Enclitic Rhythms in the Vergilian Hexameter

Author(s): Peter L. Smith

Source: Phoenix, Vol. 36, No. 2 (Summer, 1982), pp. 124-143
Published by: Classical Association of Canada
Stable URL: http://www.jstor.org/stable/1087672
Accessed: 02-11-2015 13:12 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/1087672?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Classical Association of Canada is collaborating with JSTOR to digitize, preserve and extend access to Phoenix.

http://www.jstor.org

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC RHYTHMS IN THE VERGILIAN HEXAMETER

PETER L. SMITH

PERSISTENTLY TROUBLESOME QUESTION IN LATIN PHONOLOGY has been

the accentuationof encliticcomposites.'One consequenceof thisproblem

is the uncertaintyit may cause in the pronunciationof classical dactylic
hexameter,where the separable enclitics (-que, -ve, and -ne) are to be
foundin abundance. The presentstudy attemptsto review and clarify
the areas ofphoneticcontroversythrougha metricalanalysisof all enclitic
rhythmsin the poetryof Vergil.Particularattentionwill be given to the
putative regressionof accent that may occur when an encliticis followed
by elision.
Since late antiquity,therehas always been some theoreticalconfusion
about the precisenatureof the Latin accent shiftcaused by the addition
of these commonencliticparticles.Most modernscholars readily agree
that, in the classical period at least, all finalheavy syllables must have
receiveda syntacticalaccent beforethe enclitic,regardlessof the isolate
accentuationof the base-word: multtumque,
multosque,multaeque,imperietc.2This principlecould be inferredfromthemetrical
umque,imperi6que,
evidence of Augustan Latin literatureeven if it were not confirmedby
Roman grammatical theorists.3The elements of uncertaintyoccur in
three main areas: (1) the accentuationof encliticcompositesin which
the base-wordends in a lightsyllable (multaque,limindque,armentdque,
Saturn'dque,etc.); (2) the potentialtendencyforelisionto cause a regression of accentwithinthe encliticcomposite;and (3) the possibledynamic
effectof theverse-ictusupon the accent of the composite.
The firstproblemof the trochaicand dactylic base-wordshas arisen
because of conflictbetweenthe testimonyof some imperialRoman grammarians, who appear to have advocated the accentuation multdque/
'The subject is reviewed with precisionand clarity by W. Sidney Allen in Accentand
Rhythm(Cambridge 1973) 158-161; see also Allen's summarycommentsin Vox Latina
(Cambridge 1965) 87-88. Important linguistic issues are considered by R. Whitney
Tucker, "Accentuation Before Enclitics in Latin," T,APA 96 (1965) 449-461. These
workswill be cited in a shortenedform.I am prepared to challenge Tucker's overstated
conclusion, "that in the second century B.C. at least, the place of the word-accentwas
unaffectedby the addition of an enclitic" (460); but he does offerpersuasive evidence
that substantial phonetic development must have occurred in the centuries between
Plautus and Seneca.
2This application of the term "syntactical" is taken from W. S. Allen, Accent and
Rhythm158; the terminologywill remind us that, in actual Latin usage, enclitic composites were accented only withina phrase or sentence(cf. Allen 25).
3For the most comprehensive collection and analysis of ancient theories on Latin
accent, see Friedrich Schoell, De accentu linguae latinae veterumgrammaticorumtestimonia (Acta societatisphilologaeLipsiensis 6; Leipzig 1876).
124
PHOENIX,Vol. 36 (1982) 2.
This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC

