)spendQI//
+#4)ni$ht =G4)two

then two nights passed Ilit St!ey spent two ni$htsTJ

Itats002.004.0F7J

eather)!erbs are sli$htly more comple3. "hey are usually found with ;1S a$reement e.$. u)0y') S;S)rainedT, which looks 'ust like the impersonal a$reement found in -n$lish it rained. <owe!er on askin$ a nati!e speaker S hat rainedMT, the reply is in!ariably kungwa, which is ambi$uous in meanin$ between SskyT and SGodT. Kungwa is also found as the e3plicit sub'ect of weather !erbs on ele!en occasions in the corpus, and so it is hard to be sure in cases of anaphoric a$reement that the sub'ect is truly non) referential. -3amples such as %87* below, which is taken from a topic)stimulation session on li$htnin$, su$$est that the sub'ect is at least sometimes referential, since it has a marked topic sub'ect referent, encoded by a left)dislocated +1. %87* saa kuma ;0kungwa, da'a)
or more
/C)1sky

moreo!er 30)rainQ/6S

u)0y')

or the sky, it rains

Itats002.007.0;2J

+e!ertheless, e!en if sin$ular person a$reement is understood to refer to the sky or to God, this interpretation does not seem possible with plural person a$reement. )0y') S;1)rainedT is also acceptable, and in that case it was not possible to elicit a potential plural referent. "hus this construction is another e3ample of person a$reement with a non)referential sub'ect. "his non)referential use of person a$reement seems to be a Slast resortT in the

;95

absence of a suitable $ender to choose, althou$h we may wonder why

=G9

$ender

a$reement does not occur in such e3amples, since that is what is normally found in en!ironments where the controller has no specific $ender %L5.8*. In other (enue)#on$o lan$ua$es such as Sesotho %.emuth 4990*, e3pleti!e sub'ects do occur with weather !erbs, in which case the locati!e noun class sub'ect prefi3 hoc occurs. If, as .emuth ar$ues, this prefi3 is entirely !oid of semantic content, then weather)!erb a$reement in Sesotho is an instance of neutral $ender a$reement22. =lmost all lin$uists would classify such sub'ects as non)topical %e.$. .ooley 2007K90, 6ambrecht 4998K480*. "here is a parallel between the #icipu data bein$ considered here and e3pleti!e)it constructions in -n$lish. In both cases the minimal codin$ that occurs with topical referents %person sub'ect a$reement on the one hand, unstressed pronoun on the other* is bein$ used as an apparently SdummyT sub'ect , about as far remo!ed from a suitable topic as could be ima$ined. =n alternati!e approach is to say that the pronoun does in fact ha!e some meanin$, usually that it denotes some hi$hly abstract settin$, perhaps always at least semi)acti!e in our consciousness. -3amples of such analyses include (olin$er %4977*, 6akoff %4997* on e3pleti!e)there, and 6an$acker %4994K;5F*. <owe!er this e3planation does not wash for #icipu, since acti!ated inanimate ob'ects ne!er tri$$er person a$reement unless they are topics. It is probably best 'ust to treat e3pleti!es and weather !erbs as another kind of neutral a$reement, a!ailable as an alternati!e to the usual +#9 a$reement for a different kind of atypical controller.

5.-

4rogressions

"he pre!ious section showed that both animacy and discourse topicality are rele!ant for the $enderNperson alternation across the corpus as a whole. <ere I will look at what happens in indi!idual anaphoric chains, and consider whether any patterns can be discerned with respect to a$reement. /ecall from L2.;.2.2 that the definition of discourse topic in .ooley:s conception depends on the para$raph. = discourse topic is a referent which thematically inte$rates the discourse schema of a para$raph, in that each step of the discourse schema for the para$raph is understood to be about that referent. .iscourse topics of one para$raph do not automatically carry o!er into the ne3t, and as
22 ^Y{o[ffYeopqrYq^}qe^eYfn^}nq{Y{}wpYn{Yp[wpqnYrq{^n^Yn^~Yn^qr\nYXyXXXYq^~qr a$reement does occur with weather !erbs, althou$h this is referential a$reement e.$. Soutside is coldT rq{^n^Yn^~Yn^qr\nYnr}qYopnoYp[wpqnYfnw{Yq|fqo[\qY{}jqwo{YnfoeqopqrV

;97

we saw in L2.;.2.F the acti!ation le!el of referents is reset across para$raph boundaries , this was e!ident both from psycholin$uistic e3periments and from the increased codin$ wei$ht reDuired to refer to a pre!iously)acti!ated referent once a boundary is crossed. e also obser!ed that in lan$ua$es as di!erse as #hinese, Japanese, &oine Greek, and Mby- Guarani %as well as se!eral (enue)#on$o lan$ua$es* there are within) para$raph pro$ressions from $reater to lesser codin$ wei$ht which reflect the %re)*establishment of the referent as the topic of the new discourse unit. "his section will show that in #icipu we find a similar within)para$raph pro$ression from le3ical +1 s $ender)marked anaphors s person)marked anaphors.

=(1

*ithin.paragraph progressions

/ecall from L9.4 that in $eneral only instances of anaphoric a$reement are bein$ considered in this chapter, rather than $rammatical a$reement, since the latter does not $enerally permit a choice of feature. "his decision limits the number of a$reement pro$ressions that will be found in the corpus. 0or e3ample, the switch in %89* below was not counted as a Spro$ressionT for the purposes of this analysis, since the $ender a$reement prefi3 ma0 is an instance of $rammatical a$reement. "his kind of switch is e3tremely common in the corpus. %89* see ma)0gai
then
+#8)sword

ma0'i )sa)nu)

A-4)standQI//

until

e)'i u)0para)1re)

30)stalkQ/6SR;1.1/2

then the sword waited until it spied on them

Isamy004.057J

#onseDuently the only candidates for pro$ressions are cases where a le3ical +1 is followed by an anaphor chain of two or more person markers without interruption by another le3ical +1. "he !ast ma'ority of these are consistent in their markin$, either stickin$ to $ender)markin$ throu$hout, or to person)markin$ throu$hout. 2f $reater si$nificance here are the remainin$ 5F Smi3edT e3amples. "here is a clear preference for mi3ed chains of anaphors to start off with $ender)markin$ and then pro$ress to person) markin$. "here are F8 chains of this kind in the corpus, 49 of which occur in the topic) stimulation te3ts, and ;F in the rest of the corpus. In contrast, there are only 2 chains in the re!erse direction %1-/S2+ s G-+.-/*. "here are 9 SdoubleT chains, in!ol!in$ more than one chan$e of feature %e.$. G-+.-/ s 1-/S2+ s G-+.-/ s 1-/S2+*. "he mean len$th of the 5F anaphor chains is se!en person markers, with chains

;99

ran$in$ in si>e from two anaphors %the lo$ical minimum* to 29. "he median len$th is four markers. 2f more interest for this study is the interaction between number of pro$ressions and animacy. "able 59 shows that pro$ressions from $ender to person a$reement can be found for referents at any point in the animacy hierarchy. 0able ]b- eistribution o agreement progressions according to animacy Animacy <uman Spirit +on)human folktale character =nimal Inanimate no. o' -7/57# s P7#06/ "rogressions 9 ; 20 47 5

"he hi$h fi$ures in the cells for folktale characters and animals reflect the intermediate inherent topicality of these referents. /eaders are in!ited to inspect the folktale in =ppendi3 =, which contains se!eral pro$ressions. <umans are at the top end of the inherent topicality scale, and so anaphoric $ender)markin$ is !ery rare, despite the hi$h number of references to humans in the corpus o!erall. =s a result the possibility of pro$ressions startin$ out from $ender)markin$ is limited. Inanimate ob'ects, on the other hand, are at the bottom end of the inherent topicality scale, which means that person)markin$ is rare, and similarly there is less chance of a pro$ression endin$ up with person)markin$. "here is also a correlation between the discourse topicality of the referent and its animacy, as shown in "able 70. 0able _a- eistribution o and discourse topicality Animacy <uman Spirit +on)human folktale character =nimal Inanimate
G)Ne)R

&)RdN agreement progressions according to animacy 5iscourse1to"ical /on1discourse1to"ical F 2 49 47 5 %50* %57* %90* %400* %400* ; 4 2 0 0 %80* %;;* %40* %0* %0*

e should not be surprised at the direction of this correlation, since we are only considerin$ anaphor chains in!ol!in$ both $ender and person a$reement. "he $eneral ;99

case is for human referents not to tri$$er anaphoric $ender a$reement at all. So if $ender a$reement does occur, then we mi$ht e3pect it to be for a special reason , and in the case of humans, bein$ non)topical is e3ceptional2;. =nimals and inanimates, on the other hand, occur lower down the animacy hierarchy, and so for these, $ender a$reement is the norm. 1erson a$reement with animals or inanimates is e3ceptional, indeed it is only possible at all if the referent is discourse)topical. 0olktale characters ha!e properties in common with both humans and non)humans, and they are intermediate in terms of the percenta$e of $enderNperson a$reement pro$ressions that in!ol!e discourse topics. 1ro$ressions ha!e been obser!ed for nouns of all classes e3cept
+#7.

/ecall from

LF.2 that nouns from $enders 7 and 7N9 are often abstract nouns, or kinds of trees. Such nouns are neither animate nor $ood candidates for discourse topics, and so they rarely tri$$er person a$reement. +#7 nouns can tri$$er person a$reement, but the corpus would probably ha!e to be a lot bi$$er to find e3amples of anaphoric chains pro$ressin$ from
=G7

to ;1S. "he remainder of this section will proceed by considerin$ fi!e pro$ressions %or

series of pro$ressions* in some detail. "hen the nature of the SboundaryT at which $ender)markin$ becomes person)markin$ will be e3amined, to see if there are any $eneral statements that can be made with respect to where it mi$ht fall. 0inally the e3ceptional anaphor chains will be considered, since they also show interestin$ properties.
=(11 )xamples of progressions

"he e3cerpt in %89*, repeated from %7*, pro!ides a strai$htforward instance of a within) para$raph pro$ression in!ol!in$ a human referent. "he referent is first introduced into the discourse by the +#9 noun 707a SpersonT, but then immediately Sre)introducedT usin$ the
+#8

noun m0 SchildT. "his noun is the sub'ect of its clause, and so automatically
=G8

tri$$ers

$ender a$reement on the !erb yaa Sarri!eT. "he ne3t three sub'ect prefi3es

a$ree in $ender, but the fourth, on the !erb ka a, a$rees only in person.

2; "his is especially true since we are countin$ indi!idual pro$ressions, not referents, and typically a topical referent will $i!e rise to many more anaphor chains in a te3t than a non)topical one.

800

%89* ana)

when ;S)climbQ/6S)10E then certain

u)0kum a)0na) : see 'i na) m0


/C41c!ild

707a

/C)1"erson

ya*a0na) :

arri!eQ/6S)10E

'i na)

certain

ma)0ya*a0na)

=G8)arri!eQ/6S)E-+"

e,1k0ke)eke :

62#R+#9)bicycle

when he had climbed 1 then a certain person arrived 1 a certain child arrived on a bike 1 ana)
when

ma)0ya*a0na)0na) :

A-4)arri!eQ/6S)E-+")10E

m0aya)

A-4)comeQ/6S

m)0plu)

A-4)putPdownQ/6S

0ke)eke

+#9)bicycle

u)0ka a)0na) :

30)takeQ/6S)E-+"

mo)0kulo)otu*
+#8)basket

mo)0to) :
=G8)one

when he arrived 1 then he came and he put down the bike and he took 1 one basket 1 "he second e3ample shows a plural referent pro$ressin$ from $ender to person a$reement. Most plurals in #icipu tri$$er =G2 a$reement, which as mentioned in L7.F.4 can be hard to distin$uish from ;11 person a$reement. "here is no such problem in the followin$ e3ample, since the rele!ant noun m0Nm0uu SchildNchildrenT has 8NF $ender. <ere we find anaphoric $ender a$reement within the same intonation unit as the initial +1, and then person a$reement in the ne3t intonation unit. %F0* m0uu see
/C%1c!ild

m1mo0ni

A-%E/C41water A-%1t!at

mi 0 lle)

see

unless

mi 0 ka) a)

A-%)takeQI//

ka)0rayi ) ka,1707a :
+#4)life

=G4R+#9)person

unless 3P)lea!eQI//

a0la)ha)

707a :

+#9)person

those water babies then they take the li e o a person 1 then they leave a person Ii.e. lea!e them lifelessJ 1 Itats00F.002.488J =fter two relati!ely simple illustrations, we now consider a lon$er stretch of te3t, this time taken from the end of a folktale about an anthropomorphised sword. "he passa$e in %F4* contains three pro$ressions in!ol!in$ the sword, one from f28*, and two full
6-I#=6 +1 G-+.-/

1-/S2+

%a)

G-+.-/

1-/S2+

pro$ressions %i)o and p)32F*. "hese

repeated pro$ressions show that the three)sta$e pro$ression is not confined to the introduction of the referent. Instead $ender a$reement may re)occur after a stretch of person)marked anaphors, pro!ided a le3ical +1 inter!enes. "hese multiple pro$ressions su$$est that it is not merely anthropomorphism that is $o!ernin$ the a$reements %see L9.5.4*.

28 In this para$raph the first reference to the sword is by means of $ender a$reement. See the discussion of e3ception 4 in L9.F.4.;. 2F 1ara$raph %$)h* contains only person)marked anaphors.

804

%F4* a. O2& N b. when A-4)tired N c


A-41couldn:t

carry on N

d. then A-41wait a bit N e. 30)spied on them N f. if 301stalked them, perhaps they would for$et N * . /C4A =G8)waited a bit 301spied on them 301spied $. Othen t!e sword @ma)0gai on them 301spied on them 301spied on them N h. perhaps those people would for$et N i. '. O2& those people, them, they weren:t for$ettin$ N because N
A-41P#6

k. if they for$ot, t!at sword@/C4A, if they didn:t put scabbard N l. perhaps when they were to stand to$ether N m. 301would come and 301cut them up N n. 2& they didn:t want Ithis to happenJ N o. they wanted to carry on $oin$ far off N p. O2&, some time IlaterJ N D. they were lookin$ for somethin$ N r. t.

back in the

2&, perhaps they would throw somethin$ at t!e sword@/C4A N 301was 'umpin$ Duickly 301was followin$ them Duickly here 2& N

s. when they threw at A-41P#6, A-41$rew in an$er N u. because when they didn:t come to$ether as two people N !. 301was feelin$ an$er 301was feelin$ an$er 301was feelin$ an$er N w. 301had to wait N 3. because, 301didn:t $et two people in one place N Isamy004.054J "he followin$ e3ample a$ain in!ol!es multiple pro$ressions, this time with a $eneric topical referent, the crocodile. %F2* a. b. c Oand A-*1P#6 t!e crocodile @ku)0yupu). /C*A N
A-*1is

a fearful thin$ for people N

any water where there is said to be a crocodile@/C*A inside N 802

d. if a crocodile@/C*A is sensed to be there inside N e. f. $. h. i. '. l. 2&, people won:t %.* people will be afraid to enter that water N since, t!e crocodile@/C*A N
A-*)$rabs A-*)can A-*)will

people N

$rab a person N kill N

in the water N when someone enters N

k. because a person in the water doesn:t ha!e wei$ht N m. 30.P#6 when 30)comes and 30)takes someone and 30)pulls N n. 30)causes Ithe personJ to sink in the water doin$ pullin$ the person there in the water sinkin$ N o. and a person, in the water N p. he will not find it difficult to die N D. since, water doesn:t ha!e a branch which a person will hold N r. s. t. much less a person can $et a foothold N O2& 30.P#6 N how 301rules all the water N

u. 301eats a person fish, person, whene!er 301has the chance to do for person 301takes the chance N25 !. and 30.P#6 is in the water N w. if water took me from here to "un$an &ae N 3. inside one minute N y. >. if 301runs N 301arri!es there N

aa. ri$ht away N ab. like how a motorbike runs on the $round N ac. it:s like that its speed is inside the water N ad. O2&, when it becomes e!ident that t!e crocodile@/C*A e3asperation in the water N ae.
A-*1starts =G9)reaches

takin$ people N

af. 2& whene!er someone enters inside the water 301spots the person from far off N
25 "he meanin$ of this line is obscure, but the inde3in$ is accurate.

80;

a$. in 'ust a minute 301arri!es there N ah. attempt)30.P600 when 301arri!es to hold the person N ai. 301causes Ithe personJ to sink in the water N a'. 301kills N ak. and when 301kills 301will eat N Itats002.005.042J In the first and third para$raphs there are strai$htforward pro$ressions from 6-I#=6 +1 s
G-+.-/

1-/S2+.

"he persistence of le3ical mentions from %a)f* is probably due to the

$eneric nature of the referent bein$ discussed, but beyond the $reater density of le3ical mentions, $eneric referents do not seem to beha!e any differently from specific indi!iduals with respect to the feature alternation , both $ender and person a$reement are found, and the restrictions on the nature of possible pro$ressions are the same. "he second para$raph has no instances of $ender)markin$ at all. "his is because the first mention of the crocodile in this para$raph is not a le3ical +1, but is an independent personal pronoun, and $i!en the unidirectional pro$ression from
1-/S2+ G-+.-/

a$reement bein$ ar$ued for here, there should be no opportunity for $ender) e can make three further obser!ations about this e3ample in passin$. 0irst,

markin$. =nd indeed this is what we find here. despite the fact there is no le3ical +1 denotin$ the crocodile in the second para$raph, there is still an increase in codin$ wei$ht in line %s* compared to the pre!ious mention in %n*. "he use of a marked topic structure %the left)dislocated independent personal pronoun evi S;S.1/2T* is reDuired because of the resettin$ of acti!ation le!els that occurs across a para$raph boundary. "hat a pronoun is possible here rather than a full +1 deri!es from the fact that the crocodile is a topic of the discourse as a whole, and by !irtue of this it retains a certain le!el of acti!ation throu$hout the te3t %e.$. =nderson et al. 499;*. "he second obser!ation is to note that the other instance of an independent personal pronoun in the second para$raph %line !* occurs for a different reason. In this case the acti!ation le!el of the crocodile is hi$h and no para$raph boundary has been crossed. Instead the form of the referrin$ e3pression is determined by the syntactic construction, in this case a non)!erbal clauseK

808

%F;* evi

30.P#6 more

kuma $i 0 ) me'i )

+#5)inside

ti 1mo0ni

=G5R+#8)water

and it is in the water

Itats002.005.0;;J

"his is not true o!ercodin$ since in this particular syntactic en!ironment %locati!e non)!erbal clauses* there is no predicate with a slot for a$reement morpholo$y. "herefore a free pronoun is the minimal codin$ in this case. 0inally, $oin$ back to the first para$raph, there is one more e3ample of an o!ercoded personal pronoun %line m*. "his occurs because of yet another reasonK in this case, the full pronoun is a response to the competin$ referent someone introduced in the pre!ious line. "he fifth and final illustration in this subsection also in!ol!es two pro$ressions, this time in consecuti!e para$raphs. "he referent concerned is the mass noun ;0gogoro S$in, +#9T. In both para$raphs the initial referrin$ e3pression is an appositional series of +1s evi ;0gogoro Sit gin]. %F8* a. c e. f. O2& N
A-))it

b. but 30.P#6 gin @;0gogoro. /C)A N has a lot of faults N d. there are thin$s that 301causes N 301causes a lot of thin$s N Obecause you see 30.P#6 gin@/C)A, if you are drinkin$ A-)1P#6 N

$. if you don:t ha!e $ood thin$s while you are eatin$ N h. inshaAllahu there are conseDuences N i. '. because 30.P#6 when you drink 30.6BL too while drinkin$ N because 30.P#6 301won:t build you up Itats007.004.048J In #icipu, appositional controllers of this kind seem to be eDui!alent to the second +1 %i.e. the le3ical +1* with respect to a$reement. 0or e3ample, they tri$$er $ender rather than person a$reement when they occur in sub'ect position, as demonstrated in %FF*.

80F

%FF*

;0sa'a)

+#9)time

vi 0lle)

=G9)that

t0e)ne

=G5)which

t0i )

=G5)#21

evi )

30.P#6 there

ka)0 ara

/C11oldFman

ko0ttu)

A-111P.P600

ka0aya)0na)

A-1)comeQ/6S

ko)0$i yo)

A-1)$etQ/6S

;0ma)kaami ) le*e...l
+#9)position

that time how was it that he our father came and got an in luential position there... Is!b$004.009J In line with this obser!ation, composite +1s of this kind can head a
G-+.-/

1-/S2+

pro$ression as in %F8f)'*. #onseDuently it seems best to analyse them as carryin$ the feature G-+.-/ as well as 1-/S2+. -3ample %F8* is also of interest because of the contrast between the near)homophonous =G9 pronoun and ;1S ob'ect clitic in lines %f* and %i* %see L7.;.4*. <ere are the rele!ant clauses in #icipuK %F5* n ni )
if and 2S)#2+" drink

;0yuu s)

v0i ) :

A-)1P#6

i you are drinking it 1 Itats007.004.049J %F7* ni ) ni ) si *ivi )... B0s*1vi ) Itats007.004.022J 2bser!e how the i !owel from the person)marked clitic spreads leftward onto the !erb in %F7*, but that this fails to happen for the otherwise)identical $ender pronoun in %F5*. It should be noted that the multiple pro$ressions discussed here are not the norm. 0or this section I chose three e3amples of multiple pro$ressions in consecuti!e or near) consecuti!e para$raphs, since they could be usefully discussed without ha!in$ to reproduce !ery lar$e sections of te3ts. +ormally a$reement pro$ressions are distributed more disparately %remember there are only 5F e3amples in the entire corpus*. In total there were thirteen referents which took part in multiple pro$ressions, althou$h often these pro$ressions would be separated by many other consistent anaphor chains %e.$. all $ender, or all person*.
=(1# The boundar- between gender and person agreement

and 2S)drinkM#;0E30.P#6

when you drink it...

"he point at which $ender a$reement pro$resses to person a$reement in anaphor chains is not random, and two $eneral obser!ations can be made. 0irst, the len$th of time taken to pro$ress to person a$reement depends on the conditions discussed in L9.8. So for 805

e3ample for the lower animates, the same reluctance to display person a$reement at all manifests itself in the fact that when person a$reement does occur, it is often towards the end of a lon$ chain of $ender)marked anaphors. (y the same token, anaphoric chains with human referents, if they display $ender)markin$ at all, usually pro$ress to person)markin$ early in the chain. Secondly, e!en thou$h any number of consistently)marked anaphors may occur in a sin$le chain without any pause, the switch from $ender to person markin$ rarely occurs within an intonation unit. -3ample %F9* is typicalK %F9* ana) $i )0kundu ti )0ya*a0na),
when
/C+1!yena =G5)arri!eQ/6S)10E

ti 0 ) dukwa)0na)

A-+)$oQ/6S)E-+"

ti 0 ) yo*o

A-+)$oQ/6S

ti 0 ) jo*olo)0no) :

A-+)checkQ/6S)E-+"

u)07a*a

30)findQ/6S

$e 0na)ata)naata) :
+-G

+#9)smallPspider

when hyena arrived, he came and checked 1 he didn't see spider 1 Isaat004.009.4;0J 2ut of the F8 G-+.-/ s 1-/S2+ switches, only 9 take place within an intonation unit, and in each of these cases there is a constituent separatin$ the person)marked anaphor from the pre!ious $ender)marked one, as in %F9*. %F9* ma)0kka$i )
/C41oldFbitc!

maK1ma0waa m0aya)
A-4E/C41dog

a)0hangayau

+#2)driedPpiece

he1t0ii*

=G2R+#5)shit

w0aya)

=G8)comeQ/6S

ma)0dukwa) ma)0hu8 o)0no


A-4)$oQ/6S

3s)comeQ/6S 30)snap16=#sQ/6S)/-S

u)0gi ti )lsa)0nu)

A-4)sweepQ/6S)E-+"

o1mo0ni :

62#R+#8)water

then the old bitch went and swept up dried pieces o shit, then she broke them up in the water 1 Isaat002.002.2F2J
=(1% )xceptions to the G)?$)! M P)!02? directional constraint

=t the be$innin$ of this section I noted that as well as the F8 anaphoric chains pro$ressin$ from $ender to person a$reement, there were 44 anaphoric chains that did not fit that pattern. -3planations can be offered for some of these e3ceptions, but others remain mysterious. "his subsection discusses the e3ceptions in turn, arran$ed with what are 'ud$ed to be the more con!incin$ e3planations first. )xception 1 "his e3ception has a strai$htforward e3planation. It in!ol!es a Sre!ersionT from person) to $ender)markin$ midway throu$h a folktale about a sword. "he re!ersion took place after a one)minute interruption to the narrati!e, caused by a motorbike $oin$ past the compound where the recordin$ was takin$ place. "he interrupted chain of reference to 807

the sword starts with the le3ical +1 in line %b*. Gender)markin$ in %c* is followed by person)markin$ %c, $* until the interruption at %h*. hen the speaker restarted, the sword was referred to usin$ a $ender sub'ect prefi3 %'* rather than the e3pected le3ical +1. %50* a. c Obecause of that,
A-4)was

* . /C4A, b. t!at sword @ma)0gai carryin$ on with speed, 30)was carryin$ on with speed hemmin$ them in, d. but they didn:t want to come to$ether. e. f. because, they wanted to frustrate A-41P#6,27

$. but 301wouldn:t $et people to 301kill. h. there:s a motorbike. onecminute break i. '. k. l. O2&, 2& when A-41tired A-41tired A-41tired,
A-41couldn:t, A-41had

to wait a bit. Isamy004.0F;J

m. then 30)stalked them,

"he reason for resumin$ reference with a $ender a$reement marker rather than a nominal is not alto$ether clear, but certainly the absence of minimal codin$ %;1S a$reement* can be e3plained by the passa$e of time and subseDuent shift in attention since the pre!ious reference. "he acti!ation of the referent was no doubt presumed to ha!e decayed in the consciousness of the hearer beyond the point where minimal codin$ was appropriate, and a hea!ier kind of codin$ %i.e. $ender)markin$* was therefore reDuired to re)acti!ate it. )xception # = further e3ception can be e3plained by appealin$ to the sub'ecti!e nature of the notion of intrinsic interest. /eferents may only be of intrinsic interest %and hence discourse topics* to some of the discourse participants. In the followin$ con!ersation about clothin$ %54*, an old man and his son are tryin$ to remember the word for a particular
27 "he $ender)markin$ in %f* is of course also Sout)of)turnT , see the miscellaneous e3ceptions discussed after -3ception 5.

