Anda di halaman 1dari 54

Systemic Functional Grammar: A First Step into the Theory by Christian Matthiessen & M. A. K. Halliday [c. !!

"# $ Matthiessen & Halliday %please do not copy or &uote 'ithout authors( permission) . *nto systemic+,unctional theory o, -rammar . General: %le.ico)-rammar & the study o, -rammar %(-rammatics() This is an introductory account o, a particular theory o, -rammar/ namely systemic+ ,unctional theory. Grammar is one o, the subsystems o, a lan-ua-e0 more speci,ically/ it is the system o, 'ordin-s o, a lan-ua-e. *t is a phenomenon that can be studied/ 1ust li2e li-ht/ physical motion/ the human body/ and decision+ma2in- processes in bureaucracies0 and 1ust as in the case o, these and other phenomena under study/ 'e need theory in order to interpret it. So ,or instance/ the physical phenomenon o, the atom has been interpreted theoretically in terms o, 3emocritus( theory/ 4uther,ord(s theory/ 5ohr(s theory/ and so on. 6e distin-uish bet'een the phenomenon itsel, %the atom) and 7arious theoretical models o, it. 6hat 2ind o, thin- the atom is thou-ht to be 'ill o, course 7ary considerably as 'e mo7e ,rom one theory to another. 3emocritus( atom 'as 7ery di,,erent ,rom 5ohr(s atom/ in that it 'as indi7isible/ not a con,i-uration o, subatomic particles0 that is/ 3emocritus( theory allo'ed us to see much less o, the atom than 5ohr(s theory does. A 'ell+2no'n e.ample o, the 'ay theory determines ho' 'e interpret phenomena is li-ht. 8i-ht can be interpreted either as particle or as 'a7e0 there are t'o alternati7e theories. *n this case/ the alternati7es turn out to be complementary/ in the sense that each re7eals somethin- about li-ht that 'e need to account ,or. This situation is &uite typical in science: 'e need complementary theoretical perspecti7es to account ,or the rich di7ersity o, properties 'e unco7er in the phenomena bein- studied. Grammar as a phenomenon o, study is thus interpreted accordin- to di,,erent theories. So as to maintain the distinction bet'een -rammar and theories o, -rammar/ 'e shall call theory o, -rammar -rammatics. The distinction is analo-ous to that bet'een lan-ua-e and lin-uistics/ or bet'een society and sociolo-y. The di,,iculty is that people o,ten use the same term ,or both the phenomenon and its study: e.-. 'e spea2 o, the 9-rammar o, :n-lish9 %the phenomenon) but also o, 9traditional -rammar9 %one theory o, the phenomenon). 6e could clari,y this situation i, 'e called the second 9traditional -rammatics9. ;ur concern here is thus 'ith systemic+,unctional -rammatics0 and 'e shall illustrate ho' it can be used in the study o, -rammar 'ith e.amples ,rom the -rammars o, Chinese/ :n-lish/ and <apanese. Grammar %as a phenomenon) is part o, lan-ua-e0 it is the 9system o, 'ordin-s9/ as 'e put it abo7e. 5ut ho' it is conceptuali=ed 'ill depend on our -rammatics. *n the history o, thin2inabout lan-ua-e in the 6est/ there ha7e been t'o some'hat di,,erent theoretical perspecti7es. 5oth ha7e their ori-ins in Ancient Greece0 there ha7e been many 7ariations/ but 'e can still trace these t'o strands o, thin2in- today. *n one/ lan-ua-e is a set o, rules rules ,or speci,yin- structures0 so -rammar is a set o, rules ,or speci,yin- -rammatical structures/ such as the construction o, a transiti7e sentence 'ith (7erb > ob1ect(. This perspecti7e is that o, lo-ic and philosophy/ e.-. in the ,ore-roundin- o, the sentence as the basic unit o, lan-ua-e/ or-ani=ed on a lo-ical model into Sub1ect > ?redicate. Since the sentence is the basic unit/ it is studied in isolation. *n the other 7ie'/ lan-ua-e is a resource/ a resource ,or ma2inmeanin-s0 so -rammar is a resource ,or creatin- meanin- by means o, 'ordin-. This perspecti7e is that o, rhetoric and ethno-raphy/ e.-. in the ,ore-roundin- o, te.t %discourse)

as the basic unit o, lan-ua-e/ or-ani=ed accordin- to the rhetorical conte.t. Since te.t is the basic unit/ the sentence is studied in its discourse en7ironment. The 2ind o, -rammatics that is usually presented in school is a diluted 7ersion o, the (-rammar as rule( type o, theory. *t presents rules o, -rammar in terms o, 'ords in sentences/ 'ith 'ords ser7in- ,unctions such as Sub1ect/ ?redicate/ ;b1ect/ and Ad7erbial. As a theory/ it ,alls ,ar short o, the demands that are no' bein- made on -rammatical theories. ;n the one hand/ it ta2es o7er too much ,rom the :uropean lan-ua-es it 'as ,irst applied to/ startin- 'ith Gree2 and 8atin0 hence it is o, limited 7alue in interpretin- the -rammars o, non+:uropean lan-ua-es such as Chinese/ <apanese/ *ndonesian/ Ta-alo-/ Thai/ @ietnamese or the lan-ua-es o, other re-ions and continents. ;n the other hand/ it builds in too little o, the o7erall -rammatical system o, lan-ua-e. *t allo's us to see only a small ,ra-ment o, -rammar and does not pro7ide us 'ith a 'ay o, interpretin- the o7erall or-ani=ation o, the -rammar o, a lan-ua-e as a system o, in,ormation. At this sta-e in history 'e need a richer theory o, -rammar to meet the challen-es o, the a-e o, in,ormation/ e.-. in education %ho' to or-ani=e and -i7e access to 2no'led-e) and in computation %ho' to achie7e the automatic processino, te.t). 6e are also in a position to learn more about -rammar than2s to technical inno7ations: the tape recorder allo's us to store and e.amine spo2en lan-ua-e/ and the computer allo's us to manipulate 7ast amounts o, te.t %spo2en or 'ritten) ,or the purpose o, -rammatical study. Systemic+,unctional theory is one response to these demands. The theory 'as ,irst de7eloped in 'or2 on the -rammar o, Chinese0 and it has been used in educational and computational conte.ts ,rom an early sta-e. Anli2e the theory o, -rammar that is still the recei7ed tradition in school/ systemic+,unctional -rammatics ta2es the resource perspecti7e rather than the rule perspecti7e0 and it is desi-ned to display the o7erall system o, -rammar rather than only ,ra-ments. 6e hope to brin- this out in the discussion 'hich ,ollo's. .B Grammar as resource0 systems & their reali=ation in structure 6e use lan-ua-e to interact 'ith one another/ to construct and maintain our interpersonal relations and the social order that lies behind them0 and in doin- so 'e interpret and represent the 'orld ,or one another and ,or oursel7es. 8an-ua-e is a natural part o, the process o, li7in-0 it is also used to (store( the e.perience built up in the course o, that process/ both personal and collecti7e. *t is %amon- other thin-s) a tool ,or representin- 2no'led-e/ or/ to loo2 at this in terms o, lan-ua-e itsel,/ ,or constructin- meanin-. Grammar is (part o,( this resource. 5ut the relation o, -rammar to other (parts( o, the lin-uistic system is not a part to 'hole relation0 rather/ it is a symbolic one. Grammar is a resource ,or creatin- meanin- in the ,orm o, 'ordin-s. 8et us illustrate this point by re,erence to one broad area o, semantics and -rammar/ an area that 'e shall characteri=e as interpersonal: this is one o, three such -eneral areas/ the other t'o bein- ideational and te.tual. *n interactin- 'ith one another/ 'e enter into a ran-e o, interpersonal relationships/ choosinamon- semantic strate-ies such as ca1olin-/ persuadin-/ enticin-/ re&uestin-/ orderin-/ su--estin-/ assertin-/ insistin-/ doubtin-/ and so on. The -rammar pro7ides us 'ith the basic resource ,or e.pressin- these speech ,unctions/ in the ,orm o, a hi-hly -enerali=ed set o, clause systems re,erred to as M;;3. This is the -rammar as system/ its paradi-matic or-ani=ation.

A system/ in this technical sense/ is a point o, choice. *n the -rammars o, Chinese/ :n-lish/ and <apanese/ the most -eneral choice in mood is that bet'een (indicati7e( and (imperati7e( clauses. These t'o are the options or terms in the system. The ,ollo'in- e.amples illustrate the contrast bet'een (indicati7e( and (imperati7e( in :n-lish: Systemic option %term) (indicati7e( :.ample the spyC*Cyou cameC comesC'ill come in ,rom the cold0 'ho came in ,rom the coldD0 didCdoesC'ill the spyC*Cyou come in ,rom the coldD [Eou# come in ,rom the coldF

(imperati7e(

Any -rammatical choice can be represented as a system 'ith t'o or more alternati7e terms or ,eatures/ as sho'n -raphically in Fi-ure .

Fi-. : A system This -raphic representation sho's %i) the system name %M;;3 TE?:)0 %ii) the terms ,rom 'hich one has to be chosen %(indicati7e(C(imperati7e()0 %iii) the condition under 'hich the choice is a7ailable/ the entry condition %(clause(). The ,ull set o, con7entions ,or the systemic representation is -i7en in the Appendi.. Ho' do 'e 2no' that this system is part o, the -rammar o, :n-lishD There are three parts to the ans'er. %i) *, 'e loo2 at the 'ordin- o, the e.amples -i7en in the table abo7e/ 'e can see that there are systematic di,,erences bet'een the (indicati7e( ones and the (imperati7e( ones. The ,ormer ha7e a Finite 7erb/ 'hereas the latter do not0 and the ,ormer ha7e a Sub1ect/ 'hereas the latter may or may not ha7e one/ it is typically absent. %ii) *, 'e loo2 at the system itsel, to consider 'hat choices are a7ailable ,or (indicati7e( clauses/ 'e ,ind that they ha7e a choice in tense %(pastCpresentC,uture()/ e.pressed throu-h the Finite 7erb0 and also in person/ e.pressed throu-h the Sub1ect. *n contrast/ i, 'e loo2 at the system to consider the choices that are a7ailable ,or (imperati7e( clauses/ 'e ,ind that they ha7e no choice in tense and the Sub1ect can %in principle) only be the addressee/ (you(. %iii) *, 'e loo2 at the distinction in meanin- that the system ma2es/ 'e ,ind that the choice has to do 'ith the nature o, 'hat is bein- ne-otiated in the dialo-ue: either in,ormation % (indicati7e(/ e.-. 3id the spy come in ,rom the coldD Ees/ he did.)/ or -oods+&+ser7ices %(imperati7e(/ e.-. Come in ,rom the coldF ;K.). These three parts to the ans'er illustrate three -eneral an-les o, approach to any system in the -rammar: %i) (,rom belo'(/ %ii) (,rom around(/ and %iii) (,rom abo7e(/ see Fi-ure B. %6e return to this point belo' in Section G.G.) 6e no' e.plore the system ,rom di,,erent an-les/ be-innin- (,rom belo'(/ ,rom the point o, 7ie' o, ho' the systemic contrast is created in the 'ordin-.

Fi-. B: ?erspecti7es on a system %i) Systemic contrasts are created by some aspect o, the 'ordin-: the terms o, the system are di,,erentiated by means o, -rammatical structure %e.-. the absence 7s. presence o, an element o, structure such as Sub1ect)/ by means o, -rammatical or le.ical items %e.-. the -rammatical item 2a in <apanese indicatin- interro-ati7e clauses)/ or/ as a ,urther step/ by means o, a phonolo-ical ,eature %e.-. risin- 7s. ,allin- intonation). 6e say that systemic terms/ or ,eatures/ are reali=ed %e.pressed/ coded) by aspects o, the 'ordin-. The choice in the M;;3 system bet'een (indicati7e( and (imperati7e( is reali=ed structurally: only indicati7e clauses normally ha7e a Sub1ect. 6e can indicate the presence o, the -rammatical ,unction Sub1ect in indicati7e clauses as in Fi-ure G.

Fi-. G: System 'ith associated reali=ation statement The arro' represents the relation o, reali=ation: the ,eature (indicati7e( is reali=ed by the presence o, the ,unction Sub1ect/ stated as >Sub1ect. The di,,erent types o, reali=ation statement are summari=ed in the Appendi.. The presence o, Sub1ect is one step in the speci,ication o, the ,unction structure o, an indicati7e clause/ i.e. o, the or-ani=ation o, the clause as a con,i-uration o, ,unctions. %ii) 6hen 'e come to e.plore (,rom around(/ 'e ,ind that/ throu-h their entry conditions/ a number o, systems come to-ether as an inter+related set/ called a system net'or2. 6e can illustrate a-ain ,rom the -rammar o, M;;3. The choice bet'een (indicati7e( and (imperati7e( is the most -eneral one in this area o, the -rammar0 but each alternati7e leads to ,urther

choices. For instance/ indicati7e clauses are either (declarati7e( %they rode horses) or (interro-ati7e( %did they rode horses0 'ho rode horses)0 declarati7e clauses are either (unta--ed( %e.-./ they rode horses) or (ta--ed( %e.-./ they rode horses/ didn(t they)/ and interro-ati7e clauses are either o, the 'h+ type %e.-./ 'ho rode horsesD) or the yesCno type %e.-. did they ride horsesD). See Fi-ure H.

Fi-. H: Iet'or2 o, M;;3 systems %reali=ation statements in bo.es) *n the dia-ram in Fi-ure H/ the -rammatical resources are represented as a net'or2 o, inter+ connected systems/ each o, 'hich is a choice point. The systems in the net'or2 are ordered ,rom le,t to ri-ht/ startin- 'ith the most -eneral option and mo7in- to'ards more speci,ic ones: i, (clause(/ then (indicati7e( or (imperati7e(0 i, (indicati7e(/ then (interro-ati7e( or (declarati7e(0 i, (declarati7e(/ then (ta--ed( or (unta--ed(0 i, (interro-ati7e(/ then (yesCno( or ('h+(. This is the scale o, delicacy %de-ree o, detail/ speci,icity/ -ranularity). *n the e.ample in Fi-ure H/ each entry condition is a simple ,eature/ (clause(0 but entry conditions can also be comple.es o, ,eatures/ in7ol7in- con1unction andCor dis1unction. Such ,eatures li2e'ise are al'ays terms in other systems. 8et us illustrate dis1unction in an entry condition. Consider a-ain the M;;3 -rammar o, Fi-ure H. *t has one system/ M;;3 TAG/ 'hose entry condition is (declarati7e(. Ho'e7er/ this system is actually not restricted to declarati7e clauses0 it is also open to imperati7e ones %e.-./ [you# saddle the horses/ 'ill you0 let(s saddle the horses/ shall 'e). Conse&uently/ 'e need to be able to state 9i, either (declarati7e( or (imperati7e(/ then (ta--ed(C(unta--ed(9. That is/ 'e need a dis1uncti7e entry condition: see Fi-ure J.

Fi-. J: 3is1uncti7e entry condition The same systemic ,eature or comple. o, ,eatures may occur as the entry condition to more than one system in the system net'or2. *n this case/ the systems are simultaneous. For e.ample/ the primary M;;3 system %M;;3 TE?:) is simultaneous 'ith the system ?;8A4*TE/ the choice bet'een (positi7e( and (ne-ati7e( clauses: see Fi-ure K.

Fi-. K: Simultaneous systems T'o simultaneous strands in a system net'or2 de,ine a t'o+dimensional paradi-m. *t is o,ten use,ul to present e.amples in the ,orm o, a matri. table/ 'ith one system represented by the columns and another by the ro's. Thus M;;3 TE?: and ?;8A4*TE intersect as ,ollo's: M;;3 TE?:: indicati7e ?;8A4*TE: positi7e ne-ati7e The spy came in ,rom the cold. The spy didn(t come in ,rom the cold. imperati7e Come in ,rom the coldF 3on(t come in ,rom the coldF

Such matrices can be used in probin- the accuracy o, a comple. system net'or2: i, it is not possible to ,ind e.amples ,or one or more o, the cells o, a matri./ this means that the system net'or2 predicts a combination o, systemic terms that does not e.ist. 8et us summari=e 'hat 'e ha7e sho'n about the concepts o, system and structure/ and the relation bet'een them. These concepts theori=e the a.es o, or-ani=ation in lan-ua-e/ the paradi-matic and synta-matic. The systemic/ paradi-matic/ a.is is primary in the particular sense that it de,ines the o7erall or-ani=ation o, the -rammar o, a lan-ua-e0 and the structural/ synta-matic/ a.is is secondary in the particular sense that it is speci,ied locally in the en7ironment o, the 7arious terms o, the systemic a.is. Fi-ure " sho's the intersection o, the t'o a.es in the -rammar o, M;;3/ 'ith the systemic a.is pro7idin- the o7erall or-ani=ation. This bi,urcation into the paradi-matic a.is and the synta-matic a.is ma2es it possible ,or the system to relate both to 'hat is abo7e and to 'hat is belo'/ in other 'ords/ both to 'hat the system reali=es and to 'hat it is reali=ed by. %iii) 8oo2in- at the system ,rom abo7e/ 'e are as2in- 'hat it means: in other 'ords/ 'hat semantic ,eatures are bein- reali=ed by this particular set o, options in the -rammar. As already noted/ in the case o, M;;3 the meanin- has to do 'ith the ne-otiation o, speech+ ,unctional roles in dialo-ue: 'ith basic cate-ories such as statement and &uestion %e.chan-e o, in,ormation)/ command and o,,er %e.chan-e o, -oods+&+ser7ices)/ and the comple. net'or2 o, 7ariable and more delicate cate-ories o, 7erbal interaction. 6e shall not pursue the semantic analysis here0 but 'e may note that the resources and methods ,or representinsemantic cate-ories are ,ormally identical 'ith those used in the le.ico-rammar. .G :.ample: M;;3 The -rammar o, a lan-ua-e is a 7ery rich and comple. system0 the -rammatics must brinout that richness and comple.ity/ and not obscure it. This means reco-ni=in- the di,,erent 7ectors alon- 'hich the comple.ity is ordered/ and e.plorin- one step at a time.

Here 'e ha7e introduced only one (corner( o, the -rammar/ and only in the most -eneral terms: the primary systems o, M;;3/ as these are ,ound in :n-lish. 5ecause -rammar is 7ie'ed as a resource rather than as a set o, rules/ it is interpreted in systemic+,unctional theory as a system net'or20 this represents the -rammatical potential a7ailable to the lan-ua-e user. The system net'or2 allo's us to map out the o7erall or-ani=ation o, the -rammar o, a lan-ua-e/ 'ith delicacy as the main principle ,or orderin- the 7arious systems relati7e to one another. Iaturally such net'or2s soon -et 7ery lar-e0 in the systemic -rammars o, :n-lish stored in computers/ there are some'here around LLL systems. 6e ha7e illustrated such a map o, the -rammar o, :n-lish 'ith ,ra-ments ,rom the M;;3 -rammar/ as in Fi-ures H and J. The partial :n-lish M;;3 -rammar 'e ha7e presented is a systemic+,unctional description o, one particular lan-ua-e/ cast in the theoretical terms o, systemic or-ani=ation 'ith associated structural reali=ations. That is/ 'hile the type o, or-ani=ation embodied in the system net'or2 is part o, the theory/ and is a -eneral ,eature common to all lan-ua-es/ the particular systemic ,eatures and structural reali=ations are part o, our descripti7e interpretation o, :n-lish. They are not part o, the -eneral systemic+,unctional theory o, -rammar %see ,urther Section G.G belo').