RHYTHMS

125

limlndque,4 and the evidence of classical Latin poetry,which seems to

supportthe accentuationmultdque/llzmndque.This puzzle has attracted
its share of scholarlyattention.The early years of the twentiethcentury
witnessed a flurryof activity,followinga trail-blazingarticle by Carl
Wagener.5A most noteworthycontributionwas made by the American
F. W. Shipley, who approached the question from the evidence of
Republican Latin proseclausulae.6Shipley's argumentspointclearlyand
logicallyto the conclusionthat Cicero accentedencliticcompositesby the
normal principlesof the penultimatelaw (i.e., multu'mque,
mult6sque,
multaeque;but muiltque,rather than multaque). He made the further
importantdiscoverythat Cicero normallyexcluded dactylic composites
such as crimlndqueor testimonidqueunless an accent shift could be
avoided by one of three phonetic expedients: elision (cr1imnaqu[e]),
or syncope(cet[e]raqu,peric[u]laque).
synizesis(omnjdque,testim6njdque),
No subsequentresearcherhas successfullychallengedthese conclusions.7
have
Accordingly,English-speakingclassicistsof the last half-century
been generallyprepared to discount the testimonyof the late Roman
grammarianson thisaspectoftheproblem.The modernlinguistictreatises
in English offerbasically consistentguidance, despite some areas of uncertainty(e.g., Kent favoursliminaque,whereas Allen admits the possibilityof liminaque);8 and accentuationssuch as magnaque have been
advocated by influentialliteraryscholars.9This is not to say that the
mistsof confusionhave been entirelydispelled.The revisedprinciplehas
40n this question, the ancient grammarians'testimonyis neitherclear nor consistent;
cf. Tucker 451.
5"Betonung der mit que, ve, ne zusammengesetztenW6rterim Lateinischen," Neue
PhilologischeRundschau 1 (1904) 505-511. See also H. J. Edmiston, "The Question of
the Coincidence of Word-accentand Verse-ictusin the Latin Hexameter," CR 17 (1903)
458-460; Charles B. Newcomer, "The Effect of Enclitics on the Accent of Words in
Latin," TAPA 37 (1906) xxvii-xxviii.
6Shipleypublished several studies on this subject between 1909 and 1913; his research
culminated in "Preferredand avoided combinations of the enclitic que in Cicero (considered in relation to questions of accent and prose rhythm)," CP 8 (1913) 23-47.
7Years later, Shipley's work was admired and accepted by Ernst Kalinka, in his
definitivesurvey of metricalscholarship for Bursians 7ahresbericht
256 (1937) 87: "Ich
gestehe, dass diese AusfiihrungenShipleys mich iiberzeugt haben." It is interestingto
note that most recentcomputeranalyses of hexameterrhythmhave accented all enclitic
composites by the inflexibleapplication of the penultimate law, ignoringthe possible
effectof elision. This is the method followed by Wilhelm Ott in his excellent series,
Materialen zu Metrikund Stilistik(Tiibingen 1973-). The principleis sensibly defended
by Nathan A. Greenberg in "Metrical Shape, Initial Stress, and Crosstabulation,"
Revue [InternationalOrganization forAncient Languages Analysis by Computer] (1978,
No. 3) 10-11.
8Roland G. Kent, The Sounds of Latin3 (Baltimore 1945) 68; W. S. Allen, Vox Latina
87-88, Accentand Rhythm159.
9For example, L. P. Wilkinson, "The Augustan Rules for Dactylic Verse," C2 34
(1940) 30, n. 3.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