809

item of clothin$ worn in the past. Since the item in Duestion was a discourse topic for them, it is referred to usin$ person a$reement %see lines k, m, r, and t*. hen the father calls his wife to confirm the correct word, she uses a $ender)marked pronoun %line s* , presumably because the referent held no particular interest for her at that time, and so from her point of !iew was not a topic. %54* a. SonK (ack then the old man told us there was a t!ing@i )0ri . /C3A what was A-31P#6 called, what was A-31P#6 called, mokuruM kai? I for$et. there was somet!ing@/C3A that women wear, men too there was somethin$ women wea), men wear. what did they sayM it:s not iyayiba, what did they sayM what they called A-31P#6, A-3)was similar to the koyoyu, or did they call that kokompoM kai? I for$et the name of A-31P#6. no like for women for women. yes.

b. 0atherK what thin$. c SonK d. 0atherK isn:t it iyayibaM women wore them. e. SonK

f. 0atherK no? kokompo is not like kebente. $. SonK i. SonK h. 0atherK womenM '. 0atherK ah? it:s not these clothes that the =!ai are wearin$ here, like these skirts. k. SonK m. SonK what was the name)30.P600 what was the name)30.P600 l. 0atherK hmmM n. 0atherK hey "awi? o. MotherK yesM p. 0atherK t!e t!ing@/C3A that the =!ai wear they say kakaaliM D. MotherK yes r. SonK s. MotherK but what do they call 30.6BL in #icipu.
A-31P#6

is kakaali.

t. 0atherK it:s that in #icipu for sure. but that time there was no wo!en clothin$, only 3P.P#6. Isayb004.;25J 809

)xception % "he ne3t e3ample is similar to -3ception 4, in that it in!ol!es a double G-+.-/ s 1-/S2+ pro$ression without any inter!enin$ le3ical +1. "his time, howe!er, there is no delay or para$raph boundary which mi$ht cause the acti!ation le!el of the referent to decay. Instead the reason for the repetition seems to be poetic , the second time the pro$ression occurs %d* it is simply a formal repetition of the first %c*29, and it should not be too surprisin$ that the normal rules of participant reference are suspended in such situations. =r$uably this should be re$arded as a sin$le repeated pro$ression, and should not really count as an e3ception. "he referent in Duestion is ko)0yo)ngoli ) Skind of lar$e black ant, +#4T. %52* a. then t!e ant @ko)0yo)ngoli ), he put B inside N b. of the seeds N c d.
A-11kept /C1A

came out 333 he brou$ht out t!e ant@/C1A

on $atherin$, them to the side N


A-11kept

A-11kept

on $atherin$ seeds while 301transportin$

on $atherin$ seeds while 301transportin$ them to the side until 301had $athered them all N

e. all N f. 301left the sand there separate, 301left the seeds separate N Isaat002.002.489J )xception & "he ne3t e3ample is more difficult to e3plain in terms of discourse topicality or para$raph structure. It comes from earlier on in the folktale from which %50* abo!e was taken, a story about a sword chasin$ two people. "he sword is a discourse topic throu$hout the te3t, and the passa$e Duoted below forms a complete para$raph. "he problem is that after se!eral person)marked anaphors %b)c*, the chain re!erts to $ender a$reement %line e* and remains in that state until the end of the para$raph %f, '*.

29 "he intonational contours imposed o!er the two statements are almost identical, which presumably adds to the poetic effect.

840

* . /C4A =G8)was tirin$ Duickly A-41tirin$ Duickly %5;* a. 2&, t!e sword @ma)0gai A-41tirin$ Duickly, b. 301was followin$ the people because if it was 'ust one person, 301wouldn:t kill him. c 2&, 301wanted two people, 2& 301was followin$ this one 301would follow that one it 301would follow this one,
A-4)was

d. they wouldn:t come to$ether, e. f. tirin$, e!en $et the people,


A-4)didn:t

$. assembled as two, h. far less killin$ them. i. because, '.


A-4)couldn:t

assemble them, Isamy004.0;8J

k. as two people

2ne solution to this problem is su$$ested by the distribution of information within the para$raph, and that is to consider the abo!e passa$e as consistin$ of two para$raphs rather than one. "he first $roupin$ %a)d* would then be part of the o!erall narrati!e schema, with the second %e)k* bein$ simply a recapitulation of the first. "his reanalysis would lea!e us with two anaphoric chains, the first pro$ressin$ from $ender to person a$reement as e3pected for a discourse topic, and the second with consistent $ender a$reement. "he functional moti!ation for the return to $ender markin$ in %e* would be the processin$ shift that has been ar$ued to take place across para$raph boundaries %L2.;.2.8*. "here are conceptual reasons to take this approach, since both the indi!idual $roupin$s e3press what seems to be the same discourse schema, comprisin$ of %i* the sword tirin$, %ii* its desire to assemble the two people into one place, and %iii* its failure to do so. Anfortunately the formal si$nals do not support such a para$raph di!ision. 6ine %d* has a non)terminal pitch contour %#hafe 4998KF9)52*, indicated by the comma. Secondly, there is only a short pause of 0.2Fs between %d* and %e*, compared to 0.7s before %a* and 0.Fs after %k*. 0inally, the increase in amplitude characteristic of the start of new para$raphs is lackin$ entirely from %e*. =n alternati!e, and perhaps preferable, reason for the une3pected use of $ender codin$ in line %e* mi$ht be the switch of sub'ect that took place in the pre!ious 844

intonation unit. It is possible that the increased competition arisin$ from the occurrence of the two people as sub'ect in line %d* led the speaker to au$ment the codin$ used for the sub'ect of line %e*. "his codin$ techniDue has been obser!ed e3perimentally by =nderson et al. %499;*, who found that -n$lish sub'ects une3pectedly used le3ical +1s rather than pronouns within para$raphs, 'ust when there was a competin$ referent. =$ainst this e3planation it could be ar$ued that ;1s codin$ would ha!e been sufficient to unambi$uously identify the sword, since neither of the two humans are e!er referred to independently in this story. <owe!er the effect of competin$ referents on speakers: choices of referrin$ e3pression is not limited to cases of logical ambi$uity, as =nderson et al. obser!ed. In their e3periments the mere acti!ation of an additional referent was sufficient to cause hea!ier codin$ than was Slo$icallyT necessary for disambi$uation. It may be that somethin$ similar is happenin$ in this e3ample, and that the use of sli$htly) more)than)minimal codin$ in the form of $ender a$reement in line %e* is a small price to pay to help the hearer determine the correct sub'ect referent. )xception ( "his e3ample is similar to -3ception 8, in that there is an une3pected occurrence of $ender a$reement midway throu$h a para$raph %line d*. * . /C4A, %58* a. t!is sword @ma)0gai b. c
A-4)came

and A-4)increased,

30)increased in shi!erin$.

d. when A-4)increased in shi!erin$, e. 30)came and 30)left. f. when it 30)left, $. 30)went and 30)spotted someone. h. up on a horse. Isamy002.0;5J "his time, howe!er, there is no competin$ referent, and the para$raph cannot be di!ided into two. +either %a)c* nor %d)h* form a coherent para$raph on their own, either conceptually or formally. +e!ertheless there is some kind of di!ision between %c* and %d*, as well as between %e* and %f*. "his can be seen in the form of an terminal intonational contour %represented by a full stop*, and by the fact that %d* and %f* both packa$e up the e!ent)line information introduced in the pre!ious few intonation units 842

and present it as now)presupposed information in a dependent clause. It may be that as well as bein$ sensiti!e to complete discourse units %i.e. para$raphs*, the #icipu $enderNperson alternation is also, to some de$ree, sensiti!e to smaller structures. .ooley refers to the incomplete buildin$)blocks of para$raph schemas as Smicro)le!el unitsT %2007K48, F9)54, see also p.85 fn. 82 on tail)head linka$e*, while #hafe %4998K4;9)48F* uses the term SsentenceT to refer to the structural unit correspondin$ to .ooley:s conceptual definition %see also <inds 4979K4F0*. .ooley %2007K50)54* obser!es that non)topical referents usually ha!e to ha!e their acti!ation status renewed after crossin$ the boundary of a micro)le!el unit, e!en within the same para$raph. #onseDuently speakers typically resort to hea!ier codin$ in such cases. =lthou$h the sword in %58* abo!e is most definitely a discourse topic, it may be that the sli$ht increasin$ in codin$ in line %d* is ne!ertheless related to the crossin$ of such a boundary. )xception 9 "his e3ception in!ol!es an anaphoric chain of two sub'ect a$reement markers, in the re!erse order with respect to the
G-+.-/

s 1-/S2+ pro$ression. "he $rasscutter in %5F* is

only an incidental participant in the folktale bein$ described, not a discourse topic. "he only e3planation I can think of is based on the fact that $rammatical person a$reement occurs on the first !erb aya ScomeT %recall from L7.5.4.4 that $rammatical person a$reement is possible with
+#9

controllers before !owel)initial !erbs*. It may be that

this instance of phonolo$ically)enabled person a$reement influences the subseDuent markin$ on the !erb yaa SdoT, and normal ser!ice %i.e. $ender)markin$* is only resumed in the ne3t intonation unit. %5F* a. a grasscutter @d0di yo). /C)A ;S)came and 30)did a hole. b. a place that A-))entered. I"idipo, saat002.008.02FJ "he remainin$ fi!e e3ceptions in!ol!e person)markin$ in one syntactic position, followed by une3pected $ender)markin$ in a different syntactic position %e.$. a person) marked sub'ect prefi3 followed by a $ender)marked pronoun in the complement of E1*. 2ne solution to this would be to propose that syntactic positions !ary with respect to the rate at which the
G-+.-/

1-/S2+

pro$ression takes place, allowin$ for the co)

84;

occurrence of multiple semi)independent %yet unidirectional* pro$ressions. "his is similar to what #orbett %4994K280* assumes when he restricts his predictions about pro$ressions from syntactic to semantic a$reement to parallel targets %see L9.9*. = much bi$$er corpus would be reDuired to test whether it is really necessary to weaken the claim about the nature of the pro$ression in this way, by makin$ it dependent on syntactic position. In any case, the fact that there are une3plained e3ceptions should not be too disturbin$. Gi!en the comple3ity of the factors in!ol!ed in the $enderNperson alternation, and the indeterminacy of the notion of discourse topic, it would be surprisin$ if there had been no e3ceptions at all. =s pointed out in L2.;.;, 'ud$ements about discourse topicality are Duite different to $rammaticality 'ud$ements. Speakers !ary in eloDuence and often do put thin$s badly, and so we should not be surprised if hypotheses relatin$ to well)put lan$ua$e are not reflected 400 in the data. I ha!e de!oted a considerable amount of space to discussin$ the e3ceptions here, and the fact that most of them appear to some e3tent to be moti!ated $i!es further support to the hypothesis that le3ical +1s, $ender)marked anaphors and person)marked anaphors occur in an ordered pro$ression within para$raphs. .iscussion of the e3ceptions has also been useful in that it has demonstrated the rele!ance of some of the concepts discussed in L2.; such as acti!ation le!el and cost %-3ception 4*, intrinsic interest %-3ception 2*, and competin$ referents %-3ception 8*.

=(#

Text.level progressions

In addition to the within)para$raph pro$ressions which ha!e 'ust been discussed, a priori it is at least possible that there mi$ht be discernible pro$ressions in the way discourse participants are coded at a hi$her le!el. .ooley:s %2007* conception of discourse topic is not limited to the para$raph but is applicable to any discourse unit. In particular there can be $lobal topics whose span is the whole te3t. #onseDuently we mi$ht e3pect to find a hi$her)le!el pro$ression superimposed on top of the intra) para$raph ones, and in fact we do. "here are se!eral candidates for te3t)le!el pro$ressions in #icipu. 2ne of the clearest such te3ts is a topic)stimulation te3t concernin$ witches %tats002.008*. =s well as the witch %the main participant*, for most of this te3t %units 49) 79 out of 99* a spirit %u)0pepi , also SwindT,
+#7*

is also on)sta$e. In the first part of the

848

te3tual span of the spirit %units 49)84*, which is lar$ely concerned with the witch:s acti!ities in securin$ its ser!ices, person markers denotin$ the spirit are e3clusi!ely $ender)marked. <owe!er once the spirit has been set loose to trouble its !ictim, it is thereafter %units 82)79* $ender)marked only in those en!ironments in which $ender) markin$ is obli$atory , otherwise it is e3clusi!ely person)marked. In this te3t there are not actually any indi!idual anaphoric chains showin$ the kind of pro$ressions that were discussed in L9.F.4. <owe!er the te3t as a whole clearly shows a hi$her)le!el pro$ression. = number of other te3ts show similar effects. "he re!erse state of affairs, with person)markin$ concentrated towards the be$innin$ of a te3t and $ender)markin$ predominatin$ towards the end does not occur in any of the te3ts in the corpus29. "hese obser!ations are su$$esti!e of a hi$her le!el of discourse or$anisation, and would merit in!esti$ation on a lar$er scale, not only in #icipu but also in the other est &ain'i lan$ua$es that ha!e this alternation between $ender and person a$reement. 0urther e!idence that this mi$ht be a fruitful line of research comes from the sin$le "sureshe folktale included in =$amalafiya %2007*. "he two main characters are a fro$ and one another %unidentified* animal. "here is a hi$h incidence of sub'ect markers;0 in this story, and both main participants are coded predominantly with $ender sub'ect a$reement towards the be$innin$ of the tale and person sub'ect a$reement towards the end.

5.2

Alternative explanations:

In the course of this chapter three a$reement conditions ha!e been proposed in!ol!in$ the noun class, animacy, and discourse topicality of the controller referent. "he positin$ of three independent conditions further e3acerbates an analytical problem which was complicated enou$h by the end of chapter 7. It is therefore temptin$ to find some means of reducin$ the number of conditions.

=91

Animac-

2ne way of doin$ this is su$$ested by the fact that, as in many other cultures, animals in
29 "here is an ob!ious reason why this should be the caseK we do not tend to start talkin$ about participants which are inherently interestin$ to us, lose interest in them, and yet at the same time carry on talkin$ about them %which of course is necessary for there to be any anaphoric chains at all*. ;0 It is not clear from the accompanyin$ analysis whether they should be re$arded as affi3es or pronouns , they are written as free forms by =$amalafiya, but as affi3es by (oett$er and (oett$er %4957*.

84F

#icipu folktales tend to be anthropomorphised if they are principal characters. +ot only can they occur as the sub'ect referents of predicates which usually reDuire human referents %e.$. think, speak, cook etc...*, they are also coded usin$ person a$reement more often than animals outside of this $enre. In the same way, it could be ar$ued that when, say, a shrub is cross)referenced by person a$reement in a more prosaic te3t, then this is also a case of anthropomorphism;4. If this e3planation were satisfactory, then we could do away with discourse topicality alto$ether, and claim that person a$reement is bein$ used because the referent is bein$ conceptualised as especially animate. =s an illustration of how that mi$ht work, consider the followin$ e3tract from an -n$lish con!ersation pro!ided by <ewin$s and <ewin$s %200FK247*, with the addition of bold type for what I ha!e assumed to be references to the do$K he:s a funny old stick +e! is...he says you see t!at dog he:s $ot a lit), there:s a little terrier dog moochin$ round and it:s a nice little do$ actually it:s about this bi$...B belon$s to his dau$hter Jackie...so he says see t!at dog down there and I said yeah...I couldn:t see it like I:m sittin$ there ha!in$ me hair cut like doin$...and he:s, and he:s standin$ there in front of me and he says you see t!at dog down there and I say yeah...and he said uh...you wouldn:t think !e could $et !is head in that cup would you...he:s $ot a mu$, a tea mu$ on the thin$...and I looked at the mu$ and you know it:s an ordinary mu$ and I said...no he said !e can:t...he said all that hair he said sticks out he said when it:s smoothed back and he says... !is, !is head:s narrow he said +ote how earlier on in the e3cerpt the neuter pronoun it is used to refer to the do$, whereas in the latter part of the e3cerpt the masculine pronouns he and his occur. In this case the
=G-+"

semantic role assi$ned by the !erb comple3 could get encoura$es

anthropomorphism, which then results in masculine codin$. "he earlier references usin$ the neuter pronouns are the sub'ects of descripti!e clauses. <owe!er the #icipu alternation seems to be Duite different in nature. In particular, it is possible to $et repeated pro$ressions from le3ical +1 s $ender a$reement s person a$reement for the same referent in a sin$le te3t, as shown in L9.F.4.4, seemin$ly independent of the le3ical semantics of the !erb %see L9.5.2 below*. 2nce we think of an animal or inanimate referent as ha!in$ human characteristics, we tend not to $o back to thinkin$ of it as inanimate in subseDuent sentences. Moreo!er chan$es from $ender to
;4 =lternati!ely it could be e3plained in terms of &uno:s empathy %e.$. &uno 4975*. See Siewierska %2008K209* for discussion. "he ar$uments a$ainst anthropomorphism below apply eDually well to empathy.

845

person a$reement within the same sentence %rather than intonation unit* are common in #icipu, but a chan$e from it to him in the same sentence for -n$lish would be stylistically odd;2.

=9#

Transitivit-

I ha!e not yet coded the corpus used for this study for the le3ical semantics of the !erb or for any other aspect of transiti!ity %as characterised by <opper and "hompson 4990*, althou$h this has been su$$ested to me as a potentially rele!ant factor by more than one lin$uist. =n inspection of the two passa$es discussed in L9.8.2 shows that there is certainly no deterministic link between the choice of a$reement features and transiti!ity. +e!ertheless it remains an open Duestion as to whether there is any si$nificant effect of transiti!ity, beyond the tri!ial correlation which would result from the fact that discourse topical or animate referents are more likely to be en$a$ed in transiti!e acti!ities than non)topical or inanimate ones.

5.3

#evised flo$chart

<a!in$ re!iewed all the a!ailable facts about the alternation between person and $ender a$reement, we are now in a position to re!ise the flowchart that was presented in 0i$ure ;5 %L7.5.4.;* by addin$ the three a$reement conditions discussed in this chapter. "his has been done below;;.

;2 +umerous e3amples like 8y dog had its mu>>le on because !e doesn:t like other do$s can be found usin$ Goo$le but they sound odd to me. ;; /ecall that these flowcharts are intended purely as an elucidatory aid. In particular, they are not intended to represent processin$ nor te3t)$eneration models.

847

Start

)naphoric agreement<

Vo8e"-initia" >erb<

= Pre-re@uisites satis%ie6 %or person agreement

9>a"uate inter-re"ate6 noun c"ass? animacy? an6 6iscourse topica"ity con6itions

Is contro""er gen6er 7? 7'2? or 7'3<

; Person agreement %a>oure6< Copy contro""er %eatures Cou"6 be gen6er or person agreement

= Speci%y person agreement Stop

+igure Ja- +inal revision o lowchart modelling gender1person alternation "here is an asymmetry between $ender and person a$reement in the flowchart, in that only person a$reement is e3plicitly Shard)codedT. Gender a$reement does not need to be specified at all, since it only occurs when there is a $ender)marked antecedent, and so can be represented simply by the blind copyin$ of features;8. "his asymmetry accounts for the direction of the a$reement pro$ressions. Since $ender a$reement is not e3plicitly specified, it cannot come from anywhere other than the controllin$ referent. =ssumin$ that in a chain of anaphors each tar$et $oes on to function as the controller of the ne3t anaphor in a chain, then once the chain has pro$ressed to person a$reement there is no way for it to return without re)mentionin$ the le3ical item. -3ceptional cases %especially
;8 "his account does not work for the instances of 411 person a$reement tri$$ered by le3ical +1s discussed in L7.5.4.4 %i.e. those in the form we are the people who wecare...*.

849

e3ceptions 4, 8, and F in L9.F.4.; abo!e, where there is reason to belie!e the acti!ation status of the referent mi$ht ha!e decayed* are not considered to be a$reement here, e!en thou$h a$reement morpholo$y is in!ol!ed. Instead it is su$$ested they are bein$ used deictically %see L9.9 below*. "he asymmetry in the flowchart also means that person a$reement is bein$ treated as marked, whereas $ender a$reement is unmarked. A priori the opposite position mi$ht ha!e seemed more probable, for three reasons. 0irstly, $ender a$reement is far rarer than person a$reement in the #icipu te3t corpus. "his is not surprisin$ $i!en the cross) lin$uistic association of anaphora, animacy and topic. Secondly, all lan$ua$es ha!e the cate$ory person %Siewierska 2008K9)4;*, whereas not all ha!e $ender. 0urthermore person a$reement on the !erb is fairly common in the world:s lan$ua$es, bein$ found in 29 of (ybee:s sample of F0 lan$ua$es %(ybee 499FK;0*. (y contrast, $ender a$reement as a !erbal inflection is less common, occurrin$ in only ei$ht of these lan$ua$es. -!en when $ender a$reement does occur, it is more often optional and dependent upon pra$matic conditions %Stassen 4997K;F);5*. "he third reason deri!es from the fact that $ender sub'ect a$reement in #icipu implies third person sub'ect. So the chan$e from $ender a$reement to person a$reement in!ol!es an increase in the number of features marked on the !erb from I1-/S2+J to both I1-/S2+J and IG-+.-/J. +e!ertheless, the only practical approach is to assume that $ender a$reement is the default. If we restrict our attention to (enue)#on$o lan$ua$es, it is clear that there is nothin$ marked about $ender a$reement in the $roup as a whole. Moreo!er, it seems impossible to $i!e a coherent account of $ender a$reement as marked in #icipu, since in that case the persistence of $ender a$reement would be in!ersely correlated to the discourse salience of the controller referent. =s we saw in L2.2.8.2, whene!er there is !ariation cross)lin$uistically, it is the presence of a$reement rather than its absence that correlates positi!ely with animacy or topicality.

5.5

8ender( person and coding $eight

e mi$ht ask why it should be person a$reement that inde3es referents hi$h in inherent and discourse topicality, whereas $ender a$reement is the more strai$htforward indicator of $rammatical a$reement. Gi!en that some lin$uists ha!e su$$ested that the primary function of $ender is to keep track of participant reference %e.$. #ontini)Mora!a 2002, #orbett 2005, see also .ooley 2007K9F*, it mi$ht seem that $ender a$reement is 849

balin$ out 'ust when we need it most , i.e. when there is no lon$er a nearby formal ShintT in the form of a le3ical +1. <owe!er we need to consider more carefully the process by which addressees search for possible referents of anaphors. It is not 'ust a case of searchin$ back for Srecent mentionsT in the te3t, as pointed out by Schwar>) 0riesel %2007*. Sometimes a referent achie!es a pri!ile$ed position within the discourse unit, in the sense that the discourse unit is inte$rated around it and it holds intrinsic interest for the speaker, i.e. it is a discourse topic. In that case the use of third)person markin$ will be enou$h for the hearer to uniDuely identify the intended referent, e!en when they are inanimate and ha!e competition from non)topical human referents, pro!ided that selectional restrictions and world knowled$e permit this identification to be made. e should not be surprised by thisG after all, lan$ua$es which allow null anaphora like #hinese %6i and "hompson 4979* successfully make use of the notion of discourse topic to disentan$le considerably more ambi$uity than is present in #icipu te3ts. Similarly a human referent can achie!e this pri!ile$ed status by !irtue of its inherent topicality. In the case of competin$ referents, none of which ha!e achie!ed the status of discourse topicality, $ender a$reement is useful to distin$uish which one is intended %as noted by <eath 499;, (osch %499;KF7*, #orbett %4994K;20);22*, #ontini) Mora!a 2002, and others*. e can look at this phenomenon in two ways , one way is to think of it as a in^[q{ono[e^YeYopqY|r[^w[|fqYeYqwe^eigVYe^wqr^[^Ywe~[^Yvq[poYZ[\]^Y{}q{oq~ that the followin$ principle is at workK -3pend only as much ener$y on a task as is reDuired for its performance %499;K49*. More $enerally, Grice:s %497FK8F* second ma3im of Duantity states do not make your contribution more informati!e than is reDuired. If person a$reement is sufficient to identify the referent, then there is no need to supply increased codin$ wei$ht in the form of $ender a$reement. =s .ooley %2007K408* puts it, the more the conceptual structure of a discourse unit Spoints toT an element and makes it conceptually accessible to the addressee, the less need there will be to Spoint toT it lin$uistically. (osch %499;KF7* refers to this conceptual accessibility as the salience of the referent %see also =riel 4990*;F. hile codin$ wei$ht can be understood phonolo$ically in the case of the post) !erbal alternation , the person)marked forms which occur in this en!ironment are
;F "he two scales of Sinherent topicalityT and Sdiscourse topicalityT in 0i$ure ;9 could ha!e been combined into a sin$le dimension of Sentity salienceT. <owe!er this notion is impossible to operationalise without recourse to the independent factors that comprise it.