Fi-. ": The systemic %paradi-matic) and structural %synta-matic) a.es intersectinAs a descripti7e -enerali=ation about interpersonal -rammar/ 'e can assume that all lan-ua-es ha7e a system o, M;;3: i.e. -rammatical resources ,or the interaction bet'een spea2er and addressee/ e.pressin- speech ,unctional selections in dialo-ue. Further/ the semantic cate-ories o, -i7in- in,ormation %statement)/ demandin- in,ormation %&uestion)/ and demandin- ser7ices or -oods %commands) are 7ery li2ely enacted in the -rammars o, all lan-ua-es. Ho'e7er/ the or-ani=ation o, the M;;3 system/ and the reali=ations o, the 7arious options/ di,,er ,rom one lan-ua-e to another. For instance/ the de-ree to 'hich there is a distinct -rammatical cate-ory correspondin- to commands is 7ariable: there may or may not be a distinct ,orm o, the imperati7e/ and e7en 'here there is/ there are usually many other possible reali=ations. Similarly/ 'hile all lan-ua-es probably ha7e a basic opposition bet'een statements and yes+no &uestions %polarity &uestions)/ 'hich it is o,ten %thou-h not uni7ersally) possible to e.press by means o, the distinction bet'een ,allin- and risin-

intonation/ &uestions demandin- a speci,ic element o, in,ormation %other than the 7alue o, the polarity) may be -rouped systemically either 'ith statements or 'ith yes+no &uestions. *t is easy to see 'hy: they are li2e statements in that their polarity is certain/ but at the same time they are li2e yes+no &uestions in that they demand in,ormation. 3i,,erent lan-ua-es or-ani=e their M;;3 -rammars around di,,erent -enerali=ations in this 'ay. Furthermore/ lan-ua-es di,,er considerably 'ith respect to more delicate options/ such as those concerned 'ith ho' interactants position one another in dialo-ue %e.-. by indicatin- e.pected responses) and 'ith ho' they assess the in,ormation bein- e.chan-ed %e.-. by indicatin- de-ree o, probability or source o, e7idence). At the least delicate end o, the -rammar/ Chinese/ :n-lish/ and <apanese ha7e similar M;;3 systems. All three distin-uish (indicati7e( 7s. (imperati7e( clauses/ and 'ithin the ,ormer/ (declarati7e( 7s. (interro-ati7e(/ 'ith one interro-ati7e subtype ,or &ueryin- elements and another ,or &ueryin- polarity. :.amples are tabulated belo': CH*I:S: indic. decl. Tailan- shan- .ue &u :IG8*SH Taro is -oin- to school <A?AI:S: Taroo 'a -a22oo e i2imasu Taroo 'a do2o e i2imasu 2aD Taroo 'a -a22oo e i2imasu 2aD Ga22oo e i2eF

interr-. elemental Tailan- dao nali &uD 6here is Taro -oin-D polar imper. Tailan- shan- .ue &u maD Shan- .ue &uF *s Taro -oin- to schoolD Go to schoolF

5ut 'hile the three M;;3 systems are con-ruent up to the point in delicacy sho'n in the table abo7e/ they also di,,er ,rom one another in more delicate terms. For instance/ in %Mandarin) Chinese/ (polar( interro-ati7es are ,urther di,,erentiated accordin- to the spea2er(s e.pectation re-ardin- the polarity o, the proposition: they are biased %positi7e or ne-ati7e) or unbiased0 e.-. (3o you 'ant itD(/ positi7e bias Ii yao maD/ ne-ati7e bias Ii buyao maD/ unbiased Ii yao buyaoD. :n-lish has only the biased ,orms: positi7e %semantically neutral) did you see himD0 ne-ati7e %semantically/ positi7e bias) didn(t you see himD :n-lish has no unbiased ,orm/ other than the hi-hly mar2ed %peremptory) did you see him or notD The basic M;;3 system 'e ha7e discussed is concerned 'ith %i) the nature o, the commodity bein- e.chan-ed %in,ormation 7s. -oods+&+ser7ices) and %ii) the orientation o, the e.chan-e %-i7in- 7s. demandin-). 5ut there are other aspects o, the e.chan-e that may be -rammaticali=ed in this part o, the -rammar/ in particular aspects o, the tenor o, the relationship bet'een the interactants en-a-in- in the e.chan-e/ i.e. bet'een spea2er and addressee. *n <apanese/ this area is perhaps more hi-hly codi,ied in the -rammar than in either Chinese or :n-lish. For instance/ alon-side the 9plain9 imperati7e %as in HanaseF (Tal2F)/ there are also polite options ,or situations 'here the spea2er is superior to the addressee %as in Hanashi+nasaiF) or in,erior to the addressee %as in Hanashite+2udasaiF). The elaboration o, the -rammar o, <apanese in the areas o, politeness and honori,ication is 'ell+ 2no'n. *t is an important characteristic o, the -rammatical system/ one that ma2es 7ery -ood sense in terms o, the interpersonal meta,unction. At the same time/ 'e ha7e to reco-ni=e that the -rammars o, both Chinese and :n-lish also ha7e created considerable potentials ,or enactin- a 'ide ran-e o, subtly di,,erent tenor relationships. These potentials are perhaps not immediately ob7ious because they rely to a lar-e e.tent on a cryptic ,eature o, the system/

7i=. -rammatical metaphor. Thus alon-side the con-ruent Come in ,rom the coldF/ there are also 7arious metaphorical 7ariants 'here the command is reali=ed not as an imperati7e clause but as i, it 'as a statement or a &uestion. For e.ample: *(d li2e you to come in ,rom the cold0 * 'ant you to come in ,rom the cold0 you shouldCmustC'ill come in ,rom the cold0 6ouldCCould you come in ,rom the cold. Such e.pansions o, the system are o, course characteristic o, <apanese as 'ell. 6hat -enerali=ations can be made about the reali=ation o, systemic options in moodD M;;3 options are typically reali=ed in 7arious 'ays/ includin- intonation %direction o, pitch mo7ement)/ mood particles/ relati7e se&uence o, elements %usually in7ol7in- a ,inite 7erb)/ and special 7erbal cate-ories. *t seems that interpersonal systems in -eneral tend to be reali=ed by some prosodic mode o, e.pression0 and the reali=ations o, M;;3 that 'e ,ind across lan-ua-es can o,ten be sho'n to be prosodic %e.-./ interpersonal mood particles that ser7e as 1uncture prosodies). These particulars are not/ o, course/ part o, the -eneral theory o, -rammar/ they are empirical descripti7e -enerali=ations co7erin- a number o, di,,erent lan-ua-es. And here Chinese/ :n-lish/ and <apanese illustrate nicely a -eneral principle o, crosslin-uistic similarity. 6hile their basic mood systems are con-ruent 'ith one another/ their systemic contrasts are created in di,,erent 'ays/ deployin- some'hat di,,erent subsets o, the reali=ational resources. The basic patterns are tabulated belo' %lea7in- out reali=ation by intonation/ 'hich is used by all three lan-ua-es): M;;3 TE?: :n-lish Chinese <apanese %?redicator: 9imperati7e9 7erb+ ,orm) declarati7e Sub1ectMFinite interro-ati7e imperati7e indicati7e > Mood %Sub1ect/ Finite)

*I3*C. TE?: :n-lish Chinese <apanese *IT. TE?: :n-lish Chinese

> Ie-otiation N 2aMO polar FiniteMSub1ect > Ie-otiation N maMO ;4: > ?redicatorB: ne-ati7e0 ?redicatorM?red.B elemental > 6h0 OM6hMFinite > 96h9

<apanese

> 96h9

As the reali=ational table indicates/ :n-lish di,,ers ,rom Chinese and <apanese in its mood structure. *t has a Mood element/ 'hich consists o, Sub1ect > Finite. This Mood element plays a central role in the reali=ation o, mood options/ in terms o, both its presence and its internal or-ani=ation. *n the unmar2ed cases tabulated abo7e/ it is present in (indicati7e( clauses %e.-. Mood: Eou 'illC6ill you > come in ,rom the cold)/ but not in (imperati7e( ones %e.-. Come in ,rom the cold). Further/ (declarati7e( clauses are distin-uished ,rom (polar( ones by the relati7e orderin- o, Sub1ect > Finite/ Sub1ect M Finite %'ill you) and Finite M Sub1ect %you 'ill)/ respecti7ely. The si-ni,icance o, the Mood element in :n-lish is also sho'n e.-. in ta-s/ 'here the Mood element is pic2ed up at the end o, the clause as the Moodta-/ consistin- o, Ta-,inite M Ta-sub1ect %e.-. Eou 'ill come in ,rom the cold/ 'on(t youD) Ieither Chinese nor <apanese has a distinct Mood element. *t ,ollo's that they do not rely on the se&uence o, Sub1ect > Finite in reali=in- mood options. *n ,act/ neither lan-ua-e has a separate ,unction Finite in the mood structure o, the clause. Chinese has no system o, 7erbal ,initeness at all/ and <apanese does not separate out ,initeness ,rom the rest o, the 7erbal -roup in its clausal structure as :n-lish does. *nstead/ both lan-ua-es deploy mood particles at the end o, the clause ser7in- the ,unction 'e ha7e called Ie-otiation/ since it determines the clause(s ne-otiatory 7alue in dialo-ic interaction. The di,,erence is that <apanese Ie-otiation N 2a is a property o, (interro-ati7e( clauses in -eneral/ 'hereas Chinese Ie-otiation N ma is a property o, (polar( interro-ati7es in particular. %Chinese also has another type o, (polar( interro-ati7e/ 'here the ?redicator is repeated 'ith a ne-ator as in shi bu shi. 6e re,erred to it abo7e 'hen 'e discussed di,,erences in more delicate mood systems.) *n ,act/ these mood particles are part o, more e.tensi7e sets o, interpersonal particles in both lan-ua-es/ includin- ne/ ba in Chinese and ne/ yo in <apanese0 the closest e&ui7alent o, the :n-lish option o, ta--in- a clause is a particle o, this 2ind. The -enerali=ation is that the -rammars o, Chinese and <apanese pro7ide the resource ,or indicatin- ho' the spea2er intends the addressee to ta2e hisCher mo7e in the dialo-ue as sChe is about to (hand o7er( to the addressee. %Such interpersonal particles are common around the lan-ua-es o, the 'orld0 ,or e.ample/ 'e ,ind them clause+initially in Arabic %hael/ a) and in French %est+ce &ue)/ and 'e ,ind them in 7arious %South+):ast Asian lan-ua-es/ e.-. in Korean/ Thai %clause ,inal mP/ rii0 na etc.)/ @ietnamese %clause ,inal phon-/ a/ u/ chu/ di etc.). Such particles may also reali=e options in interpersonal systems ha7in- to do 'ith tenor/ such as the se. o, spea2er and addressee and the status and po'er relations bet'een them.) The ,unction Sub1ect is not re,erred to in the table abo7e in the reali=ation statements o, Chinese and <apanese. 3oes this mean that these lan-ua-es do not ha7e a Sub1ect0 or that their Sub1ects deri7e ,rom di,,erent meta,unctionsD As2in- these &uestions is in ,act not the best 'ay o, e.plorin- the -rammars o, Chinese and <apanese. The cate-ory o, Sub1ect 'as posited in the description o, :n-lish and other lan-ua-es0 and in :n-lish its interpersonal nature is 7ery clear once you be-in to study dialo-ue. Ho'e7er/ this does not mean that 'e should -o loo2in- ,or Sub1ect in Chinese/ <apanese or any other lan-ua-e 'e interpret systemic+,unctionally. 4ather/ 'e should as2 more abstract &uestions that are less li2ely to pre1ud-e the ans'er. Thus 'e as2 ho' the clause in Chinese or <apanese is or-ani=ed as an interacti7e mo7e in a dialo-ue/ as an e.chan-e bet'een spea2er and addressee0 and 'e can -o on to as2 'hether there are elements in the clause that are -i7en some special status in this interaction/ as 'hen an element is -i7en the status o, bein- the point o, in,ormation demanded ,rom the addressee in an (elemental interro-ati7e( clause. *n :n-lish/ the Sub1ect is such an element: it is the element -i7en the status o, modal responsibility0 that is/ it is

responsible ,or the success o, the clause as an interacti7e mo7e. This is perhaps easier to see in (imperati7e( clauses than in (indicati7e( ones0 but it applies to both types. *n an (imperati7e(/ modal responsibility means responsibility ,or complyin- 'ith or re,usin- to comply 'ith the command/ as in: 5eha7e yoursel,F/ 5e politeF0 3on(t be ,ooled by his pleasant demeanourF0 5e -uided by your parentsF * 'illC* 'on(t. As these e.amples indicate/ modal responsibility is &uite distinct ,rom actorhood0 it can e7en be assi-ned to an element in a passi7e clause. *t is also in the en7ironment o, an (imperati7e( clause that 'e are perhaps most li2ely to ,ind a similar type o, status in Chinese and <apanese %and in other lan-ua-es as 'ell): they both -i7e one element in the clause the special interpersonal status o, responsibility ,or complyin- 'ith the command in an (imperati7e( clause. The &uestion is then 'hether there is an indicati7e 7ariant o, this status o, modal responsibility assi-ned to a clausal element in dialo-ue/ or some other status o, special interpersonal si-ni,icance. Since this 'ould re&uire a len-thy e.ploration/ 'e lea7e the issue open. B. :.pandin- the %dimensions o,) le.ico-rammatical space Fi-ure Q locates the M;;3 (corner( o, the -rammar in relation to the other most -eneral systems/ accordin- to the t'o dimensions o, meta,unction and ran2. Meta,unction re,ers to the di,,erent modes o, meanin- construed by the -rammar0 ran2 re,ers to the di,,erent (si=es( o, the -rammatical units %layers o, constituency). 6e shall discuss each o, these concepts in turn %Sections B. and B.B)/ ,ollo'ed by a short account o, the third dimension/ that o, delicacy %Section B.G).

Fi-. Q: The 7ie' o, the -rammar so ,ar/ relati7e to e.pansion by meta,unction and ran2

B. % ) 5y meta,unction: ,rom M;;3 to T4AIS*T*@*TE & TH:M: 6e be-in our mo7e ,rom the M;;3 corner o, the system located at the intersections o, (clause( ran2 and (interpersonal( meta,unction by mo7in- alon- the dimension o, meta,unction. B. . The three meta,unctions 8et us introduce these meta,unctions in t'o steps. The -rammar creates meanin- 'ithin t'o hi-hly -enerali=ed meta,unctions that relate to phenomena outside lan-ua-e: %i) interpersonal and %ii) ideational. %i) The interpersonal meta,unction is concerned 'ith the interaction bet'een spea2er and addressee%s) ++ the -rammatical resources ,or enactin- social roles in -eneral/ and speech roles in particular/ in dialo-ic interaction0 i.e. ,or establishin-/ chan-in-/ and maintainininterpersonal relations. ;ne o, its ma1or -rammatical systems is M;;3/ the -rammaticali=ation o, speech ,unction that 'e ha7e already met. %ii) The ideational meta,unction is concerned 'ith (ideation( ++ the -rammatical resources ,or construin- our e.perience o, the 'orld around us and inside us. ;ne o, its ma1or -rammatical systems is T4AIS*T*@*TE/ the resource ,or construin- our e.perience the ,lu. o, (-oin-s+ on(/ as structural con,i-urations0 each consistin- o, a process/ the participants in7ol7ed in the process/ and circumstances attendant on it. For e.ample: [8ocation:# in the open -lade [Actor:# the 'ild rabbits [?rocess:# danced [Accompaniment:# 'ith their shado's. These t'o meta,unctions orient to'ards t'o (e.tra+lin-uistic( phenomena/ the social 'orld and the natural 'orld0 'e construe the natural 'orld in the ideational mode and to enact the social 'orld in the interpersonal mode. For instance/ 'e can construe a picture o, 'hat can participate in an action %ideational) and 'e can enact 'ho -i7es orders to 'hom %interpersonal). *n addition/ there is a third meta,unction/ intrinsic to lan-ua-e %that is/ orientin- to'ards the phenomena created by lan-ua-e itsel,/ 7i=. meanin-s)/ the te.tual meta,unction. %iii) The te.tual meta,unction is concerned 'ith the creation o, te.t/ 'ith the presentation o, ideational and interpersonal meanin-s as in,ormation that can be shared by spea2er and listener in te.t un,oldin- in conte.t. ;ne o, the ma1or te.tual systems is TH:M:/ the resource ,or settin- up a local conte.t ,or a clause by selectin- a local point o, departure in the ,lo' o, in,ormation %or perhaps rather (s'ell o, in,ormation(/ since it is not a uni,orm ,lo'). Thus the spatial 8ocation is -i7en thematic status in the e.ample analysed ,or T4AIS*T*@*TE abo7e: [Theme:# in the open -lade [4heme:# the 'ild rabbits danced 'ith their shado's. The role o, the te.tual meta,unction is an enablin- one. *t ser7es to enable the presentation o, ideational and interpersonal meanin- as in,ormation that can be shared: it pro7ides the spea2er 'ith strate-ies ,or -uidin- the listener in hisCher interpretation o, the te.t. As Fi-ure Q su--ests/ the three meta,unctions are simultaneous0 this simultaneity applies to both a.es o, or-ani=ation/ the systemic and the structural.

%i) Systemically/ this means that M;;3 %interpersonal)/ T4AIS*T*@*TE %ideational)/ and TH:M: %te.tual) are simultaneous strands 'ithin the system net'or2 o, the clause: see Fi-ure !. That is/ the meta,unctions are mani,ested as clusterin-s in the o7erall system net'or2 o, the clause %and other -rammatical units). The ,i-ure sho's a ,ra-ment o, the :n-lish net'or20 similar simultaneous strands are ,ound in Chinese and <apanese/ althou-h/ as 'e shall see belo'/ M;;3 and TH:M: relate in some'hat di,,erent 'ays in the three lan-ua-es/ and the operation o, the system o, @;*C: in mappin- structural ,unctions ,rom the di,,erent meta,unctions onto one another is also 7aried. Around the lan-ua-es o, the 'orld/ 'e can e.pect considerable 7ariation in these systems 'hich relate the di,,erent meta,unctions to one another. There are lan-ua-es 'hich ha7e no e&ui7alent o, the @;*C: system 'e ,ind in e.-. Chinese/ :n-lish/ and <apanese0 and 'here lan-ua-es ha7e both @;*C: and TH:M:/ 'e ,ind 7ariation in the di7ision o, labour bet'een them/ in particular in ho' the choice o, an unmar2ed thematic status is achie7ed. Further/ systems deri7in- ,rom the di,,erent meta,unctions may also be distributed alon- the ran2 scale %constituency hierarchy0 see belo') in di,,erent 'ays/ particularly across the ran2s o, clause/ 7erbal -roup and 7erb.