126

PHOENIX

had only partial acceptance in France, where therehas always been resistance to English and German theoriesof Latin accent.'0Elsewhereon
the European continent,the time-honouredconservativedoctrineis still
at times proclaimed,perhaps under the continuingauthorityof the old
German handbooks."
The second problem is the possible effectof elision (synaloephe)on
Latin word-accent.In addition to Shipley's articles,a numberof studies
have pointed to elision as a factorthat may controlor modifyenclitic
accent and usage.'2 In a wider context,an importantschool of modern
scholarship,typifiedby Eduard Fraenkel,13revived the old theorythat
all elided wordswill undergoa regressionof word-accentthroughthe loss
of the final syllable. The weight of recent researchhas effectively
discredited this once-populardoctrine,as can be seen fromthe work of
Jean Soubiran and W. S. Allen.14Nonetheless,even the most scepticalof
modernscholarsappear preparedto concedethatencliticelisionmay cause
a regressionfromthe syntactical accent of the composite (colorimque)
to the isolate accent of the base-word(col6remqu[e]).
The thirdand interrelatedprobleminvolvesthequestionofverse-ictus,
a topic thathas always bristledwithcontroversy.15
If theLatin hexameter
'OThe conservativepositionis upheld by Max Niedermann,Pricis de phon6tique
du latin3(Paris 1953) 14-15. Referenceis made below (note 35) to Jean
historique
Soubiran'sexaminationof trochaiccomposites.Some French-speaking
scholarsseem
moreinclinedto doubtthe ancientgrammarians
on thissubject;see J. Hellegouarc'h,
Le monosyllabe
dansl'hexamrtre
latin(Paris 1964)266 and n. 1; E. Lienard,"Reflexions
surl'accentlatin,"Hommages
a MarcelRenard(Collection
Latomus101; Brussels1969)
554 and n. 3.
"For example,G. BernardiPerini,L'accentolatino2(Bologna1967)38-43. Even after
thestandardGermanhandbooksstilldefertotheimperialgrammarians;
modernrevision,
Grundriss
dervergleichenden
derindogermanischen
Grammatik
e.g., Brugmann-Delbriick,
Handbuchder
Sprachen2(Berlin-Leipzig1967 [1930]) 1.975-976; Sommer-Pfister,
lateinischen
Laut- und Formenlehre4
(Heidelberg1977) 1.217-218;withsome reservaLaut- und Formenlehre'
tions,Manu Leumann,Lateinische
(HandbuchderAltertums2.2.1; Munich1977)240.
wissenschaft
"G. Eskuche,"Die Elisionenin den 2 letztenFussendes lateinischen
Hexameters,
vonEnniusbis Walahfridus
Strabo,"RhM 45 (1890) 236-264,385-418;W. M. Lindsay,
"The SaturnianMetre.II," AJP 14 (1893) 313; W. M. Lindsay,EarlyLatin Verse
Maro AeneisBuchVI4 (Stuttgart1957),
(Oxford1922)34-35; E. Norden,P. Vergilius
AnhangXI.9, 456; R. D. Williams,"The Effectof elided-queon WordAccentin the
Hexameter,"PACA 47 (1950) 31; Nils-OlaNilsson,"EncliticanachzweikurzenSilben
im Latein,"Eranos52 (1954) 195-216.See also L. P. Wilkinson,
GoldenLatin Artistry
(Cambridge1963) 233, 235, 236, and D. S. Raven, Latin Metre(London 1965) 32.
"Iktus und Akzentim lateinischen
(Berlin1928) 268-269.
Sprechvers
14JeanSoubiran,L'Elision dans la po6sielatine(Paris 1966) 457-480; W. S. Allen,
AccentandRhythm
159-161.
"6Thescopeof thispaperdoes notpermitme to reviewthefundamental
controversy
overLatinaccentand ictus,whichhas extendedfromthetimeofBentleyto thepresent
it to say that I findmyselfin close agreement
withthe positionof W. S.
day. Suffice

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC RHYTHMS

127

has an audiblepulse-beat,as manybelieve,can thataural expectation

influence
theaccentuation
ofencliticcomposites?
May it possiblybe the
actual cause of accentregression
beforeencliticelision?The question
assumesparticularrelevancein any enquiryon fourth-foot
a
harmony,
circumstance
overlookedby W. F. JacksonKnightin hispioneerstudy
of Vergilianhomodyne.16
A strongcase can be made forfourth-foot
ofictusand accentin suchVergilianversesas these(which
coincidence
are markedto showhomodyne17):
6mnesint.ntique
Aen.2.1
oratenebant
Canticueere
6bruepuppis den. 1.69
fncute
ufmuentissubmersasque
In an attemptto resolvetheseuncertainties,
insofaras theypertainto
I have undertaken
a metricalclassification
of all
Augustanhexameter,
encliticrhythms
in theentireVergiliancorpus,excludingtheAppendix
thehazardsofreaching
Vergiliana.
ThoughI acknowledge
objectiveconclusionson encliticaccentfromtheevidenceofhexameter
poetry,I stop
shortofacceptingR. Whitney
Tucker'sgloomypronouncement:
"Little,
ifanything,
can be learnedon thissubjectfromtheformalverseof the
GoldenAge."18
It is obviouslytrue,as Tuckerinsists,thatwecannotrely
on ourownphoneticintuition
in thisarea: it is hardlyadequateto state
thata poetusessuch-and-such
an accentuation
because"it soundsright
to me." Moreover,thosewho tryto proveaccentualtheoriesfromthe
evidenceof classicalLatin hexameter
mustalwaysremember
that the
of
factors
and
Augustanpoetswereconstrained
by
metre,syntax, style
in VoxLatina83-94,and developedfurther
in AccentandRhythm,
Allen,as summarized
GoldenLatin Artistry
passim.See also L. P. Wilkinson,
(above,n.12) 118-132and 221236. I have not yet had access to the 1980 University
of TorontoPh.D. thesisby S.