820

bound ob'ect clitics, while the $ender)marked forms are free pronouns , this cannot be the case for the two paradi$ms of sub'ect prefi3es. =s we obser!ed in L7.F.4 they are !ery similar in phonolo$ical wei$ht, to the point of considerable ambi$uityV Y Z[\]^ %499;* obser!es that there are se!eral independent scales in!ol!ed in the concept of codin$ wei$ht, and 6e!insohn %2000K257ff* and Seifart %200FK289)2F0* stress the role of the semantic specificity of the anaphoric de!ice. <owe!er it should be clear from the discussion so far %especially L5.; on antecedentless a$reement* that !ery little, if any, meanin$ can be attached to either $ender or person markers, so they cannot be ranked in terms of semantics. Instead, as mentioned in L9.7 they can be ranked in terms of the number of features they e3pound , two for the $ender markers, one for the person markers. I am not aware of any lan$ua$es outside (enue)#on$o showin$ this kind of SfeatureT codin$ wei$ht hierarchy , but if one were to be set up then the #icipu markers would be the predicted way round, with the hi$her number of features correlatin$ with lower accessibility. Inspection of the a!ailable te3ts in the other &ain'i and 1lateau lan$ua$es which ha!e this alternation su$$ests that the same
=G/--M-+" 6-I#=6

+1 s

G-+.-/

1-/S2+ =G/--M-+"

pro$ression holds for these lan$ua$es too, althou$h this

reDuires further in!esti$ation %see chapter 9*. "he other way to think of the phenomenon is that the use of an une3pected codin$ strate$y actually functions as a si$nal to the hearer that, as well as selectin$ the most appropriate referent, they should also ele!ate this referent to the status of discourse topic i.e. to inte$rate their understandin$ of the current discourse unit around that referent. =ccordin$ to this second understandin$ then, the switch to person)markin$ causes rather than reflects an instance of discourse topicality. It should be possible to test this hypothesis e3perimentally %a$ain see chapter 9*. "his !iew of person a$reement as pickin$ out a conte3tually)salient referent calls to mind the distinction between anaphoric and deictic reference discussed in L2.2.4. /ecall that a$reement morpholo$y may either be %i* tri$$ered by an antecedent controller in the discourse, or %ii* used to encoura$e the hearer to infer an appropriate referent, which may or may not ha!e been mentioned pre!iously in the te3t. ould it make sense to think of $ender a$reement bein$ ScontrolledT by the antecedent +1, with person a$reement morpholo$y bein$ a deictic si$nal to the addressee to encoura$e them to search for an appropriate referentM

824

"here is e!idence to su$$est that $ender)marked person markers in an anaphoric chain headed by a le3ical +1 show true a$reement rather than deictic reference. =s we ha!e seen, $ender a$reement is rarely found far from an antecedent , for three reasonsK %i* speakers do not tend to talk about inanimate or non)topical referents for a lon$ time, %ii* when they do, they tend to repeat the +1, and %iii* in the few cases when lower animates actually do achie!e the status of discourse topics, they may pro$ress on to person a$reement. #orbett %4994K288)28F* comments that some lin$uists ar$ue that there is no such thin$ as anaphoric a$reement, and that all pronouns deictically refer to the referent of the antecedent, without mediation from their antecedents. <owe!er if this were the case for #icipu then we mi$ht e3pect all $ender)marked person markers to display the same freedom with respect to a$reement morpholo$y as we find for deictic reference %recall the SstoneT e3ample in L2.2.4*. Instead what we find is that chains of $ender)marked anaphors remain consistent as to the $ender they show. "his is e!en the case if the initial referrin$ e3pression is a superordinate, non)basic term such as i 0 ) ri Sthin$T. "he lack of !ariation in this and the other person)marker chains in #icipu supports #orbett:s %4994K28F* contention that anaphoric a$reement cannot be reduced to deictic reference. "he relati!e influence of the controllin$ antecedent and the referent itself is presumably scalar rather than discrete in nature. "he further remo!ed the tar$et is from the last occurrence of the antecedent the less stron$ we can e3pect the influence of the controllin$ e3pression to be. In contrast, the referent will remain in at least semi) acti!ation and indeed its influence can be e3pected to $row with time if it continues to inte$rate the discourse;5 and becomes more entrenched as a discourse topic within the para$raph. =s discussed abo!e, at a certain point in the codin$ pro$ression this prominence may enable the speaker to lea!e off usin$ anaphoric $ender a$reement and start usin$ person a$reement, in which case we could ar$ue that reference is no lon$er bein$ mediated by the controller element in the te3t;7.
;5 /ecall from L2.;.2.2 that this can happen without the topic bein$ mentioned , for e3ample if e!ents are reported from the topic referent:s point of !iew. ;7 "he use of the controllerNtar$et terminolo$y actually obscures what happens when a$reement morpholo$y is processed by the hearer. "he speaker knows what the referent of the person marker is, and in the case of anaphoric a$reement she chooses the appropriate form dependin$ on syntactic properties of the controller. "he hearer:s task is the opposite. Startin$ from the features marked on the anaphor, the selectional restrictions imposed by the Shost predicationT %#ornish 2007K22*, and his own world knowled$e, the hearer must choose the referent which will result in the hi$hest amount of te3tual coherence. hether a referent is brou$ht to prominence by the decayin$ influence of a te3tual antecedent or by the increasin$ de$ree to which it inte$rates the discourse is essentially irrele!ant to

822

<owe!er it would be wron$ to simply eDuate $ender)marked person markers with a$reement and person)marked person markers with deictic reference. "he most we can say is that $ender)marked person markers are more stron$ly associated with a$reement and less likely to be in!ol!ed in deictic reference. "here cannot be a deterministic link between the type of a$reement and the type of reference, since we ha!e seen e3amples in L5.; of $ender a$reement morpholo$y bein$ used with indisputable deictic reference, and in L7.5.4.4 of person sub'ect a$reement prefi3es co)occurrin$ with true +1 sub'ects. (efore lea!in$ person markers and $oin$ on to other tar$ets which share the $enderNperson alternation, it is worth comparin$ this alternation in #icipu to #orbett:s theoretical distinction between syntactic and semantic a$reement. #orbett %4994K2;9) 284* identifies certain constraints on their occurrence, in particular the followin$ corpus)le!el constraintK 0or any particular tar$et type, the further it is remo!ed from its controller, the $reater the likelihood of semantic a$reement %#orbett 4994K280*. In support of this claim #orbett $i!es e3amples of pro$ressions from le3ical +1 s {g^onwo[wYnrqqiq^oYY{qin^o[wYnrqqiq^oY[^Yf~Y^f[{pYn^~Yp[wpqnYXXyUaX 282, 2F0*, which of course e!oke the #icipu pro$ressions from le3ical +1 s $ender a$reement s person a$reement. <owe!er #icipu person a$reement cannot be considered to be semantic a$reement since it does not in!ol!e any semantic features , as we saw in L9.8, referents of all kinds can be inde3ed with person a$reement markers. In L2.2.8.2.; we saw that the -ast &ain'i lan$ua$e =mo also has two paradi$ms of independent pronouns, one inflected for $ender and the other for person. #orbett %4994K287* identifies the =mo personal pronoun series as displayin$ semantic a$reement , implicitly treatin$ the ;1SN;11 a$reement pattern as additional minor tar$et $enders %#orbett 4994K450* only a!ailable for human referents. <owe!er if nouns with any kind of referent can tri$$er person a$reement, then there is no lon$er any semantic basis for a$reement. #orbett %2005K4FF* remarks that "he distinction between syntactic and semantic a$reement links to Steele:s definition in that the co!ariance in!ol!es a Ssemantic or formal propertyT of the controller. ith #icipu person a$reement no semantic property of the controller is in!ol!ed. "herefore it should not be considered as a kind of $ender
the hearer:s main task.

82;

a$reement, and the only feature !alue that the antecedent and anaphor ha!e in common is I1-/S2+ ;J. "he pro$ressions discussed in L9.F can be seen as followin$ a path from canonical $rammatical a$reement, throu$h non)canonical anaphoric a$reement, endin$ up with person markin$ which sometimes may not e!en be Sa$reementT at all, but rather deictic reference reDuirin$ pra$matic inference on the part of the hearer. "his kind of inference can only be successful if the intended referent has sufficient inherent or discourse topicality.

5.6

The gender;person alternation on other agreement targets


1-/S2+ M=/&-/S

"his chapter and the last ha!e been concerned with

, pronouns,

pronominal clitics, and sub'ect a$reement prefi3es. <owe!er in chapter 8 we saw that the alternation between $ender and person a$reement actually applies to four additional a$reement tar$etsK the article, the demonstrati!es, the interro$ati!e Duantifier 0e)ne SwhichT, and the copula;9. Siewierska %2008K48F* notes that cross)lin$uistically person a$reement is !ery rarely found on tar$ets other than predicates, possessed nouns, and adpositions, and certainly the $rammatical cate$ory of SpersonT may not seem to make much sense when applied to the tar$ets 'ust identified. hile person)marked w0e)ne SwhichT or wu0mpa) SthisT readily co)occur with a noun to form an +1 with the feature !alue I1-/S2+ ;J, the meanin$ of Swhich youMT or Sthis meT is less clear. 0urthermore, as we saw in L8.8.F.; the SpersonT forms of the article %wu0na) %s$.* and a0na) %pl.** only occur in indefinite +1s, which are of course incompatible with first and second persons. It could be ar$ued that the formal identity of the SpersonT a$reement markers across tar$ets is simply a coincidence, and that a semantically)empty term such as uc a$reement should be used instead. +e!ertheless the term Sperson a$reementT is retained here because %i* it is less clumsy, %ii* in the case of the article, plural controllers tri$$er ac prefi3es 'ust like ;11 !erb a$reement, and %iii* to a limited e3tent, the conditions identified abo!e %animacy, topicality,
+#9

controller* also influence the a$reement

alternation on these more e3otic tar$ets. In the rest of this section I will consider the four tar$ets in turn. It should be stressed that there are far fewer rele!ant e3amples than for the person markers, and accordin$ly the analysis is sketchier and in need of refinement based on additional research.
;9 "here is dialectal !ariation here , in "ikula person a$reement does not seem possible on the article, the demonstrati!es, or 0e)ne.

828

=>1

The article
hen it

/ecall from L8.8.F.; that the article na) can occur either before or after the noun.

appears pre)nominally %e.$. ka0na) ka)0 ara;9 Sa certain old manT* it indicates an indefinite but specific +1, whereas post)nominally %e.$. ka)0 ara ka0na) Sthe old manT* it marks an anaphoric %and therefore also definite* +1. =$reement on the post)nominal article is strai$htforward in that it always takes a $ender)marked prefi3 with hi$h tone. <owe!er with the pre)nominal article there is a choice between $ender %55* and person %57* a$reement. %55* o)koo
therePis

ma0na)

A-41A#(

mo)0ni

+#8)water

'u*

therePfarPoff

aRka0ppata)
62#R+#4)pit

there is some water there in the pit Itats00F.002.2F8J %57* ;0looka)$i vi 0 lle)
+#9)time =G9)that

wu0na) ma0ha)uka)$i m1d0d*


301A#(
+#8)lunatic

=G8R+#9)horse =G8)#21

m0e

back them there was this lunatic o a horse Isayb004.F04J ith a plural noun as the controller, either the ;11 a$reement prefi3 a0 %59* or the ;1S prefi3 wu0 may occur on the article %59*. "he latter e3ample shows wu0na) markin$ both sin$ular %707a* and plural %)0 w* nouns. %59* a0na)
3P1A#(

a)07a
/C

1"erson ;1)<=()drink

)0si 0 ) s

i )0$u'u)

+#;)seed =G;R+#9)k.o.Pplant

* yi 10sukudai

some people7M?89W7<PD7GDDMG7E\7<PD7GQWQM87XB>9< Itats00F.004.498J %59* ku0na) , wu0na) )0 w, h0unda) kw0aa'a


+#9)day

=G9)=/"

301A#(

A-

1t!ie' ;1)seeQ/6S 301A#(

wu0na) 707a

/C)1"erson

one day, some thieves, they saw someone Isahs004.004.004J "here is one further form of the pre)nominal article, and this is i 0 na), which does not appear to be related to any of the person sub'ect prefi3es. 6ike wu0na) it occurs with either sin$ulars or plurals. 2ther than the restriction of a0na) to plural nouns, I ha!e not found any way to predict the distribution of these three articles. "o some e3tent the difference is idiolectal, and e!en a$e)mates from the same !illa$e differ in their usa$e.
;9 Ka)0 ara Sold manT is written with its citation tone here. In actual fact there is downstep between pre) nominal articles and their head nouns Ikana), ka araJ , one further en!ironment in which nouns are found with their ScomplementT tone %L;.8.7*. If the noun precedes the article then it occurs with citation tone.

82F

If we $roup the three Snon)$enderT articles to$ether %wu0na), a0na), and i 0na)*, then the difference in distribution between the $ender and person forms of the article is actually similar to what we ha!e found for the pronouns and sub'ect a$reement markers. <ere I will concentrate on wu0na) %the distribution of i 0 na) is similar, and a0na) is rare in the corpus*. (efore the noun 707aNa)07a SpersonNpeopleT, only the three person)marked forms can occur, not the $ender)marked
=G9

article vi 0na) or the correspondin$

=G2

plural

ha0na). /ecall that in L7.5.4.4 we saw the same prohibition a$ainst 707a tri$$erin$ $ender sub'ect a$reement. In the corpus there are thirty)si3 occurrences of the article wu0na) with 707a or a)07a as the head noun, and if we discount these %since the prereDuisites for $ender a$reement ha!e not been met, and therefore there is no choice* this lea!es 58 cases of wu0na) plus noun head. It turns out that the same three conditions we identified earlier for the person)forms are also at work for a$reement on the articleK the class of the controller %+#9 or non)+#9*, the animacy of the controller, and the discourse topicality of the controller. 2f the 58 controller referents, ;5 belon$ to
+#9,

lea!in$ the other ei$ht noun


+#9

classes accountin$ for only 29 of the e3amples. In contrast there is not a sin$le instance in the corpus of a pre)nominal $ender)marked article with an controller80, althou$h they can be elicited. "he results for the other classes are $i!en in "able 74, and althou$h the fi$ures are lower than we mi$ht wish, they do su$$est that nouns from +#9 are more likely to tri$$er person a$reement than those belon$in$ to any other class.

80 "hese do occur in the "ikula te3ts, but as has already been mentioned "ikula does not offer the $enderNperson alternation for the article and the demonstrati!es, and so $ender a$reement is the only option. "herefore the counts $i!en in this subsection and the ne3t e3clude te3ts from the "ikula dialect.

825

0able _,- No. o occurrences in the corpus o gendercmarked and personcmarked prec nominal articles, according to the noun class o the controller Class o' controller /P -ender1marked article Person1marked article 4 2 ; 8 F 5 7 9 9 (otal
84

4 0 2 8 0 0 0 48 1

; 4 47 F 4 0 ;5 4 +4

hen dealin$ with such small numbers there is a dan$er of the counts bein$ skewed by Sone dayT, which accounts common con!entionalised e3pressions, such as ku0na) kwaa'a for 42 out of the 48
+#9

nouns tri$$erin$ $ender a$reement, and wuna) i 0 ) ri Sa certain

thin$, somethin$T which is responsible for all 47 +#; nouns tri$$erin$ person a$reement. e can miti$ate a$ainst this by e3cludin$ all nouns that account for more than a certain number of tokens %e.$. F*. "he re!ised count is as followsK 0able _4- No. o occurrences in the corpus o gendercmarked and personcmarked prec nominal articles according to the noun class o the controller .e*cluding nouns contributing more than ive tokens2 Class o' controller /P -ender1marked article Person1marked article 4 2 ; 8 F 5 7 9 9 (otal
84 0i$ures cannot be $i!en for marked article.
+#7

4 0 0 8 0 0 0 2 +

; 4 0 F 4 0 40 4

because the $ender)marked form is wu0na), the same as the person)

827

hile the low fi$ures in $eneral may not be terribly con!incin$, the hi$h fi$ure for

+#9

person)markin$ is at least su$$esti!e. "he +#9 nouns were of !aryin$ animacyK aka)ake S!erseT, da a) SanimalT, looka)$i StimeT, sa'a) StimeT, si ya) SsideT, 'i ri ) SkindT, 'ali )ji )ni SspiritT, and vm$i Sstran$erT82. +one of these nouns had referents which went on to become discourse topics. Mo!in$ on to animacy, nouns with human referents are more likely to tri$$er person a$reement on the pre)nominal article than $ender a$reement. .iscountin$ the
+#9

nouns, the remainin$ 49 occurrences of pre)nominal articles from "able 72 are

distributed accordin$ to the animacy of the controller as shown in "able 7;. 0able _6- No. o occurrences in the corpus o gendercmarked and personcmarked prec nominal articles according to the animacy o the controller .e*cluding nouns contributing more than ive tokens and N5` nouns2 Animacy o' controller /P -ender1marked article Person1marked article <umans 0olktale character =nimals Inanimates (otal 0 0 ; ; + 1 F 2 2 ;

It can be seen that nouns with human referents fa!our person a$reement, whereas those with animal and inanimate referents are split between the two kinds of a$reement. 6astly, if we consider the 4; non)human referents from "able 7; we also find a correlation with discourse topicality, shown in "able 78. 2f course, a non)specific indefinite article cannot encode a topic referent in its integration function %L2.;.2.5*, since the topic cannot be Sa matter of standin$ interest or concernT at the point at which it is introduced to the discourse. <owe!er discourse topics may also be si$nalled by special constructions when they are used in their access function %cf. SEI1T strate$ies of reference , .ooley and 6e!insohn 2004K449)42;*.

82 =ll but the last are <ausa loanwords , recall from LF.5 that $enders 9, 9N2, and 9N; ha!e a hi$h proportion of loans.

829

0able _J- No. o occurrences in the corpus o gendercmarked and personcmarked prec nominal articles according to the animacy o the controller .e*cluding nouns contributing more than ive tokens, N5` nouns, and nouns with human re erents2 (o"icality o' controller /P -ender1marked article .iscourse topic +on)discourse topic (otal ; ; + Person1marked article 7 0 7

"his distribution mirrors what we saw for the person markers earlier in the chapterK for non)human referents bein$ discourse topical is a necessary but not always sufficient condition for person a$reement. It seems then that the same three factors occur as conditions on a$reement on the article as for the person markers. "he fi$ures $i!en in the abo!e tables are ob!iously lower than one would like, but they are wholly in line with what we would e3pect from the earlier sections in this chapter8;.

=>#

$emonstratives

/ecall from L8.8.F.2 that demonstrati!e modifiers can occur either before or after the head noun. Just as with the article, the post)nominal demonstrati!es a$ree in $ender while the pre)nominal ones offer a choice of $ender or person. "he difference in meanin$ between the pre)nominal and post)nominal modifiers is not as ob!ious as in the case of the article %indefinite specific !s. definite*, but as we will see below there is a link with discourse topicality. =s one would e3pect in a lan$ua$e so stron$ly left)headed as #icipu, in the ma'ority of cases the demonstrati!e occurs after the noun. In the entire corpus there are only ;8 pre)nominal tokens %27 person)marked and 7 $ender)marked* compared with 894 post)nominal tokens %which are obli$atorily $ender)marked*. =s with the person markers and the article, a si>eable number of the person)marked demonstrati!es occur before the word 707a SpersonT %44 out of 27 tokens*. 07a is ne!er preceded by a $ender) marked demonstrati!e, and it is only followed by a $ender)marked demonstrati!e on one occasion, in contrast to the $eneral o!erwhelmin$ preference for post)nominal $ender)markin$. <owe!er this statistic apart, since there are so few pre)nominal demonstrati!es it is difficult to tell what distin$uishes the $ender)marked ones from the
8; It is not lo$ically possible to find codin$ pro$ressions from $ender to person, since the restriction to indefinite +1s means that the same referent cannot be coded with the pre)nominal article twice.

829

person)marked ones. "herefore I will treat both to$ether, and concentrate on what characterises their use as opposed to the post)nominal demonstrati!es. "he most ob!ious difference that I was able to elicit from speakers is that unlike the post)nominal demonstrati!es, the pre)nominals cannot be used deictically to point out referents in the speech situation. <owe!er this is lar$ely irrele!ant with respect to the corpus, since the ma'ority of tokens are bein$ used anaphorically %or cataphorically* instead. =fter discardin$ the 44 tokens in!ol!in$ 707a, one of the most strikin$ thin$s about the remainin$ 2; pre)nominal demonstrati!es is that they almost all occur with abstract nouns. "he distribution is also hi$hly $enre)dependent, with 5 tokens occurrin$ in reported con!ersations or thou$hts, 44 in prayers, and 4 in a son$ , recall from L4.F that these $enres comprise a !ery small part of the corpus. In contrast, the $enres that make up the bulk of the corpus are lar$ely de!oid of pre)nominal demonstrati!esK folktales %4 token*, inter!iewsNhistorical narrati!es %; tokens*, and topic)stimulation te3ts %4 token*. <owe!er in addition to these obser!ations, and more rele!ant for the present discussion, the referents encoded in this way are almost all discourse topics, in that the speaker is either introducin$ a ma'or topic of the subseDuent discourse, or recapitulatin$ a ma'or topic of the pre!ious discourse. "he e3amples that follow $i!e a fla!our of the kind of en!ironments in which pre)nominal demonstrati!es are found. -3ample %70* comes from the start of a con!ersation, where the speaker is e3plicitly settin$ out the topic for the subseDuent discourse. "his is deictic reference, but anchored to an abstract entity in the knowled$e of the te3t)internal speakers, rather than to the e3ternal world. %70* I#onte3tK one folktale character confronts anotherJ wu0mpa) i 0 ) ri yi 0 na) 0yo0no) a1u0yaa0wa) 1mu)
301t!is
=G;)be

yi 0 ) 8a) $e yi )0'etei_
+-G =G;)fine

/C31t!ing

=G;)/-6

2S)beQ/6S)10E

62#R+#7)do)=116R4S.1/2

this thing you are doing to me is not a ine thing# Isaat004.007.0;FJ In %74* the initial reference to the topic +1 is formally)marked as such by left) dislocation. "he anaphoric reference to the topic is made usin$ a pre)nominal demonstrati!e.