Fi-. !: Meta,unctions as mani,ested in the system net'or2 o, the clause %:n-lish) %ii) Structurally/ the meta,unctional simultaneity is mani,ested as three simultaneous strands or layers in the structure o, the clause: see Fi-ure L/ 'hich sho's the three meta,unctional perspecti7es on our earlier e.ample. The structural ,unctions ,rom the di,,erent meta,unctional strands are con,lated 'ith one another0 ,or e.ample/ Sub1ect is con,lated 'ith Actor %represented as Sub1ectCActor0 see the Appendi.). The e.ample o, structural simultaneity is ,rom :n-lish. Structures are also meta,unctionally layered in this 'ay in Chinese and <apanese/ but the or-ani=ation 'ithin each strand may be di,,erent ,rom 'hat 'e ,ind in :n-lish. 6e re,erred abo7e to the di,,erences in the interpersonal layer/ 'hich is probably 'here the main structural di,,erences lie. Fi-ure presents an e.ample ,rom <apanese o, an (elemental interro-ati7e( clause. There is no Mood element and 'e ha7e not

posited a Sub1ect ,unction/ but the clause ends 'ith the Ie-otiation ,unction 'here its ne-otiatory or interactional contribution is reali=ed. Ie-otiation is preceded by ?redicator/ the interpersonal perspecti7e on the 7erbal -roup ser7in- in the clause: the ?redicator carries assessments o, mood and polarity0 and it also carries de-rees o, (politeness( and (,ormality( %such as the di,,erence bet'een desu and the plain ,orm da). The 6h element is in the position it 'ould ha7e in an unmar2ed declarati7e clause0 6h and Theme are not con,lated (by de,ault(/ as they are in :n-lish. Chinese is li2e <apanese in this respect. The t'o are also similar in that ,unctions that are reco7erable ,rom the te.t or the conte.t ,or the addressee may be le,t implicit0 ,or e.ample/ a Theme that is continuous 'ith precedin- Themes is li2ely to be le,t implicit. This also means/ o, course/ that any structural ,unctions ,rom the other meta,unctional layers 'hich are con,lated 'ith it are also le,t implicit. Around the lan-ua-es o, the 'orld/ 'e can e.pect to ,ind considerably more 7ariation in the 'ay the three meta,unctional contributions to structure are mapped onto one another. The main 7ariable here is most probably ran2. 8an-ua-es di,,er in the 'ay that the reali=ational domains o, TH:M:/ M;;3/ T4AIS*T*@*TE/ and related systems are distributed across ran2s. For e.ample/ many lan-ua-es do much more 'or2 in the 7erb or 7erbal -roup than lan-ua-es such as Chinese and :n-lish. meta,unction: system: te.tual interpersonal TH:M: M;;3 in the open -lade Theme Ad1unct 4esidue ideational T4AIS*T*@*TE 8ocation the 'ild rabbits 4heme Sub1ect Mood Actor ?rocess Finite ?redicator 4esidue B Accompaniment Ad1unct danced 'ith their shado's.

Fi-. L: The simultaneous meta,unctions in the structure o, the clause %:n-lish) meta,unction: system: te.tual interpersonal ideational Fi-. TH:M: M;;3 T4AIS*T*@*TE Carrier Kore 'a Theme 96h9 Attribute ?redicator ?rocess Ie-otiation nan desu 2a

: The simultaneous meta,unctions in the structure o, the clause %<apanese)

B. .B *deational %at clause ran2): T4AIS*T*@*TE The ideational meta,unction en-enders resources ,or construin- our e.perience o, the 'orld around us and inside us0 the ideational system at clause ran2 is T4AIS*T*@*TE. T4AIS*T*@*TE is concerned 'ith construin- one particular domain o, our e.perience/ our e.perience the ,lu. o, (-oin-s+on(/ as con,i-urations o, a process %o, some -eneral type: material/ mental/ relational)/ the participants in7ol7ed in it %Actor/ Goal0 Senser/ ?henomenon0 Carrier/ Attribute0 and so on)/ and the circumstances attendant on it %Cause/ 8ocation/ Manner %includin- means and instrument)/ Accompaniment/ and so on).

There are t'o types o, 7ariable in systems o, transiti7ity: %i) The type o, process. %ii) The type o, participation in process. %i) The type o, process is represented in the system net'or2 in Fi-ure !. The transiti7ity system o, a lan-ua-e 'ill construe e.perience into a small set o, domains o, meanin- 'hich di,,er accordin- to the process itsel, and the nature o, the participants in7ol7ed in it. *n :n-lish/ the primary options in ?4;C:SS TE?: are (materialCmentalC7erbalCrelational(: ,or a more detailed account o, the transiti7ity system/ see Fi-ure B. This system is moti7ated by criteria %i) (,rom abo7e(/ %ii) (,rom belo'(/ and %iii) (,rom around(. Some o, these criteria are set out and illustrated in the ,ollo'in- table: ?4;C:SS %i) From TE?: abo7e cate-ory meaninmaterial doin- & Actor happeninthe company sensin?rocess Goal is -i7in4ecipient present+ in+ present %ii) From belo' structural reali=ation ?ro1ection Tense %iii) From around

a ne' teapot to my aunt

mental

Senser: ?rocess ?henomenon conscious my aunt my aunt 'ants 'ants a ne' teapot >pro1ection present them to buy a ne' teapot 4ecei7er

7erbal

sayin-

Sayer: symbol source

?rocess @erbia-e

the says company(s letter the says company(s letter relational bein- & ha7inCarrier

2ind thin-s

to my aunt to my aunt

>pro1ection present

that she is entitled to a ne' teapot

?rocess Attribute beauti,ul present

this teapot is *denti,ied

?rocess *denti,ied

this

is

the teapot the company -a7e my aunt

present

These are di,,erent clause types0 a number o, 7erbs can ser7e in more than one type/ in di,,erent senses. For e.ample/ the 7erb ma2e can ser7e in a material clause in the sense o, (produce( and in a relational clause in the sense o, (be( %or (cause to be(). Thus it made a -ood drin2 is ambi-uous bet'een material (it %e.-. the appliance) produced a -ood drin2( and relational (it %e.-. the mi.ture) 'as a -ood drin2(. Such ambi-uous instances can al'ays be probed (,rom abo7e(/ (,rom belo'( and (,rom around(. 8et us ta2e (material( and (mental( in the system o, ?4;C:SS TE?: as illustrations o, the three perspecti7es that moti7ate this system.

% ) ?4;C:SS TE?:: (material(. %i) 8oo2ed at (,rom abo7e(/ a material clause construes doin-s & happenin-s/ includin- actions/ acti7ities/ and e7ents0 con,i-urations o, a process and participants in7ol7ed that re&uire some input o, ener-y to occur and 'here one participants is li2ely to under-o a chan-e. %ii) 8oo2ed at (,rom belo'(/ a material clause is characteri=ed by particular structural con,i-urations/ such as ?rocess > Actor > Goal %> 4ecipient)/ and ?rocess > 4an-e. There is al'ays an Actor/ 'hich can be reali=ed by a nominal -roup representin- any (thin-( or e7en a non+,inite clause representin- a (macro+ thin-( %as in the boy 'ith -reen hair bro2e the 'indo'/ and the earth mo7in- bro2e the 'indo' respecti7ely)/ but not by a (meta+thin-( %a ,act/ that the earth mo7ed bro2e the 'indo' is not possible). Further options determine 'hether the process is (directed(/ in 'hich case there is a Goal as 'ell %[Actor:# the policeman [?rocess:# hunted [Goal:# the

demonstrator)/ or not %[Actor:# the policeman [?rocess:# ran). *, the process is directed/ it may be (bene,acti7e(0 and it i, is/ there may be a 4ecipient %[Actor:# the 1ud-e [?rocess:# -a7e [4ecipient:# the demonstrator [Goal:# a le-al document). %iii) 8oo2ed at (,rom around(/ a material clause is the entry condition to a number o, ,urther systems0 'e ha7e already re,erred to directedness and bene,action as t'o e.amples. *t does not lead to a system o, ?4;<:CT*;I %a system 'ith an option o, reportin- or &uotin- speech or thou-ht/ 'hich 'e ,ind 'ith 7erbal and mental clauses/ as in The paper said 9The buildin- collapsed9)0 it is thus not possible to say the earth mo7ed: 9The buildin- collapsed9: there can be a causal relation bet'een these t'o clauses %the earth mo7ed so the buildin- collapsed)/ but not a semiotic one 'here the clause the earth mo7ed 'ould pro1ect the clause 9The buildin- collapsed9 onto a more abstract plane as its content. *, 'e e.plore around ?4;C:SS TE?: but outside the T4AIS*T*@*TE systems themsel7es/ 'e ,ind that in reports o, present time/ there is an unmar2ed correlation 'ith di,,erent T:IS: selections ,or the di,,erent process types. *n material clauses/ the unmar2ed tense is the present+in+present rather than the simple present/ as in The cat(s 'a7in- its tail rather than The cat 'a7es its tail. %The simple present is used to construe a di,,erent time ,rame/ such as -eneric or habitual time/ as in The cat 'a7es its tail 'hene7er it(s uncertain.) This systemic association bet'een ?4;C:SS TE?: and T:IS: is semantically moti7ated: processes are phenomena that un,old in time and hence ha7e a tense system0 but di,,erent process types ha7e di,,erent temporal pro,iles and hence di,,erent unmar2ed present tense selections. %B) ?4;C:SS TE?:: (mental(. %i) 8oo2ed at (,rom abo7e(/ a mental clause construes sensin-/ perception/ co-nition/ intention/ and emotion0 con,i-urations o, a process o, consciousness in7ol7in- a participant endo'ed 'ith consciousness and typically a participant enterin- into or created by that consciousness. %ii) 8oo2ed at (,rom belo'(/ a mental clause is characteri=ed by a particular structural con,i-uration/ ?rocess > Senser > ?henomenon. There is al'ays a Senser/ 'hich is reali=ed by a nominal -roup denotin- a bein- endo'ed 'ith consciousness %e.-. she in she sa' them crossin- the road). *t is thus much more constrained than the Actor0 in ,act/ it is the most constrained o, all the participants in any o, the process types. *n contrast/ the ?henomenon can be not only any 2ind o, thin- or macro+thin-/ but also a meta+ thin- %as in she sa' them/ she sa' them crossin- the road/ she sa' [the e7idence# that they had crossed the road). %iii) 8oo2ed at (,rom around(/ a mental clause leads to a system o, ?4;<:CT*;I. A mental clause can pro1ect the content o, consciousness/ (thou-ht( or (ideas(/ as another/ separate clause %as in He thou-ht R the moon 'as a balloon). Such a clause is not a participant 'ithin the mental clause0 ,or e.ample/ it cannot ser7e as the Sub1ect in a passi7e 7ariant %'e do not -et That the moon 'as a balloon 'as thou-ht by him). Further/ unli2e a material clause/ a mental clause does not lead to a bene,acti7e option %there is no He thou-ht me R the moon 'as a balloon0 e.amples such as He thou-ht to himsel, R 9The moon is a balloon9 are not prototypical/ but are (mental as i, 7erbal(/ inner speech). 6ith respect to T:IS:/ the unmar2ed selection ,or present time is the simple present rather than the present+ in+present %,or e.ample/ He thin2s the moon is a balloon rather than He is thin2in- that the moon is a balloon). T4AIS*T*@*TE/ then/ o,,ers a net'or2 o, inter+related options ,or representin- di,,erent types o, e.perience/ our e.perience o, the material 'orld/ o, the 'orld o, our inner consciousness/ o, the 'orld o, symboli=ation/ and so on. The criteria ,rom abo7e/ ,rom belo'/ and ,rom around 'hich 'e ha7e illustrated to-ether moti7ate the ?4;C:SS TE?: system in the -rammar o, transiti7ity. That is/ in our description o, this area o, the -rammar/ these types yield the most po'er,ul -enerali=ations. 5ut their di,,erences in the o7erall system are not immediately ob7ious. There are no o7ert mar2ers di,,erentiatin- the process

types0 ,or e.ample/ there are no transiti7ity particles at the end o, the clause reali=in- the selection in process type %as 'e illustrated ,or M;;3 in Section .G abo7e)/ and there are no di,,erences in 7erbal morpholo-y. The process types are co7ert systemic types in the transiti7ity system/ in many cases/ cryptotypes in 6hor,(s terminolo-y. 6e reco-ni=e that they are (in the system( e.plorin- them ,rom the three perspecti7es 'e ha7e illustrated. 6hen 'e e.plore them in this 'ay/ 'e see ho' the o7erall system is (a,,ected( by their presence/ ho' it (reacts( to their presence. For e.ample/ 'e ,ind that the T:IS: system (reacts( to the distinction bet'een the material and non+material process types. 6hor, called such properties reactances. 6e ha7e e.empli,ied some reactances to ?4;C:SS TE?: such as T:IS: and ?4;<:CT*;I. ;thers include classes o, 7erb that can ser7e as the ?rocess in clauses o, the di,,erent process types/ and a set o, reactances outside the ideational meta,unction. For e.ample/ the te.tual meta,unction includes the option o, substitution 'hereby one piece o, 'ordin- is substituted ,or by a particular substitute ,orm %such as nominal one and 7erbal do in :n-lish) to present that in,ormation as continuous %in the en7ironment o, contrast/ as in 6hich to'el 'ould you li2eD The red one/ please). The 7erbal substitute do %toC'ith) can only be used in material clauses/ not in mental/ 7erbal or relational ones. Thus 'e can -et (material( 6hat the company did 'ith the teapot 'as -i7e it to my aunt/ but not (mental( 6hat my aunt did 'ith the teapot 'as 'ant it/ (7erbal( 6hat my aunt did 'ith the story 'as tell it/ and (relational( 6hat my aunt did 'ith the director 'as be her. 4eactants are o,ten outside the meta,unctional domain o, the system they (react to(/ and e7en 'hen they ,all 'ithin the same meta,unction/ they can be a considerable distance a'ay ,rom the system they (react to(. Chinese and <apanese seem to ha7e the same primary ?4;C:SS TE?: system as 'e ha7e 1ust illustrated in :n-lish. They di,,er in the 2inds o, reactance that pro7ide e7idence ,or the di,,erent process types. For e.ample/ the temporal issues are di,,erent ,or :n-lish and Chinese since :n-lish construes time in the process on a tense model 'hereas Chinese construes it on an aspect model. They also di,,er in more delicate process types. For e.ample/ both Chinese and <apanese brin- possession and e.istence closer to-ether than :n-lish does. *n <apanese/ material/ mental/ 7erbal and relational clauses di,,er ,or e.ample 'ith respect to patterns o, postpositional mar2in-/ options in 7oice and the resultati7e construction/ and pro1ection. There is al'ays one participant mar2ed by the postposition -a %or 'a i, it is thematic)0 and there may be one or t'o more participants mar2ed by the postpositions o %or 'a i, it is thematic) or ni %or ni 'a i, it is thematic) or le,t 'ithout a postposition i, the clause is an unmar2ed relational one. Thus in an acti7e material clause the Actor is mar2ed by -a/ the Goal by o and the 5ene,iciary by ni: Sensei -a (teacher( Actor 'ata2ushi ni (*( 5ene,iciary hon o (boo2( Goal 2udasaimashita (-i7e( ?rocess

(The teacher -a7e me a boo2( 'hereas the Attribute or @alue o, an unmar2ed relational clause is 'ithout a postposition: 6atashi 'a * sensei (teacher( desu (be(

Carrier (* am a teacher(

Attribute

?rocess

?4;<:CT*;I is an option ,or mental and 7erbal clauses. They can pro1ect a clause as the (content( o, the mental or 7erbal processin- and the pro1ected status o, this clause is mar2ed by to/ 2a or the li2e. For instance/ the ,ollo'in- e.ample is a combination o, t'o clauses/ a pro1ectin- mental one o, thin2in- and another one representin- the idea pro1ected by thin2in-: 6atashi 'a (*( S [mental# T [pro1ected idea# Senser Manner+ means ?rocess Ie-otiation (* thin2 * 'ill -o by bus( @erbal clauses are similar 'ith respect to pro1ection0 but in addition/ they can ha7e a 4ecei7er %mar2ed by ni) representin- the addressee o, a mo7e in dialo-ue. Around the lan-ua-es o, the 'orld/ 'e can e.pect more 7ariation in process type than is e7ident 'ith Chinese/ :n-lish or <apanese. *t seems plausible that 'e 'ill ,ind prototypical material/ mental/ and relational process types in the transiti7ity systems o, most lan-ua-es/ but there 'ill be considerable 7ariation in ho' they construe more intermediate cate-ories %such as beha7ioural and e.istential) precisely because o, their more indeterminate status. The central reason ,or reco-ni=in- 7erbal processes as a distinct type in e.-. Chinese/ :n-lish and <apanese is their ability to pro1ect &uoted or reported clauses. Ho'e7er/ some lan-ua-es such as Ta-alo- may use a relational strate-y %c,. :n-lish his statement 'as that the moon is a balloon) 'hile others may enact pro1ection interpersonally as 9mood pro1ection9 'ith a special &uotati7e mood. And 1ust as 'e ,ind considerable 7ariation in the reali=ation o, mood types across lan-ua-es %as illustrated abo7e)/ 'e 'ill ,ind considerable 7ariation in the o7ert mar2ers and co7ert reactances o, process types. The criteria 'e tabulated abo7e are thus speci,ic to :n-lish %as 'e already noted 'ith respect to unmar2ed present tense selection). %ii) The second ma1or 7ariable is the mode o, participation in the process ++ ho' participants a,,ect one another throu-h their in7ol7ement in a process. The interpretation in traditional -rammar is in terms o, o, the concept o, transiti7e deri7ed mainly ,rom material clauses. 6e ha7e in ,act already alluded to it: it is concerned 'ith 'hether the Actor impacts another participant %the Goal) throu-h the process/ transiti7e/ or not/ intransiti7e. Since this model is oriented to'ards one type o, clause/ it leads to an interpretation in 'hich process types are ha7e to be di,,erentiated. That is/ 'hile the material model operates 'ith an Actor potentially impactin- a Goal/ once the description is broadened/ other process types ha7e to be reco-ni=ed: these are mental/ 7erbal/ and relational %in e.-. Chinese/ :n-lish/ French/ <apanese and Ta-alo-). basu de (bus( i2oo (-o( to omou (thin2( ?rocess