Traverse, entitled Ictus Metricus: Phonological, Historical and ComparativeStudies in

Greekand Latin Metrics.

16Accentual
in Vergil(Oxford1939). Knight'slack of precisionin defining
Symmetry
his own rulesforword-accent
and the problemof elided encliticsare bothnotedby
NathanA. Greenberg
in "Vergiland the Computer:Fourthfoottexturein Aeneid1,"
Revue[International
forAncientLanguagesAnalysisby Computer]
Organization
(1967,
No. 1) 1-16. See also E. D. Kollmann,"Remarkson the Structure
of the Latin Hexa46 (1968)301-302.
meter,"Glotta
17I use the convention
of markingthe hexameter
ictus by a subscriptdot, the main
word-accent
by an acute, and the putativesecondaryaccent by a grave; cf. L. P.
GoldenLatin Artistry
Wilkinson,
(above, n. 12). AveryWoodward,one of the earliest
studentsof fourth-foot
ora tenebant
texture,describedthe cadence intentique
as "a
threefold
blow"(referring
to ictus-accent
explicitly
harmony);see "The FourthFoot in
Vergil,"P. 15 (1936) 129. Thoughwritingin fullawarenessand approvalof Miss
Woodward'spaper,Knightcouldactuallyquotesubmersasque
obruepuppisas an instance
of deliberateheterodyne
in Vergil(previous
(fourth-foot
clash), AccentualSymmetry
note)24.
'"Tucker456.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

128

PHOENIX

that had greaterinfluenceon word-placementthan any considerationsof

stressaccent.'1 Still, thereis much that we can learn about accent from
the hexametersofVergiland Ovid, ifwe focusour attentionfirston those
verse-positionswhere thereis a preponderanttendencyforictus-accent
harmony. Then, by inference,we may be justified in making some
cautious hypothesesabout accentual rhythmsin other sections of the
hexameterline.
My methodology,therefore,involved correlatingand tabulating the
encliticcompositesas distinctmetricalpatterns,by verse-positions.20
My
classificationwas limited to occurrencesin the Eclogues,Georgics,and
Aeneidof the threestandardlightenclitics-que,-ve,and -ne,whichhave
commonphoneticproperties.I excluded monosyllabicand pyrrhicbasewords (chieflyformslike mequeor meaque and the conjunctionsatque,
namqueand neque),sinceit was obvious that theycould not enterinto the
My only otherexclusionswerefixedcomquestionof accent regression.21
pounds such as deniqueand undique. In the tables that summarizeeach
group of patterns, the figurein parentheses is the total number of
Vergilianoccurrences.22
I present my conclusions in the form of hypotheses,rather than
establishedproofs.These conjecturesmay be summarizedas follows:
1. In Vergilianhexameter,encliticcompositesare typicallydeployed
so as to exploit the natural coincidenceof verse-ictusand word-accent
that must occur when the encliticis not elided. Despite some opinionto
the contrary,the principleis probably not violated by compositeswith
a trochaicbase-word(such as armaque).
2. For the frequentpatternsof Vergilianencliticelisionin thesiwithin
the 2nd, 3rd, and 4th feet (and the rare cases withinthe 5th foot), the
principleof harmonyis maintained by the combinedphoneticeffectof
the hexameterictus and the natural syntacticalword-accent.There is no
clear evidenceto sustain a theoryof accent regressionin thesesituations.
'9See, for example, E. G. O'Neill, Jr., "Word-Accents and Final Syllables in Latin
Verse," TAPA 71 (1940) 335-359. Though I disagree with several of O'Neill's arguments, I do accept his principle that we must not "put the accentual cart before the
metrical horse" (358).
20Such tasks of classificationare now substantiallyless onerous since the publication
of Wilhelm Ott's metrically coded Reverse-Index, Rickliufiger Wortindexzu Vergil
(Tiibingen 1974), to which I acknowledge my indebtedness. Like Ott, I use the text of
R. A. B. Mynors, P. Vergili Maronis Opera (Oxford 1969).
21The monosyllabic type must produce harmonyif the enclitic is unelided or if it is
elided in thesi,and it will produce clash if the encliticis elided in arsi. The pyrrhictype,
of which there are merelynine examples in all of Vergil,can be accommodated only by
encliticelision in arsi, which causes accentual conflictin the precedingfoot.
22In his Index VerborumVergilianus (New Haven 1930), M. N. Wetmore lists 4350
light enclitics (-que 4168, -ve 102, -ne 80), including their occurrencesin the poems of
the Appendix. The total of my tabulated enclitics forthe three major poems, afterthe
exclusions as noted, is 3809.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC

RHYTHMS

129

3. In Vergil'sinfrequentuse of elision in arsi afterthe 5th footof the

hexameter,a coincidenceof ictus and accent is still apparent. The Vergilian evidence does not allow us to say whetheror not thisphenomenon
should be describedas accent regression.
4. When a Vergilianencliticis elided in arsi at the end of the 1st or
afterthe 6th foot,the in4th feetof the hexameter,or hypermetrically
fluenceof the metrical pulse-beat establishes a harmonyof ictus and
accentin the 1st,4th, and 6th feetrespectively.For spondaic feetat least,
thisconjecturerequiresthe hypothesisofa regressionfromthesyntactical
to the isolate word-accent.Though he was not the firstLatin poet to
encliticrhythm,Vergildeveloped it into
experimentwith the fourth-foot
a distinctivestylisticpattern,foundno fewerthan 74 timesin his mature
work (and imitatedby Ovid and later poets).
5. The self-consciousartificeof the -que/-que coordination(with the
first-que prolongedin arsi), to be found17 timesin Vergiland 12 times
in Ovid, has a rhythmicalkinshipto parallel patternsof accent regression. Though thisis a literarydevice that bends the normalrulesofLatin
prosodyand phonology,it too was designedto be read witha harmonyof
ictus and accent.
In the followingexamination of Vergilian enclitic rhythms,I have
arrangedthe evidence by the metricalshape and positionof the various
base-words.The fivesectionsofmyanalysisare notdesignedto correspond
numericallywith the hypothesespresentedabove, thoughmy rationale
forthese hypotheseswill become apparent.

Table I presents a substantial majority (69.9%) of all 3809 Vergilian

encliticrhythmstabulated-2663 compositeswith a spondaic, anapestic,
or iambic base, of which689 (25.9% of 2663) show elisionin thesi.
If we accept the orthodoxtheoryof encliticaccent, it is evident that
all the unelided patterns (group A) will receive a syntacticalaccent on
thepenultof the composite,that thispenultmustinevitablycoincidewith
a metricalarsis,and thatthecompositemusttherefore
producea harmony
of ictus and accent withinthe constraintsof dactylichexameter.Vergil's
sensitivityto thisnaturalharmonymay be inferredfromthelargenumber
of encliticcompositesthat are placed so as to end at the fifthtrochee:
when we add the unelided fifth-foot
figuresfromTable I to those shown
below in Table II, we findthat Vergilhas a total of 1251 such patterns
(32.8% of all rhythmstabulated). Expressingthe statisticaldata another
way, we can state that roughlyone Vergilianverse in ten, on average,
has an unelidedencliticat the fifthtrochee.23
Now thereweresurelyother
considerationsof metreand syntaxthat influencedthisfeatureof Vergil's
2aThe 1251 cases represent9.69% of the 12,913 verses in Mynors's OCT.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