8;0

%74*

I#onte3tK the ine things Ii 0 ) ri Sthin$s, +#;TJ that you eat 1 a ter you eat it then you smoke 1 3 you see tobacco spoils 1 wu0lle) i0 ) ri yi Rka0raa
301t!at
/C31t!ing =G;R+#4)eat

those thing of eating Itats007.002.090J #ontrast %74* with %72*, where the fire is the sole discourse topic, and the +1 with the postnominal demonstrati!e is non)specific and non)topical. %72* I#onte3tK a description of what fire does. No matter how big the collection o cornstalk stacks, or o grass or whatever...J ni )1d0duwa)0nu) u0laa, u)0laa wu0u0ra)a i0 ) ri yi 0lle) po*
andR2S)putQ/6S)/-S
+#7)fire +#7)fire =G7)0A")eatQI// /C31t!ing A-31t!at

all

when you set ire, ire will consume all o that thing Itats002.002.0F9J In %7;* the speaker is recapitulatin$ the topic of the current discussion, which was the fact that his children ha!e re'ected #icipu and turned to <ausa instead. "his time the +1 with the pre)nominal demonstrati!e is formally marked as a discourse topic by left) dislocation from the main clause. %7;* I#onte3tK inter!iew about the loss of #icipu lan$ua$e and cultureJ a)maa ti 0 mpa) "i 0 ) kg ti 0na) m0uu mi 0 ttu) n)0ka a)0n0na) :
but

o)tu)

A-+1t!is

41.1/2

gaanu)kwa)

/C+1?ausa =G5)/-6 =GF)child

understandQ/6S

$e :
+-G

=GF)41.12SS

=GF)takeQ/6S)10E)E-+"

but this Hausa that our children have brought 1 we don't understand 1 I"ikula, sa$b004.;;7J -3ample %78* is similar to %7;* in that the reference is backward)lookin$. "he speaker was asked to $i!e the reason for the enmity e3istin$ between two di!isions of the =cipu, and after e3plainin$ that marria$e practices were behind the Duarrel, he summarised his e3planation as followsK %78* le*e
there because with

s )

ni )

$i 0 ) 'i ta)ni *

ke0lle)

A-11t!at /C11enmity

ka)0gaa a

+#5)marria$e

t0i )

=G5)#21 =G5)causeQ/6S)/-S)10E

ti )0yo*o0nu)0na)

there it's because o marriage it caused that enmity I"ikula, sami004.822J =lthou$h the enmity had been mentioned in the discourse prior to this point, and so in some sense this is anaphoric reference, it is not of the same kind as the post)nominal e3ample %72*, where there is no discontinuity in the structure of the discourse. In %78* 8;4

the speaker is takin$ a step back from the main e!ent)line of the narration to pro!ide a summary of what he has 'ust said88. hen the discourse topic of the e!ent)line, the enmity, is introduced into the new mental space %0auconnier 4998* set up to process the summary in %78*, it is encoded usin$ a pre)nominal demonstrati!e. "he followin$ e3ample is similar to %78* in that it comes from a summin$ upK %7F* I#onte3tK Summin$ up an inter!iew on the history of the =kula. ' it weren't or all you elders remaining here...J hu0u05a)a $e ke0lle) ka)0na) ayi )
;1)0A")ha!eQI//
+-G A-11t!at /C11story

they wouldn't have that story I"ikula, sa$b004.592J <owe!er instead of anaphoric reference to an +1 such as ka)0gaa a SenmityT, here the e3ample in!ol!es the reification of the precedin$ discourse. "his is back$round information rather than fore$round information %an Se!aluation non)e!entT in Grimes: 497F terminolo$y*. =s was the case for %78*, when the referent of ke0lle) ka)0na) ayi ) Sthat storyT is introduced into the new mental space reDuired for this e!aluation of what has $one before, we find a pre)nominal demonstrati!e. "his kind of transfer of referents between mental spaces characterises most of the pre)nominal e3amples. =ll instances of pre)nominal demonstrati!es encode discourse topics, apart from si3 occasions when wu0mpa) SthisT occurs with a time word, as in Sthis dayT. =ll of these are within #hristian prayers and this may well wu0mpa) kw0aa'a be a calDue from a formal <ausa style. It was noted in L8.8.F.2 that the position of &iswahili demonstrati!es relati!e to their head nouns is said to be dependent on the topicality of the referent. "he passa$e is worth Duotin$ %6yons 4999K44F*K .emonstrati!es can also encode the fact that a referent is the current topic of the discourse. It is often difficult to distin$uish such topic demonstrati!es from anaphoric ones, since a topic is likely to ha!e been 'ust mentioned...(ut a particularly clear case is pro!ided by Swahili, in which topic is e*pressed by the position o the demonstrative, and anaphoric reference Duite differently. e ha!e seen abo!e that Swahili makes a two)way distance contrast and has an anaphoric demonstrati!e. "hese forms occur post)nominally, but the deictic hc and cle forms can appear precnominally, and then they indicate that the re erent is the current topic IMy italics , S.M.J.

88 i.e. he is encodin$ a non)e!ent rather than an e!ent in Grimes: %497F* terms.

8;2

It seems that the position of the demonstrati!e in #icipu ser!es a similar purpose 8F, althou$h what the conseDuences are of usin$ a person)marked pre)nominal demonstrati!e rather than a $ender)marked one remains to be seen. I will end this subsection with a short obser!ation about the use of demonstrati!e pronouns. hile the $ender)marked demonstrati!e modifiers can also occur as pronouns %L8.8.F.2*, the person)marked ones cannot , instead a dedicated demonstrati!e personal pronoun %e0mpe), e0lle), and so on , see L8.8.;.8* must be used. 0mpe) in particular is interestin$ because it can readily be used deictically to refer to any kind of animate or inanimate ob'ect, much more so than the anaphoric pronoun evi . It may be that this is related to the fact that referents that are bein$ e3plicitly pointed out are often of ma3imum current interest to the speaker, and therefore $ood candidates to become discourse topics.

=>%

0e)ne EwhichF

"he wh)word 0e)ne SwhichT also offers a choice between $ender and person a$reement. 0e)ne usually occurs before the head noun, but it may also appear afterwards. =s with the article and demonstrati!e, person a$reement cannot occur post)nominally. Most of the e3amples in the corpus are found with person a$reement in the uni!ersal Duantifier construction %L8.9* e.$. saa w0e)ne looka)$i Sall the timeT. 2nce these e3amples are remo!ed from consideration, there are only twel!e e3amples %9 $ender) and ; person) marked* left in the corpus, and this is simply too small a number to deduce anythin$ further.

=>&

Copula

"he followin$ tables illustrate the two paradi$ms in!ol!in$ the copula85.

8F <ausa also has a contrast between pre)nominal and post)nominal demonstrati!es, but <ausa pre) nominals differ from #icipu ones in that they are associated with both new information and e3ophoric reference %+ewman 2000K4F0)4F2*. 85 /ecall from L8.;.; that the copula occurs in both identificational and predicate nominal constructions. It is the former kind that is of interest here, since the copula only a$rees in person if the +1 controller is a personal pronoun, which bein$ referential cannot function as a predicate.

8;;

0able _T- Noun class pronouns plus copula %repeated from L5.2.4F* Class 4 2 ; 8 F 5 7 9 9 Pronoun k0i h0i y0i m0i m0i t0i w0i v0i kw0i Co"ula k0e) h0e) y0i ) m0e) m0i ) t0i ) w0i ) v0i ) kw0i ) (ranslation it's it it's them it's it it's it it's them it's it it's it it's it it's it Ie.$. ko0ggo)m o) SbatTJ Ie.$. o0ggo)m o) \batsTJ Ie.$. i )0nama) SmeatTJ Ie.$. ma)0di ya ShareTJ Ie.$. n)0di ya SharesTJ Ie.$. $i )0kooto) SdrumTJ Ie.$. u)0yaa SroadTJ Ie.$. $0$') SsheepTJ Ie.$. ku)0la$i S$irlTJ

0able _]- &ersonal pronoun1copula comple*es %repeated from "able 45* 0ingular 1 3 am i ) i )v vi ) evvi ) it's me it's you .sg.2 it's him1her Plural otti ) i )8o yi ) ere) it's us it's you .pl.2 it's them

"he morpholo$ical structure of the first) and third)person entries in "able 75 is not alto$ether transparent, in contrast to the tidy $ender)marked paradi$m in "able 7F. "he person)marked forms were called Sfocused independent personal pronounsT in L8.;.;.4, but they can also be thou$ht of as showin$ the copula a$reein$ in person. "his structure is still transparent in the second)person entries, and it seems likely that the other entries are also deri!ed from a structure analo$ous to the $ender)marked constructions in "able 7FK an independent personal pronoun, plus the #) form of the correspondin$ person sub'ect a$reement prefi3 Iin boldJ, plus the copular stem i . ) 0able __- duggested historical derivation o the personcmarked copula orms 0ingular 1 amu m i ) i )v v i ) 3 evi w i ) s am i ) s i )v vi ) s evvi ) Plural otu t i ) i )8o y i ) ere h i ) s otti ) s i )8o yi ) s ere)

"he sin$ular entries would perhaps be more con!incin$ if v was substituted for m and w, which calls to mind
=G9

neutral $ender a$reement %L5.8* rather than person

a$reement, while the ;11 deri!ation would be entirely unmoti!ated were it not for the rest of paradi$m. <owe!er there is no problem with the second)person cells, and the 411 deri!ation looks reasonable.

8;8

"he distinction between $rammatical and anaphoric a$reement which was applied to !erbs in L7.5 is also rele!ant to the copula, at least in the third person. Gender) marked copulas can be thou$ht of as takin$ part in $rammatical a$reement, since the sub'ect +1 is obli$atoryK %75* %m0i * m0e)
A-41P#6 A-41C6P

le*e

there

it's there I3)0kk' Sname of a lake, +#8TJ

I"ikula, sa$b004.587J

"he ;1S copulas, on the other hand, usually take part in anaphoric a$reement, as in %77*, althou$h e3plicit sub'ect +1s seem possible in restricted circumstances %e.$. %79* with a proper name*, as was the case for the person sub'ect prefi3. %77* ')pi vi )_ 'pmaT1vi )
holdQIM1R;S.1/2

evvi _ ) evvi )
30.C6P

stop him# it s him# Isaat004.009.4;2J %79* {u)gun$i evvi ) le*e


there

Iname of lakeJ 30.C6P

Rugunci is there Itats00F.002.4;7J ith respect to choice of a$reement feature, this usually depends on the choice of sub'ect e3pression. If the speaker chooses a $ender)marked +1 or pronoun then the copula must a$ree in $ender, as in %75*. If not, then the copula almost always a$rees in person. 2n rare occasions, in the case of plural person a$reement the e3pected comple3 form ere) is replaced by the e !ariant of the ;1S pronoun followed by an une3pected $ender)marked copula, in which case the pronoun is interpreted as ha!in$ plural meanin$87. "here is only one e3ample in the corpus %cf. h0i h0e) in "able 7F*K %79* e
;S.1/2

h0e)

=G2)#21

it's them Ia)0yupu) Scrocodiles, +#2TJ

Itats00F.002.2F0J

"he crocodiles were discourse)topical at this point in the te3t, and it may be that this %coupled with their relati!ely low animacy* accounts for the hybrid nature of %79*. =s with all the tar$ets bein$ discussed in this section, more research is reDuired to impro!e
87 2therwise e is always understood as sin$ular.

8;F

on the current sketchy account. (efore lea!in$ this section it is worth e3pandin$ on the similarities between the two !ariants of the ;1S pronoun, e and evi , and the person)marked copula evvi . ) It is likely that the form evvi ) is actually the result of the renewal of copular encliticisation, with evi bein$ the now fully)$rammaticalised result of a pre!ious deri!ation, in!ol!in$ the coalescence of the ori$inal ;1S pronoun e and the copula. =part from the fact that the postulated ori$inal e still e3ists, further e!idence comes from the possessi!e pronouns %L8.8.F.4.4*. In the "irisino dialect the ;1S possessi!e pronoun is 0evi , ) as in %90a*. In "ikula howe!er, the correspondin$ form is 'ust e, as in %90b*. It seems likely that "ikula preser!es the ori$inal form. %90* %a* ka)0taari
+#4)stone

ke0evi )

=G4);S.12SS

his1her stone I"irisinoJ %b* ka)0taari


+#4)stone

ke0e

=G4);S.12SS

his1her stone I"ikulaJ "he su$$ested deri!ation has parallels in other =frican lan$ua$es, where independent pronouns also show a fossilised pronoun plus copula construction %e.$. the Mande lan$ua$e (okobaru and se!eral 2motic lan$ua$es , Siewierska 2008K2FF)2F7*.

5.1<

)hapter summary

=fter preliminary sections on methodolo$y %L9.2* and participant reference in #icipu %L9.;*, we saw in L9.8 that the notions of both inherent topicality %specifically, animacy* and discourse topicality are reDuired to account for the alternation between $ender and person a$reement in #icipu. +either on its own is sufficient, since unindi!iduated inanimate referents %e.$. tobacco* can tri$$er person a$reement if they are sufficiently discourse)topical, and humans which are not discourse topics can tri$$er $ender a$reement where person a$reement mi$ht otherwise be e3pected. =s well as these semantic and pra$matic conditions on a$reement, there is also a morphosyntactic condition , we saw that +#9 controllers are more likely to tri$$er person a$reement than nouns from other classes. 0urthermore, there is e!idence for an effect of natural $ender, and it may be that the animacy hierarchy in #icipu should distin$uish between male and female humans.

8;5

#oncernin$ the properties that the theoretical notion of topic must ha!e in order to account for the alternation, we found that the notion of intrinsic interest introduced in L2.;.2.2 is important. Sections of a discourse, and in some cases entire te3ts89, may be inte$rated semantically by a referent in which the interlocutors ha!e only a passin$ interest. <owe!er it does not follow from the repeated mentions of such referents that they will always inte$rate the te3t thematically. Just talkin$ about somethin$ is not enou$h to make it topical, and if our conception of topic is restricted to the number and freDuency of mentions within te3ts then we will not be able to make use of this notion to account for e3amples such as those discussed in L9.8.2. Similarly a theory of topic restricted to the sentence such as 6ambrecht:s %L2.;.4* cannot account for the fact that in se!eral of the e3amples in this chapter %see L9.8.5*, the thin$ that an indi!idual utterance is about is inde3ed with Snon)topicalT $ender a$reement rather than StopicalT person a$reement. "o understand this alternation in #icipu we must consider the importance of the referent to the wider discourse, rather than 'ust the presuppositions and assertion rele!ant at one snapshot in time. In L9.F we saw that $ender and person e3ponents are distributed accordin$ to a codin$ pro$ression from le3ical +1 s $ender a$reement s person a$reement. In L9.5 we briefly considered possible alternati!e e3planations for the two paradi$ms of a$reement. Section 9.7 brou$ht to$ether the analyses in chapters 7 and 9 and presented the !arious factors in!ol!ed in flowchart form. It was ar$ued that $ender a$reement should be considered the default, with person a$reement occurrin$ in specific circumstances. In L9.9 it was su$$ested that the codin$ pro$ression identified in L9.F mi$ht be re$arded as a pro$ression in codin$ wei$ht, and the relationship of the two kinds of a$reement to anaphora and dei3is was discussed. 0inally in L9.9 we in!esti$ated the other tar$ets which show the $enderNperson alternation. e found that the factors rele!ant for sub'ect prefi3es and pronouns also appeared, to some e3tent, to be rele!ant for the article. 0or demonstrati!e modifiers the position of the demonstrati!e seems to depend on these factors, but there was not enou$h data to in!esti$ate the difference between $ender and person a$reement.

89 0or e3ample the match report discussed in L2.;.2.2.

8;7

Chapter > Conclusion


In this thesis I ha!e described the #icipu $ender system, focusin$ on the alternation between $ender and person a$reement in discourse. 1art I introduced the #icipu lan$ua$e and the research methodolo$y used here %chapter 4*, before settin$ out the rele!ant theoretical conte3t %chapter 2*. 1art II pro!ided a sketch phonolo$y %chapter ;* and $rammar %chapter 8*. "he scope narrowed in 1art III where I described the #icipu noun class system, concentratin$ on noun classification and deri!ed nominals in chapter F, and on a$reement in chapter 5. 1art IE turned to the ma'or research Duestions of this thesisK chapter 7 in!esti$ated the pre)reDuisites for both $ender and person a$reement, while chapter 9 looked at the rele!ant conditions when there was a choice between the two different kinds of a$reement. "he conclusion is di!ided into three sections. In L9.4 I summarise the main findin$s of 1art IE and re!isit the research Duestions set out at the be$innin$ of the thesis. In L9.2 I discuss the contribution the research makes to lin$uistic science. 0inally in L9.; I su$$est topics for further study, arisin$ both from 1art IE and more $enerally.

6.1

Main findings

1art IE was de!oted to the alternation between $ender and person a$reement on person markers %i.e. a$reement prefi3es, pronominal clitics, and pronouns*, considered from the point of !iew of the research conte3t set out in chapter 2. In chapter 7 I in!esti$ated the phonolo$ical, morpholo$ical, and syntactic properties of the fi!e different paradi$ms of #icipu person markers, and concluded that the I+.-1-+.-+" G-+.-/ %noun class* 1/2+2A+S and
I+.-1-+.-+" 1-/S2+=6 1/2+2A+S

are, as the names su$$est, free pronouns rather than


1-/S2+)M=/&-. 2(J-#"

bound morphemes. "here is a separate paradi$m of post)!erbal


#6I"I#S,

occurrin$ in complementary distribution with the independent personal

pronouns. "hese ob'ect clitics ha!e reduced !ersions that predominate in certain syntactic en!ironments %particularly before the clausal ne$ator $e*. In addition there are two paradi$ms of sub'ect a$reement prefi3es , one marked for G-+.-/, and one marked for 1-/S2+. /easons were $i!en in L7.9 for considerin$ these as two separate paradi$ms rather than a sin$le comple3 one. (oth of these sets of prefi3es must be considered ambi$uous a$reement markers accordin$ to the typolo$y of (resnan and Mchombo %4997* and Siewierska %4999*, but the $ender sub'ect prefi3es are closer to the 8;9

S$rammaticalT end of the $rammaticalisation cline than the person sub'ect prefi3es, since person a$reement only rarely occurs with an e3plicit sub'ect +1. +either type of a$reement seems as far alon$ the cline as the correspondin$ kinds in #entral &ambari. "he properties of the fi!e paradi$ms are summarised in "able 79, and 0i$ure 84 represents the relati!e status of the two sub'ect a$reement paradi$ms in #entral &ambari and #icipu with respect to $rammaticalisation. 0able _`- dummary o properties o 5icipu person markers .repeated rom 0able ],2 Paradigm Independent 1-/S2+ Independent G-+.-/ 2b'ect clitics Sub'ect 1-/S2+ 8or"!. -rammatical >eatures -rammatical Ana"!oric status 'unction agreementH agreementH ord ord #litic 1refi3 =ny =ny 2b'ect Sub'ect 1erson Gender 1erson 1erson +o +o +o Sometimes, depends on sub'ect +1 Asually, depends on sub'ect +1 Bes Bes Bes Bes

Sub'ect G-+.-/

1refi3

Sub'ect

Gender

Bes

=naphoric #entral &ambari #icipu

s Person

=mbi$uous

s -ender

Grammatical -ender

Person

+igure J,- Relative positions o person and gender agreement in 5entral 3ambari and 5icipu on (resnan and Ichombo1diewierska's agreement marker typology .repeated rom +igure 6_2 In L2.2.8 I discussed cross)lin$uistic e3amples of !ariation in a$reement and considered the kind of conditions that are typically said to be rele!ant when there is a choice. "his led on to the discussion of topicality in L2.;, where e!idence was $i!en that cross) lin$uistically the concept of discourse topicality has a role to play in influencin$ $rammatical structure. In chapter 9 we saw that discourse topicality is a condition on a$reement on #icipu person markers, to$ether with the animacy and the noun class of the controller. 1erson a$reement is more likely if the controller referent is discourse) topical or hi$hly animate, or if the controller +1 is from noun class 9. Gender

8;9

a$reement is more likely in the re!erse scenariosK if the controller referent is non) discourse)topical, low in animacy, or if the controller is from some other noun class. "he situation with respect to discourse topicality and animacy can be dia$rammed as in 0i$ure 82.

Person

&iscourse topica"ity Gen6er Inherent topica"ity


+igure J4- ^ariation in agreement according to inherent and discourse topicality, repeated rom +igure 6` "he e3amples in L9.8 showed that the rele!ant notion for the topicality a$reement condition is discourse topic rather than sentence topic %L9.8.5*, and that inanimate or animal referents with a hi$h de$ree of referential density are not inde3ed with person a$reement, unless they are also of intrinsic interest to the speaker %in other words they thematically inte$rate the te3t as well as semantically inte$rate it*. In L9.8.F.; the effect of natural $ender was discussed, with male referents seemin$ly more likely to tri$$er person a$reement than females %at least for male speakers*. Section 9.F showed that if, when the conditions mentioned abo!e are applied to a particular referent, they work out so that either $ender or person a$reement is possible, then there is a !ery stron$ tendency for $ender a$reement to precede person a$reement in mi3ed anaphoric chains. "hus there is a codin$ pro$ression from
=+=1<2/I# G-+.-/ =G/--M-+" 6-I#=6 +1

=+=1<2/I# 1-/S2+ =G/--M-+",

which recalls the other

codin$ pro$ressions mentioned in L2.;.2.F. 0inally, preliminary in!esti$ation of the other a$reement tar$ets which show the $enderNperson alternation %L9.9* su$$ests that the same three conditions %animacy, discourse topicality, and noun class* also apply. 880

"he research Duestions set out in L4.4.4 are repeated and summarily answered belowK

hat are the a$reement tar$ets in #icipuM hich of these a$reement tar$ets inflect for more than one a$reement feature paradi$mM

=re the syntactic en!ironments in which the different paradi$ms occur mutually)e3clusi!eM If not, what are the factors that influence the a$reement paradi$m when there is a choiceM If topicality is one of these factors, what kind of topicality is itM Is the data better e3plained by theories of Ssentence topicT or of Sdiscourse topicTM

"he a$reement tar$ets in #icipu were set out in L5.2. Inside the +1 almost all modifiers a$ree in $ender, includin$ numerals, ad'ecti!es, wh)words, demonstrati!es, the associati!e constructionNpossessi!e pronouns, the articleNrelati!iser, and the Duantifiers ni nni *i SonlyT and 0m ) SanotherT. 2utside the +1, a$reement is found on !erbs, the copula and ne$ati!e copula, pronounsNob'ect clitics, and demonstrati!e ad!erbs when used as predicates. Se!eral different tar$ets inflect for either
G-+.-/

or

1-/S2+K

!erbs, pronounsNob'ect

clitics, the article, demonstrati!es, 0e)ne SwhichT, and the copula. -ach of these tar$ets can occur in syntactic constructions where there is a choice of feature. In the case of the sub'ect prefi3es, ob'ect clitics, and pronouns, there are three conditions influencin$ the choiceG the noun class of the controller +1, and the animacy and topicality of the controller referent. =s mentioned abo!e, the notion of discourse topic e3plains the data better than sentence topic.

6.

)ontribution to linguistics

"he sketch phonolo$y and $rammar %1art II* and description of a$reement %1arts III and IE* offered here is the first descripti!e work on #icipu, a member of the se!erely under) studied &ain'i $roup. Similarly, the corpus submitted to the -ndan$ered 6an$ua$es =rchi!e at S2=S contains the only e3tant audio!isual recordin$s of the lan$ua$e. =s described in L4.8, this corpus consists of si3 hours of time)ali$ned interlinearised te3ts %appro3imately 42,000 clauses*, to$ether with recordin$s of almost two thousand 884

le3emes. =lthou$h I was compilin$ it in order to document a pre!iously undescribed lan$ua$e of uncertain endan$erment, this electronic corpus pro!ed to be an indispensable tool for both the descripti!e work in 1art II and the study on a$reement in 1arts III and IE. In addition to this documentary and descripti!e work, the main theoretical contribution of this study is towards our understandin$ of %i* incorporated pronouns and %ii* a$reement, in particular the factors rele!ant to !ariation in a$reement. I will discuss both of these below, and also comment on the relationship between discourse analysis and typolo$y. It is not surprisin$ that the !arious &ain'i and 1lateau lan$ua$es mentioned in L2.2.8.2.; ha!e not made it into the literature on incorporated pronouns, since they are poorly known and !ery little has been published on them. = further complication is the persistent est =frican lin$uistic practice of writin$ prefi3es as separate words, so that e!en when lan$ua$es such as =mo %=nderson 4990a* make it into the typolo$ical literature, they are not reco$nised as ha!in$ more than one series of incorporated pronouns. It is to be hoped that the analysis of another &ain'i lan$ua$e presented in this thesis will lead to a re)e!aluation of the automatic association that is so often made between anaphoric a$reement and topicality. =s was stressed in L2.2.F, not e!erythin$ that is introduced into a discourse automatically becomes a topic %otherwise topicality loses its independent e3planatory power*. Some referents are transient in discourse, but nonetheless stay around lon$ enou$h to reDuire some sort of codin$ that neither sets them up as a newNcontrasti!e referent, nor marks them out as a discourse topic. In #icipu $ender a$reement markers ser!e this purpose, whereas person a$reement markers function as indicators of discourse topics %particularly in the case of inanimates and animal referents*. =s we ha!e seen both $ender and person a$reement markers can be in!ol!ed in anaphoric a$reement , in fact anaphoric reference to non)topical inanimates or animals is always by means of $ender a$reement. #onseDuently the #icipu data supports #uly:s %2000* contention that topicality is an independent dimension alon$ which incorporated pronouns can be placed4, and su$$ests that Siewierska:s %2008* obser!ation that dependent markers do not !ary with respect to their discourse function should be re)e!aluated %see L2.2.F.4*.
4 It is not possible to $o as far as #uly does for "akelma ckwa when he analyses it purely as a discourse topic marker. =s we saw in chapter 9 person a$reement can inde3 non)topical referents if they are sufficiently animate.