5ut there is an alternati7e to the type o, model that 'as reco-ni=ed in traditional -rammar ++ the er-ati7e model o, transiti7ity. This represents a process not in terms o, impact but in terms o, causation. There is al'ays %in all process types) one participant that is most closely associated 'ith the process/ the Medium %since it is the medium throu-h 'hich the process is mani,ested)0 and the basic option is 'hether to represent the combination o, Medium > ?rocess as bein- e.ternally caused by an A-ent or not. So the combination (door > open(/ can be represented as %say) [Medium:# the door [?rocess:# opened/ 'ithout speci,yin- 'hat brou-ht the occurrence about/ or as [A-ent:# the 'ind [?rocess:# opened [Medium:# the door/ 'ith a speci,ication o, the A-ent brin-in- about the occurrence. A clause 'ith ?rocess > Medium 'ithout the A-ent is 2no'n as (middle(/ and a clause 'ith an A-ent %e.plicit or implicit) is 2no'n as (e,,ecti7e(. These t'o transiti7ity models do not represent mutually e.clusi7e sets o, phenomena/ but rather complementary perspecti7es on the same set o, phenomena. *n any -i7en lan-ua-e/ some areas 'ill display more ,eatures o, the transiti7e and others more ,eatures o, the er-ati7e. The balance bet'een them is clearly a ma1or point o, 7ariation in transiti7ity systems around the 'orld. :n-lish is a typically mi.ed system. The er-ati7e pattern/ 'ith the contrast bet'een (middle( and (e,,ecti7e(/ is ,ound 'ith all the process types e.cept ,or 7erbal processes/ 'hich are (middle( only(0 ,or e.ample: ?4;C:SS TE?: AG:ICE middle %Medium > ?rocess) material mental 7erbal relational The door opened She li2ed the ne' musical She told them a story He 'as mad He 'as Henry She made him mad She called him Henry e,,ecti7e %> A-ent) The 'ind opened the door The ne' musical pleased her

6e can no' see that the e.istence o, pairs o, mental clauses such as she li2ed the ne' musical : the ne' musical pleased her can be accounted ,or by re,erence to the system (middle( 7s. (e,,ecti7e( o, the er-ati7e model. They mani,est the same er-ati7e pattern as 'e ,ind in material and relational clauses. They di,,er in one respect. 5oth middle and e,,ecti7e mental clauses ha7e the same set o, participants/ Senser > ?henomenon0 and the di,,erence lies in the assi-nment o, er-ati7e roles: in the e,,ecti7e the ?henomenon is construed as an A-ent brin-in- about the Senser(s sensin- %emotion o, pleasure in our e.ample)/ 'hereas in the middle it is construed non+a-enti7ely. Middle and e,,ecti7e thus constitute t'o complementary perspecti7es on mental processes0 they can be seen ,rom t'o di,,erent an-les/ either as the Senser en-a-in- in sensin- 'hich ran-es o7er %or creates) a ?henomenon/ or as the ?henomenon brin-in- about sensin- 'hich impin-es on a Senser. Around the lan-ua-es o, the 'orld/ the de-ree to 'hich one o, the t'o models o, transiti7ity dominates may be di,,erent/ and 'e can see this 7ariation in the increasin- ,ore-roundin- o, the er-ati7e model in the history o, :n-lish. Chinese and :n-lish are 7ery similar in the balance bet'een the models in their transiti7ity systems0 but Chinese does not ha7e a

systematic contrast bet'een (middle( and (e,,ecti7e( mental clauses/ these e.ist only in the middle type. <apanese and :n-lish also appear to be 7ery similar in the balance o, the transiti7e and the er-ati7e. There is 7ariation across transiti7ity systems beyond 'hat 'e ha7e su--ested so ,ar. ;n the one hand/ there may be yet other transiti7ity models. *n his interpretation o, the transiti7ity o, Ta-alo-/ Martin %to appear) identi,ies a transiti7e pattern 'here di,,erent process types are distin-uished/ and a complementary one 'hich construes a clause nucleus consistin- o, the ?rocess and one participant/ the Medium/ throu-h 'hich it is actuali=ed. There may or may not be another participant0 i, there is/ it is either dra'n into the clause nucleus or repelled by it. The common theme seems to be that transiti7ity systems embody a complementarity bet'een t'o perspecti7es on e.perience: one in 'hich happenin-s are distin-uished into di,,erent types/ the other in 'hich they are treated as all ali2e. B. .G Te.tual %at clause ran2): TH:M: The te.tual meta,unction en-enders resources ,or presentin- interpersonal and ideational meanin-s as in,ormation or-ani=ed into te.t that can be on-oin-ly e.chan-ed bet'een spea2er and listener. This in7ol7es transitions in the de7elopment o, te.t %con1uncti7e relations) and the assi-nment o, di,,erent te.tual statuses %thematicity/ ne's'orthiness/ continuity and contrast/ reco7erability). These transitions and statuses enable the e.chan-e o, in,ormation0 the spea2er is -uidin- the listener in interpretin- the un,oldin- te.t. At clause ran2 the ma1or te.tual system is TH:M:. TH:M: is a resource ,or or-ani=in- the interpersonal and ideational meanin-s o, each clause in the ,orm o, a messa-e. :ach clause 'ill occur at some particular point in the un,oldin- o, the te.t0 this is its te.tual en7ironment. The system o, TH:M: sets up a local en7ironment/ pro7idin- a point o, departure by re,erence to 'hich the listener interprets the messa-e. 6ith this system the spea2er speci,ies the place in the listener(s net'or2 o, meanin-s 'here the messa-e is to be incorporated as rele7ant. The local en7ironment/ ser7in- as point o, departure/ is the Theme0 'hat is presented in this local en7ironment is the 4heme. The clause as a messa-e is thus a con,i-uration o, t'o thematic statuses/ Theme > 4heme. *n :n-lish/ thematic status is e.pressed by position in se&uence. Theme is reali=ed by initial position and 4heme is reali=ed by non+initial position: e.-. [Theme:# *n "! [4heme:# <ohn Macarthur arri7ed in Sydney. There are a number o, thematic options/ includin- %i) the choice bet'een (mar2ed theme( %as in the e.ample abo7e) and (unmar2ed theme( %the Sub1ect in a declarati7e clause: [Theme:# <ohn Macarthur [4heme:# arri7ed in Sydney in "! .)0 %ii) the option o, theme predication %e.-./ *t 'as <ohn Macarthur 'ho arri7ed in Sydney in "! )/ typically to identi,y a particular theme out o, a%n implicit) set o, potential candidates0 and %iii) the option o, theme identi,ication/ ,ore-roundin- some part o, the messa-e by means o, nominali=ation %e.-./ 'hat <ohn Macarthur did in "! 'as arri7e in Sydney). Since TH:M: is a te.tual resource/ it relates the clause to the o7erall de7elopment o, te.t in conte.t in particular: here the te.t in &uestion 'as bein- de7eloped as a chronolo-ical se&uence. 6e ha7e 1ust presented the or-ani=ation o, the te.tual or-ani=ation o, the clause as a con,i-uration o, t'o discrete constituents/ Theme M 4heme. This ma2es it possible to sho' ho' they map onto ,unctional elements 'ithin the other meta,unctional strands o, the clause. Ho'e7er/ all te.tual statuses are really de-rees o, prominence0 'hat 'e ha7e here is a cline/ a -radual mo7e ,rom thematic prominence to non+prominence. 6e can thus construe the clause

as a ('a7e( in the ,lo' o, in,ormation/ startin- 'ith a thematic pea2 and mo7in- into a thematic trou-h. Such 'a7e+li2e or periodic or-ani=ation is the mode o, e.pression en-endered by the te.tual meta,unction: see ,urther Section G. %ii) belo'. Thematicity is one o, a set o, te.tual statuses or 2inds o, prominence. The clause also displays a complementary 2ind o, prominence/ de-ree o, ne's'orthiness. This is a cline ,rom -i7en in,ormation to ne' in,ormation/ represented as a con,i-uration o, Gi7en > Ie'. ?rominence as ne's is reali=ed by intonational prominence: 'hile the mo7ement o, pitch in a tone -roup %intonation unit) is a continuous contour/ there 'ill be some ma1or mo7ement/ e.-. a ma1or rise or a ma1or ,all0 and this ma1or mo7ement is prominent a-ainst the bac2-round o, the mo7ement o7erall. Clause and tone -roup are not necessarily co+e.tensi7e0 one clause may be reali=ed by more than one tone -roup/ and one tone -roup may reali=e more than one clause. This in ,act re7eals the e.istence o, another -rammatical unit alon-side the clause/ the in,ormation unit. This unit is reali=ed by the tone -roup0 and it is the domain o, the system o, *IF;4MAT*;I F;CAS/ reali=ed by Gi7en > Ie'. *n the unmar2ed case/ a clause is co+e.tensi7e 'ith an in,ormation unit/ so that Theme > 4heme and Gi7en > Ie' complement one another 'ithin the domain o, a sin-le clause. 6hile Theme is reali=ed se&uentially/ Ie' is not0 it is reali=ed intonationally. Conse&uently/ thematicity and ne's'orthiness are independent 7ariables. *n the unmar2ed case/ the Ie' is mapped onto the last element 'ithin the 4heme that has a le.ical content. Conse&uently/ the unmar2ed messa-e is a combination o, t'o te.tual 'a7es: Theme shadin- into 4heme and Gi7en shadin- into Ie'/ 'ith Theme ,allin- 'ithin Gi7en and Ie' ,allin- 'ithin 4heme. See Fi-ure G ,or an e.ample %assumin- a moment in the in,ormation ,lo' correspondin- to ('here did <ohn Macarthur -o in "! D(). *n "! TH:M: *IF;4MAT*;I F;CAS Theme Gi7en <ohn Macarthur arri7ed 4heme R U++++ Ie' in Sydney

Fi-. G: Theme > 4heme and Gi7en > Ie' in unmar2ed combination *n the unmar2ed case/ the clause thus un,olds ,rom prominence as Theme to prominence as Ie'. From the listener(s point o, 7ie'/ sChe is thus -i7en -uidance as to 'here to inte-rate the messa-e 'ith hisCher interpretation o, the te.t so ,ar and 'hat to =oom in on as the main point. This s'ell o, in,ormation 'ithin the clause also tends to correlate 'ith a mo7e 'ithin the elements o, the clause ,rom elements 'hose re,erents are presented as speci,ic and so reco7erable to the listener to elements presented as non+speci,ic and so non+reco7erable to the listener. 5ut/ in :n-lish/ choices in 4:F:4:IC: 7ary independently o, thematicity and ne's'orthiness/ so the pattern 1ust described is only a tendency. The distribution o, in,ormation in the clause 'e ha7e 1ust discussed also helps e.plain the te.tual aspect o, @;*C:/ (acti7eCpassi7e(/ in :n-lish. From a te.tual point o, 7ie'/ this system pro7ides alternati7e options ,or participants as unmar2ed Theme and unmar2ed Ie'. Thus ,or e.ample in an (e,,ecti7e( clause %see Fi-ure B abo7e)/ the acti7e 7ersion 'ill ha7e A-entCSub1ect as unmar2ed Theme and Medium as a candidate ,or unmar2ed Ie'/ 'hereas

the passi7e 7ersion 'ill ha7e MediumCSub1ect as unmar2ed Theme and A-ent %i, present) as a candidate ,or unmar2ed Ie'. Thus @;*C: has to be understood in relation to the te.tual meta,unction. All lan-ua-es 'ill ha7e te.tual resources ,or or-ani=in- the presentation the clause as a messa-e/ assi-nin- di,,erent te.tual statuses to di,,erent parts o, the clause. 6hen 'e e.plore the clause ,rom a te.tual point o, 7ie'/ 'e ha7e to study it as a messa-e in the un,oldin- te.t/ a messa-e that is ad1usted to/ and 'hich helps create/ the (,lo' o, in,ormation( in the te.t. 5ecause this 7ie' has only rarely been ta2en in re,erence -rammars/ there is lac2 o, in,ormation on te.tual systems. Io doubt a -ood deal o, insi-ht into 7arious lan-ua-es is lost throu-h claims that they are 9,ree 'ord order9 lan-ua-es since te.tual options are o,ten reali=ed by se&uence %as 'e indicated ,or :n-lish abo7e) and throu-h absence o, accounts o, intonation systems. 5ut te.tual statuses can also be indicated by other means such as adpositions %e.-. the preposition an- in Ta-alo- and the postposition 'a in <apanese). There is 7ariation in ho' te.tual systems relate to clausal systems ,rom the other meta,unctions. ;n the one hand/ lan-ua-es di,,er in ho' they relate the te.tual meta,unction to the interpersonal one. As 'e ha7e indicated/ unmar2ed Theme in :n-lish is determined by the mood type %declarati7e: Sub1ect/ 'h+interro-ati7e: the 6h+element/ yes+no interro-ati7e: Finite M Sub1ect/ and imperati7e: ?redicator)0 but many lan-ua-es do not orient TH:M: to M;;3 in this 'ay: this is true o, Chinese/ <apanese and Ta-alo-. ;n the other hand/ lan-ua-es di,,er in ho' they relate the te.tual meta,unction to the ideational one. As 'e ha7e indicated/ :n-lish has a system o, @;*C: ,or -i7in- participants di,,erent te.tual potentials. Many other lan-ua-es ha7e a similar system/ but they may ta2e up the passi7e option less ,re&uently than :n-lish does or restrict the system o, 7oice more in relation to transiti7ity. 5ut a lan-ua-e may also achie7e the mappin- bet'een te.tual systems and transiti7ity roles 'ithout a separate 7oice system. For instance/ Ta-alo- has a -eneral system ,or selectindi,,erent participants and circumstances as Theme0 but there is no separate system o, 7oice. *n :n-lish/ Theme -enerali=es across the interpersonal and ideational meta,unctions and it may also contain contributions ,rom 'ithin the te.tual meta,unction itsel,/ con1uncti7e and continuati7e parts o, the Theme. ;ther lan-ua-es may separate out ideational Themes %i.e. thematic participants and circumstances)/ -i7in- them a clearly distinct status/ as in Ta-alo%'here ideational Themes are mar2ed by an- ) and <apanese %'here ideational Themes are mar2ed by 'a). There is also 7ariation in the di7ision o, labour amon- di,,erent te.tual systems. 6e su--ested that there is a tendency in :n-lish ,or Themes also to be -i7en and speci,ic. This tendency may be stron-er in a lan-ua-e 'here here is no obli-atory mar2in- o, speci,icity 'ithin nominal -roups/ as in Chinese. *n such lan-ua-es there may be a closer relationship bet'een (participant trac2in-( in discourse and te.tual systems 'ithin the clause than there is in :n-lish. B. .H *deational: lo-ical %at clause ran2): C;M?8:V*IG The meta,unctional components o, the -rammar discussed in B. . to B. .G ha7e one si-ni,icant ,eature in common: their structural re,le.es are in the broadest sense con,i-urational. That is to say/ the structural reali=ation o, selections in the systems o, transiti7ity/ o, mood/ and o, theme is some or-anic con,i-uration o, distinct ,unctions/ li2e Actor > ?rocess > Goal/ or %Sub1ect > Finite) > %?redicator > Ad1unct). 6e ha7e pointed out that these are not al'ays clearly de,ined/ or bounded/ as se-mental constituents0 the critical

characteristic that they share is that o, or-anic solidarity/ each part ,ul,ils a distincti7e ,unction 'ith respect to the 'hole. There is one ,urther component in the -rammar/ one 'hose structural re,le. is o, a di,,erent order: this is 'hat 'e re,er to as the lo-ical meta,unction. There are in e7ery lan-ua-e systems o, lo-ical relations: relations such as (and( and (or( and (i, ... then( and (because ... so(/ 'hich construe the lin2s bet'een one piece o, the discourse and another. These systems are reali=ed not by con,i-uration but by iteration: one clause bonded 'ith another clause/ or one -roup or phrase 'ith another -roup or phrase. The characteristic ,eature o, these relationships is that they do not create closure0 each element %each clause/ each -roup/ and so on) can al'ays be ,ollo'ed by another one o, the same. 6e re,er to these structures as comple.es: clause comple.es/ -roup comple.es/ and so on. :ach bond in such a comple. is called a ne.us. Here/ as else'here in this chapter/ 'e shall direct our attention to the clause/ 'hich means/ in the present section/ to the clause comple.. The -eneral ,orm o, systems o, this 9comple.in-9 2ind/ 2no'n as recursi7e systems/ is sho'n in Fi-ure H0 a set o, options .CyC... is combined 'ith a simultaneous choice o, (stop( or (-o round a-ain(. As ,ar as the clause comple. is concerned/ the ,ormer is itsel, a combination o, t'o simultaneous systems: % ) interdependency and %B) lo-ical+semantic relation. The system o, interdependency speci,ies the relati7e status o, the t'o clauses in a ne.us. %A prototypical ne.us consists o, 1ust t'o elements0 'e shall assume this ,or purposes o, discussion.) The t'o may ha7e e&ual status/ neither bein- dependent on the other %hence each in principle independent)0 this relationship is paratactic. ;r/ one may dependent on the other/ 'here the relationship is hypotactic. *n our notation/ parata.is is sho'n by Arabic numerals/ hypota.is by letters o, the Gree2 alphabet. The system o, lo-ical+semantic relations speci,ies 'hat its name su--ests: the particular 2ind o, lo-ical interconnection. *t is important to stress that 9lo-ical9 here re,ers to the lo-ic o, natural lan-ua-es/ a common+sense lo-ic characteri=ed by ,le.ibility and 9,u==9. This is/ o, course/ the ultimate source o, lo-ic in its ,ormal and symbolic sense0 but since such systems o, lo-ic are deri7ed ,rom natural lan-ua-e/ not the other 'ay round/ it is not 7ery pro,itable to try and interpret natural+lan-ua-e lo-ic as an imper,ect copy o, a lo-ic that has been desi-ned. The basic distinction in the :n-lish system/ in the lo-ical+semantic relations o, the clause comple./ is bet'een the t'o types e.pansion and pro1ection. *n a ne.us related by e.pansion/ the secondary clause pic2s up the messa-e o, the primary one and e.pands on it. *t may do this in any one o, three 'ays: by elaboratin-/ by e.tendin-/ or by enhancin-. :laboratin- means sayin- the same thin- o7er a-ain/ either by direct repetition %the limitin- caseF) or/ more 9elaborately9/ by re'ordin- it/ clari,yin- it/ or -i7inan e.ample. This is the relationship that is si-nalled by e.pressions such as in other 'ords/ that is to say/ ,or instance0 or by abbre7iations such as i.e./ e.-. and 7i=. :.tendin- means addin- somethin-/ 7aryin-/ replacin- or ta2in- a'ay/ e.pressed by and/ or/ instead/ e.cept/ and alsobut in its ad7ersati7e sense. :nhancin- means -i7in- some ,urther in,ormation that is related in a systematic 'ay by a semantic ,eature o, %typically) time/ cause/ condition/ or concession: here 'e ,ind con1unctions such as %and) then/ %and) so/ in that case/ other'ise/ ne7ertheless/ and but in its concessi7e sense. All o, these may be combined 'ith both types o, interdependency/ parata.is and hypota.is. 6e ha7e illustrated so ,ar 'ith paratactic

con1unctions0 but enhancin- relations/ in particular/ are o,ten construed hypotactically/ 'ith con1unctions such as 'hen/ because/ i,/ unless/ althou-h. *n a ne.us related by pro1ection/ the secondary clause is instated by the primary clause as 'hat somebody said %locution) or thou-ht %idea). This relationship is the 9direct and indirect speech and thou-ht9 o, our traditional -rammars. Here also the interdependency may be paratactic %9direct9) or hypotactic %9indirect9)0 in other 'ords/ pro1ection/ li2e e.pansion/ may combine 'ith either o, the t'o relationships in status. An e.ample o, a clause comple. is -i7en in Fi-ure H. The system net'or2/ sho'in- 1ust these ,irst steps in delicacy/ is sho'n in Fi-ure J.