PHOENIX

130

TABLE I
RHYTHMS

ENCLITIC

IN VERGIL:

BASE-WORD

A. WithoutElision(1974):
. caelumque (1262) -2--(18)
-I-

2. sociosque

B. WithElisionin thesi(689):
1. caelumqu(e) (423) -2-t

-I-t
2. sociosqu(e) (238)
'l-t
-l-t

A SPONDEE,

-3--(295)-4--(102)

(544) --2- -(13) -33- -(152) -4-

3. uirumque (168) -2--(29)

3. uirumqu(e)

IS

(28)

-3--

-2-t

(8)

-3-t

- (61) --5- -(312) -66-55-- (30)

(7)

(91) -4-t
(9)

OR IAMB (2663)

-5--(847)

(1) -3- t (145) -4-t(269)

- 3-t

ANAPEST,

-4-t

-5-t

(8)

(144) -5-t

(3)

(6)

-6- -(102)

(11)

style; but therecan be little doubt that he foundthe strongsyntactical

accentof the encliticcompositeone pleasingway in whichto reinforcethe
characteristicrhythmof thehexametercadence,wherefifth-foot
harmony
has been shown to approximate 99%.24

We should notice,too, the significantnumberof Vergilianhexameters

(102) that end with an encliticcompositeof threesyllables; here again
thereis an opportunityto create an emphatic coincidenceof ictus and
accent at a verse-positionwhere Vergilianharmonyapproaches 100%.
The rhythmicaleffectis even morestronglymarkedin the 60 of these 102
verseswherethe finalword is paired with a precedingencliticin a polysyndeticcoordinationsuch as caelumquediemque(Aen. 1.88) or hominumque deumque(Aen. 1.229)-a stylisticpracticethat has oftenbeen noted
by commentatorsas an epic mannerismimitativeofHomer and Ennius.25
The six examples of four-syllableenclitic verse-endingsin group A.2
24Statisticson fifth-foot
harmonywill vary in accordance with the researcher'sstand

on accentualcontroversies.
The combinedfigure
forVergil's5thand 6thfeetis givenas

99.5% by E. H. Sturtevant, The Pronunciationof Greekand Latin2 (Philadelphia 1940)

184. Ott (above, n. 7) shows99.47% forfifth-foot

harmonyin Aeneid1, and 99.11%

for Aeneid 6.

25SeeNordenon Aen. 6.336,R. G. Austinon Aen.4.83. All of these60 pairsextend

fromthehephthemimeral
caesurato theendoftheverse;43 beginwitha spondaicbase-

word and 17 with an anapestic; 57 are -que/-que,2 are -ne/-ne(Aen. 1.308, 12.321), and

1 is -ve/-ve
wereanalyzedbyH. Christen(Aen. 9.211). Thesepatternsofpolysyndeton
Hexametrikern
sen, "Que -que bei den romischen
(bis etwa 500 n. Chr.)," Archiv
fur
lateinischeLexikographieund Grammatik15 (1908) 165-211. Christensen (169) shows

67 Vergilian-que/-quepairsin this position,but his countincludesthe hypermetric

patternsthat I considerunderTable IV below.He also showsthatOvid farexceeded
with246 casesof-que/-que
afterthehephthemimeral
Vergilin hisuse ofthismannerism,