882

If topicality is indeed an independent dimension to the $rammatical !s. anaphoric a$reement distinction, then this raises the Duestion as to whether grammatical a$reement markers can also !ary in their discourse function. In #icipu there is rarely a choice between $rammatical $ender a$reement and $rammatical person a$reement. "he only en!ironment in which this choice materialises is when there is both an +#9 sub'ect +1 and a !owel)initial !erb stem %L7.5.4.4*. In that scenario it seems likely that both animacy and topicality play a role, althou$h there are too few e3amples in the corpus to be certain. In #entral &ambari, howe!er, the two kinds of a$reement are only in contrast when there is an e3plicit sub'ect +1 %L7.5.9*. It seems likely, $i!en #ro>ier:s %4998* characterisation of person a$reement as topic a$reement, that $rammatical a$reement markers can !ary as to their discourse function in 'ust the same way as anaphoric a$reement markers. "he "sureshe te3t mentioned in L2.2.8.2.; su$$ests that the lan$ua$e patterns with #entral &ambari in this respect. =s was obser!ed in L2.2.8, recent studies on a$reement ha!e acknowled$ed the importance of topicality with respect to !ariation in a$reement. "he data from #icipu presented here ar$ue for an open mind as to the proper domain of StopicT. hate!er its merits on other $rounds, we saw in L9.8 that 6ambrecht:s theory of information structure cannot distin$uish between inanimate and animal referents that tri$$er only $ender a$reement, and those inde3ed by anaphoric chains that pro$ress to person a$reement. /ather it is the property of bein$ discourse)topical %L2.;.2* that enables inanimates or animals to tri$$er person a$reement. Similarly we saw a difference between non)discourse)topical and discourse)topical adult humans, in that $ender a$reement appears to persist across intonation boundaries in the former but not in the latter. "herefore #icipu pro!ides an e3ception to 6ambrecht:s contention that discourse topic %as opposed to sentence topic* has little to do with the $rammatical form of sentences %L2.;.4*. 6ambrecht:s dismissal of discourse topic with respect to morphosynta3 may be a little premature, $i!en there ha!e been so few cross)lin$uistic studies on discourse topic. Ideally, of course, typolo$ists interested in the morphosyntactic effect of discourse) pra$matic properties such as topicality would base their conclusions on comparati!e discourse analysis of a broad ran$e of lan$ua$es. <owe!er this is not currently possible, partly due to the practical limit on the number of lan$ua$es a typolo$ist can know well,

88;

and partly due to the lack of interlinearised te3t collections in a wide ran$e of lan$ua$es %see Myhill %2004K45;* for discussion*. "he te3ts collected and made a!ailable as part of this pro'ect are one contribution towards offsettin$ this dearth %which is particularly se!ere for +i$er)#on$o lan$ua$es*, but a !ast amount of work still needs to be done in this area. #onseDuently typolo$y is almost always of the $rammar)e3traction !ariety % lchli 2007*. If the lin$uistic effects of topic and focus are mentioned at all in reference $rammars, it is not usually done to the e3tent that would make cross)lin$uistic comparison possible for someone who did not already know the lan$ua$e %Myhill 2004K452)45;*. 2ften discussion is limited to elicited paradi$ms, and as (earth %4999K484* has pointed out, lin$uistic de!ices which appear to be strai$htforward indicators of topicNfocus in elicited paradi$ms may turn out to beha!e !ery differently in natural te3t2. "ypolo$ists of course do more than 'ust browse reference $rammars, and typolo$y is also informed by syntactic theory, in particular typolo$ically)oriented frameworks such as 6e3ical)0unctional Grammar. "hus the effect of an ine!itable reliance on elicited material at the e3pense of more authentic forms of communication is reinforced by the adoption of insi$hts about the structure of the sentence $leaned from $enerati!e theories of synta3 such as 60G. "his would not be a problem if the domain of inDuiry of typolo$y was restricted to the sentence. <owe!er this is not the case, and instead mainstream typolo$ists share a $eneral ideolo$ical compatibility with discourse analysts, based on the !iew that the study of lan$ua$e should be based upon analysis of empirical data %Myhill 2004K454*. Studies such as the present one, which in!ol!e a detailed te3tual analysis of a$reement in a sin$le lan$ua$e, ser!e to miti$ate the ine!itable skewin$ of typolo$ical theory towards the domain of the sentence. I mentioned abo!e that it is not currently possible for typolo$ists to base their theories of topic)sensiti!e phenomena such as a$reement on comparati!e discourse analysis, which of course be$s the Duestion what can be done to make it possibleM "he creation of interlinearised te3t collections of minority and endan$ered lan$ua$es is an ob!ious and ur$ent priority. (ut is it enou$h 'ust to pro!ide an interlinearised corpus and lea!e the rest to the typolo$ist of tomorrow %or fi!e hundred years: time*M 2r is there a way to furnish documentary corpora with annotation sufficiently thick %+athan and =ustin 2008* to, say, enable a typolo$ist interested in !ariation in a$reement to
2 See also McGill %2008* on focus in the Gur lan$ua$e Sisaala)1aasaal.

888

arri!e at the analysis which was deri!ed hereM =d!ances in technolo$y ha!e made the creation and access of lan$ua$e)specific corpora much easier than in the past, and recent de!elopments in documentary lin$uistics ha!e led to best practice standards for interlinearised annotation %e.$. Schult>e)(erndt 2005, (ow et al. 200;* as well as stimulatin$ interest in ontolo$ies for lin$uistic description %e.$. 0arrar and 6an$endoen 200;*. <owe!er, despite these de!elopments it remains the case that e!en the most comprehensi!e outputs of lan$ua$e documentation pro'ects remain relati!ely inaccessible to typolo$ists. Makin$ corpora of interlinearised te3ts publicly a!ailable is a step in the ri$ht direction, but without a more fine)$rained codin$ of the corpus than is usual the analysis of such corpora will be too time)consumin$ for most typolo$ists. In $eneral, it is unreasonable to e3pect lan$ua$e documenters to take into account specific areas of typolo$y when buildin$ their corpora , why should they be particularly concerned with, say, the discourse !ariation of a$reement, compared to any other kind of phenomenonM 6an$ua$e documenters face a difficult enou$h task without puttin$ yet another burden on their shoulders. <owe!er for documentary pro'ects which focus on a particular area of lin$uistic theory %such as the present work*, it is feasible and ar$uably desirable to ha!e as one of their outputs a documentary corpus which has been e3plicitly marked up, in such a way that the patternin$ of lin$uistic structures rele!ant to their analyses can be replicated by others who ha!e less lan$ua$e)specific knowled$e and time to spare than the ori$inal researcher. "o take the present study as an e3ample, ideally it should be possible for a typolo$ist to use the #icipu corpus to electronically test the claims I ha!e made about the association between animacy, topicality;, noun class 9, and a$reement features. =s well as makin$ the ori$inal lin$uist:s work more accountable8, this approach can of course be of help in the initial sta$es of de!elopin$ and testin$ hypotheses about the lan$ua$e. 1ro!ided the corpus is coded to an appropriate le!el of detail, then it will also be of use in cross)lin$uistic in!esti$ations. 0or e3ample in a cross)lin$uistic study on !ariation in a$reement, !ariables such as definiteness, specificity, animacy, topicality,
; "his is contro!ersial , see Myhill %200;* for the point of !iew that topicality is too !a$ue a notion to be of any use for cross)lin$uistic discourse analysis. .ooley:s definition of topic is cross)lin$uistically applicable but still sub'ecti!e , topics cannot be read directly from the te3t, but instead reDuire an introspecti!e consideration of thematicity. 8 Just makin$ one:s corpus transparent to the philolo$ist of the future % oodbury 200;K87* does not ensure that the associated descripti!eNanalytical work is accountable %(ird and Simons 200;*, since there may not be anyone sufficiently interested in both %i* your te3ts and %ii* your analysis to $o to the trouble of testin$ your conclusions.

88F

focus, and precedence %and perhaps others* would all ha!e to be coded, and a means pro!ided to match up indi!idual controllers with tar$ets. "his would in!ol!e a si$nficant amount of codin$ beyond what would be immanent in e!en a well)structured interlinear corpus and accompanyin$ le3icon, but would pro!ide a means not only of findin$ out the factors that determine !ariation in indi!idual lan$ua$es, but also of testin$ more $eneral claims, for e3amples the ones #orbett %4994K2;9)280, see L9.9* makes about the nature of a$reement pro$ressions in!ol!in$ syntactic and semantic a$reement. "he details of such an approach are a matter for further in!esti$ation.

6."

Areas for further study

Se!eral of the topics discussed in 1art IE merit further research. 0irst, the effect of natural $ender on the choice of a$reement feature %L9.8.F.;* is a promisin$ topic, $i!en that se3 is $enerally held to be irrele!ant with respect to (enue)#on$o $ender systems. In particular, it is necessary to find out whether male speakers: tendency to use person a$reement for male referents and $ender a$reement for female referents is replicated or re!ersed in the speech of women. "his line of research would pro!ide a $ood opportunity for addressin$ the main weakness in the present #icipu corpus, which is the lack of representation of female speakers. Secondly, the te3t)le!el pro$ressions discussed in L9.F.2 reDuire further in!esti$ation. 2ne possible techniDue would be to ask for retellin$s of a story structured in the same way as the SwitchT te3t %L9.F.2*, with a participant of incidental importance in the first part pro$ressin$ to a character more likely to function as a discourse topic in the second. "hirdly, the in!esti$ation into the alternation on other tar$ets in L9.9 was rather sketchy and needs de!elopment. 0ourthly, now that more is understood about the $enderNperson alternation %and about the kind of e3periments likely to pro!e successful in =cipuland , see L4.8* it is possible to desi$n e3periments to pro!ide corroborati!e e!idence for the analysis proposed in chapter 9. 0or e3ample, in a story with two characters the $enderNperson inde3in$ of these referents could be manipulated to test the effect on sub'ects: perceptions about who the story is about, or how it will continue %i.e. a discourse completion task, (ardo!i)<arli$ 2002K498*. =n e3periment of this kind, if successful, would also demonstrate that person a$reement %or the lack of it* functions as an acti!e 885

indication to the hearer as to the discourse status of the inde3ed referent %thus causin$ him to re)e!aluate the way he is construin$ the discourse*, rather than merely passi!ely reflectin$ what is already known %see L9.9*. =reas mentioned in 1art III that reDuire further research include the nature of the
+#7

infiniti!e %LF.8.4, is it truly a mi3ed cate$oryM*, the distinction in meanin$ between


+#9N=G9

the !arious different de!erbal nominalisations %LF.8*, the

$eminatin$ prefi3es

%LF.F.7, L5.4.;*, hybrid nouns %L5.;, are there any others beside ma0ga)ji ) SpriestT, and what are their a$reement possibilities with respect to #orbett:s =$reement <ierarchyM*, tri$$er)happy a$reement %L5.F*, and the possibility of $ender resolution with the most distant con'unct %L5.5*. In addition to impro!in$ the analyses set out in the present work, the study of a$reement in other &ain'i lan$ua$es should be a hi$h priority. It is well)known that despite the !ast number of lan$ua$es on the continent, =frican lan$ua$es punch below their wei$ht when it comes to their contributions to theory)buildin$F. "his is in part due to the bias in typolo$ical samplin$ methods5, but also because of the dearth of detailed descriptions of =frican lan$ua$es. "his is especially true for the &ain'i branch of (enue)#on$o. #oncernin$ $ender and person a$reement, we saw hints in L2.2.8.2.; that the alternation found in #icipu and #entral &ambari may in fact be present in most &ain'i lan$ua$es. In order to test this, more detailed descriptions of a$reement in these lan$ua$es are necessary , and in fact are desirable for other reasons, since e!en the little that is known about them su$$ests they ha!e interestin$ a$reement systems. 0or e3ample, in the +orthwest lan$ua$es %e.$. At)ma:in, Smith 2007* the noun class prefi3 becomes a suffi3 when the noun functions as a sub'ect, in which case the $ender sub'ect prefi3 on the !erb does not occur. "hus it seems as if a former a$reement prefi3 has been reinterpreted as a nominal suffi3. "he e3istence of a separate paradi$m of person a$reement pro!ides an opportunity to test the current status of the SflippedT marker. It was mentioned in L2.2.8.2.; that in 1on$u, in addition to the $ender and person sub'ect prefi3es familiar from #icipu, there are also what appear to be portmanteau sub'ect prefi3es, combinin$ both features, thus leadin$ to a three)way contrast. Studies on these a$reement systems would do!etail nicely with much)needed documentary pro'ects. Many of the other topics touched on in this thesis warrant further research. Just a
F 1honolo$ical theory bein$ the ma'or e3ception %especially with respect to tone and autose$mental phonolo$y*. 5 See Mous %200;* on the loss of lin$uistic di!ersity in =frica.

887

few that stand out are $emination and its implications for syllable structure %L;.4.9, LF.F.7*, the e3treme tonal downstep in complement position %L;.8.7*, the beha!iour of loanwords with respect to !owel harmony %L;.F.2*, the number of nasalisation contrasts %L;.5.2*, the neutralisation of le3ical contrasts caused by the backwards spreadin$ of the i !owel from the ;1S ob'ect clitic %L;.7.;*, the orderin$ possibilities in the +1 %L8.8.F.5*, the morpholo$ical status of the pluractional and causati!e infi3es %L8.5.4*, and the connection between the %homophonousM* applicati!e and anticausati!e %L8.5.2)8.5.;*. 0orty)four years a$o #arl <offmann wrote in the introduction to one of the first modern works on a &ain'i lan$ua$e "he publication of this wordlist hardly needs 'ustification %495FK7*. Sadly, e!en now it remains true that a sin$le published paper would be a step forward for most &ain'i lan$ua$es. +e!ertheless there has been a recent resur$ence of interest in the $roup, and I hope that this thesis will be a $ood foundation, and perhaps e!en a moti!ation, for other lin$uists as they take up the study of #icipu and other &ain'i lan$ua$es.

889

Appendix A Two Cicipu texts


"he corpus used for this study is a!ailable as audio)enabled interlinearised te3t at www.cicipu.or$. "wo sample te3ts are presented here.

The bloodthirsty s$ord =samy<< >


"his folktale was told by "irisino speaker Markus Mallam Babani in Galadima !illa$e on Fth 0ebruary 2009. "he main character is a sword, references to which are in bold. +ote the pro$ressions from
;7GICA; /P

-7/57#

P7#06/

%L9.F*. "his te3t has a hi$her

incidence of le3ical +1 mentions than is usual, which of course pro!ides $reater opportunity for multiple occurrences of such pro$ressions within a sin$le te3t. "he minor character %the horseman* is consistently inde3ed by person a$reement since the antecedent 707a SpersonT offers no other choice %see L7.5.4.4*. 1ara$raph boundaries ha!e been indicated with a O, based on the formal and conceptual e!idence described in L2.;.2.8. +ote the correlation with the discourse marker to) S2&T.

mi so)oni mi so)oni #
story story

story story mi i mi )i# Iformulaic replyJ a)ya


comeQIM1

ti 0yo)o0no)#

41)$oQI//)E-+"

5ome let's go hoi ho)i# Iformulaic replyJ * O wu0na) ma)0gai


;S)=/"

/C41sword

m0e)

=G8)#21 "21

* go : ma)0gai

/C41sword

me0lle) :

A-41t!at

nce there was a sword 1 that sword 1 evi


30.P#6 or

saa w0e)ne

;S)which

s0sa'a) :

+#9)time

u)0si )0ta'a

30)<=()want

mo)0hi'i ) :

+#8)blood

it all the time 1 it needed blood 1 * ma)0gai


/C41sword

me0lle)

A-41t!at

kuma :
and

and that sword 1

889

saa w0e)ne
or

;S)which

s0sa'a)

+#9)time

u)0si )0ta'a

30)<=()want 30)seeQI//

w0i *nda)

u)0'ugu)

+#7)spill

mo)0hi'i )

+#8)blood

a,1i 08aa :

62#R+#;)$round

all the time it needed to see the spilling o blood on the ground 1 * O to) ma)0gai
2&
/C41sword A-41A#(

ma0na) kuma :
and

mo)0yo*,

A-4)beQ/6S

n)

with sharp much

) : ka)ra gei

3 and the sword 1 it was very sharp 1 s )da) n1u)0$i yo)


because whenR30)$etQ/6S

707a,

+#9)person

wu0u0k)8)

30)0A")cutQI//

707a :

+#9)person

because when it got someone, it would cut them 1 k)0kk8i


+#4)slice

ka)0yapu), wu0na) looka)$i


=G4)two

;S)=/"

time%+#9*

k)0kk8i ko)0nosi ) :
+#4)slice =G4)four

into two slices, sometimes our slices 1 * O to) ma)0gai


2&
/C41sword

me0lle) : kaakulle),
A-41t!at

thatPday

mo)0$i yo)

A-4)$etQ/6S

$e
+-G

mo)0hi'i ) :

+#8)blood

3 that sword 1 that day, it didn't get any blood 1 * ma)0gai


/C41sword

ma)0hwa*ara) ku)07i )7a'u), u)0yuu


=G8)startQ/6S +#9)shi!er

30)#2+" lookPfor

)l)

) : mo)0hi'i

+#8)blood

the sword started shivering, it was looking or blood 1 u)0hwa*ara)


30)startQ/6S

tu)na)ni : u0yaa
+#7)do

thinkin$%+#9*

it started thinking 1 wu0u0d)n)


30)0A")followQI// where)/-.A1 30)$oQI// 30)$etQI//

ha)na0ha)na u0yo)o

u0$i )yo)

mo)0hi'i )

+#8)blood

ka'a) :
now

where it could ollow to go and get blood now 1 s )da) u)0ta'a)


because 30)likeQ/6S

$e
+-G

) mo)0hi'i

+#8)blood

mi Ri0vooto) : u)0ta'a)
=G8R+#;)$oat

30)likeQ/6S

$e
+-G

maRma0waa :
=G8R+#8)do$

because it didn't like the blood o goats 1 it didn't like Ithe bloodJ o the dog 1 u)0ta'a)
30)likeQ/6S

$e maRn0naa : u)0ta'a)
+-G =G8R+#9)cow

30)likeQ/6S

$e
+-G

mRku0t :

=G8R+#9)chicken

it didn't like Ithe bloodJ o the cow 1 it didn't like Ithe bloodJ o the chicken 1 see
unless

gok ma1707a :
"21

=G8R+#9)person

s )da) maR707a
because

=G8R+#9)person

ma)08a*a

=G8)surpassQ/6S

n1u)0jii :

withR+#7)!alue

only Ithe bloodJ o man 1 because Ithe bloodJ o man was more precious 1 ma)08a*a
=G8)surpassQ/6S

kuma n)
and

with

ka)0ri )ma*i :

+#4)pleasure

it was also more pleasurable 1

8F0

* O to) ma)0gai
2&

/C41sword A-41t!at

me0lle)

m0a+ya)

=G8)comeQ/6S A-4)startQ/6S

ma)0hwa*ara) : ku)07i 7a'u :


+#9)shi!er

3 then that sword then it started 1 shivering 1 m0a+ya)


A-4)comeQ/6S

mo)0'ungo)

A-4)riseQ/6S

* ma)0hyaa

A-4)sayQ/6S

to) :
2&

then it got up and it said 3 1 na)ha u0yo)o


letQIM1 30)$oQI// lookPfor place%+#7*

)l)

'asu)

wu0na) 707a
=G7)/-6

+#9)person

u)0yo0no) :

;S)beQ/6S)10E

let it go and search or where people are 1 ana) u)0yo*o0no) : 'asu)


how 30)$oQ/6S)10E place%+#7*

* wu0na) u)0hyaa0na)
=G7)/-6

30)sayQ/6S)10E 30)$oQI// 30)findQI//

u0yo)o

u07a)a

707a :

+#9)person

when it went 1 to the place it had thought it would go and ind someone 1 a)07a*a,
;1)findQ/6S abandonedPhabitation%+#9*

kongo :

aa a07a :
+-G

+#2)person

only an abandoned habitation was ound 1 there was nocone there 1 * O ma)0gai
/C41sword

me0lle) : m0a+ya)
A-41t!at

A-4)comeQ/6S

mo)0'oosu) :

A-4)a$ainQ/6S

u)0'ese)nu)

30)addQ/6S

ku)07i )7a'u) :
+#9)shi!er

that sword 1 it again 1 it shivered even more 1 looka)$i


time%+#9*

n0na)

=G9)/-6

me)0'ese)nu)

=G8)addQ/6S

ku)07i )7a'u) : w0a+ya)


+#9)shi!er

30)comeQ/6S 30)passQ/6S

u)0'wa*a :

when it shivered even more 1 then it le t 1 ana) u)0'wa*a0na) :


how 30)passQ/6S)10E 30)$oQ/6S 30)spotQ/6S ;S)=/"

u)0yo*o

u)0hangu)

wu0na) 707a :

+#9)person

g0ge8u) aRd0d* :
+#9)up

62#R+#9)horse

when it le t 1 it went and spotted someone 1 up on a horse 1 * O ma)0gai


/C41sword

me0lle)

A-41t!at

m0a+ya)

=G8)comeQ/6S A-4)sayQ/6S

* ma)0hyaa

to)

2& now

ka'a) : wu0u0$i )yo)

30)0A")$etQI//

mo)0hi'i ) :

+#8)blood

then that sword then it said 3 now 1 it would get blood 1 to)
2& and

kuma e)see

ofPcourse

* ma)0gai

/C41sword =G8)knowQ/6S +-G

ma)0lapa)

$e d0d*

+#9)horse

a)kwai

therePis

i )0la8i )

+#;)speed

) : gei

much

3 and o course the sword didn't know the horse was very ast 1 * ma)0gai
/C41sword

me0lle) : m0a+ya)
A-41t!at

A-4)comeQ/6S

* ma)0hyaa

A-4)sayQ/6S -#6

ka*i#

that sword 1 then it said kai wu0u0$i )yo)


30)0A")$etQI//

) mo)0hi'i

+#8)blood

s )

because 30)spotQ/6S

u)0hangu)

707a :

+#9)person

ka'a) wu0u0$i )yo)

now 30)0A")$etQI//

mo)0hi'i ) :

+#8)blood

it would get blood because it spotted someone 1 now it would get blood 1

8F4

O ana) u)0hangu)0na)
how

30)spotQ/6S)10E

707a :

+#9)person

* ma)0gai

/C41sword

m)'p)

=G8)holdQ/6S

i )0la8i ) :

+#;)speed

when it spotted someone 1 the sword took o 1 707a


+#9)person +#9)horse

d0d*

w0i +nda)

;S)seeQ/6S

* : ma)0gai

/C41sword

a1u0waya 'asu)
62#R+#7)come

place%+#7*

we0evi ) :

=G7);S.12SS

the horseman saw the sword 1 coming towards him 1 * s )da) ma)0gai
because
/C41sword A-41t!at

me0lle),

707a

+#9)person

n0na)

=G9)/-6

po* w0i nda)0na)


all

;S)seeQ/6S)10E

m0e) :

A-41P#6

because that sword, whoever saw it 1 n ni )


if when ;S.1/2

e)vi )

707a

+#9)person

t0to)

=G9)one

aa saa 707a :
+-G

or

+#9)person

i it was %ust one person and nocone else 1 wu0u0la)pa)


;S)0A")knowQI//

* ma)0gai

/C41sword

me0lle)

A-41t!at

) : ma)0yuu ta)'a) mo)0hi'i


=G8)#2+"

like

+#8)blood

he would know that sword wanted blood 1 de)ge)le*e wa+0aya),


then ;S)comeQ/6S ;S)hitQ/6S

u)0vasa)

d0d*

+#9)person =G9);S.12SS

ve0evi ) :

u)0'wa*a

;S)passQ/6S withR+#;)speed

n1i )0la8i ) :

then, he slapped his horse 1 he went o $uickly 1 * ma)0suma) O ma)0gai


/C41sword =G8)ranQ/6S

ma)0suma)

A-4)ranQ/6S

ma)0suma)

A-4)ranQ/6S

ma)0suma) ma)0suma) ma)0suma)#

A-4)ranQ/6S A-4)ranQ/6S A-4)ranQ/6S

the sword ran it ran it ran it ran it ran it ran * ana) ma)0gai
how
/C41sword

ma)0suma)0na) : to)
=G8)ranQ/6S)10E

2& 30)knowQ/6S

u)0lapa)

$e 707a
+-G

+#9)person

d0d*

+#9)horse

kuma :
and

when the sword ran 1 3 it didn't know that the horseman 1 u)0yo*o,
;S)$oQ/6S how

ana) d0d*

+#9)horse

07aru)wa)0na)

=G9)stepPo!erQ/6S)10E +#4)stone

ka)0taari :

he went, when the horse stepped over a rock 1 * ma)0gai


/C41sword

ana) mo)0yo*o0no),
how

A-4)$oQ/6S)10E

e)vi )

30.P#6 ;S)comeQ/6S 30)comeQ/6S 30)comeQ/6S 30)$oQ/6S

w0aya)

wa+0aya)

w0aya)

u)0yo*o :

the sword when it went, it then came and went 1 u)0du'o)


30)collideQ/6S

ka)0taari : de)ge)le*e w0aya)


+#4)stone

then

30)comeQ/6S 30)cut16=#sQ/6S)/-S

u)0kt)ls)0nu) :

it collided with the rock 1 then it olded up 1 to)


2& before 30)rise)=+"I#QI//

ka)ahi u0'u)ngo)0wo) :

u0gu)tu)

30)$oPbackQI//

u)0lapi )la) :

+#7)prepare

3 be ore it could get up 1 and recover 1

8F2

707a

+#9)person +#9)horse

d0d*

* yaa

doQ/6S

u)0'wi *i :

+#7)distance

the horseman was ar o 1 * O to) ma)0gai


2&
/C41sword

me0lle) : ma+0aya)
A-41t!at

A-4)comeQ/6S

mo)0do*onu)
A-4)sitQ/6S

le*e : u)0'p) :

there 30)holdQ/6S

3 that sword 1 then it sat down there 1 it started 1 kuRu0yaa


=G9R+#7)do

tu)na)ni

thinkin$%+#9* 30)holdQ/6S

u)0'p)

) : mi0isi

+#F)cryin$

te)0ene

=G5)which =G5)#21

t0i )

evi )

30.P#6 ;S)0A")doQI//

) : wu0u0yaa

thinking it started crying 1 what would it do i )nduu saa w0e)ne


herePis or ;S)which

u)0lenji : see
+#7)day

unless andR30)$etQ/6S

n1u)0$i yo)

mo)0hi'i )

+#8)blood =G8R+#9)person

maR707a :

you see every day 1 it had to get human blood 1 to)


2& today 30.P#6 ;S)0A")doQI//

anna) e)vi )

) wu0u0yaa

te)0enel

=G5)which

3 today what would it do * O de)ge)le*e ma)0gai


then
/C41sword

me0lle) : ma+0aya)
A-41t!at

A-4)comeQ/6S

mo)0do*onu)
A-4)sitQ/6S

le*e :

there

then that sword 1 then it sat down there 1 * ma)0gai


/C41sword

me0lle) : me)0'ese)nu)
A-41t!at

A-4)addQ/6S

$e
+-G

kuma u)0nu :
and

+#7)work

that sword 1 it could no longer work 1 s )da) kaakulle) : u)0$i yo)


because thatPday 30)$etQ/6S

) $e mo)0hi'i
+-G

+#8)blood

maR707a :

=G8R+#9)person

because that day 1 it didn't get human blood 1 * kaakulle) ma)0gai


thatPday
/C41sword

me0lle)

A-41t!at

mo)0ko*o :

=G8)dieQ/6S

that day that sword died 1 mi so)oni mi so)oni ju)ngo)nu*#


story story closeQIM1

story story close Iformulaic endin$J

The loss of Ti,ula =sagb<<1>


"his te3t was recorded in Garkuwa !illa$e near Maburya on the 9th 0ebruary 2009. "he speaker is Garkuwa (awa, a member of the =kula chieftaincy and speaker of the endan$ered "ikula dialect %L4.2.;*. "his unsolicited lament is an e3cerpt from a much lon$er recordin$ about the history of the =kula.