Fi-. H: Analysis o, clause comple. ,rom casual con7ersation %S7art7i2 & Wuir2/ !QL)

Fi-. J: System net'or2 ,or the clause comple. %e.cludin- more delicate options) B.B %B) 5y ran2: ,rom clause to phrase and -roup Constituency is built on the part+'hole relation0 it presupposes a 'hole o, 'hich 'e identi,y constituent parts. 6holes 'hich display an or-anic constituency structure are called -rammatical units. Anits ha7e ynta-matic inte-rity: they are ,ully accounted ,or by their structures/ and they are not structurally mi.ed 'ith other units. Grammatical units are identi,iable in ,unctional terms. This means that %i) they are the points o, ori-in o, system net'or2s %such as those o, transiti7ity and mood in the clause) and %ii) they ,unction as constituents in their entirety. 6e can arri7e at ,unctionally determined units i, 'e adopt a ran2+based type o, constituency. 4an2 orders units into a hierarchy accordin- to their constituency relation: the hi-hest+ ran2in- units consist o, units o, the ran2 immediately belo'/ these units consist o, units at the ne.t ran2 belo'/ and so on/ until 'e arri7e at the units o, the lo'est ran2/ 'hich ha7e no internal constituent structure. 4an2 is thus a theory o, the -lobal distribution o, the units o, the -rammar. The :n-lish -rammatical ran2 scale is clause -roupCphrase 'ord morpheme That is/ a clause consists o, -roups/ a -roup o, 'ords/ and a 'ord o, morphemes. %For more on the phrase/ see belo'.) For instance/ the ran2ed constituency structure o, ne'born cal7es

are easy prey is as sho'n in Fi-ure K. Fi-ure " sho's the analysis o, a clause 'ith systemic ,eatures/ ,unction structure and preselections o, -roup ,eatures.

Fi-. K: 4an2+based constituency ran2: a.is: meta,un: S2ies 'ill be clear to o7er the partly cloudy rest o, Cali,ornia

clause

system

te.tual:

[unm2d theme/ unm2d culmination/ non+con1uncted0

interpers.: indic.: decl.: unta--ed/ temporal/ non+interactant/ positi7e0 ideational: relational: ascripti7e & intensi7e & middle & locati7e ...# structure te.tual: Theme 4heme ?redicator Complement Ad1unct 4esidue Attribute [nm -p# ... 8ocation [prp phr# ...

interpers.: Sub1ect Finite Mood ideational: Carrier ?rocess -roupCphrase system structure etc. [nm -p# ... [7b -p# ...

Fi- ": Systemic and structural analysis o, clause 'ith preselections at ran2 belo' 6e can treat the type o, constituency tree abo7e as the norm: all constituents o, a units o, the ran2 ne.t belo'. Ho'e7er/ the theory also needs to allo' ,or ran2shi,t/ 'hereby a unit one ran2 ser7es as i, it 'ere a unit o, a lo'er ran2/ i.e./ it is do'nran2ed. For instance/ a clause may ser7e as i, it 'ere a -roup as in %double+barred s&uare brac2ets/ [[ ##/ mar2 the ran2shi,ted clause): They(d send you a bill ,or a percenta-e o, [[ 'hat they are 'orth ## 4an2shi,ted units di,,er ,rom ran2in- ones in 7arious 'ays/ both in their o'n ma2e+up and in the selections that are open to them. For e.ample/ a ran2shi,ted clause is typically not

a7ailable ,or ar-ument/ it cannot be con,irmed or denied. *t is thus important that the theory should distin-uish bet'een ran2in- units %units ,unctionin- accordin- to their ran2) and ran2shi,ted ones %units ser7in- as i, they 'ere units o, a lo'er ran2). The meta,unctional or-ani=ation o, the -rammar that 'e illustrated abo7e ,or the clause applies to the other ran2s as 'ell. For e.ample/ the nominal -roup has ideational systems o, TH*IG TE?:/ C8ASS*F*CAT*;I/ :?*TH:S*S and WAA8*F*CAT*;I/ interpersonal systems o, ?:4S;I and ATT*TA3:/ and te.tual systems o, 3:T:4M*IAT*;I %c,. Fi-ure ! belo'). 5ut the 'ay the meta,unctional contributions map structurally one onto another 7aries0 in particular/ -roups are or-ani=ed both as or-anic 'holes and as lo-ical comple.es. Fi-ure B belo' sho's an e.ample o, an :n-lish nominal -roup. 8an-ua-es di,,er both 'ith respect to the number o, ran2s and 'ith respect to the di7ision o, -rammatical labour bet'een the di,,erent ran2s. For e.ample/ Chinese and :n-lish do ,airly little -rammatical 'or2 at 'ord ran2/ and @ietnamese e7en less. *n contrast/ many lan-ua-es ,a7our 'ord ran2 as the domain o, reali=ation ,or e.-. nuclear transiti7ity and modality. 8an-ua-es also di,,er 'ith respect to the nature o, the ran2 that is intermediate bet'een 'ords and clauses. 5oth Chinese and :n-lish (deri7e( the units o, that ran2 ,rom both ends/ as it 'ere: -roups are e.pansions o, 'ords %-roups o, 'ords/ 'ith a Head and Modi,iers) 'hereas phrases are contractions o, clauses %mini+clauses/ 'ith a con,i-uration o, ?rocess > 4an-e). The preposition is thus a 7erbal 2ind o, 'ord/ as is sho'n by :n-lish prepositions such as re-ardin-/ concernin-). *n Chinese this principle is e7en more pronounced0 items such as =ai ser7e either in phrases or in clauses: 'e can interpret the items in phrases as a class o, 7erb/ postpositi7e 7erb %c,. Fi-ure BL belo'). <apanese also has phrases/ but the phrasal relation comes a,ter the nominal -roup %i.e. nominal -roup > 'a/ -a/ o/ ni/ o/ 2ara/ made etc.)/ 1ust as the ?rocess o, a clause comes at the end o, the clause0 the phrasal relation is a post+position rather than a pre+position. Some lan-ua-es ha7e both phrases and nominal a,,i.es ,or reali=in- the ,unction ser7ed by the nominal -roup/ morpholo-ical cases/ o,ten usin- cases alone ,or participants and preposition or postposition %adposition) > nominal -roup mar2ed by case ,or circumstances %as is the tendency in German). ;ther lan-ua-es tend to use case+mar2ed nominal -roups ,or both/ as Finnish does. Eet other lan-ua-es ha7e no phrases at all/ but dra' on lo-ical se&uences o, dependent 7erbs instead to brin- certain participants or circumstances into the clause %e.-. A2an). 8an-ua-es may also use the ?rocess o, the clause as the site ,or mar2in- transiti7ity roles/ as Ta-alo- does ,or the Theme o, a clause. As -rammars e7ol7e/ there is a tendency ,or items to mo7e do'n the ran2 scale/ becominphonolo-ically reduced in the process. For e.ample/ pronouns %'ord ran2) may slide do'n the ran2 scale to become pronominal a,,i.es %morpheme ran2) ser7in- as parts o, 7erbs/ and au.iliaries %'ord ran2) may similarly be reduced in ran2 to become a,,i.es %morpheme) ran2 ser7in- as parts o, 7erb to indicate tenseCaspect/ modality and the li2e. As an intermediate step/ such items may be cliticised to other elements be,ore they become bound morphemes. This do'nran2in- o7er time is one aspect o, -rammatici=ation/ a process 'hereby cate-ories become more ti-htly inte-rated into the -rammatical system and the le.ico-rammatical system creates ne' meanin-s 'ithin some -rammatical subsystem. Another aspect o, -rammatici=ation is reduction in delicacy: see belo'. B.G %G) 5y delicacy: ,rom more -eneral to more particular

4an2 can be interpreted as a principle ,or distributin- a le.ico-rammatical system into a number o, di,,erent domains or units or-ani=ed into a constituency hierarchy/ or hierarchy o, or-anic 'holes and their parts %a 9holarchy9/ as it has been called 'ith re,erence to biolo-ical and other non+semiotic systems). This ,actorin- o, the o7erall system into subsystems accordin- to ran2 ma2es the o7erall system both simpler and more po'er,ul. *t ma2es the o7erall system simpler precisely because it is ,actored or partitioned into subsystems that are relati7ely independent o, one another and interact throu-h preselection rather than ('irin-( in a system net'or2. :ach subsystem thus has its o'n domain o, responsibility. *t ma2es the o7erall system more po'er,ul because since each subsystem has its o'n domain o, responsibility/ the di,,erent subsystems are in principle ,reely 7ariable 'ith respect to one another so that the o7erall potential o, the le.ico-rammatical system is the total intersection o, all possible ,eatures 'ithin all subsystems. This total intersection is/ in ,act/ in,inite since/ 'hen a system is ran2ed %i.e. ,actored into subsystems accordin- to ran2)/ its potential can e.pand throu-h ran2shi,t %see abo7e): ,or e.ample/ a clause can ser7e as i, it 'ere a -roup or 'ord/ thus openin- up the ,ull clausal subsystem at -roup or 'ord ran2. As an or-ani=ational principle/ ran2 is reasonable easy to detect %althou-h/ in lin-uistics/ it has sometimes been con,used 'ith other principles o, or-ani=ation/ notably strati,ication: sometimes morphemes ha7e been 'ron-ly thou-ht to consist o, phonemes instead o, beinreali=ed by [se&uences o,# phonemes)0 ran2 represents a ,airly o7ert or e.plicit 2ind o, order/ that o, a 'hole to its parts/ and it is e7en re,lected partially in many 'ritin- systems. Ho'e7er/ le.ico-rammar is also or-ani=ed in a more co7ert or implicit 2ind o, 'ay. 6e ha7e already re,erred to this 2ind o, or-ani=ation: the orderin- o, the systems o, a system net'or2 in a relation o, delicacy. For e.ample/ the systems ?4;C:SS TE?:/ TE?:+;F+5:*IG/ and 4:8AT*;I TE?: in Fi-ure B are ordered in increasin- delicacy. This 2ind o, le.ico-rammatical order is more co7ert in that it is not directly re,lected in the 'ordin- o, a -rammatical unit0 rather/ it is a more abstract 2ind o, order that is imposed on the systems 'hose options that 'ordin- reali=es. 3elicacy is a 7ery simple yet po'er,ul principle o, or-ani=ation. *t orders systems on a cline ,rom the most -eneral systems o, options to the most speci,ic ones0 and at the same time/ it orders reali=ations o, these options accordin- to their systemic en7ironment. This means that the reali=ational properties o, a clause or any other -rammatical unit can be (placed( in the system so that it applies only to the appropriate subset o, units. For e.ample/ only (yes+no interro-ati7e( clauses ha7e the reali=ational property o, Finite precedin- Sub1ect %i.e./ Finite M Sub1ect)0 it does not apply to interro-ati7e clauses in -eneral/ nor to indicati7e clauses in -eneral/ nor to ma1or clauses in -eneral. 5y the same to2en/ i, (indicati7e( clauses ha7e the reali=ational property o, ha7in- an e.plicit Sub1ect %i.e/ > Sub1ect)/ then all more delicate options accessible ,rom (indicati7e(/ such as (yes+no interro-ati7e( also ha7e that property %c,. Fi-ures H and " abo7e). That is/ reali=ational properties are inherited alon- the cline o, delicacy ,rom less delicate to more delicate. 3elicacy can thus be interpreted as a -eneral principle ,or or-ani=in- le.ico-rammar/ 1ust li2e ran20 more speci,ically/ it is a principle ,or distributin- in,ormation in le.ico-rammar accordin- to ta.onomic domain o, application. 5ut delicacy is/ in ,act/ more than an orderin- o, systemic options and/ by implication/ the reali=ation statements associated 'ith them. *t is also the principle accordin- to 'hich the t'o (parts( o, le.ico-rammar/ le.is %7ocabulary) and -rammar are related. 8oo2ed at ,rom the point o, 7ie' o, -rammar/ le.is is most delicate -rammar0 and loo2ed at ,rom the point o, 7ie' o, le.is/ -rammar is least delicate %most -eneral) le.is. *n other 'ords/ the systemic options o, the more -eneral systems in the system net'or2 %such as

(declarati7eCinterro-ati7e(0 ('h+Cyes+no(0 (materialCmentalC7erbalCrelational(0 (e.istentialCe.pandin- relational(0 (intensi7eCpossessi7eCcircumstantial(0 (speci,icCnon+speci,ic() are reali=ed by -rammatical structure ,ra-ments %e.-. Sub1ect M Finite/ ?rocess > :.istential) or -rammatical items %e.-. interpersonal particles 2a/ ne/ yo in <apanese or ma/ ne/ ba in Chinese0 determiners such as theCthisCthat and au.iliary 7erbs such as do/ be/ ha7e in :n-lish)/ 'hereas the more delicate options are reali=ed by le.ical items %e.-. le.ical 7erbs beCrepresentCmeanCindicateCsymboli=e0 and le.ical nouns manCboyC'omanC-irl). As 'e ha7e noted/ delicacy is a cline/ so there are re-ions intermediate bet'een -rammar and le.is/ such as prepositions in :n-lish and phase in Chinese. Such intermediate re-ions ser7e to re7eal the -radual mo7e bet'een -rammar and le.is alon- the scale o, delicacy. And one aspect o, the semo-enic process o, -rammatici=ation is the mo7e o7er time o, items ,rom le.is to -rammar as they are -enerali=ed0 ,or e.ample/ it is common in lan-ua-es ,or some le.ical items o, motion to be -enerali=ed in delicacy to ser7e as -rammatical items reali=in- options in tense systems %c,. :n-lish -oin- to/ French 7enir de/ S'edish 2omma att) and ,or some le.ical items o, material manipulation: -rabbin-/ ta2in- to be -enerali=ed in delicacy to ser7e as -rammatical items mar2in- Goals %under certain conditions0 c,. Chinese ba/ ori-inally a le.ical 7erb (ta2e(). Fi-ure Q -i7es a 7ery simple e.ample o, the mo7e to'ards le.ical delicacy in the system o, ?4;C:SS TE?: in :n-lish 'ithin relational clauses.

Fi-. Q: To'ards le.ical delicacy in intensi7e ascripti7e relational clauses Grammar and le.is are ne7er totally di7orced ,rom one another. ;n the one hand/ small le.ical sets are o,ten associated 'ith little (local -rammars(/ i.e. delicate 7ariations in -rammatical potential/ as is the case 'ith the le.ical set o, percepti7e processes in :n-lish %see/ notice/ -limpse/ espie/ hear/ o7erhear/ ,eel/ taste/ smell0 sense/ e.perience)/ 'hich can combine 'ith di,,erent 2inds o, ?henomenon than co-niti7e or a,,ecti7e processes. %For e.ample/ 'e can say * sa' somebody crossin- the street but not * thou-ht somebody crossinthe street.) ;n the other hand/ collocations bet'een small le.ical sets typically occur bet'een le.ical items reali=in- closely bonded -rammatical ,unctions such as ?rocess > Medium %nei-h > horse/ bar2 > do-0 a-e > 'ine/ mature > cheese)/ ?rocess > 4an-e %'rea2 > ha7oc/ do > dance/ ma2e > mista2e0 -o > madCcra=yCinsaneCbananasCbon2ers0 ,all > illCsic2)/ ?rocess > Manner %re-ret > deeply/ understand > completely)/ and Facet > Thin- %-a--le > -eese/ school > ,ish/ ,loc2 > birds)0 and in Chinese also :7ent > 4esult and Measure > Thin-.

*n the description o, the le.ico-rammatical systems o, 7arious lan-ua-es/ delicacy has pro7ed a help,ul conceptual resource ,or mana-in- comple.ity. 8e.ico-rammatical systems are e.plored and mapped out at a ,airly lo' de-ree o, delicacy so that the o7erall distribution and or-ani=ation o, the system can be established. This o7er7ie' that is limited in delicacy pro7ides the map that can -uide subse&uent e.cursions into more delicate systems. This also applies to the mo7e ,rom -rammar to le.is: le.ical or-ani=ation can be in7esti-ated in terms o, the cate-ories o, the -rammatical part o, the system0 that is/ -rammar construes the -eneral parameters in terms o, 'hich le.ical distinctions are made. G. System and te.t G. Function and ran2 The -rammatical system o, e7ery natural lan-ua-e can be summarily presented as a ,unctionCran2 matri./ 'here ,unction is used in the sense o, meta,unction %see Section B. ). So ,or :n-lish 'e can construct such a matri. as in Fi-ure !. ideational ran2 clause [class# comple.es %clause lo-ical e.periential T4AIS. %process type) M;;3 TH:M: interpers. te.tual

%coh

C;H 4:8

M;3A8*TE CA8M*I+ AT*;I ?;8A4*TE @;*C: M*I;4 M;;3 %ad1unct type) F*I*T:+ I:SS C;I<AIC+ T*;I

phrase [prep.#

phrase

M*I;4 T4AIS. %circ. type) *IT:4+ T:IS: 3:?:I3:ICE %parata.isC hypota.is) 8;G*CA8+ S:MAIT*C 4:8AT*;I %e.pansionC pro1ection) :@:IT TE?: AS?:CT %non+,inite)

-roup [7erbal#

-roup

@;*C: 3:*CT+ *C*TE

4:F :88 SA5 T*;

[nominal#

M;3*F*C+ TH*IG AT*;I TE?: C8ASS*F*+ CAT*;I :?*TH:S*S WAA8*F*+ CAT*;I

?:4S;I ATT*TA3:

3:T:4M*I+ C;I AT*;I T*;

[ad7erb.#

M;3*F*C+ WAA8*TE AT*;I %circ. type)

C;MM:IT C;I<AIC+ %ad1. type) T*;I

'ord

'ord)

3:4*@A+ T*;I *IF;. TAV*S comple.es ACC:IT+ AAT*;I simple.es

%3:I;T+ AT*;I)

%C;II;T+ AT*;I) K:E *IF;. F;CAS

in,o. unit

in,o. unit comple.