caesura.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC RHYTHMS

131

(Aen. 5.300, 6.11, 9.344, 9.574, 9.767, 10.505) are anomalous patterns
used forspecial effect.They do maintain a rhythmicalharmonyin the
sixth foot,but they violate the regularinner-metrical
principlesof the
hexametercadence by requiringan awkwardelision in orderto establish
a precedingfifth-foot
harmony(mentemanimumque,Aen. 6.11); without
this elision, they can be expected to cause an extraordinaryfifth-footh
clash (. . . at socii multogemitulacrimisque,Aen. 10.505).26
This awkwardelisionjust mentionedexplains the low numberof fifthfootexamplesin Table I.B. In contrastto 1189 fifth-foot
occurrencesof
unelided compositesof the types caelumque,sociosque,or uirumque,we
findhere only 11 instancesof these types with fifth-foot
elision. Vergil's
thesisof his dactylichexameter
generaldistasteforelisionin the fifth-foot
was not a phenomenonof word-accent,but a principleof inner metric
(see O'Neill, n. 19 above). Enclitic elision at this point in the line will
produce an exotic pattern such as pactosquehymenaeos(Aen. 4.99) or
pictique Agathyrsi(den. 4.146). The same metricalprinciplelimits the
use ofthelongerelidedcompositeslistedbelowin Table II (B. 3 and note),
each found once only: inconcessosquehymenaeos,Aen. 1.651; Lacedaemoniosquehymenaeos,den. 3.328; and LycaoniumqueErichaeten,Aen.
10.749. Of the 14 examplesof thisencliticelisionin total, 8 involve some
formof the word hymenaei(Aen. 1.651, 3.328, 4.99, 6.623, 7.344, 7.358,
11.217,11.355), 4 a Greekpropername (Aen. 4.146, 6.445, 9.344, 10.749),
and 2 a Greek noun (elephanto,Geo. 3.26; orichalco,Aen. 12.87). There is
still accentual harmony,it should be noted;27but the metricalpatternis
a clear departurefromthe accepted Augustan norm,and it is thus used
only in formulaicGreek expressions.28
The main lesson to be learned fromthe finaltwo feetis that Vergil's
ear was well attuned to the natural syntacticalaccent of encliticcomposites. Beyond question,therefore,
he musthave sensed a similarlymarked
rhythmelsewherein the verse. Indeed, in 23 verses he joined coordinate
pairs of unelidedencliticsin the4th and 5th feet,just as he employedthe
caelumquediemquetype in the finalcadence.
If Vergiland his Roman readerswere accustomedto a strongrhythmical harmonyforthe unelided patternsin group A, would theyhave been
"That theunusualcadencementem
is an echoof Enniusand Lucretiusis
animumque
shownby Nordenon den. 6.11 (cf.AnhangIX.2, 439). The anomalousrhythm
of 4en.
10.505is notedby R. D. Williams,The AeneidofVirgil,Books7-12 (London1973)354.
"In thefifth-century
VergilianpapyrusfromOxyrhynchos,
publishedin Papiri Greci
e Lafini1 (1912) 47 and plate 12,a superscript
macron(herean accentmark?)suggests
fifth-foot
harmonyin den. 4.99: pact6s[que
an acute accentappearson the
hymenaeos;
unelidedcompositein den. 4.68: totdquev[agaltur.See CliffordH. Moore, "Latin
Exercisesfroma GreekSchoolroom,"
CP 19 (1924) 322-325.