8F;

0napa)

2S)knowQ/6S 41.1/2

o)tu)

a)0$i pu)

+#2)#ipu

h0e :

=G2)#21

ti )0yo*

41)beQ/6S

a0ma0kuu :

62#R+#8)mountain

you know we are Acipu 1 we are at the mountain 1 to)


2& now

ka'a), -i )0$i pu) tu)0kusu)


+#5)#ipu

=G5)remainQ/6S +-G

$e, "i )0kg

+#5)<ausa

tu)0'usa) :

=G5)takeQ/6S

3 now, 5icipu no longer remains, Hausa has taken 1 )0kg


+#2)<ausa

a)0'usa)1tu)

;1)takeQ/6SR41.1/2

m0uu :

+#F)child +#2)<ausa

)0kg

a)0'usa)1tu)

;1)takeQ/6SR41.1/2

m0uu :

+#F)child

0he Hausas have taken our children rom us 1 0he Hausas have taken our children rom us 1 ) a)05a)a $e a,1u0yaa
;1)be
+-G a4-1/-D0do

-i )0$i pu) :
+#5)#ipu

they Ithe childrenJ don't speak 5icipu 1 to), i )nduu i 8 o)


2& herePis 21.1/2 21)comeQ/6S 21)teachQI//

ya+0aya)

i 0ko)yu)ko)

-i )0$i pu) kuma :


+#5)#ipu

and

3, here you are, you've come to teach 5icipu 1 ) to), Kungwa yaa
2& God%+#9* doQI// 21)ser!eQI//R41.1/2

i 0ra)asa)1tu) :

) Kungwa yaa

God%+#9* doQI// 21)canQI//

i 0gu)ya) :

3, may God enable you to serve us 1 God make you able 1 m0uu
+#F)child

mi 0ttu)

=GF)41.12SS

kuma n0na)
and

=GF)/-6

o)0yo0no)

;1)beQ/6S)10E

aR0kg,

62#R+#2)<ausa

a0gu)ya) : -i )0$i pu) :


;1)canQI//
+#5)#ipu

and our children who are amongst the Hausas, may they be able to speak 1 5icipu 1 to) mi 0na) mi )0yo0no)
2&
=GF)/-6 =GF)beQ/6S)10E

R0kg,

62#R+#2)<ausa

a)05a)a $e aRu0koyu)ko ye
;1)be
+-G 62#R+#7)teach

;1.1/2

-i 0$i pu) :
+#5)#ipu

3 those that are amongst the Hausas, they aren't teaching them 5icipu 1
farPless

e)lle

a0gu)ya) :

;1)canQI//

so how can they learn 1 i 77i )ni )


permission 21.1/2 21)comeQ/6S 21)takeQ/6S

i )8 o)

ya0aya)

i )0ka a)

-i )0$i pu) m0uu


+#5)#ipu

+#F)child

mi 0ttu) :

=GF)41.12SS

even i you come and take 5icipu to our children 1 a0ssa)


+#2)$randchild =G2)41.12SS

ho0ttu)

a)0guya)

;1)canQ/6S

$e :
+-G

teke)

useless

v0i :

=G9)#21

our grandchildren can't Ispeak #icipuJ 1 it's in vain 1 a)05a)a $e aRu0damu0wa


;1)be
+-G 62#R+#7)speakQ=116 +#F)child

m0uu

mi 0ttu)

=GF)41.12SS

n)

with

-i )0$i pu) :
+#5)#ipu

they don't speak to our children in 5icipu 1 8F8

a)05a)a $e
;1)be
+-G

aRu0tepu0we

62#R+#7)talkQ=116

m0uu

+#F)child

n)

with

-i )0$i pu) :
+#5)#ipu

they don't talk to children in 5icipu 1 vesi ) 707a


all
+#9)person =G9)/-6

n0na)

u)0yo0no)

;S)beQ/6S)10E

R0kg :

62#R+#2)<ausa

everyone who is amongst the Hausas 1 u)05a)a $e aRu0ta'a


;S)be
+-G 62#R+#7)want

u0ko)yu)ko) m0uu
;S)teachQI//

+#F)child

-i )0$i pu) :
+#5)#ipu

he doesn't want to teach the children 5icipu 1 o)tu)


41.1/2 now thePday

ka'a) kwaakuna) tu)0ko*o0no)

41)dieQ/6S)10E

-i )0$i pu) ti )0kt) :

+#5)#ipu =G5)finishQ/6S

us now, the day we die 5icipu is inished 1 to), v0i nda),


2& 2S)seeQ/6S 41.1/2 41)likeQ/6S

otu

ti )0ta'a)

$e
+-G

'i ni ) :

likePthat

3, you see, us we don't want that to happen 1 to)


2& how ;1)canQ/6S)10E

ana) a)0guya)0na)

$e
+-G

-i )0$i pu), tu0u0da)ma)


+#5)#ipu

41)0A")speakQI//

$e
+-G

k0kg :

+#9)<ausa

since they don't know 5icipu, we can't have secrets rom the Hausa 1 to)
2& formerly

da*a

hi na) n)
"21

when 41)meetQ/6S when

ti )0kud) :

n)

ti )0ta'a)

41)likeQ/6S

ti 0yu)wwo)

41)turnParoundQI//

ke)0re*e :

+#4)lan$ua$e

3 in the past when we met 1 when we wanted to switch language 1 see ti 0da)ma)
then 41)speakQI// with

n)

-i )0$i pu) k0kg


+#5)#ipu

+#9)<ausa

0napa)

=G9)knowQ/6S

$e :
+-G

then we spoke in 5icipu and the Hausa didn't know 1 to), ka'a) a)0naha)
2& now ;1)lea!eQ/6S

-i )0$i pu), )0'p)


+#5)#ipu

;1)holdQ/6S

ti )0kg

+#5)<ausa

ti )0to) :

=G5)one

3, now they le t 5icipu, they've taken %ust Hausa 1 to), vu0u0yu)wwo)


2& 2S)0A")turnParoundQI//

ke)0re*e

+#4)lan$ua$e

te)0enel

=G5)which

vu0u0yu)wwo)

2S)0A")turnParoundQI//

ke)0re*el

+#4)lan$ua$e

3, how can you switch language will you switch language vu0u0yu)wwo)
2S)0A")turnParoundQI//

$e ke)0re*e :
+-G

+#4)lan$ua$e

you won't switch language 1 m0uu


+#F)child

mi 0ttu)

=GF)41.12SS

ka'a), o)0doowQu)sRu)
now

;1)disappear#=ASsQ/6S

ka)0tii

+#4)head

ki 0ive) :

=G4);1.12SS

our children now, they're causing themselves to disappear 1

8FF

v0i nda),

2S)seeQ/6S

a)0si )0ta'a

;1)<=()need

$i )0dami

+#5)whisperin$

s )

because when 2S)$oQ/6S

ni )

)yo*o

aRka0daada)
62#R+#4)court

you see, they need whispering Ii.e. secret talkJ because when you go to court ni )1t0t*
withR+#9)nei$hbour

va0avu)

=G9)2S.12SS

-i pu) :

#ipu%+#9* 2S)seeQ/6S

v0i nda)

mo)0ni

+#8)water

mu0u0ri )i :

=G8)0A")eatG;S.1/2QI//

with your neighbour a 5ipu 1 you see water will eat him Ii.e. he:s in troubleJ N see
unless 2S)turnParoundG;S.1/2QI// with

vu0yu)wwi )

n)

-i )0$i pu) : 8aana) :


+#5)#ipu

myPfriend

) ti 0yaa

41)doQI//

ka)7a :

likePthis

unless you turn to him in 5icipu 1 my riend 1 let's do this 1 ka)7a :


likePthis

wu*0uto)0wo)1tu) :

;S)$oPout)=116R41.1/2

a)0naha)0nu)

;1)lea!eQ/6S)/-S

t0i )

=G5)1/2

8u)m um#

completely

this will get us out 1 they've completely le t the language# "i )0kg
+#5)<ausa

ti )0to),

=G5)one

"i )0kg

+#5)<ausa

t0i )

=G5)#21

ti )0ni nni *i :
=G5)alone

%ust Hausa, it's %ust Hausa alone 1 'u0u0da)ma) $e 707a n) "i )0kg : ti )0kg +-G +#9)person with +#5)<ausa one can't speak against people in Hausa 1 v0i nda)
2S)seeQ/6S if when ;1)doQ/6S

n da)

* a)0yaa

-i )0$i pu), n1i )0la8i )


+#5)#ipu

withR+#;)speed

y0i ,

=G;)#21

you see i they could speak 5icipu, $uickly, ) see i 0hyaa


then 2S)sayQI//

t0t*

+#9)nei$hbour =G9)2S.12SS

va0avu)

ka)rka)ta le*e

swer!eQIM1 there riseQIM1 41)runQI//

'u)ngo ti 0su)ma)#

then you tell your neighbour swerve there, get up and let's run# 'asu)
place%+#7*

wu0mpa) wu)08a*a
=G7)this

=G7)surpassQ/6S

u)0'usu)

+#7)power

wu0ttu) :

=G7)41.12SS

ma)0sa*a

M)should

u0su)ma) :
+#7)run

this place is too much or us 1 we should run 1 to)


2&

-i )0$i pu) ti )0to)


+#5)#ipu

=G5)one

t0i

=G5)#21

ka'a) : -i )0$i pu) ti )0to)


now
+#5)#ipu

=G5)one =G5)#21

t0i :

a*a ti 0na) :
+-G

=G5)=/"

it's %ust one language now 1 it's %ust one language Ii.e. <ausaJ 1 not any other 1 tu0uwa)
41)feelQ/6S

$e
+-G

ka)0ri )mai

+#4)pleasure

ko1m0uu

=G4R+#F)child

mi 0ttu) :

=GF)41.12SS

we're not happy with our children 1 m0uu


+#F)child

mi 0ttu)

=GF)41.12SS

n)0yadda)0nu)1tu) :

=GF)abandonQ/6S)/-SR41.1/2

our children abandoned us 1

8F5

i )nduu ye)

herePis ;1.1/2 yonder

'yu*u

aRu0mata m0uu :

62#R+#7)bear +#F)child

m0uu

+#F)child

mi )0guya)

=GF)canQ/6S

$e
+-G

-i )0$i pu) :
+#5)#ipu

there they are outside giving birth to IotherJ children 1 the children don't know 5icipu 1 to)
2& where benefit%+#9*

han

am<aani

vi 0lle) :

=G9)that

3 where's the use in that 1 a)0'aasa)


;1)forbidQ/6S

) $e ka)0si llu)m, a0yaa


+-G +#4)Islam

;1)doQI//

ka)0si llu)m :
+#4)Islam

'slam was not orbidden 1 let them practice 'slam 1 a)maa saa ya)anu u0ko)yu)ko)
but or who ;S)teachQI//

m0

+#8)child

-i )0$i pu) tu0u0ree


+#5)#ipu

=G5)+#7)town

we0evi ) :

=G7);S.12SS

but everyone should teach his child the 5icipu o his own town 1 ka)07aamani ka0aya)
+#4)future =G4)comeQ/6S

tu)0gwede) : a)maa tu0uwa)


41)thankQ/6S but

41)feelQ/6S

$e ka)0ri )mai
+-G

+#4)pleasure

ka1k0e) :

=G4R=G4)1/2

the uture has come we're grate ul 1 but we're not happy with it 1 o)tu)
41.1/2

a)07a

+#2)person

ha0na)

=G2)/-6

tu)0kusu)0na) :

41)remainQ/6S)10E

us the people that remain 1 oo, tu0uwa)


yes 41)feelQ/6S

$e ka)0ri )mai
+-G

+#4)pleasure

ka1k0e),

=G4R=G4)1/2

saa

e!en little

) : $i mmai

yes, we're not happy with it, not a bit 1

8F7

Appendix / Contributors
"e3ts from the followin$ speakers were recorded, annotated, and used in this study. -ach of them has kindly $i!en their permission for the recordin$s and annotations to be distributed without restriction. +ame =mos (ako "imothy (i$i Ma$a'i Ga'ere &uyama Garba .amina Garba .aniel <ussaini Sale Ibrahim Can'uma Ibrahim Babani I$e Ga'ere Ishaya =udu Mallu &uusu Markus Mallam Babani Se3 =$e4 .ialect M M M M M M M M 0 M M M 80 "irisino FF "irisino F0 "irisino FF "irisino 49 "irisino 4; "irisino ;0 "irisino ;0 "irisino ;F "irisino ;0 "irisino 50 "irisino ;0 "irisino #ontributions 0olktales, riddles, topic)stimulation ordlists, <istorical account Son$s Introduction to the #ipu kin$ "opic)stimulation, pear story 0olktales, news stories ordlists, news stories ordlists, elicitation Son$s 1rayers, riddles, pear story Greetin$s, discussion of chieftancy ordlists, folktales, riddles, prayers, sermon, topic)stimulation, pear story, elicitation <istorical account, topic)stimulation Son$s, topic)stimulation /iddles ordlists, riddles, topic)stimulation, elicitation "opic)stimulation Son$s Son$s Son$s <istorical account ordlists Son$s <istorical account, pear story

Mohammadu Mallam Murna Musa Musa (i$i Musa Can'uma Musa Mallam Ma$a'i /ahi Amar Sara &wande Ibrahim "ani &uyama "enii Ma$a'i "imothy Can'uma "ontono Can'uma Babani Galadima

M 0 M M M 0 0 0 0 M 0 M

20 "irisino 2F "irisino 2F "irisino ;0 "irisino 50 "irisino 29 "irisino 2F "irisino ;0 "irisino 7F "irisino 80 "irisino FF "irisino 7F "irisino

4 =ll a$es are appro3imate, and are relati!e to the commencement of my fieldwork in 2005.

8F9

.o$o "imbidii Garkuwa Maburya (awa Malli Idanbanya Mohammed Musa Saani Cantanyi =yuba Sani Ibrahim akiso Ishiaku Ibrahim "abitha Markus Amar Ma$a'i

M M M M M M M M 0 M

80 "ikula 90 "ikula 7F "ikula ;F "ikula FF "ikula ;0 "idipo 50 "idipo 20 "idipo 2F "idipo 50 "i>oriyo

ordlists <istorical account <istorical account <istorical account, topic)elicitation <istorical account, con!ersation 0olktales, elicitation ordlists -licitation 1ear story ordlists

8F9

Appendix C Person and gender agreement on person markers referring to humans@ animals@ and inanimates
In L9.8 it was stated that all person markers can a$ree anaphorically in either person or $ender, irrespecti!e of the animacy of the controller referent. "he e3amples from the te3t corpus below demonstrate this, althou$h it should be remembered that some combinations are rarer than others. "he 49 combinations ha!e different !alues for a$reement feature %32 , person, $ender*, animacy le!el %3 ; , human, animal, inanimate*, and morphosyntactic position %3 ; , sub'ect prefi3, post)!erbal ob'ect, and SelsewhereT , see chapter 7*. Sub'ect, human, personK %4* u)0yuwo)
30)fallQ/6S

he ell Itapf004.00F.020J Sub'ect, human, $enderK %2* ;0looka)$i n0na)


+#9)time =G9)/-6

ko)0'ungo)0no)
A-1)riseQ/6S)10E

when he Ika)0 ara :old man, +#4:J grew up Sub'ect, animal, personK %;* u)0yo*
30)beQ/6S with

Isayb004.09FJ

n)

;0am<aani so)sai
+#9)benefit

!ery

it Ika)0ra)kumi :camel, +#4:J has many bene its Sub'ect, animal, $enderK %8* ka)0ra)kumi n)
+#4)camel

Itats008.004.042J

ko)0do*oho),
A-1)disappearQ/6S

ko)0si 0 ) to*ono)
A-1)<=()comePhomeQ/6S

$e
+-G

when

the camel when it goes missing, it doesn't come home Itats004.002.0;;J Sub'ect, inanimate, personK %F* u0la)ngwa) 707a)
30)spoilQI//

;0hu)uhu Itats005.002.0;5J

+#9)person +#9)lun$

it Ita0a a) Stobacco, +#5TJ spoils a person their lungs

850

Sub'ect, inanimate, $enderK %5* n1u)0lenji he)0he'we)


andR+#7)day
A-

)dryQ/6S

in the a ternoon they Io)0gi no S$roundnuts, +#2TJ dry 1ost)!erbal ob'ect, human, personK %7* elle) e0lle) * ) u)hyiivi * u)0hyaa1vi )

Ieaim005.4888J

;S)that ;S)sayQ/6SR30.P#6

that one said to him the man Isaat004.00;.048J 1ost)!erbal ob'ect, human, $enderK %9* a)0si 0 ) hyaa
;1)<=()say

k0e),
A-11P#6

m) 0panda),

Oaahi )ya)

) Solder siblin$, +#4TJ, ' orget, dahiya they call her Ika)0maya Isaat002.009.00;J 1ost)!erbal ob'ect, animal, personK %9* saa saa $$i *ivi ) $0$a*a1vi ) ha)raawa) ;0ha)raawa)
+#9)fodder

4S)for$etQ/6S InameJ

e!en 2S)$i!eQ/6SR30.P#6

even though you give it Ika)0ra)kumi Scamel, +#4TJ odder 1ost)!erbal ob'ect, animal, $enderK %40* kada) a0hu)na)
1/2<

Itats004.002.0;8J

v0i )
A-)1P#6

;1)killQI//

they are not to kill it Ikwaaro) Screature, +#9TJ 1ost)!erbal ob'ect, inanimate, personK %44* ) a)si )hyiivi a)0si 0 ) hyaa1vi )
;S)<=()sayR30.P#6

Ieaim009.0;;J

ko)pogo ko)0pogo
+#4)cochlospermumPtinctorium

they call it kopogou Ia kind of treeJ Isaat004.00;.048J

854

1ost)!erbal ob'ect, inanimate, $enderK %42* n1o)0'i so)


whenR;1)cookQ/6S

m0e)

A-41P#6

when they drink it Imo)0yoo Sbeer, +#8TJ -lsewhere, human, personK %4;* e)re
3P.P#6

Itats009.004.008J

m0uu

+#F)child

they were children Isayb004.098J -lsewhere, human, $enderK %48* m0


+#8)child

me0lle)

=G8)that

da'a) m0i
1=/"

A-41P#6

m0e)

=G8)#21

ma)0ya8u)kwa)0na)

=G8)spreadPoutQ/6S)10E

that child it's her that multiplied Itats008.004.009J -lsewhere, animal, personK %4F* a)0si 0 ) ka a
;1)<=()take

i )0tangi ) n1e)vi
+#;)thin$

withR30.P#6

they carry things with it Ika)0ra)kumi Scamel, +#4TJ -lsewhere, animal, $enderK %45* n)0si 0 ) yaa
4S)<=()do

Itats008.004.009J

i )0tuma)

+#;)farm

n)

with

k0e

/C11P#6

' arm with it Ika)0ra)kumi Scamel, +#4TJ -lsewhere, inanimate, person =+. -lsewhere, inanimate, $enderK %47* ko0osi )
+#4)eye

Itats008.004.00;J

evi

30.P#6

pa*a, ku)0$i ye)


here

+#9)hand

kw0i

A-*1P#6

kw0i )

=G9)#21

pa*a

here

the eye it is here, the hand it is here Ieab$004.092J

852

!eferences
=$amalafiya, =pollos. 2007. = preliminary sketch of the "sureshe lan$ua$e. Anpublished manuscript. =$amalafiya, =pollos. 2009. .e!elopment of the /eshe lan$ua$e. 1aper presented at the 4st orkshop on est &ain'i lan$ua$es, &onta$ora. =ikhen!ald, =le3andra B. 2000. 5lassi iers- a typology o noun categoriHation devices. 23fordK 23ford Ani!ersity 1ress. =ikhen!ald, =le3andra B. n.d. Gender and noun class in one lan$ua$eK 1aumar in typolo$ical perspecti!e. -lectronic ms. I=!ailable onlineK httpKNNwww.aikhen!aldlin$uistics.comNima$esNpdfsNpublicationsN1aumari 20+#6.pdfJ =issen, Judith. 200;. .ifferential codin$, partial blockin$, and bidirectional 2". In 1awel +owak #orey BoDuelet %eds.*, (erkeley Linguistics dociety, 29, 4)45. (erkeley 6in$uistics Society. =nderson, =., S. #. Garrod =. J. Sanford. 499;. "he accessibility of pronominal antecedents as a function of episode shifts in narrati!e te3t. uarterly !ournal o )*perimental &sychology ;F, 827)880. =nderson, Stephen #. 4990a. "he noun class system of =mo. In 6arry M. <yman %ed.*, Noun classes in the Grass ields (antu borderland %Southern #alifornia 2ccasional 1apers in 6in$uistics 9*, 4FF)479. 6os =n$elesK Ani!ersity of Southern #alifornia. =nderson, Stephen #. 4990b. "he noun class system of +y$embn)(amileke. In 6arry <yman %ed.*, Noun classes in the Grass ields (antu borderland %Southern #alifornia 2ccasional 1apers in 6in$uistics 9*, ;9)F5. 6os =n$elesK Ani!ersity of Southern #alifornia. =nderson, Stephen /. 4975. 2n the notion of sub'ect in er$ati!e lan$ua$es. In #harles +. 6i %ed.*, dub%ect and topic, 4)28. +ew BorkK =cademic 1ress. =riel, Mira. 4990. Accessing nouncphrase antecedents. 6ondonK /outled$e. =sher, +icholas. 2008. .iscourse topic. 0heoretical linguistics ;0, 45;)204. =ustin, 1eter &. 2008. #litics in Sasak, eastern Indonesia. -lectronic ms. I=!ailable onlineK httpsKNNeprints.soas.ac.ukNF9NJ =ustin, 1eter &. Joan (resnan. 4995. +on)confi$urationality in =ustralian lan$ua$es. Natural Language and Linguistic 0heory 48%2*, 24F)259. 85;

(aker, Mark #. 4995. 0he &olysynthesis &arameter. 23fordK 23ford Ani!ersity 1ress. (ardo!i)<arli$, &athleen. 2002. 1ra$matics and second lan$ua$e acDuisition. In &aplan, /obert (. %ed.* 0he * ord handbook o applied linguistics, 492)492. 23ford, 23ford Ani!ersity 1ress. (arlow, Michael. 4992. A situated theory o agreement %2utstandin$ .issertations in 6in$uistics*. +ew BorkK Garland. (earth, "homas. 4999. "he contribution of =frican lin$uistics towards a $eneral theory of focus. Apdate and critical re!iew. !ournal o A rican Languages and Linguistics 20%2*, 424)4F5. qnropYxpein{VYUWVYg^onVY ^YzqrqY}r{qYYZurnr~Yp[f[||{e^Yq~{VY 0he (antu languages, 424)482. 6ondonK /outled$e. (erman, /uth =. .an I. Slobin. 4998. Relating events in narrative- a crosslinguistic developmental study. <illsdale, +JK 6awrence -rlbaum. (ertho, J. 49F2. =peru d:ensemble sur les dialectes de l:ouest de +i$ria. (ulletin de lu'+AN .'nstitut +ranais de luA ri$ue Noire2 48, 2F9)274. (iber, .ou$las. 4998. 1ra$matic roles in #entral Somali narrati!e discourse. dtudies in A rican Linguistics 4F, 4)25. (ickel, (althasar. 200;. /eferential density in discourse and syntactic typolo$y. Language 79%8*, 709)7;5. (ickel, (althasar, (ernard #omrie Martin <aspelmath. 2008. 0he LeipHig glossing rules. 5onventions or interlinear morpheme by morpheme glosses. 6eip>i$K Ma3) 1lanck Institute for -!olutionary =nthropolo$y. (ickel, (althasar Johanna +ichols. 2007. Inflectional morpholo$y. In "imothy Shopen %ed.*, Language typology and syntactic description, 2nd edn, ; !ols, !ol. III, 459)280. #ambrid$eK #ambrid$e Ani!ersity 1ress. (innick, /obert I. 4994. 0ime and the verb- a guide to tense and aspect. 23fordK 23ford Ani!ersity 1ress. (ird, Ste!en Gary Simons. 200;. Se!en dimensions of portability for lan$ua$e documentation and description. Language 79%;*, FF7)F92. (leek, (leek, . <. I. 4952. A comparative grammar o douth A rican languages. &art '. <. I. 4959. A comparative grammar o douth A rican languages. &art ''- the phonology. 6ondonK "r@bner.