Fi-. !: Function+ran2 matri. ,or :n-lish There are ,our ,eatures to note about this ,i-ure. %i) :ach cell is the location %point o, ori-in) o, a system net'or2/ indicated by the name in small capitals. 6here t'o names are -i7en %e.-. M;;3 and M;3A8*TE)/ this means that it is use,ul on semantic -rounds %i.e. (,rom abo7e(0 c,. Section .B/ G.G) to reco-ni=e t'o %or more) subsystems0 but since these share the same address in the matri./ they ha7e a common point o, ori-in and thus ,orm a sin-le net'or2. As 'e ha7e seen/ a system net'or2 is a theory o, options: it de,ines a topolo-ical re-ion in terms o, the meanin-s that can be construed by the -rammar. The net'or2 is open+ended in delicacy: 'hate7er distinctions are displayed/ it 'ill al'ays be possible to add more. %ii) Structural considerations are embodied in the notion o, ran2: each ran2 is a structural unit 'hose constituent ,unctions are typically ,ul,illed by classes o, members o, the ran2 belo'/ clauses are composed o, %,unctions ,illed by classes o,) -roups and phrases/ and so on. Compare an or-anic (ran2 scale( in biolo-y: or-an+tissue+cell. 5ut beyond this/ structure plays no part in determinin- the topolo-y0 hence the matri. accommodates 7arious alternati7es to a strict constituent structure. Some o, this 7ariation appears to be systematic: thus/ :n-lish and at least some other lan-ua-es display a tendency ,or: e.periential systems interpersonal systems te.tual system lo-ical systems to be reali=ed 9 9 9 se-mentally %by constituency) prosodically %by colourin-) periodically %by oscillation) serially %by iteration)

%iii) The matri. has the indeterminacy that is a property o, all -rammatical representations. Thus the structural mani,estations o, ran2 are not &uite identical across the di,,erent columns. For e.ample/ in :n-lish the interpersonal -rammar displays a layerin- o, structure intermediate bet'een clause and phraseC-roup %the Mood > 4esidue re,erred to in Section .G)0 and in the te.tual meta,unction there is a parallel structure in the ,orm o, the in,ormation unit %locus o, Gi7en > Ie')/ 'hich in unmar2ed association 'ith the clause but not identical to it. 8i2e'ise there is indeterminacy across the columns: a system may ori-inate 'ith one meta,unction/ but be acti7ated ,rom 'ithin another0 ,or e.ample/ :n-lish @;*C:/ 'here %a) 'hat is possible is determined e.perientially/ but %b) 'hich option is ta2en up deri7es ,rom its association 'ith the te.tual systems o, *IF;4MAT*;I and TH:M:.

%i7) :ach net'or2 -enerates its o'n structural (output(/ in terms o, se-mental con,i-urations or other types o, structure0 the latter may ho'e7er still be represented in con,i-urational ,orm: e.-. the periodic mo7ement/ or oscillation/ bet'een t'o 2inds o, prominence %9in,ormation ,lo'9) in the :n-lish clause/ represented as a t'o+,old con,i-uration o, Theme > 4heme and Gi7en > Ie'. Thus an item under description 'ill typically ha7e multiple representations. For e.ample/ a clause in :n-lish/ or Chinese/ or <apanese %and possibly in all lan-ua-es) 'ill be represented three times/ that is/ as a mappin- onto each other o, three di,,erent structures/ one deri7in- ,rom each meta,unction. Fi-ure BL sho's such a multilayered structure. &iu autumn+ 'ind ,enbX shYyZ chui+ blo' do'n lu[ =Xi on d\+ -round shXn-

mar2er lea7es o, Goal

clause: %i) material & middle: e7enti7e & locati7e & %ii) declarati7e & positi7e & nonmodal & %iii) unmar2ed theme & unmar2ed in,ormation ,ocus Actor Sub1ect Theme 4heme Ie' IG Classi,ier Thinnoun noun mar2er IG Thinnoun @G :7ent 7erb phrase 4esult+ Minor+ ati7e process 7erb Minor+ ran-e Goal ?rocess 8ocatio n Ad1unct

Compl+ ?redicato ement r

pre+ IG positi7e 7b Thin- Facet noun post+ positi7 e noun

Fi-. BL: Multilayered structure o, a Chinese clause *n some lan-ua-es at least it may be help,ul to represent certain structures both e.perientially/ as se-mental con,i-urations/ and lo-ically/ as iterati7e comple.es: ,or e.ample/ the :n-lish nominal -roup: see Fi-ure B . Iote that each line in the analysis is accountin- ,or a di,,erent 7ector o, a-nation: that is/ it is e.pressin- the 'ay the item is systemically related to certain other items.

]S ^ the _U t'o ]U

++++++ TU

+++++++ S -ol,ers Thin-

]T

-reatest pro,essional

o, Wuali,ier: Minor ?rocess

all time

3eictic Iumerati7e :pithet Classi,ier

Minor 4an-e

Fi-. B : :.periential and lo-ical analysis o, the :n-lish nominal -roup G.B System and instance 6hat 2ind o, acti7ity/ then/ is -rammaticsD 6hat is a -rammarian doin-/ in 'or2in- on the -rammar o, a lan-ua-eD 3oin- -rammar means establishin-/ and e.plainin-/ the principles that lie behind the 'ordin-s o, a natural lan-ua-e: in traditional terms/ its synta./ morpholo-y/ and le.is %7ocabulary). This di7ision into three parts is a methodolo-ical one: in 6estern lin-uistics/ di,,erent methods e7ol7ed ,or synta. %-rammar abo7e 'ord ran2)/ morpholo-y %-rammar belo' 'ord ran2)/ and le.icolo-y %relations and meanin-s o, 'ords). As ,ar as the phenomena themsel7es are concerned/ they constitute a sin-le stratum %c,. Section H. belo'). *n many lan-ua-es the distinction bet'een synta. and morpholo-y is unnecessary e7en at a methodolo-ical le7el/ since the 'ords themsel7es are in7ariant %9there is no morpholo-y9) and hence no special methods are needed to account ,or -rammar belo' 'ord ran2 %e.-. Chinese/ @ietnamese). 9:stablishin- and e.plainin- the principles9 means reco-ni=in- that the -rammar is not random0 it ,orms a system. %This is o, course an assembla-e o, di,,erent systems/ but 'e can still re,er to the totality as a system.) 5ut the tas2 o, -rammatics is not 1ust to describe the system0 it is also to relate the system to the instance/ or rather %since these are not distinct steps) to describe the system as it relates to actual instances o, lan-ua-e %re,erred to as te.t). This relation o, system to te.t/ 2no'n as instantiation/ is not as simple as it ,irst appears0 in ,act/ it is hi-hly problematic. System and te.t are not t'o di,,erent thin-s0 they are the same thin- seen ,rom di,,erent perspecti7es/ di,,erent standpoints o, the obser7er. A help,ul analo-y may be that o, 9climate9 and 9'eather9: there is only one set o, phenomena here/ not t'o/ but 'e call it 'eather 'hen 'e e.perience it instantially/ as meteorolo-ical 9te.t9/ and climate 'hen 'e are ta2in- a lon-term perspecti7e in order to establish and e.plain the principles/ the meteorolo-ical 9system9 that lies behind. Te.t is meanin-,ul only because it is the instantiation o, a systemic potential: this is 'hat 'e mean by sayin- that spea2er and listener must share a 9command o, the lan-ua-e9/ an unconscious a'areness o, the interstratal patterns %ho' ,orms o, 'ordin- reali=e meanin- and are reali=ed in sound) and o, the topolo-y o, each stratum %ho' 'hat is said contrasts 'ith 'hat mi-ht ha7e been said but 'as not). 5y the same to2en/ our concept o, system is 7alid only because it is instantiated in te.t: each instance 2eeps ali7e the potential/ on the one hand rein,orcin- it and on the other hand challen-in- and chan-in- it. This dialectic o, te.t and system is 'hat 'e understand by a li7in- lan-ua-e.

A lan-ua-e 9li7es9 in the sense that it is a dynamic open system/ 'hich is maintained in e.istence by constantly chan-in- in interaction 'ith its en7ironment. The system is probabilistic %li2e the climate)0 'hen 'e say that there is a choice/ say in M;;3/ bet'een (declarati7e( and (interro-ati7e(/ 'e mean 9'ith a certain de-ree o, probability attached9. %The probability is li2ely to 7ary/ o, course/ in di,,erent ,unctional conte.ts: there are local as 'ell as -lobal probability pro,iles.) Hence each instance o, an indicati7e clause in,initesimally perturbs the probabilities0 and no' and a-ain the e,,ect 'ill be catastrophic/ brin-in- a ne' system into e.istence or else eliminatin- an old one %as happened in :n-lish 'hen the probability o, in,ormal sin-ular thou as opposed to polite or plural you dropped -radually to =ero). The net'or2 representation o, a systemic -rammar is a 'ay o, modellin- the potential so as to allo' ,or its on-oin- e7olution. ;ne aspect o, this e7olution is -rammatici=ation: instantial patterns in te.ts may -radually become part o, the -rammatical potential. *t is a ,eature o, semiotic systems %since they are social0 see Section H belo') that an instance carries 7alue. Thus a -i7en te.t may be particularly hi-hly 7alued: ,or e.ample/ a political mani,esto/ or a literary arti,act. *t ,alls to -rammatics/ there,ore/ to interpret a te.t not only as a 9'indo' on the system9 but also as an ob1ect in its o'n ri-ht/ e.plainin- not merely 'hy it means 'hat it does but also 'hy it is 7alued as it is. *n conte.ts o, lan-ua-e education/ ,or e.ample/ the analysis should sho' 'hy one piece o, a learner(s 'ritin- is more e,,ecti7e than another. Much o, the 'or2 o, the -rammatics/ ho'e7er/ may be located precisely at some intermediate point on the instantiation cline: ,or e.ample in the study o, codes and o, re-isters. 4e-isters are ,unctional %or 9diatypic9) 7arieties o, a lan-ua-e that e7ol7e in di,,erent conte.ts o, use: ,ormal or casual/ technical or non+technical/ more openended or more closed. They are not 97ariants9 in the sense o, alternati7e reali=ations o, a common meanin- at some hi-her stratum. Codes are subcultural 7ariants/ the di,,erent discourses o, youn- and old/ males and ,emales/ di,,erent classes or casts 'ithin a society0 they are di,,erent semantic styles typically associated 'ith some common -enerali=ed conte.t o, use. Codes are hard to study precisely because their instantial status is indeterminate: should they be represented as recurrent classes o, instances or as subsystems o, the o7erall systemD To use our meteorolo-ical analo-y/ it is not clear 'hether codes should be treated as lon-term 'eather trends or as minor 7ariations in the climate. %The ans'er is/ in ,act/ as both0 but that is 1ust 'here the di,,iculty arises.) *n 'or2in- on the -rammar o, a lan-ua-e/ one tries to mo7e ,reely alon- the instantiation cline. 6e are able to do this more easily once 'e ha7e a corpus/ a treasury o, instances that 'e can come bac2 to all the time ,or rene'al o, connection. *t is this that constitutes the -rammarian(s essential data ban2. The corpus is not a substitute ,or a theory0 it does not 9contain all the ,acts9/ li2e hidden 'ords buried in a 1umble o, letters. Facts/ and the principles behind them/ ha7e to be construed0 but they can be construed much more reliably on the &uantitati7e ,oundation o, a modern computeri=ed corpus. 6e can represent the cline o, instantiation dia-rammatically as in Fi-ure BB.

Fi-. BB: *nstantiation At the present sta-e o, 2no'led-e/ 'e usually try to 'rite a -rammar ,or the system o, a lan-ua-e as a 'hole/ notin- as 9special cases9 le.ico-rammatical 7ariation due to re-ister or code. The reason ,or this is that such 7ariation is lar-ely &uantitati7e0 it in7ol7es the relati7e ,re&uency o, terms in systems/ rather than the presence or absence o, an entire system. For e.ample/ there are certain re-isters in 'hich ,irst and second person are e.tremely rare0 but 'e 'ill probably not 'ant to 'rite a separate -rammar ,or those re-isters/ 'ith the person system entirely le,t out. 6hen it comes to semantic representation/ since codes are in ,act semantic 7ariants/ they can be best e.plained by re,erence to a common semantic system/ as in Hasan(s studies o, socio+semantic code 7ariation amon- mothers and children in Sydney. 4e-isters on the other hand ha7e no 9hi-her+le7el constant9 and so may be more e,,ecti7ely represented as distinct semantic subsystems. G.G 3escripti7e principles As 'ith any scienti,ic theory/ the basic concepts %thin-s li2e stratum and meta,unction/ reali=ation/ instantiation/ and delicacy and ran2) are not empirically 7eri,iable0 they ,orm an abstract ,rame'or2/ or model/ set up as a means o, e.plorin- and e.plainin-. %An abstract model o, this 2ind is itsel, a semiotic system: see Section H belo'.) Such a model is all the time bein- re,ined and elaborated in use. 5y contrast 'ith these -eneral theoretical cate-ories/ the descripti7e cate-ories are set up by re,erence to speci,ic ,eatures o, -i7en lan-ua-es/ and hence in principle are 7eri,iable. That is to say/ they can be characteri=ed in 'ays 'hich ma2e it possible to decide 'hether or not somethin- is an instance/ and 'hether or not some cate-ory is present in the system at all. T'o 2inds o, considerations are typically in7ol7ed. %i) Most -enerali=ations made in describin- -rammar are 7alid only 'ith some de-ree o, probability. They allo' us to identi,y 9e.ceptions9/ ,or 'hich 'e then see2 a ,urther -enerali=ation0 and so on. For instance: 'e say that the :n-lish Moodta- consists o, Ta-,inite > Ta-sub1ect/ 'here the ,ormer repeats the Finite operator/ and the latter repeats the Sub1ect %each -rammatici=ed in deictic ,orm/ as in Mary 2ne'/ didn(t sheD). Typically/ the ta- does

repeat the Finite0 but 'e also come across instances li2e She(ll li2e ,airy tales/ does sheD 'here Finite 'ill is replaced by does. The meanin- is (* e.pect she li2es ,airy tales0 'ill you con,irm my e.pectationD(. This contrasts 'ith She(ll li2e ,airy tales/ 'illC'on(t sheD/ 'hich 'ould mean (3oCdon(t you share my e.pectationD(0 and it is an instance o, a more delicate subclass 'hereby a modal Finite in the clause is replaced by a temporal one in the ta- %c,. *t must ha7e been cro'ded in there/ 'as itD). And 'here the ta- system in -eneral ,a7ours re7ersin- the polarity %positi7e R ne-ati7e/ or ne-ati7e R positi7e)/ in this subsystem the local probability is ,or the polarity to remain constant/ thou-h in both cases the other/ mar2ed option also occurs. Such probabilistic -enerali=ations can still be tested/ by re,erence to &uantitati7e patterns in the corpus. They are si-ni,icant because they enable us to de,ine the meanin- o, doinsomethin- by contrast 'ith the meanin- o, not doin- it/ or o, doin- somethin- else. Certainty then appears as the limitin- case o, probability/ the point 'here no ,urther meanin- is created. Thus i, the second part o, the -enerali=ation abo7e %9repeat the Sub1ect9) holds true in all instances/ there is no more delicate choice point here at 'hich more semantic space is opened up. 6e then -o bac2 to the corpus to chec20 'e ,ind 5ut my husband heard it too/ didn(t youD and 'e ha7e to decide 'hether that ,its the pattern or not. This leads into the ne.t principle to be considered. %ii) Clearly the ans'er to 'hether you in the ta- in the last e.ample is a repetition o, the Sub1ect or not depends on ho' 'e are loo2in- at it. *, 'e are loo2in- at it (,rom belo'(/ at ho' it is reali=ed in ,orm/ then it is not0 the pronominal o, my husband must be he. *, 'e are loo2in- at it (,rom abo7e(/ at the meanin- 'hich is bein- reali=ed/ then it is0 you is ,unctionin- anaphorically and the t'o are co+re,erential. ?robably the spea2er turned to'ards her husband 'hile mo7in- into the Ta-0 but note that 'e do not need any in,ormation o, this 2ind ,rom outside the te.t/ it is the te.t itsel, that construes the meanin- and the conte.t o, situation ,or us. 5ecause 'e 2no' the M;;3 system o, :n-lish/ includin- the principle o, the Ta- %based on the meanin- o, Sub1ect as the modally responsible element in the proposition)/ 'e are able to interpret the instance by locatin- it in its place in the meaninpotential. And here 'e are adoptin- the third perspecti7e/ loo2in- at it (,rom around(. As 'e sa' in Section / it is a critical ,eature o, systemic -rammatics that the -rammarian has trinocular 7ision/ loo2in- at any phenomenon ,rom each o, these three stratal perspecti7es. 6e may choose to pri7ile-e one or another0 but all are ta2en into account/ and since they 'ill typically con,lict the optimum description ,or any particular occasion 'ill almost al'ays be a compromise. Traditionally -rammarians ha7e be-un by loo2in- (,rom belo'(/ because this is the most ob7ious 'ay in: 'e as2 &uestions li2e 96hat is the meanino, 'a in a <apanese clauseD9/ ,irst identi,yin- a ,orm %this then becomes the -rammatical (,act()/ and then as2in- 'hat this ,orm means. 5ut in a ,unctional -rammatics such as systemics/ relati7ely -reater priority is accorded to the perspecti7e (,rom abo7e(/ 'here the &uestion is one such as 9Ho' does the <apanese clause construe the ,lo' o, in,ormationD9. *nterestin-ly/ the perspecti7e that seems to be most o,ten i-nored is that (,rom around(/ 'here 'e construct the paradi-matic en7ironment: the set o, options that constitute the local -rammatical potential. *n this e.ample/ 'e 'ould be settin- up the net'or2 o, systems that constitute the te.tual resources o, the clause: on the analo-y o, :n-lish/ the systems o, TH:M: and o, *IF;4MAT*;I/ and their reali=ation throu-h the structural elements o, Theme > 4heme and Gi7en > Ie'.