858

concord section '- the noun. 6ondonK "r@bner. (lench, /o$er M. 4999. "he status of the lan$ua$es of #entral +i$eria. In Mathias (ren>in$er %ed.*, )ndangered languages in A rica, 497)205. &lnK &ppe Eerla$. (lench, /o$er M. n.d. a. "he est &ain'i lan$ua$es of northwestern +i$eriaK #omparati!e wordlist. -lectronic ms. I=!ailable onlineK httpKNNwww.ro$erblench.infoN6an$ua$e20dataN+i$er)#on$oN(enue)#on$oN est 20&ain'iN est20&ain'i)subclassification.pdfJ (lench, /o$er M. n.d. b. "he (enue)#on$o lan$ua$esK a proposed internal classification. -lectronic ms. I=!ailable onlineK httpKNNwww.ro$erblench.infoN6an$ua$e20dataN+i$er)#on$oN(enue)#on$oN GeneralN(enue)#on$o20classification20latest.pdfJ (le!ins, Juliette. 2008. )volutionary &honology. #ambrid$eK #ambrid$e Ani!ersity 1ress. (oett$er, -. E. (oett$er. 4957. "sureshe Grammar. Anpublished ms. (olin$er, .wi$ht. 4977. Ieaning and orm. 6ondonK 6on$man. (olin$er, .wi$ht. 4979. 1ronouns in discourse. In "almy Gi!n %ed.*, eiscourse and synta* %Synta3 and semantics 42*, 299);09. +ew BorkK =cademic 1ress. (ond, 2li!er. 2009. Infrastructure reDuires a foundationK a base for the canons of ne$ation. 1aper presented at #reatin$ Infrastructure for #anonical "ypolo$y, Ani!ersity of Surrey. (osch, 1eter. 499;. Agreement and anaphora %#o$niti!e Science Series*. 6ondonK =cademic 1ress. (ow, #atherine, (aden <u$hes Ste!en (ird. 200;. "owards a $eneral model of interlinear te3t. &roceedings o )I)Le "orkshop 4aa6- eigitiHing and annotating te*ts and ield recordings, -ast 6ansin$, Michi$an I=!ailable onlineK httpKNNwww.lin$uistlist.or$NemeldNworkshopN200;Nbowbadenbird)paper.pdfJ. (resnan, Joan. 2004. Le*icalc+unctional synta*. 23fordK (lackwell. rq{^n^Yen^YYe^^[YVYn^qr\nVYXVYewno[\qY[^\qr{[e^Y[^Yp[wpqnyYnYwn{qY{o}~gY of factori>ation in $rammar. Linguistic 'n$uiry 20%4*, 4)F0. rq{^n^Yen^YYniYVYwpeijeVYXVYxe|[wY|re^e}^Yn^~Ynrqqiq^oY[^Yp[wpqnV Language 5;%8*, 784)792. (resnan, Joan 6ioba Moshi. 4990. 2b'ect asymmetries in comparati!e (antu synta3.

85F

Linguistic 'n$uiry 24%2*, 487)49F. (ybee, Joan. 499F. Iorphology- a study o the relation between meaning and orm %"ypolo$ical Studies in 6an$ua$e 9*. =msterdamK John (en'amins. (ybee, Joan, /e!ere 1erkins #hica$o 1ress. #ahill, Michael. 2000. "onal associati!e morphemes in 2ptimality "heory. d/ "orking &apers in Linguistics F;, ;4)70. #=1/2 /esearch 2ffice. 499F. 3ingdoms at war. JosK #=1/2 Media. #hafe, #hafe, allace %ed.* 4990. 0he &ear stories- cognitive, cultural, and linguistic aspects allace. 4990. Some thin$s that narrati!es tell us about the mind. In (ruce &. o narrative production. +orwood, +JK =ble3. (ritton =nthony .. 1elle$rini %eds.*, Narrative thought and narrative language, 79)99. <illsdale, +JK 6awrence -rlbaum. #hafe, allace. 4998. eiscourse, consciousness, and time. #hica$oK Ani!ersity of #hica$o 1ress. #hambers, Mary. 2009. "hich way is up Iotion verbs and paths o motion in 3ubokota, an Austronesian language o the dolomon 'slands . 1h. dissertation, School of 2riental and =frican Studies. pn|in^ Y p[rfqg Y Y zq{ie^~ Y V Y zqrjg{p[rqV Y XXV Y n}inrbV Y ^ Y zq{ie^~ Y V Y .erbyshire Geoffrey &. 1ullum %eds.*, Handbook o AmaHonian languages, !ol. ;, 454);F2. (erlinK Mouton de Gruyter. #hilds, G. "ucker. 4998. =frican ideophones. In 6eanne <inton, Johanna +ichols John J. 2hala %eds.*, dound symbolism, 479)208. #ambrid$eK #ambrid$e Ani!ersity 1ress. #homsky, +oam. 4994. Lectures on Government and (inding. .ordrechtK 0oris. #hun$, Sandra. 4999. 0he design o agreement- evidence rom 5hamorro. #hica$oK Ani!ersity of #hica$o 1ress. #inDue, Gu$lielmo. 200F. .eri!in$ Greenber$Ts Ani!ersal 20 and its e3ceptions. Linguistic 'n$uiry ;5%;*, ;4F);;2. #lements, G. +. 2000. 1honolo$y. In (ernd <eine .erek +urse %eds.*, A rican languages- an introduction, 42;)450. #ambrid$eK #ambrid$e Ani!ersity 1ress. illiam 1a$liuca. 4998. 0he evolution o grammartense, aspect, and modality in the languages o the world. #hica$oK Ani!ersity of

855

#lements, G. +. Samuel Jay &eyser. 499;. 5^ &honology- a generative theory o the syllable. #ambrid$e, Mass.K MI" 1ress. #omrie, (ernard. 4994. Language universals and linguistic typology. 23fordK (lackwell. #omrie, (ernard. 4999. "opics, $rammaticali>ed topics, and sub'ects. &roceedings o the +ourteenth Annual Ieeting o the (erkeley Linguistics dociety, 25F)279. (erkeley, #alifornia. #omrie, (ernard. 200;. hen a$reement $ets tri$$er)happy. 0ransactions o the &hilological dociety 404%2*, ;4;);;7. #onnell, (ruce .. /obert 6add. 4990. =spects of pitch reali>ation in Boruba. &honology 7, 4);0. #ontini)Mora!a, -llen. 2002. % hat* do noun class markers meanM In =msterdamK John (en'amins. #ook, Ei!ian J. Mark +ewson. 4995. 5homsky's /niversal Grammar- an introduction, 2nd edn. 23fordK (lackwell. #orbett, Gre!ille G. 4994. Gender. #ambrid$eK #ambrid$e Ani!ersity 1ress. #orbett, Gre!ille G. 2000. Number. #ambrid$eK #ambrid$e Ani!ersity 1ress. #orbett, Gre!ille G. 200;a. =$reementK terms and boundaries. In 409)422. #orbett, Gre!ille G. 200;b. =$reementK the ran$e of the phenomenon and the principles of the Surrey .atabase of =$reement. 0ransactions o the &hilological dociety 404%2*, 4FF)202. #orbett, Gre!ille G. 200Fa. "he canonical approach in typolo$y. In Hy$munt 0ra'>yn$ier, =dam <od$es .a!id S. /ood %eds.*, Linguistic diversity and language theories, 2F)89. =msterdamK John (en'amins. #orbett, Gre!ille G. 200Fb. +umber of $enders. In Matthew .ryer, Martin <aspelmath, .a!id Gil (ernard #omrie %eds.*, 0he "orld Atlas o Language dtructures, 425)429. 23fordK 23ford Ani!ersity 1ress. #orbett, Gre!ille G. 2005. Agreement. #ambrid$eK #ambrid$e Ani!ersity 1ress. #orbett, Gre!ille G. 2007. Gender and noun classes. In "imothy Shopen %ed.*, illiam Griffin %ed.*, 0he Role o Agreement in Natural Language %"e3as 6in$uistics 0orum F;*, allis /eid, /icardo 2the$uy +ancy Stern %eds.*, dignal, meaning, and message, ;)58.

857

Language typology and syntactic description, 2nd edn, ; !ols, !ol. III, 284)279. #ambrid$eK #ambrid$e Ani!ersity 1ress. #reissels, .enis. 2000. "ypolo$y. In (ernd <eine .erek +urse %eds.*, A rican languages- an introduction, 2;4)2F9. #ambrid$eK #ambrid$e Ani!ersity 1ress. #roft, #roft, illiam. 4994. dyntactic categories and grammatical relations- the cognitive illiam .. =lan #ruse. 2008. 5ognitive Linguistics. #ambrid$eK #ambrid$e organiHation o in ormation. #hica$oK Ani!ersity of #hica$o 1ress. Ani!ersity 1ress. #rothers, John. 4979. "ypolo$y and uni!ersals of !owel systems. In Joseph <. Greenber$ %ed.*, /niversals o human language, !ol. 2, 9;)4F2. StanfordK Stanford Ani!ersity 1ress. #ro>ier, .a!id <. 4998. A study in the discourse grammar o 5ishingini. 1h. dissertation, Ani!ersity of Ibadan. #ro>ier, .a!id <. /o$er M. (lench. 4992. An 'nde* o Nigerian Languages, 2nd edn. .allasK SI6. #ruttenden, =lan. 4997. 'ntonation, 2nd edn. #ambrid$eK #ambrid$e Ani!ersity 1ress. #uly, #hristopher. 2000. =n incorporated topic marker in "akelma. In Miriam (utt "racy <olloway &in$ %eds.*, &roceedings o the L+G aa 5on erence, /niversity o 5ali ornia, (erkeley, 5;)94. StanfordK #S6I 1ublications. .ahl, sten Ei!eka Eelupillai. 200F. "he 1erfect. In Martin <aspelmath, Matthew .ryer, .a!id Gil (ernard #omrie %eds.*, 0he "orld Atlas o Language dtructures, 274)272. 23fordK 23ford Ani!ersity 1ress. .alby, .a!id. 4958. 1ro!isional identification of lan$ua$es in the 1oly$lotta =fricana. dierra Leone Language Review ;, 9;)90. .alrymple, Mary Irina +ikolae!a. 200F. +onsub'ect a$reement and discourse roles. In * ord "orking &apers in Linguistics, &hilology, and &honetics 40, 74)92. I=!ailable onlineK httpKNNwww.lin$)phil.o3.ac.ukNdownloadN2 1200F.pdfJ .ancy$ier, (arbara -!e Sweetser. 200F. Iental spaces in grammar- conditional constructions. #ambrid$eK #ambrid$e Ani!ersity 1ress. .aniel, Michael -dith Mora!csik. 200F. "he associati!e plural. In Martin <aspelmath, Matthew .ryer, .a!id Gil (ernard #omrie %eds.*, "orld Atlas o Language dtructures, 4F0)4F;. 23fordK 23ford Ani!ersity 1ress.

859

.an>i$er, -!e. 499;. Ianual or the dpace dtimuli 3it ,.4. +i'me$enK Ma3 1lanck Institute for 1sycholin$uistics. .a!is, Stuart. 4999. 2n the representation of initial $eminates. &honology 45%4*, 9;) 408. .e olf, 1aul 1olydoor. 4974. 0he noun class system o &rotoc(enuec5ongo. "he <a$ueK Mouton de Gruyter. .emuth, &atherine. 4990. 6ocati!es, impersonals and e3pleti!es in Sesotho. 0he Linguistic Review, !ol. 7, 2;;)289. .ordrechtK 0oris. .emuth, &atherine Mark Johnson. 4999. Interaction between discourse functions and a$reement in Setawana. !ournal o A rican Languages and Linguistics 44, 24);F. .enny, J. 1eter #het =. #reider. 4995. "he semantics of noun classes in 1roto (antu. In #olette #rai$ %ed.*, Noun classes and categoriHation, 247,2;9. =msterdamK John (en'amins. .ettweiler, Stephen <. 2000. Eowel harmony and neutral !owels in #:6ela. !ournal o "est A rican Languages 29%4*, ;)49. .ettweiler, Stephen <. n.d. #:6ela $rammar sketch. -lectronic ms. .ettweiler, Stephen <. Sonia G. .ettweiler. 2002. Sociolin$uistic sur!ey %le!el one* of the &amuku lan$ua$e cluster. -lectronic ms. httpKNNwww.sil.or$NsilesrN200;Nsilesr200;)00;.pdf .in$emanse, Mark. 2005. "he semantics of (antu noun classification. -lectronic ms. httpKNNwww.mpi.nlNMembersNMark.in$emanseN1ublications .i3on, /obert M. . =le3andra B. =ikhen!ald. 2000. Introduction. In /obert M. . .i3on =le3andra B. =ikhen!ald %eds.*, 5hanging valency- case studies in transitivity, 4)29. #ambrid$eK #ambrid$e Ani!ersity 1ress. .i3on, /obert M. . =le3andra B. =ikhen!ald. 2002. "ord- a crossclinguistic typology. #ambrid$eK #ambrid$e Ani!ersity 1ress. .ooley, /obert =. 2007. -3plorations in discourse topicality. d'L )lectronic "orking &apers 2007)040. httpKNNwww.sil.or$NsilewpN2007Nsilewp2007)040.pdf .ooley, /obert =. Stephen <. 6e!insohn. 2004. AnalyHing discourse- a manual o basic concepts. .allasK SI6 International. .ryer, Matthew. 200F. 1osition of interro$ati!e phrases in content Duestions. In Martin <aspelmath, Matthew .ryer, .a!id Gil (ernard #omrie %eds.*, "orld Atlas o

859

Language dtructures, ;79);94. .u (ois, John . 4997. "he discourse basis of er$ati!ity. Language 5;%8*, 90F)9FF. .uchan, Judith 0., Gail =. (ruder 6ynne -. <ewitt. 499F. eei*is in narrative- a cognitive science perspective. <illsdale, +JK 6awrence -rlbaum. .uff, -. #. 4972. Ga>etteer of the &onta$ora 1ro!ince. In =. <. M. &irk)Greene %ed.* GaHetteers o the Northern provinces o Nigeria, 4)72. 6ondonK 0rank #ass. I2ri$inally published in 4920 by -!ans, +icholas. 4999. pronounsK e!idence from aterlow %6ondon*J. (inin' Gun)wok. dprachtypologie und hy ar$ument affi3es in polysynthetic lan$ua$es are not

/niversalien orschung F2, 2FF)294. -!ans, +icholas. 2002. "he true status of $rammatical ob'ect affi3esK e!idence from (inin' Gun)wok. In +icholas -!ans <ans)J@r$en Sasse %eds.*, &roblems o polysynthesis, 4F)F0. (erlinK =kademie Eerla$. 0alk, Behuda +. 2005. 2n the representation of case and a$reement. In Miriam (utt "racy <olloway &in$ %eds.*, 0he &roceedings o the L+G a] 5on erence. StanfordK #S6I 1ublications. 0arrar, Scott .. "erence 6an$endoen. 200;. = lin$uistic ontolo$y for the Semantic eb. GL0 'nternational 7%;*, 97)400. 0auconnier, Gilles. 4998. Iental spaces- aspects o meaning construction in natural languages, 2nd edn. #ambrid$eK #ambrid$e Ani!ersity 1ress. 0auconnier, Gilles. 4997. Iappings in thought and language. #ambrid$eK #ambrid$e Ani!ersity 1ress. 0er$uson, #harles =. Michael (arlow. 4999. Introduction. In Michael (arlow #harles =. 0er$uson %eds.*, Agreement in natural language, 4)22. StanfordK #S6I. 0oley, illiam =. 200;. Genre, re$ister and lan$ua$e documentation in literate and preliterate communities. In 1eter &. =ustin %ed.*, Language documentation and description, !ol. 4, 9F)99. 6ondonK <ans /ausin$ -ndan$ered 6an$ua$es 1ro'ect. 0ortescue, Michael .. 4998. "est Greenlandic. (eckenham, &entK #room <elm. 0o3, (arbara =. 4997. =naphora in popular written -n$lish narrati!es. In /ussell S. "omlin %ed.*, 5oherence and grounding in discourse, 4F7)478. =msterdamK John (en'amins. 0rancis, <artwell S., Michelle 6. Gre$ory 6aura =. Michaelis. 4999. =re le3ical

870

sub'ects de!iantM &apers rom the 6Tth regional meeting o the 5hicago Linguistics dociety, !ol 4., 9F)97. #hica$oK #hica$o 6in$uistics Society. 0rawley, illiam. 4992. Linguistic semantics. <illsdale, +JK 6awrence -rlbaum. Gerhardt, 6udwi$. 4999. &ain'i and 1latoid. In John (endor)Samuel %ed.*, 0he Nigerc 5ongo languages, ;F9);75. 6anham, Md.K Ani!ersity 1ress of =merica. Gernsbacher, Morton =nn. 4998. #o$niti!e responses to lin$uistic topic chan$es in discourse. In Morton =nn Gernsbacher %#hair*, 5rosscdisciplinary approaches to language processing. &roceedings o the si*th annual con erence o the 5ognitive dcience dociety, 9;)99. Gernsbacher, Morton =nn. 499F. Surface information loss in comprehension. 5ognitive &sychology 47, ;28);5;. Gi!n, "almy. 4975. "opic, pronoun, and $rammatical a$reement. In #harles +. 6i %ed.*, dub%ect and topic, 489)499. +ew BorkK =cademic 1ress. Gi!n, "almy. 499;. "opic continuity in discourseK an introduction. In "almy Gi!n %ed.*, 0opic continuity in discourse- a $uantitative crossclanguage study %"ypolo$ical Studies in 6an$ua$e ;*, F)84. =msterdamK John (en'amins. Gordon, /aymond G., Jr. 200F. )thnologue- languages o the world, 4Fth edn. .allasK SI6 International. Greenber$, Joseph <. 49FF. dtudies in A rican linguistic classi ication. +ew <a!enK #ompass 1ublishin$. Greenber$, Joseph <. 495;. Some uni!ersals of $rammar with particular reference to the order of meanin$ful elements. In Joseph <. Greenber$ %ed.*, /niversals o language, 7;)44;. #ambrid$e, M=K MI" 1ress. Grice, <. 1aul. 497F. 6o$ic and con!ersation. In 1eter #ole Jerry 6. Mor$an %eds.*, dpeech acts %Synta3 and semantics ;*, 84)F9. +ew BorkK =cademic 1ress. Grimes, Joseph -. 497F. 0he thread o discourse. "he <a$ueK Mouton de Gruyter. Grimes, Joseph -. 4979. &apers on discourse. .allasK Summer Institute of 6in$uistics. Grine!ald, #olette 0rank Seifart. 2008. +oun classes in =frican and =ma>onian lan$ua$esK towards a comparison. Linguistic 0ypology 9%2*, 28;)29F. Gunn, <. .. 0. 1. #onant. 4950. &eoples o the middle Niger region, Northern Nigeria. 6ondonK International =frican Institute. <aiman, John. 4979. #onditionals are topics. Language F8, F58)F99.

874

<alliday, M. =. &. /uDaiya <asan. 4975. 5ohesion in )nglish. 6ondonK 6on$man. <aspelmath, Martin. 4997. 'nde inite pronouns. 23fordK 23ford Ani!ersity 1ress. <aspelmath, Martin. 2002. /nderstanding morphology. 6ondonK =rnold. <aspelmath, Martin. 2008. #oordinatin$ constructionsK an o!er!iew. In Martin <aspelmath %ed.*, 5oordinating constructions, ;);9. =msterdamK John (en'amins. <ayes, (ruce. 4999. #ompensatory len$thenin$ in moraic phonolo$y. Linguistic 'n$uiry 20%2*, 2F;);05. <eath, Jeffrey. 497F. Some functional relationships in $rammar. Language F4, 99)408. <eath, Jeffrey. 499;. /eferential trackin$ in +un$$ubuyu %=ustralia*. In John <aiman 1amela Munro %eds.*, dwitch re erence and universal grammar- proceedings o a symposium on switch re erence and universal grammar, "innipeg, Iay ,b`, %"ypolo$ical Studies in 6an$ua$e 2*, 429)489. =msterdamK John (en'amins. <edin$er, /obert. 4990. "he noun classes of Aks J eJ. In 6arry M. <yman %ed.*, Noun classes in the Grass ields (antu borderland %Southern #alifornia 2ccasional 1apers in 6in$uistics 9*, 4)25. 6os =n$elesK Ani!ersity of Southern #alifornia. <eine, (ernd. 4992. =frican noun class systems. In <ans'akob Seiler #hristian 6ehmann %eds.*, Apprehension- eas sprachliche )r assen von Gegenstnden, &art '- (ereich und rdnung der &hnomene, 499,245. "@bin$enK Gunter +arr. <endrikse, =. 1. 4997. Systemic polysemy in the Southern (antu noun class system. In <ubert #uyckens (ritta -. Hawada %eds.*, &olysemy in 5ognitive Linguistics, 49F)242. =msterdamK John (en'amins. <ewin$s, =nn Martin <ewin$s. 200F. Grammar and conte*t- an advanced resource book. 6ondonK /outled$e. <immelmann, +ikolaus 1. 4999. .ocumentary and descripti!e lin$uistics. Linguistics ;5, 454)49F. <inds, John. 4979. 2r$ani>ational patterns in discourse. In "almy Gi!n %ed.*, eiscourse and synta* %Synta3 and semantics 42*, 4;F)4F9. +ew BorkK =cademic 1ress. <inds, John. 4998. "opic maintenance in Japanese narrati!es and Japanese con!ersational interaction. eiscourse &rocesses 7, 85F)892. <ockett, #harles 0. 49F9. A course in modern linguistics. +ew BorkK Macmillan. <offmann, #arl. 495;. "he noun)class system of #entral &ambari. !ournal o A rican

872

Languages 2%2*, 450)459. <offmann, #arl. 495F. = wordlist of #entral &ambari. !ournal o "est A rican Languages 2%2*, 7);4. <offmann, #arl. 4957. =n outline of the .akarkari noun class system and the relation between prefi3 and suffi3 noun)class systems. In G. Manessy %ed.*, La classi ication nominale dans les langues nqgroca ricaines, 2;7)2F9. 1arisK #+/S. <offmann, #arl. 4972. = note on !owel contraction in #entral &ambari. Research Notes .eepartment o Linguistics and Nigerian Languages, 'badan2 F%2N;*, 7;)94. <opper, 1aul J. Sandra =. "hompson. 4990. "ransiti!ity in $rammar and discourse. Language F5, 2F4)299. <opper, 1aul J. Sandra =. "hompson. 4998. "he discourse basis for le3ical cate$ories in uni!ersal $rammar. Language 50, 70;)7F2. <ume, -li>abeth, Jennifer Muller =one !an -n$elenho!en. 4997. +onmoraic $eminates in 6eti. &honology 48%;*, ;74)802. <yman, 6arry M. 4990. Noun classes in the Grass ields (antu borderland %Southern #alifornia 2ccasional 1apers in 6in$uistics 9*. 6os =n$elesK Ani!ersity of South #alifornia. Ja$$ar, 1hilip J. 499;. Some dimensions of topic)+1 continuity in <ausa narrati!e. In "almy Gi!n %ed.*, 0opic continuity in discourse- a $uantitative crossclanguage study %"ypolo$ical Studies in 6an$ua$e ;*, ;5F)828. =msterdamK John (en'amins. Ja$$ar, 1hilip J. 499F. +actors governing the morphological coding o re erents in Hausa narrative discourse. 1h. dissertation, Ani!ersity of #alifornia, 6os =n$eles. Ja$$ar, 1hilip J. 4999. .iscourse)deployability and indefinite +1)markin$ in <ausaK a demonstration of the uni!ersal S#ate$oriality <ypothesisT. In Graham 0urniss 1hilip J. Ja$$ar %eds.*, dtudies in Hausa language and linguistics in honour o +. ". &arsons, 8F)54. 6ondonK &e$an 1aul. Ja$$ar, 1hilip J. 2004. Hausa. =msterdamK John (en'amins. Jelinek, -loise. 4998. -mpty cate$ories, case, and confi$urationality. Natural Language and Linguistic 0heory 2, ;9)75. Ji, Shao'un. 2002. Identifyin$ episode transitions. !ournal o &ragmatics ;8, 42F7) 4274.