The analo-y 'ith :n-lish may not hold/ o, course0 and here a-ain it is the trinocular perspecti7e that is 7ital. There is no ob1ecti7e criterion ,or decidin- ho' much ali2e t'o phenomena must be ,or us to call them by the same name: in the last resort &uestions such as 9*s there a passi7e in that lan-ua-eD9 or 9Ho' do 'e reco-ni=e the Sub1ect in this lan-ua-eD9 are &uestions about 'hether to trans,er labels in comparin- one lan-ua-e 'ith another. There is nothin- 'ron- 'ith ma2in- predictions about one lan-ua-e on the basis o, 'hat is 2no'n about others0 this is a normal 'ay o, proceedin-. 5ut one has to -uard a-ainst ,oistin- the cate-ories alon- 'ith the labels. The systemic approach 'ould be/ rather/ to as2 a &uestion such as 9*s there a system 'hich redistributes the participants into di,,erent te.tual statusesD9 %as :n-lish Actor/ Goal/ etc. are redistributed into di,,erent patterns o, thematic and in,ormation structure). *, there is/ 'e call it the @;*C: system0 and then i, it sets up some 2ind o, unmar2edCmar2ed opposition/ 'e may call the unmar2ed term 9acti7e9 and the mar2ed term 9passi7e9. Similarly/ instead o, 9Ho' do 'e reco-ni=e the Sub1ectD9 'e mi-ht as2 93oes the -rammar incorporate an element ha7in- a speci,ic ,unction 'ith re-ard to 7alidatin- a propositionD9/ i, so/ 'e 1usti,iably re,er to this as the 9Sub1ect9. All these steps depend on adoptin- the same (trinocular( stratal perspecti7e. H. ?erspecti7es beyond le.ico-rammar *n this ,inal section/ 'e shall brie,ly conte.tuali=e 'hat 'e ha7e discussed. First 'e consider -rammar itsel, in its %stratal) en7ironment in lan-ua-e in conte.t/ and then 'e conte.tuali=e our o'n -rammatics/ systemic theory/ 'ith respect to other 2inds o, -rammatics and 'ith respect to lan-ua-e and other 2inds o, systems. H. Grammar as a stratal subsystem o, lan-ua-e in conte.t Grammar is/ as 'e put it abo7e/ the system o, 'ordin-s o, a lan-ua-e. More precisely/ it is the system ,or reali=in- meanin-s in 'ordin-s. The systems o, meanin-s is semantics/ so 'e can say that semantics is reali=ed by -rammar. For e.ample/ the interpersonal resource o, mood in the -rammar reali=es the interpersonal resource o, speech ,unction in the semantics0 thus a declarati7e clause reali=es a mo7e in dialo-ue that -i7es in,ormation. 4eali=ation is a relation that orders 'hole subsystems o, lan-ua-e relati7e to one another in symbolic abstraction0 'e say that it strati,ies these subsystems. Semantics and -rammar are strati,ied/ 'ith semantics as the hi-her stratum and -rammar as the lo'er one. *n a similar 'ay/ 'ordin-s are reali=ed by soundin- %or 'ritin-)0 that is/ -rammar %the system o, 'ordin-) is reali=ed by phonolo-y %the system o, sound) %or by -rapholo-y/ the system o, 'ritin-). For e.ample/ delicate options 'ithin the interpersonal resource o, mood in the -rammar are reali=ed by distinctions in tone %direction o, pitch mo7ement) in the phonolo-y0 thus a declarati7e clause is/ unmar2edly/ reali=ed by ,allin- tone. The three strata or stratal subsystems semantics/ -rammar and phonolo-y %-rapholo-y) ma2e up the lin-uistic system. ;, the three/ -rammar is the stratum that is purely internal to the lin-uistic system/ slotted in bet'een t'o other lin-uistic strata. Grammar thus has to be or-ani=ed in such a 'ay that it can both ser7e to reali=e semantics and at the same time be reali=ed by phonolo-y. *n contrast/ both semantics and phonolo-y ha7e to inter,ace 'ith systems outside the lin-uistic system. Fi-ure BG represents the stratal or-ani=ation o, the lin-uistic system and illustrates it 'ith the interpersonal e.ample indicated abo7e: statement is reali=ed by declarati7e in neutral 2ey/ 'hich is reali=ed by ,allin- tone. As the dia-ram sho's/ the terms at the di,,erent strata/ (statement(/ (neutral(/ and (,allin-(/ enter into di,,erent systemic relationships 'ithin each

stratum. This multiplication o, systemic relationships is part o, 'hat -i7es po'er to a strati,ied system. *n the dia-ram/ 'e ha7e sho'n each stratal subsystem 'ithin a circle0 the circle represents the stratum to 'hich it belon-s. The circles increase in si=e 'ith the mo7e ,rom phonolo-y 7ia -rammar to semantics/ to indicate that these systems increase %both in si=e and in comple.ity) 'ith the mo7e to a hi-her stratum. ;ur e.ample comes ,rom the interpersonal meta,unction %at clause ran2/ in the -rammar)0 but the stratal principle it illustrates also applies to the te.tual/ e.periential/ and lo-ical meta,unctions. Ho'e7er/ there is one respect in 'hich the e.ample is not representati7e o, the -eneral principle o, interstratal relations. *t su--ests that the interpersonal meta,unction and/ by implication/ all the meta,unctions are pro1ected throu-hout the 'hole lin-uistic system to or-ani=e each stratum. Ho'e7er/ the -eneral principle is that the meta,unctions are mani,ested as or-ani=ational principles 'ithin semantics and -rammar but not 'ithin phonolo-y. 6hy should this be the caseD The e.planation is &uite strai-ht,or'ard. Semantics and le.ico-rammar are the content strata o, lan-ua-e/ 'hereas phonolo-y %or -rapholo-y) is the e.pression stratum. The meta,unctions are modes o, meanin-/ not modes o, soundin- or 'ritin-. 5ein- modes o, meanin-/ the meta,unctions are mani,ested in the or-ani=ation o, the t'o strata concerned 'ith meanin-/ the content strata/ semantics and le.ico-rammar. Fi-ure BH sho's the meta,unctional or-ani=ation o, the content strata to-ether 'ith the other -lobal dimension o, or-ani=ation/ strati,ication. *t also represents the stratal en7ironment o, the 'hole lin-uistic system/ the conte.t in 'hich the system is embedded. 6e 'ill return to conte.t presently/ but let us ,irst say a ,e' more 'ords about meta,unction and strati,ication. The meta,unctional or-ani=ation o, both semantics and le.ico-rammar is the most central aspect o, the -eneral stratal principle that semantics and le.ico-rammar are natural in relation to one another: le.ico-rammar is the lo'er o, these t'o content strata0 it is not an autonomous ,ormal system. 8e.ico-rammatical or-ani=ation/ both systemic and structural/ is semantically (transparent( rather than opa&ue. For e.ample/ -i7en a -rammatical structure such as Theme > 4heme/ Mood > 4esidue/ or Carrier > ?rocess > Attribute > 8ocation/ 'e can read it as a natural reali=ation o, a semantic con,i-uration0 and -i7en a -rammatical system such as M;;3/ 'e can read it as a natural realisation o, a semantic system such as S?::CH FAICT*;I. %Grammatical metaphor complicates the picture0 but the metaphorical e.pansion o, the system depends on the natural relationship bet'een semantics and le.ico-rammar.) *n contrast/ the relationship bet'een phonolo-y %or -rapholo-y) and le.ico-rammar is lar-ely con7entional %arbitrary) rather than natural. For e.ample/ in :n-lish/ -i7en some phonolo-ical structure such as ;nset > ?ea2 > Coda or *ctus > 4emiss/ 'e cannot read it as con-ruent 'ith a le.ico-rammatical structure. Con7ersely/ -rammatical structures are not con-ruent 'ith phonolo-ical ones/ and -rammatical and le.ical items are similarly reali=ed con7entionally by di,,erent sound shapes. %There are local little partial iconic relationships/ but these are mar-inal seen a-ainst the bac2-round o, the o7erall system.) ;ne e.ception to the -eneral principle is intonation/ more speci,ically T;I: and T;I*C*TE at the hi-hest+ran2in- unit o, the phonolo-ical system/ the tone -roup. T;I: stands in a natural relationship to the interpersonal -rammatical system o, K:E/ and T;I*C*TE stands in a natural relationship to the te.tual -rammatical system o, *IF;4MAT*;I F;CAS. The natural relationships here are part o, the more -eneral principle that interpersonal ,eatures tend to be reali=ed prosodically and te.tual ,eatures tend to be reali=ed periodically %see Section G. abo7e). There is no comparable correspondence 'ithin the ideational meta,unction0 and all the interpersonal and te.tual -rammatical and le.ical items o, le.ico-rammar are reali=ed con7entionally. %At a more abstract le7el/ 'e can reco-ni=e the 7arious modes o, or-ani=ation 'e ,ind 'ithin the content strata/ iterati7e/

se-mental/ prosodic and periodic/ also 'ithin phonolo-y0 but here they are not directly associated 'ith modes o, meanin-.)

Fi-. BG: :.ample o, interstratal reali=ation %interpersonal meta,unction) 8et us no' return to conte.t. Conte.t is a hi-her+le7el semiotic system in 'hich lan-ua-e is (embedded(. More speci,ically/ lan-ua-e is embedded in a conte.t o, culture or social system and any instantiation o, lan-ua-e as te.t is embedded in its o'n conte.t o, situation. Conte.t is an ecolo-ical matri. ,or both the -eneral system o, lan-ua-e and ,or particular te.ts. *t is reali=ed throu-h lan-ua-e0 and bein- reali=ed throu-h lan-ua-e means that it both creates and is created by lan-ua-e. This reali=ational relationship is or-ani=ed accordin- to the principle o, ,unctional di7ersi,ication. 8i2e lan-ua-e/ conte.t is ,unctionally di7ersi,ied into three -eneral domains: ,ield/ tenor and mode. Field concerns 'hat(s -oin- on/ the social processes and the domains o, sub1ect matter created by lan-ua-e in the reali=ation o, these social processes. Tenor concerns 'ho(s ta2in- part/ the social roles and relations o, those ta2in- part in the interaction and the speech roles and relations created by lan-ua-e in the reali=ation o, these social roles and relations. Mode concerns 'hat role lan-ua-e is playin- in conte.t/ its

distance to those in7ol7ed accordin- to medium %spo2en/ 'ritten and 7arious more comple. cate-ories) and channel %,ace to ,ace/ telephonic/ etc.)/ its complementarity 'ith other social processes %,rom ancillary to constituti7e)/ and its rhetorical contribution %didactic/ instructi7e/ persuasi7e/ and so on). Field tends to be reali=ed by ideational meanin-s/ tenor by interpersonal meanin-s/ and mode by te.tual meanin-s. For e.ample/ the tenor o, the relationship bet'een the interactants in a dialo-ue correlates 'ith the ran-e o, speech+ ,unctional options they select. To illustrate ho' an instance o, lan-ua-e in conte.t is stratally and ,unctionally distributed/ 'e 'ill 1ust pro7ide a ,ra-ment o, a description o, the ,ollo'in- e.ample ,rom a 'eather report: S2ies 'ill be clear to partly cloudy o7er the rest o, Cali,ornia. The description -i7en belo' speci,ies %i) conte.tual ,eatures o, ,ield/ tenor and mode %'ith re,erence to the type o, situation the e.mple belon-s to)/ %ii) semantic ,eatures ,or the e.ample 'ithin each meta,unction %ideational: process con,i-uration/ interpersonal: mo7e and te.tual: messa-e)/ and %iii) le.ico-rammatical ,eatures ,rom -rammatical systems 'ithin the three meta,unctions and the structural strands reali=in- these ,eatures: ,unctional di7ersi,ication: strati,ication: conte.t %o, re-ister: situation type) ,ield+ideational 3issemination throu-h media: daily ne'spaper0 State o, en7ironment: natural: 'eather %present & immediate ,uture) semantics process con,i-uration o, bein-/ ascription o, -raded &uality/ ,uture time tenor+interpersonal mode+te.tual

:.pert to lay audience0 *n,ormati7e: reportin& ,orecastin-0 :.pert: impersonal/ 'ith uncertainty0 6ritten: print0 Audience: -eneral public mo7e o, -i7inin,ormation %statement)/ positi7e and temporal Accompanied by other semiotics: maps etc. messa-e o, unmar2ed in,ormation distribution/ 'ith physical ,eature as point o, departure and place as ne's unmar2ed theme & unmar2ed culmination Theme %s2ies) M 4heme %'ill be clear to partly cloudy o7er the rest o, Cali,ornia)

le.ico-rammar relational: ascripti7e & intensi7e & middle & %at clause ran2) locati7e: place Carrier %s2ies) > ?rocess %'ill be) > Attribute %clear to partly cloudy) > 8ocation %o7er the rest

ma1or: indicati7e: declarati7e: unta--ed & non+interactant & temporal & temporal Mood [Sub1ect %s2ies) M Finite %'ill)# M 4esidue [?redicator %be) M Complement

o, Cali,ornia)

%clear to partly cloudy) M Ad1unct %o7er the rest o, Cali,ornia)#

-rapholo-y

-rapholo-ical ,eatures: e- clause reali=ed as an ortho-raphic sentence/ -rammatical 'ords separated by spaces

Fi-. BH: The meta,unctional di7ersi,ication o, content 6e are thus interpretin- le.ico-rammar not as an isolated system/ but rather as an inte-ral subsystem o, lan-ua-e+in+conte.t. More speci,ically/ le.ico-rammar is the stratum that is internal to lan-ua-e/ inter,acin- 'ith both semantics and phonolo-yC -rapholo-y0 and to-ether 'ith semantics it ,orms the content system o, lan-ua-e. *t shares the meta,unctional or-ani=ation 'ith semantics0 and this meta,unctional or-ani=ation resonates 'ith the di7ersi,ication o, conte.t into ,ield/ tenor and mode. Strati,ication and meta,unction thus to-ether construct the o7erall semiotic space in 'hich the systems o, lan-ua-e in conte.t are distributed. 5ut as 'e ha7e seen 'ith le.ico-rammar/ a stratal subsystem has internal or-ani=ation0 in particular/ le.ico-rammar is or-ani=ed a.ially into systems %ordered in delicacy) and structures/ and it is or-ani=ed by ran2 into a hierarchy o, units: clauses/ -roupsCphrases/ 'ords/ morphemes %in :n-lish). *t is concei7able that a semiotic system such as lan-ua-e 'ould or-ani=e its stratal subsystems internally accordin- to 7ery di,,erent principles/ so that the internal or-ani=ation 'e ha7e discussed so ,ar 'ould be uni&ue to the stratal en7ironment o, le.ico-rammar. Ho'e7er/ the basic principle is 1ust the opposite: all stratal subsystems are or-ani=ed on the same (-rid( by a.is and ran2. That is to say/ all strata ha7e t'o modes o, a.ial or-ani=ation/ the paradi-matic mode/ net'or2s o, systems such as

'e ha7e met in -rammar/ and the synta-matic mode/ structural con,i-urations o, ,unctions0 and these system+structure cycles are distributed accordin- to a ran2ed hierarchy o, units. 6e can thus interpret a.is and ran2 as -eneral intra+stratal principles o, or-ani=ation that are mani,ested in the di,,erent stratal en7ironments o, lan-ua-e in conte.t: see Fi-ure BJ.

Fi-. BJ: A.is and ran2 as principles o, intra+stratal or-ani=ation mani,ested in the di,,erent stratal subsystems o, lan-ua-e in conte.t H.B 8ocatin- systemic theory in the 'ider 'orld Amon- current theories o, -rammar/ systemic -rammatics can be located 'ithin a broadly de,ined class o, (,unctional( -rammars that are typically characteri=ed by certain orientations: it is oriented to'ards ,unction rhetoric rather than 9 ,orms lo-ic

te.t resource meanin-,ulness

9 9 9

sentences rules -rammaticality

6ithin this -eneral class/ 'e ha7e seeen that systemics has certain particular characteristics. %i) *t is paradi-matic/ ta2in- the system %theory o, possible and probable) as its base rather than structure %theory o, composition). That is/ the paradi-matic a.is is the o7erall or-ani=inprinciple rather than the synta-matic a.is. %ii) *t is strati,ied/ 'ith the strata related by reali=ation %9metaredundancy9/ not timeCcause) and e.tended beyond lan-ua-e to the conte.t o, situation and o, culture %e.-. 7ia -enre and ideolo-y). %iii) *t is comprehensi7e/ 7aryin- in delicacy and in instantiation/ these bein- de,ined by re,erence to the o7erall system o, a lan-ua-e. %i7) *t is multi,unctional/ assi-nin- e&ual 7alue to interpersonal and te.tual as 'ell as to ideational modes o, meanin- %at both the semantic and le.ico-rammatical strata)0 and representin- these modes o, meanin- as simultaneous both in system and in structure. Ho' does our 9-rammatics9 relate to lan-ua-e itsel,D 8et us set up a simple typolo-y o, systems %'here system/ as al'ays/ stands ,or (system+&+process(): % ) %B) %G) %H) physical 9 9 9 systems > li,e N systems biolo-ical 9 9 > 7alue N social 9 systems > N systems meanin- semiotic

A lan-ua-e is ,ourth+order system/ to be in7esti-ated % ) acoustically/ %B) neuro+ physiolo-ically/ %G) culturally/ as 'ell as %H) le.ico-rammatically %or rather/ %H) in terms o, the (core( strata o, semantics/ le.ico-rammar and phonolo-y). A theory o, any domain is also a semiotic system/ one o, a dedicated 2ind: part e7ol7in-/ part desi-ned as a means o, re,lection and o, action. A theory o, lan-ua-e is thus a dedicated meta+semiotic. The prototypical semiotic system is a natural lan-ua-e: human (consciousness( can be de,ined as the ability to mean. At the same time/ a natural lan-ua-e is conte.tuali=ed amon- other semiotic systems e7erythin- that 'e denote by 9culture9: 7isual and other art ,orms0 patterns o, beha7iour/ secular and reli-ious0 social institutions/ modes o, sel,+presentation and so on. 6e can use our -rammatics to interpret these 7arious non+lin-uistic semiotic systems/ as2in- to 'hat de-ree o, speci,icity they are li2e natural lan-ua-e. *n particular/ to 'hat e.tent does our understandin- o, the problematic relation bet'een system and instance in lan-ua-e apply to semiotic systems as a 'holeD And to 'hat e.tent does lan-ua-e ,unction as a 9connotati7e semiotic9 throu-h 'hich other semiotic systems are presented %or 9reali=ed9)D 8an-ua-e does not passi7ely 9re,lect9 or 9correspond to9 some pre+e.istin- reality. 8an-ua-e constructs reality0 or rather/ 'e/ as human bein-s/ construct reality in lan-ua-e. 6e do this throu-h the meta,unctional interplay o, action and re,lection: lan-ua-e both enacts interpersonal relationships and construes human e.perience. Thus the %spea2in-) sub1ect/ the multi,aceted personae/ the hierarchies and po'er structures that 'e call society are all created

in lan-ua-e. *deolo-ies o, class/ -ender/ and the li2e are established and maintained/ and also challen-ed/ throu-h the meanin- potential o, lan-ua-e. Humans e.chan-e -oods+&+ser7ices and they also e.chan-e in,ormation: human history is an on-oin- dialectic o, material and semiotic processes. %6e can 'atch this pattern emer-in- in the ,irst ,e' months o, each indi7idual(s li,e.) *n the scienti,ic/ 9modern9 a-e all systems 'ere modelled in material terms/ 'ith physical systems ta2en as prototypical. *n the 9postmodern9 in,ormation society 'e are increasin-ly usin- semiotic models/ interpretin- e7en physical systems in terms o, an e.chan-e o, meanin-s. Wuantum physics/ in particular/ has 7eered us in that direction. This puts -rammatics s&uarely in the centre o, the sta-e0 not 1ust as theory o, -rammar but as theory o, 2no'led-e/ that is/ o, all systems construed as systems o, meanin-. *n this settin-/ -rammatics means usin- -rammar to thin2 'ith: its conte.t o, operation is as a theory 'hereby our understandin- o, lan-ua-e may be brou-ht to bear on phenomena o, any 2ind. Ieither systemic theory nor any other theory o, lan-ua-e has reached any'here near this le7el o, achie7ement. 6e are still rather ,ar ,rom understandin- e7en the basic properties o, a semiotic system. 6hat is important at this sta-e is to locate theories o, -rammar in the current intellectual conte.t/ so that as they continue to be elaborated and impro7ed they mo7e closer to/ and not ,urther a'ay ,rom/ the transdisciplinary concerns o, thou-ht,ul people as a 'hole. Appendi. : Glossary . Cate-ory. A construct or abstraction in systemic theory0 units/ ,unctions/ classes/ and so on are cate-ories o, the theory o, -rammar0 it contrasts 'ith (scale( %c,. Halliday/ !K ). B. Class. The systemic term ,or the term cate-ory in ,ormal -rammar. *t -enerali=es the traditional notion o, 'ord classes and thus applies to morphemes/ -roups/ phrases/ and clauses as 'ell as 'ords. The least delicate classes are sometimes called primary classes and ,urther di,,erentiations are secondary classes %c,. Halliday/ !K 0 !KG). G. Cohesion. The te.tual le.ico-rammatical resources ,or e.pressin- relations 'ithin te.t 'ithout creatin- -rammatical structure. The cohesi7e resources include re,erence/ substitutionCellipsis/ con1unction/ and le.ical cohesion. The term cohesion is also used in non+systemic literature/ sometimes in direct re,erence to systemic 'or2 on cohesion %particularly/ Halliday & Hasan/ !"K)/ sometimes more loosely to re,er to the te.t+ness o, a te.t. %The term has a di,,erent use in Ta-memics0 it re,ers to one o, the ,our cells in a ,our+cell ta-meme.) H. Comple.. Comple. o, -rammatical units o, any ran2 or class/ potentially lineally recursi7e0 comple.es include the traditionally reco-ni=ed cate-ories o, coordination %e.tendin- comple.es) and apposition %elaboratin- comple.es). J. Conte.t. Conte.t o, culture0 conte.t o, situation. Hi-her+order semiotic systems abo7e the lin-uistic system. Conte.t spans ,ield/ tenor/ and mode. %*n earlier 'ritin-s/ conte.t 'as used ,or 'hat is no' called semantics.) The term conte.t is also used 'idely in non+systemic literature/ sometimes in the systemic sense sometimes not. Frames/ schemata/ and scripts 'ithin co-niti7e psycholo-y and A* are similar to situation and situation types in many respects. The notions o, conte.t o, situation and conte.t o, culture ori-inate 'ith 5ronisla' Malino's2i/ an anthropolo-ist 'or2in- in

the ,irst hal, o, this century. 3oin- ,ield 'or2 in the Trobriand *slands/ he came to reco-ni=e and ar-ue ,or the importance o, conte.t in the interpretation o, te.t. His 'or2 on conte.t 'as ,urther de7eloped 'ithin lin-uistics ,irst by Firth and then by Halliday and others. Halliday & Hasan % !QJ)0 Martin % !!B: Ch. "). K. 3elicacy. The scale ,rom -eneral to speci,ic. *n a system net'or2/ delicacy corresponds to the orderin- o, systems ,rom le,t to ri-ht by means o, entry conditions. For e.ample/ the ,ollo'in- systems o, M;;3 increase in delicacy ,rom le,t to ri-ht:

". Feature. The label o, a term in a system0 it can be semantic/ le.ico+-rammatical/ or phonolo-ical. For instance/ in the system (indicati7eCimperati7e(/ there are t'o terms/ the ,eatures (indicati7e( and (imperati7e(. Feature is also used 'idely in the non+ systemic literature/ 'here it does not entail systemici=ation in a system. *t is used &uite e.tensi7ely in phonolo-y and le.ical semantics but also %increasin-ly) in -rammar/ in particular in Generali=ed ?hrase Structure Grammar and 8e.ical Functional Grammar. The term component is also used %as in componential analysis). Q. Function. Common term both in systemic and non+systemic lin-uistics. *n systemic lin-uistics/ there are three terms ,or particular types o, ,unction. %i) micro+,unction: ,unctionally de,ined constituent0 e.-. Sub1ect/ Actor/ Theme. %ii) macro+,unction: lan-ua-e use in early child+lan-ua-e/ be,ore use and meta,unction ha7e become di,,erentiated. %iii) meta,unction: -enerali=ed ,unctional principle o, lin-uistic or-ani=ation. There are three meta,unctions/ ideational %e.periential > lo-ical)/ interpersonal/ and te.tual. %Iote that there is a special use o, the term ,unction in mathematics and ,ormal semantics: such a ,unction ta2es an ar-ument and returns a 7alue.) !. Grammar. The term has the traditional sense in systemic theory. That is/ it includes synta. as 'ell as morpholo-y/ the t'o simply ha7in- di,,erent domains on the -rammatical ran2 scale. Grammar is ta2en to be the most -eneral part o, le.ico-rammar/ the resource ,or e.pressin- meanin-s. The other part o, le.ico-rammar is le.is %7ocabulary). L. Grammatics. Systemic term ,or -rammatical theory/ sometimes used to a7oid the potential ambi-uity bet'een -rammar in the sense o, -rammatical theory %as in 9Functional Grammar9) and -rammar as the phenomenon under study %as in 9the -rammar o, Hopi9). . *nstantiation.The cline bet'een the o7erall systemic potential o, lan-ua-e and the te.t %instance o, the potential). 8yin- alon- the cline o, instantiation intermediate bet'een these t'o endpoints are 7arieties o, re-ister and code. At the hi-her+le7el system o, conte.t/ the o7erall systemic potential is associated 'ith conte.t o, culture/ re-isters 'ith situation types and te.ts 'ith situations. *nstantiation also re,ers to the process o, mo7in- bet'een potential and instance/ the process o, actuali=in- the system in te.t.

B. Meta,unction. The hi-hly -enerali=ed ,unctions lan-ua-e has e7ol7ed to ser7e and 'hich are e7idenced in its or-ani=ation %and are thus intrinsic to lan-ua-e). Halliday % !K"CQ) identi,ies three meta,unctions/ the ideational/ the interpersonal/ and the te.tual. The ideational meta,unction can be ,urther di,,erentiated into the e.periential and the lo-ical subtypes. Meta,unctions are distin-uished ,rom macro,unctions and micro,unctions. Macro,unctions can be identi,ied in a child(s transition bet'een hisCher protolan-ua-e and adult lan-ua-e %c,. Halliday/ !"J)0 micro,unctions are the ,irst ,unctionsCuses o, a child(s protolan-ua-e. *n other lin-uistic theories/ ideational -rammar is o,ten treated as part o, semantics/ and re-arded as a le.ical rather than as a -rammatical phenomenon0 'hile te.tual and interpersonal -rammar tend to be assi-ned to the headin- o, pra-matics. *n systemic theory/ all three meta,unctions are located both at the le7el o, semantics and at the le7el o, -rammar: ,or e.ample/ transiti7ity is analysed as a -rammatical system/ 'hich is then interpreted in more e.planatory terms at the le7el o, semantics. The meta,unctions are summari=ed in the table belo'. G. Iet'or2. A relational type o, or-ani=ation0 a -raph. :.amples include discrimination net'or2s/ the net'or2s o, strati,icational theory/ and system net'or2s. *n systemic theory/ a net'or2 is speci,ically a system net'or2. *n this sense/ a net'or2 is an assembly o, systems ha7in- the same point o, ori-in %-eneral entry condition) such that each system is associated 'ith all the others by some combination o, simultaneity and dependency in delicacy. H. 4an2 %scale). A hierarchy o, units such as clause/ -roupCphrase/ 'ord/ morpheme or tone -roup/ ,oot/ syllable/ phoneme. The ran2 scale re,lects the basic reali=ation patterns. Functions o, the units at one ran2 are reali=ed by units at the ran2 belo'. For e.ample/ clause ,unctions are reali=ed by -roupsCphrases and -roup ,unctions are reali=ed by 'ords. *n non+systemic 'or2/ the term le7el is sometimes used. %The term ran2 'as used in a di,,erent sense in <espersen(s 'ritin-s.) J. 4an2shi,t. The situation 'here a unit o, one ran2 ser7es in the structure o, another unit as i, it 'ere a unit o, a lo'er ran2/ as in a de,inin- relati7e clause ser7in- as a ?ostmodi,ier in the structure o, a nominal -roup. K. 4eali=ation. Term in lin-uistics in -eneral ,or a symbolic relationship bet'een content and e.pression0 also e.pression/ codin-/ etc.. 4eali=ation and mutation ha7e been contrasted %c,. Gleason/ !KJ) as basic principles underlyin- -rammatical theories. Systemic -rammar is reali=ational 'hereas trans,ormational -rammar is mutational. ". 4eali=ation operator. To-ether 'ith one or more operands/ a reali=ation operator ma2es up a reali=ation statement. 4eali=ation operators are *nsert/ Con,late/ :.pand/ ;rder and ?reselect. See reali=ation statement. Q. 4eali=ation statement. A speci,ication o, a structure ,ra-ment/ such as the presence o, a ,unction or its orderin- in relation to another ,unction/ stated as a ree.pression o, a systemic ,eature or a combination o, ,eatures. A reali=ation statement consists o, one reali=ation operator and one or more operands. For e.ample/ the statement %Con,late Sub1ect A-ent) consists o, the con,lation operator Con,late and the operands Sub1ect and A-ent/ 'hich are -rammatical ,unctions. !. Stratum. A subsystem o, a particular order o, symbolic abstraction in lan-ua-e: semantics/ le.ico-rammar/ and phonolo-y are the three strata o, systemic theory.

Strata are related throu-h %inter+stratal) reali=ation0 ,or instance/ semantics is reali=ed throu-h le.ico-rammar. The earlier term in systemic lin-uistics %ta2en o7er ,rom Firth) 'as le7el %as in Firth(s le7els o, analysis)0 since le7el 'as used in other senses in non+systemic lin-uistics/ the e&ui7alent term stratum 'as ta2en o7er ,rom strati,icational lin-uistics. *n 4elational Grammar/ stratum has a di,,erent use/ more li2e the layer in a ,unction structure.

BL. Structure. A ,unction structure %or structure ,or short) is made up o, a con,i-uration o, -rammatical ,unctions such as Actor/ Sub1ect/ and Theme. :ach ,unction may be reali=ed by either a set o, -rammatical ,eatures or a set o, le.ical ,eatures. The -rammatical ,eature set constitutes a preselection o, ,eatures that ha7e to be chosen 'hen the -rammar is reentered to de7elop a ,unction ,urther. For e.ample/ the ,unction Actor may ha7e the associated preselection (nominal -roup(/ 'hich means that once the structure o, the clause o, 'hich Actor is a constituent has been ,ully de,ined/ the -rammar is reentered and Actor is de7eloped as a nominal -roup. The term ,unction structure is used inside and outside systemic lin-uistics. *t al'ays re,ers to a con,i-uration o, ,unctions/ but in certain non+systemic theories there may be only one ,unctional layer. *n systemic theory/ ,unction structure is contrasted 'ith synta-m

%Halliday/ !KK). *n 8e.ical Functional Grammar/ there is a similar contrast bet'een ,unction structure %,+structure) and constituent structure %c+structure). B . System. A system is the central cate-ory ,or representin- paradi-matic or-ani=ation at any stratum/ phonolo-ical/ -rammatical/ or semantic. *t consists o, %i) a statement o, a choice bet'een t'o or more terms/ represented by ,eatures/ %ii) and an entry condition/ 'hich speci,ies 'hen the choice is a7ailable. The entry condition is a simple ,eature or a ,eature comple.0 these ,eatures are terms in other systems. 5ecause o, their entry conditions/ systems ,orm system net'or2s. :ach term in a system may ha7e one or more reali=ation statements associated 'ith it. %The reali=ation statements speci,y structure ,ra-ments0 ,rom their 7anta-e point/ the system is li2e a (metarule(.) :.ample:

Appendi. B: Systemic con7entions [i# System net'or2

[ii# 4eali=ation statements A reali=ation statement consists o, an operator/ such as insert or con,late/ and one or more operands/ at least one o, 'hich is a -rammatical ,unction.

. ?resence o, ,unctions in the structure: the presence o, a ,unction in a ,unction structure is speci,ied by insertin- the ,unction into the structure0 the operation o, insertion is symboli=ed by (>(0 e.-. >Sub1ect/ >Mood/ etc. B. Functional constituency relations: t'o ,unctions may be related by constituency and to speci,y this constituency relationship in the ,unction structure one ,unction is e.panded by the other0 the e.pansion is symboli=ed by puttin- the e.pandinconstituent ,unction 'ithin parenthesis/ e.-. Mood %Sub1ect)/ 'hich means that Mood is e.panded to ha7e Sub1ect as a constituent ,unction. A ,unction may be e.panded by more than one other ,unction/ e.-. Mood %Sub1ect/ Finite). G. 4elati7e orderin- o, ,unctions and orderin- relati7e to unit boundaries: t'o ,unctions may be ordered relati7e to one another in the ,unction structure and this relati7e orderin- is symboli=ed by (M (0 e.-. Sub1ect M Finite/ Mood M 4esidue. The orderinmay also be relati7e to the le,t or ri-ht boundary o, a -rammatical unit %represented by O)/ e.-. O M Theme and Moodta- M O. H. Con,lation o, one ,unction 'ith another: one ,unction ,rom one perspecti7e is con,lated 'ith a ,unction ,rom another perspecti7e/ i.e. the t'o ,unctions are speci,ied as di,,erent layers o, the same constituent/ they are identi,ied 'ith one another. Con,lation is symboli=ed by (C(0 ,or e.ample/ Sub1ectCA-ent means that Sub1ect %interpersonal) and A-ent %ideational) apply to the same constituent. J. 4eali=ation o, a ,unction in terms o, ,eatures ,rom the ran2 belo': the reali=ation o, a ,unction in a ,unction structure is stated by preselectin- one or more ,eatures ,rom the unit reali=in- it0 preselection is symboli=ed by (:(/ e.-. Sub1ect: nominal -roup/ Finite & ?redicator: 7erbal -roup/ etc. Further readin5elo' 'e list a number o, boo2s and articles ha7in- to do 'ith systemic theory and description. These are or-ani=ed under ,our headin-s: . 3iscourse analysis A -eneral account o, :n-lish discourse ,rom a le.ico-rammatical and semantic point o, 7ie' is pro7ided by Martin % !!B). Hasan % !QJ) is concerned particularly 'ith poetic and other literary modes o, discourse. Halliday % !QJ) is a summary presentation o, the -rammatical systems that are li2ely to be ,ore-rounded in discourse o, all 7arieties0 Mann & Thompson % !!B) includes a number o, articles describin- a particular te.t ,rom the standpoint o, systemic and other ,unctional theories. Appendi. in Halliday % !!H) -i7es a clause by clause analysis o, a short passa-e o, spontaneous :n-lish speech. For the analysis o, in,ormal speech/ see :--ins & Slade % !!"). B. 8e.ico-rammatical descriptions There are t'o -eneral descriptions o, the -rammar o, :n-lish in systemic+,unctional terms: Halliday % !!H) presents the -rammar ,rom the structural an-le/ 'hile Matthiessen % !!J) presents it in the ,orm o, systems and system net'or2s. 5utt/ Fahey/ Spin2s & Eallop % !!J)/ Martin/ Matthiessen & ?ainter % !!") and Thompson % !!K) present the -rammar in te.t boo2 ,orm. :.amples o, accounts o, speci,ic portions o, the -rammar are: %on transiti7ity) 3a7idse % !!B/ !!K)/ Fa'cett % !Q")0 %on theme) Collins % !! )/ Fries % !!J)/ and the

7arious papers in Ghadessy % !!K)0 %on the clause comple.) Matthiessen & Thompson % !QQ)/ Iesbitt & ?lum % !QQ)0 %on intonation and -rammar) :lmenou,y % !QQ)/ Halliday % !K")0 %on tense) Matthiessen % !!K). A number o, di,,erent topics are dealt 'ith in 5erry/ 5utler/ Fa'cett & Huan- % !!K). For -rammatical descriptions o, other lan-ua-es in a systemic ,rame'or2/ see amon- others: %Chinese) Fan-/ Mc3onald & Chen- % !!J)/ Mc3onald % !!H)0 %Finnish) Shore % !!K)0 %French) Ca,,arel % !!B/ !!J)0 German %Steiner & 4amm/ in press)0 %Gooniyandi) McGre-ory % !!L)0 %<apanese) Hori % !!J)0 %?it1ant1at1ara) 4ose % !!K)0 %Ta-alo-) Martin % !!L/ !!K). <ohnston % !!B) pro7ides a meta,unctional interpretation o, the -rammar o, Auslan %Australian Si-n 8an-ua-e o, the dea,). Halliday % !!B) is a systemic analysis o, the syllable in Mandarin Chinese. Further papers on aspects o, Chinese -rammar/ mainly 'ritten in Chinese/ 'ill be ,ound in Hu % !!L) and `hu % !!G). G. Applications Systemic methods ha7e been deployed in 7arious ,ields o, application/ such as natural lan-ua-e processin-/ lan-ua-e education and child lan-ua-e de7elopment. *n natural lan-ua-e processin-/ Matthiessen & 5ateman % !! ) describe te.t+-eneration research in :n-lish and <apanese0 Fa'cett % !QQ) and Cross % !!B) e.plore di,,erent aspects o, le.ico-rammar in computational ,orm. *n the ,ield o, lan-ua-e education/ the papers in Hasan & Martin % !Q!) pro7ide a representati7e co7era-e o, the central issues/ 'hile Halliday & Martin % !!G) e.plores a -rammar+based approach to educational theory and practice. Genres as social processes are e.plored in Christie & Martin % !!")/ 'hich also includes e.tensi7e analysis o, te.ts. Hasan & ?errett % !!H) o,,ers a systemic an-le on second+lan-ua-e learninand teachin-. Systemic studies o, lan-ua-e de7elopment in early childhood include Halliday % !"J)/ ?ainter % !QH/ !!K0 see also ?ainter(s chapter in Hasan & Martin/ !Q!). For the application o, systemic theory in the study o, semiotic systems other than lan-ua-e: %art/ architecture and sculpture) ;(Toole % !!H)/ %music) Steiner % !QQ)/ %7isual ima-es) Kress & 7an 8eeu'en % !!K). H. Theory For a more comprehensi7e treatment o, systemic theory relatin- to the present account/ see Matthiessen & Halliday %,orthcomin-). Halliday % !"K/ !"Q) are selections o, earlier theoretical papers on lan-ua-e and its relation to social processes. Fa'cett % !QL) pro7ides a detailed account o, the or-ani=ation o, a systemic -rammar 'ithin a broadly co-niti7e perspecti7e. 3i,,erent aspects o, the theory are e.plored in the 7arious chapters o, 3a7ies & 4a7elli % !!B)0 Halliday & <ames % !!G) e.plores a probabilistic approach to -rammar based on &uantiti7e data dra'n ,rom a lar-e+scale corpus. Martin % !!B) presents a systemic+ ,unctional account o, -rammar in its relation to discourse semantics. The ideational ,unction o, -rammar as a theory o, e.perience is in7esti-ated in Halliday & Matthiessen %,orthcomin-). Hasan/ Cloran & 5utt % !!K) includes both discussion o, systemic theory and theory+based descriptions o, transiti7ity in 7arious lan-ua-es. Hu et al % !Q!) presents a comprehensi7e account o, systemic ,unctional theory 'ritten in Chinese. For a discussion o, lan-ua-e in relation to a -eneral theory o, systems/ see 8em2e % !!G).

Anda mungkin juga menyukai