87;

Johnston, <arry <. 4949. A comparative study o the (antu and demic(antu languages , !ol. 4. 23fordK #larendon 1ress. &ameyama, Me$umi. 499F. ero anaphora- the case o !apanese. 1h. dissertation, Stanford Ani!ersity. nonijn Y rn^w[{V Y UWV Y n^o} Y ^ei[^nf Y ier|pefegV Y ^ Y zqrq Y }r{q Y Y Zurnr~ Y 1hilippson %eds.*, 0he (antu languages, 40;)420. 6ondonK /outled$e. &ayne, /ichard S. 4998. 0he antisymmetry o synta* %6in$uistic InDuiry Mono$raphs 2F*. #ambrid$e, M=K MI" 1ress. &eenan, -dward 6. (ernard #omrie. 4977. +1 accessibility and uni!ersal $rammar. Linguistic 'n$uiry 9, 5;)400. &horoun'aia, -katerina 6iliana "olchinsky. 2008. .iscursi!e constraints on the le3ical reali>ation of ar$uments in Spanish. In /uth =. (erman %ed.*, Language development across childhood and adolescence, 9;)409. =msterdamK John (en'amins. &in$, "racy <olloway. 4997. 0ocus domains and information structure. In Miriam (utt "racy <olloway &in$ %eds.*, 0he &roceedings o the L+G b_ 5on erence. StanfordK #S6I 1ublications. &iparsky, 1aul. 4992. 0rom cyclic phonolo$y to le3ical phonolo$y. In <arry !an der <ulst +or!al Smith %eds.*, 0he structure o phonological representations .&art '2, 4;4)475. .ordrechtK 0oris. [{{qjqropYpnrfq{YYzn\[~Y~~q^VYUWVYxe^qVY ^YzqrqY}r{qYYZurnr~Yp[f[||{e^Y %eds.*, 0he (antu languages, F9)70. 6ondonK /outled$e. &lamer, Marian. 2009. .ifferential markin$ of intransiti!e sub'ects in &ambera %=ustronesian*. In <elen de <oop 1eter de Swart %eds.*, ei erential sub%ect marking, 294)299. .ordrechtK Sprin$er. &oelle, Si$ismund #olle$e. &ni$, -kkehard 1eter Siemund %with Stephan "pper*. 200F. Intensifiers and refle3i!e pronouns. In Martin <aspelmath, Matthew .ryer, .a!id Gil (ernard #omrie %eds.*, "orld Atlas o Language dtructures, 498)497. 23fordK 23ford Ani!ersity 1ress. ilhelm. 495;. &olyglotta A ricana- photomechanic reprint o the original edition, 5hurch Iissionary dociety, London ,`TJ. 0reetownK 0ourah (ay

878

&reiman, J. 4992. 1erception of sentence and para$raph boundaries in natural con!ersation. !ournal o &honetics 40, 45;)47F. &roe$er, 1aul /. 2008. AnalyHing synta*- a le*icalc unctional approach. #ambrid$eK #ambrid$e Ani!ersity 1ress. &uno, Susumu. 4975. Sub'ect, theme, and the speaker:s empathy , a ree3amination of relati!i>ation phenomena. In #harles +. 6i %ed.*, dub%ect and topic, 847)888. +ew BorkK =cademic 1ress. 6adefo$ed, 1eter Ian Maddieson. 4995. 0he sounds o the world's languages. 23fordK (lackwell. 6akoff, Geor$e. 4997. "omen, ire and dangerous things- what categories reveal about the mind. #hica$oK Ani!ersity of #hica$o 1ress. 6ambrecht, &nud. 4998. 'n ormation structure and sentence orm. #ambrid$eK #ambrid$e Ani!ersity 1ress. 6an$acker, /onald 6an$acker, /onald 6an$acker, /onald Gruyter. 6an$acker, /onald 42%2*, 48;)499. 6aoust, -. 4929. 5ours de berbre marocain. 1arisK <achette. 6ehiste, Ilse. 4979. Sentence boundaries and para$raph boundaries , perceptual e!idence. In 1. /. #lyne, Society. 6e!inson, Stephen #. 2008. .ei3is. In 6aurence /. <orn Gre$ory 6. handbook o pragmatics, 97)424. 23fordK (lackwell. 6i, #harles +. Sandra =. "hompson. 4979. "hird)person pronouns and >ero)anaphora in #hinese discourse. In "almy Gi!n %ed.*, eiscourse and synta* %Synta3 and semantics 42*, ;44);;F. +ew BorkK =cademic 1ress. 6ichtenberk, 0rantisek. 2000. Inclusory pronominals. ceanic Linguistics ;9%4*, 4);2. ard %eds.*, 0he . 0. <anks #. 6. <ofbauer %eds.*, 0he elements- a parasession on linguistic units and levels, 99)409. #hica$oK #hica$o 6in$uistics . 2004. .iscourse in #o$niti!e Grammar. 5ognitive Linguistics . 4997. +oundations o 5ognitive Grammar- theoretical prere$uisites, !ol. 4. StanfordK Stanford Ani!ersity 1ress. . 4994. +oundations o 5ognitive Grammar- descriptive . 2000. Grammar and conceptualiHation. (erlinK Mouton de application, !ol. 2. StanfordK Stanford Ani!ersity 1ress.

87F

6on$acre, /obert -. 4995. 0he grammar o discourse, 2nd edn. +ew BorkK 1lenum. 6o!elace, .a!id. 4999. #omputer printout on "su!ai. 6@pke, 0riederike. 2005. Small is beautiful , the contribution of small field)based corpora to different lin$uistic disciplines. In 1eter &. =ustin %ed.*, Language eocumentation and eescription, !ol. ;, 7F)40F. 6ondonK <ans /ausin$ -ndan$ered 6an$ua$es 1ro'ect. 6yons, #hristopher. 4999. ee initeness. #ambrid$eK #ambrid$e Ani!ersity 1ress. Mc#arthy, Michael. 2004. 'ssues in applied linguistics. #ambrid$eK #ambrid$e Ani!ersity 1ress. Mac.onell, James. 2007. = phonolo$ical and $rammatical sketch of the 1on$u lan$ua$e. -lectronic ms. McGill, Stuart. 2008. +ocus and activation in &aasaal- the particle r. M= dissertation, Ani!ersity of /eadin$. McGill, Stuart. 2007. "he #icipu noun class system. !ournal o "est A rican Languages ;8%2*, F4)90. McGill, Stuart. 2009. "he .amakawa lan$ua$e. -lectronic ms. I=!ailable onlineK httpKNNwww.cicipu.or$NpapersNdamakawa.pdf J McGill, Stuart. n.d. "he classification of #icipu. -lectronic ms. I=!ailable onlineK httpKNNwww.cicipu.or$NpapersN2009)0F)05PclassificationPcicipuPworkin$.pdf J Mc&inney, +orris 1. 4979. 1articipant identification in &a'e narrati!e. In Joseph -. Grimes %ed.*, &apers on discourse, 479)499. .allasK Summer Institute of 6in$uistics. Maho, Jouni 0. 4999. A comparative study o (antu noun classes. Gtebor$K =cta Ani!ersitatis Gothobur$ensis. Maibaum, =nita. 4979. 1articipants in Jirel narrati!e. In Joseph -. Grimes %ed.*, &apers on discourse, 20;)207. .allasK Summer Institute of 6in$uistics. Marten, 6ut>. 200F. "one cases in <erero. -lectronic ms. S2=SNAni!ersity of +amibia. Mathews, =. (. 4925a. =nthropolo$ical notes on the (usawa. +ational =rchi!es, &adunaK S+1 47N9Nk.2059. Mathews, =. (. 4925b. <istorical and anthropolo$ical report on the =chifawa. +ational =rchi!es, &aduna, S+1 47N9Nk.2059. Mathews, =. (. 49F0. "he &isra le$end. A rican dtudies 9, 488)487.

875

Mchombo, Sam =. 2008. 0he synta* o 5hichewa. #ambrid$eK #ambrid$e Ani!ersity 1ress. Meek, #harles &in$sley. 492F. 0he northern tribes o Nigeria, 2 !ols. 6ondonK 23ford Ani!ersity 1ress. Meinhof, #arl. 49;2. 'ntroduction to the phonology o the (antu languages. /e!ised, enlar$ed and translated from German %Grundriss einer 6autlehre der (antu) sprachen, 2nd edition, 4940* by +. J. !an and .r. =lice 47%2*, ;07);7;. Mithun, Marianne. 4992. Is basic word order uni!ersalM In .oris 1ayne %ed.*, &ragmatics o word order le*ibility, 4F)54. =msterdamK John (en'amins. Mithun, Marianne. 200;. 1ronouns and a$reementK the information status of pronominal affi3es. 0ransactions o the &hilological dociety 404%2*, 2;F)279. Morimoto, Bukiko. 2002. 1rominence mismatches and differential ob'ect markin$ in (antu. In Miriam (utt "racy <olloway &in$ %eds.*, &roceedings o the L+G a4 5on erence, National 0echnical /niversity o Athens, 292);48. StanfordK #S6I 1ublications. Morimoto, Bukiko. 2009. 0rom topic to sub'ect markin$K implications for a typolo$y of sub'ect markin$. In <elen de <oop 1eter de Swart %eds.*, ei erential sub%ect marking, 499)222. .ordrechtK Sprin$er. Mous, Maarten. 200;. 6oss of lin$uistic di!ersity in =frica. In Mark Janse Si'men "ol %eds.*, Language death and language maintenance, 4F7)470. =msterdamK John (en'amins. Mufwene, Salikoko S. 4990. (antu class prefi3esK inflectional or deri!ationalM 5hicago Linguistic dociety, !ol. 45, 285,2F9. Munro, 1amela. 4997. "he Garifuna $ender system. In Jane <. <ill, 1. J. Mistry 6yle #ampbell %eds.*, 0he li e o language- papers in linguistics in honour o "illiam (right %"rends in 6in$uistics 409*, 88;)852. Mouton de Gruyter. Myhill, John. 2004. "ypolo$y and discourse analysis. In .eborah Schiffrin, .eborah "annen <eidi -. <amilton %eds.*, 0he handbook o discourse analysis, 454)478. 23fordK (lackwell. armelo in collaboration with the author erner. (erlinK .ietrich /eimer.

qf}Y erVYXWVYrqqiq^oYe\qr^iq^oYwe^r}q^wqVYLingvistic 'nvestigationes

877

+athan, .a!id 1eter &. =ustin. 2008. /econcei!in$ metadataK lan$ua$e documentation throu$h thick and thin. In 1eter &. =ustin %ed.*, Language eocumentation and eescription, !ol. 2, 479)497. 6ondonK <ans /ausin$ -ndan$ered 6an$ua$es 1ro'ect. +ewman, John 0. 4979. 1articipant orientation in 6on$uda folk tales. In Joseph -. Grimes %ed.*, &apers on discourse, 94)408. .allasK Summer Institute of 6in$uistics. +ewman, 1aul. 4990. Nominal and verbal plurality in 5hadic %1ublications in =frican 6an$ua$es and 6in$uistics 42*. .ordrechtK 0oris. +ewman, 1aul. 2000. 0he Hausa language- an encyclopedic re erence grammar. +ew <a!enK Bale Ani!ersity 1ress. +ewmeyer, 0rederick J. 4999. Language orm and language unction. #ambrid$e, M=K MI" 1ress. 2nishi, Masayuki. 4998. A grammar o Iotuna .(ougainville, &apua New Guinea2. 1h. dissertation, =ustralian +ational Ani!ersity. 1almer, 0. /. 2004. Mood and modality, 2nd edn. #ambrid$eK #ambrid$e Ani!ersity 1ress. 1ayne, "homas -. 4997. eescribing morphosynta*. #ambrid$eK #ambrid$e Ani!ersity 1ress. 1erlmutter, .a!id. 499F. 1honolo$ical Duantity and multiple association. In John =. Goldsmith %ed.*, 0he handbook o phonological theory, ;07);47. 23fordK (lackwell. 1errin, Mona. 4979. hoTs who in Mambila folk stories. In Joseph -. Grimes %ed.*, &apers on discourse, 40F)449. .allasK Summer Institute of 6in$uistics. 1olinsky, Maria (ernard #omrie. 4999. =$reement in "se>. +olia Linguistica ;;%2*, 409)4;0. 1ostal, 1aul M. 4970. 2n so)called pronouns in -n$lish. In /oderick =. Jacobs 1eter S. /osenbaum %eds.*, Readings in )nglish trans ormational grammar, F5)92. altham, Mass.K Ginn. 1ra>an, #eslaus. 4977. 0he eukkawa o northwest Nigeria. 1ittsbur$hK .uDuesne Ani!ersity 1ress. _uirk, /andolph, Sidney Greenbaum, Geoffrey 6eech, and Jan S!art!ik. 499F. A

879

comprehensive grammar o the )nglish language. 6on$manK 6ondon. /ialland, =nnie. 2007. _uestion prosodyK an =frican perspecti!e. In #. Gussenho!en ". /iad %eds.*, 0ones and tunes- studies in word and sentence prosody, ;F)52. (erlinK Mouton de Gruyter. /ichardson, I. 4957. 6in$uistic e!olution and (antu noun class systems. In G. Manessy %ed.*, La classi ication nominale dans les langues nqgroca ricaines, ;7;);90. 1arisK #+/S. /i'khoff, J. 2002. 0he Noun &hrase %23ford Studies in "ypolo$y and 6in$uistic "heory*. 23fordK 23ford Ani!ersity 1ress. /osch, -leanor. 4979. 1rinciples of cate$ori>ation. In -leanor /osch (arbara (. 6loyd %eds.*, 5ognition and categoriHation, 27)89. <illsdale, +JK 6awrence -rlbaum. /owlands, -. #. 4952. +otes on some class lan$ua$es of northern +i$eria. A rican Language dtudies ;, 74)9;. Sa$na, Ser$e. 2009. +ormal and semantic properties o the G %%olaay )egimaa nominal classi ication system. 1h. dissertation, School of 2riental and =frican Studies. wpn~qjqrYxp[feYVYUWVYzqr[\no[e^VY ^YzqrqY}r{qYYZurnr~Yp[f[||{e^Yq~{VY0he (antu languages, 74)99. 6ondonK /outled$e. Schult>e)(erndt, -!a. 2005. 6in$uistic annotation. In Jost Gippert, +ikolaus 1. <immelmann Alrike Mosel %eds.*, )ssentials o language documentation, 24;)2F4. (erlinK Mouton de Gruyter. wpvnroYqn^}wYe}[{qn^YeYnopnf[qYnffuqYYpr[{tian =bry. 4997. Ma'or trends in !owel system in!entories. !ournal o &honetics 2F, 2;;)2F;. Schwar>)0riesel, Monika. 2007. Indirect anaphora in te3tK a co$niti!e account. In Monika Schwar>)0riesel, Manfred #onsten Mareile &nees %eds.*, Anaphors in te*t, ;)20. =msterdamK John (en'amins. Seifart, 0rank. 200F. LPD7G<?QV<Q?D7>9M7QGD7E\7GP>XDNj>GDM79EQ97VB>GGDG7897:8?>> 7 .North "est AmaHon2. 1h. dissertation, /adboud Ani!ersity, +i'me$en. Sel!ik, &ari)=nne. 4997. hen a dance resembles a treeK a polysemy analysis of three Setswana noun classes. In <ubert #uyckens (ritta -. Hawada %eds.*, &olysemy in cognitive linguistics, 454)498. =msterdamK John (en'amins. Siewierska, =nna. 4998. 0he passive- a comparative linguistic analysis. 6ondonK #room

879

<elm. Siewierska, =nna. 4999. 0rom anaphoric pronoun to $rammatical a$reement markerK why ob'ects don:t make it. In Gre!ille G. #orbett %ed.*, Agreement .dpecial 'ssue o +olia Linguistica 66142, 22F)2F4. (erlinK Mouton de Gruyter. Siewierska, =nna. 2008. &erson. #ambrid$eK #ambrid$e Ani!ersity 1ress. Siewierska, =nna. 200F. 1assi!e constructions. In Martin <aspelmath, Matthew .ryer, .a!id Gil (ernard #omrie %eds.*, "orld Atlas o Language dtructures, 8;8)8;7. 23fordK 23ford Ani!ersity 1ress. Sil!erstein, Michael. 4975. <ierarchy of features and er$ati!ity. In /obert M. =ustralian Institute of =bori$inal Studies. Skopeteas, S., I. 0iedler, S. <ellmuth, =. Schwar>, /. Stoel, G. 0anselow, #. 0ry M. &rifka. 2005. uestionnaire on 'n ormation dtructure ./'d2 %Interdisciplinary Studies on Information Structure %ISIS* 8*. 1otsdamK 1otsdam Ani!ersity. Smith, /ebecca .ow. 2007. 0he noun class system o utcIa'in, a "est 3ain%i language o Nigeria. M= dissertation, Ani!ersity of +orth .akota. Snider, &eith James /oberts. 2008. SI6 comparati!e =frican wordlist. !ournal o "est A rican Languages ;4%2*, 7;)422. Soua$, 6ameen. 2007. "he typolo$y of number borrowin$ in (erber. 1aper presented at the 0ifth #ambrid$e 1ost$raduate #onference in 6in$uistics. I=!ailable onlineK httpKNNwww.lin$.cam.ac.ukNcamlin$NManuscriptsN#am6in$2007PSoua$.pdf J Spencer, &atharine. 2009. &amuku lan$ua$es sur!ey report. 1aper presented at the 4st orkshop on est &ain'i lan$ua$es, &onta$ora. Stallcup, &enneth 6. 4990. +oun classes in -simbi. In 6arry M. <yman %ed.*, Noun classes in the Grass ields (antu borderland %Southern #alifornia 2ccasional 1apers in 6in$uistics 9*, 4;9)4F;. 6os =n$elesK Ani!ersity of Southern #alifornia. Stark, Janie 1. 200;. 0he 0sikimbacHausac)nglish dictionary. SalkaK &ambari 6an$ua$e 1ro'ect. Stark, Janie 1. 2008a. 0he 5ishinginicHausac)nglish dictionary. SalkaK &ambari 6an$ua$e 1ro'ect. Stark, Janie 1. 2008b. 0he 0sishinginicHausac)nglish dictionary. SalkaK &ambari . .i3on %ed.*, Grammatical categories in Australian languages, 442)474. #anberraK

890

6an$ua$e 1ro'ect. Stassen, 6eon. 4997. 'ntransitive predication. 23fordK 23ford Ani!ersity 1ress. Steele, Susan. 4979. ord order !ariationK a typolo$ical study. In Joseph <. Greenber$, #harles =. 0er$uson -dith =. Mora!csik %eds.*, /niversals o human language, !ol. 8K Synta3, F9F)52;. StanfordK Stanford Ani!ersity 1ress. Ste!ens Jr., 1hillips. 497F. "he &isra le$end and the distortion of historical tradition. 0he !ournal o A rican History 45%2*, 49F)200. Stewart, Mar'orie <. 4990. "he &isra le$end as oral history. 'nternational !ournal o A rican Historical dtudies 4;%4*, F4)70. Strawson, 1. 0. 4974. Identifyin$ reference and truth !alues. In .. .. Steinber$ 6. =. Jakobo!its %eds.*, demantics, 7F)9F. #ambrid$eK #ambrid$e Ani!ersity 1ress. "emple, 2. 4922. Notes on the tribes, provinces, emirates, and states o the northern provinces o Nigeria. #ompiled from official reports by 2. "empleG edited by #. 6. "emple. 6a$osK #MS (ookshop. "hompson, Sandra =. /obert -. 6on$acre. 499F. =d!erbial clauses. In "imothy Shopen %ed.*, Language typology and syntactic description, !ol. 2K #omple3 constructions, 474)2;8. #ambrid$eK #ambrid$e Ani!ersity 1ress. "omlin, /ussell S. 499F. 0ocal attention, !oice, and word orderK an e3perimental, cross) lin$uistic study. In 1amela .ownin$ Michael +oonan %eds.*, "ord order in discourse, F47)FF8. =msterdamK John (en'amins. "omlin, /ussell S., 6inda (. 0orrest, Min$ Min$ 1u Myun$ <ee &im. 4997. .iscourse semantics. In "eun =. !an .i'k %ed.*, eiscourse studies- a multidisciplinary introduction, vol. ,- eiscourse as structure and process, 5;)444. 6ondonK Sa$e. Ayechi, 6inda. 4994. "he functional structure of the +a!a'o third person alternation. &apers rom the 4_th regional meeting o the 5hicago Linguistics dociety, 8;8)885. #hica$oK #hica$o 6in$uistics Society. !an der (er$, <elma. 4999. Gender and person a$reement in =kusha .ar$i. In Gre!ille G. #orbett %ed.*, Agreement .dpecial 'ssue o +olia Linguistica 66142, 4F;)459. (erlinK Mouton de Gruyter. !an .i'k, "eun =. 4994. -pisodes as units of discourse analysis. In .eborah "annen %ed.*, AnalyHing discourse- te*t and talk, 477)49F. Geor$etownK Geor$etown

894

Ani!ersity 1ress. Ean Ealin, /obert .., Jr. 200F. )*ploring the synta*csemantics inter ace. #ambrid$eK #ambrid$e Ani!ersity 1ress. Eendler, Heno. 49F7. Eerbs and times. &hilosophical Review 55, 48;)50, re!ised in Heno Eendler %4957*, Linguistics in &hilosophy, 97,424. #ornellK #ornell Ani!ersity 1ress. Eisser, Marianna . 4999. "he synta3 of the infiniti!e in hosa. douth A rican !ournal o A rican Languages 9%8*, 4F8)49F. Eoelt>, 0. &. -rhard #hrista &ilian)<at>. 2004. Introduction. In 'deophones, 0. &. -rhard Eoelt> #hrista &ilian)<at> %eds.*, 4,9. =msterdamK John (en'amins. lchli, (ernhard. 2007. .escripti!e typolo$y, orG the typolo$ist:s e3tended toolkit. -lectronic ms. I=!ailable onlineK httpKNNlin$.uni)konstan>.deNpa$esNhomeNa20P44NwaelchliNwaelchli)desctyp.pdfJ ald, (en'i. 4979. "he de!elopment of the Swahili ob'ect markerK a study of the interaction of synta3 and discourse. In "almy Gi!n %ed.*, eiscourse and dynta* %Synta3 and Semantics 42*, F0F)F28. +ew BorkK =cademic 1ress. atters, John /. 4990. "he -'a$am noun class systemK -koid (antu re!isited. In 6arry <yman %ed.*, Noun classes in the Grass ields (antu borderland %Southern #alifornia 2ccasional 1apers in 6in$uistics 9*, 99)4;7. 6os =n$elesK Ani!ersity of Southern #alifornia. atters, John /. 2000. Synta3. In (ernd <eine .erek +urse %eds.*, A rican languages- an introduction, 498)2;0. #ambrid$eK #ambrid$e Ani!ersity 1ress. nooqr{Yep^YVYUWVYZrn{{[qf~{Yn^o}VY ^YzqrqY}r{qYYZurnr~Yp[f[||{e^Yq~{VY 0he (antu languages, 22F)2F5. 6ondonK /outled$e. elmers, illiam -. 497;. A rican language structures. (erkeleyK Ani!ersity of #alifornia 1ress. estermann, .iedrich <ermann. M. =. (ryan. 49F2. Languages o "est A rica- &art ''. 6ondonK 23ford Ani!ersity 1ress. illiamson, &ay. 4999. +i$er)#on$o o!er!iew. In John (endor)Samuel %ed.*, 0he Nigerc5ongo languages, 4)8F. 6anhamK Ani!ersity 1ress of =merica. illiamson, &ay /o$er M. (lench. 2000. +i$er)#on$o. In (ernd <eine .erek +urse %eds.*, A rican languages- an introduction, 44)82. #ambrid$eK #ambrid$e

892

Ani!ersity 1ress. olff, <. -kkehard. 499;. Re erenHgrammatik des Hausa %<ambur$er (eitr$e >ur =frikanistik 2*. M@nsterK 6I". oodbury, "ony. 200;. .efinin$ documentary lin$uistics. In 1eter &. =ustin %ed.*, Language eocumentation and eescription, !ol. 4, ;F)F4. 6ondonK <ans /ausin$ -ndan$ered 6an$ua$es 1ro'ect. Bip, Moira. 2002. 0one. #ambrid$eK #ambrid$e Ani!ersity 1ress. Hubin, .a!id &laus)Michael &pcke. 4995. Gender and folk ta3onomyK the inde3ical relation between $rammatical and le3ical cate$ori>ation. In #olette #rai$ %ed.*, +oun classes and cate$ori>ation, 4;9)490. =msterdamK John (en'amins.

89;

Anda mungkin juga menyukai