2"See Norden, Aeneis VI, Anhang IX.1, 438; cf. Wilkinson, Golden Latin Artistry

(above,n. 12) 225-226.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

132

PHOENIX

accentuationto the correspondingelided compolikelyto give a different

sites in group B? Enclitic elision in thesi,it is crucial to note, occurs
invariablyunder metricalconditionswhere the ictus must counterany
tendencyforaccent regression.
In a briefsummaryarticle published in 1950 (above, n. 12), R. D.
Williams made a tentative proposal for accent regressionwhenever
encliticelision occurs in the thirdthesis-he called it "elision over the
caesura"-in such a line as imminetaduersasqueaspectatdesuperarces,
Aen. 1.420 (an example fromTable I is scuta uirumgaleasque etfortia
corporauoluit,Aen. 1.101). Williams argued that we should accentuate
adue'rsasque(or galeasque),because the rhythmaduersasque(or galeasque)
"would give coincidenceof accent and ictus in both the thirdand fourth
feet,whichthecaesura normallyprevents."
It is doubtfulthat the question can be proven one way or the other.
However,some flawsin Williams'sproposal can be revealedby an examination of all third-foot
elisionsin Aeneid,Book 1.
In Aeneid 1, thereare 35 elisions at the 3rd thesis,24 on the second
long beat of a spondaic 3rd footand 11 on the firstshortbeat of a dactyl.
Of the35, 21 are encliticcomposites(14 elidedon thesecondlong,7 on the
firstshort).Of these21 elided composites,8 could producetheconsecutive
harmoniesin the 3rd and 4th feetthat Williams wished to avoid. These
harmoniesoccur under two circumstances:(a) the encliticis followedby
a monosyllable-molemqueet montis,Aen. 1.61 (a natural coordination;
cf. Aen. 1.101, 119, 458, 550, 556); or (b) it is followedby a molossusin addition to Williams's example (Aen. 1.420), we findlargo.que
umectat
in line 465. In the remaining13 cases of encliticelision at the 3rd thesis
(62% of the total in Aeneid 1), an anapestic or spondaic word followsto
the hephthemimeralcaesura, producing fourth-footconflict: saeuttque
animis (149), onerantqueau'ro(363); see also Aen. 1.11, 57, 98, 165, 476,
495, 506, 512, 514, 733, and 744. Are we likelyto read the followingelided
compositeswith accent regression?
Aen. 1.11
impulerit.tantqeneanimis caelestibusirae?
jras Aen. 1.57
sceptrate'nensmollitqueanimos et temperat
Aen. 1.476
ferturequis curruquehae'retresuptnusinani
Jrcturum
pluuiasque Hyadas geminosqueTrionesAen. 1.744
In the largergroupof 13 elisions,therefore,
Williams'sargumentdoes not
and
even
within
the
of
the
third-foot
ictus makes accent
apply;
group 8,
difficult
and
artificial:
regression
hoc metuensm1lemqueet mo.ntis
jnsuperaltos Aen. 1.61
Williams's hypothesisis furtherweakenedby the evidenceof versesthat
beginwithpaired wordsin enclisis,wherethe rhythmicharmoniesof the

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC

RHYTHMS

133

firstword create an expectationof harmonyin the second:

laetitiaquemetuque;auidi coniungeredextras Aen. 1.514
Aen. 1.566
urtute'squeuirosqueaut tantiince'ndiablli
Though it is not a typicalVergilianrhythm,thelast versemust,I believe,
show total coincidenceof ictus and accent.
My conviction that all the patterns in Table I maintain the usual
syntactical accent is strengthenedby the evidence of the fourth-foot
composites,where elision is actually the norm (424/587 occurrences=
72.2%). Here we see a characteristicrhythmin which accent regression,
in defianceof ictus,seems highlyunlikely.29For syntacticaland stylistic
reasons,Vergilis partial to encliticelisionat thispoint in the hexameter;
Aeneid 6 alone shows 36 cases. From Aeneid 6, I quote threerepresentative versesof this type,whichcan be given a natural readingonly if one
maintainsthe normalsyntacticalaccent:
Sic faturlacrimans,classiqueimmittit
habenas
Aen. 6.1
ex
his
hoste
Aen.
umeris
6.111
medioque
recepi
eripui
uoce uocans HecatencaeloqueEreboquepotentem Aen. 6.247
II

Vergilmakes regularuse of encliticrhythmsin which the base is a word

of longer shape-a molossus, choriamb, or dispondee (the last found
only afterthe penthemimeralcaesura). Table II displays a total of 697
such rhythms,of which173 (24.8%) show elisionin thesi.
It will be apparentthat the constraintsof the hexametermustproduce
at least one ictus-accentharmonyin the case of all the unelidedpatterns
(group A). For the reasons given above, I doubt that this syntactical
accent will ever shiftunder the influenceof elision in thesi (group B).
A difficultand important question, which can be only tentatively
addressed on Vergilian evidence alone, is the possible existence of a
secondaryword-accentupon the encliticcompositesof Table II.30 There
are clues that suggest its presence; but a decision on the matter must
involve some arbitraryjudgement.
29The Ennian cadence Latiumque augescereuultis (Warmington 472 = Vahlen 466)
was apparentlyquoted by Varro (De Lingua Latina, Goetz and Schoell Frag. 86, 218.810) to illustratethe syntactical accent Latizumque,even though the word is elided in the
4th thesis. This citation is seen as a clinchingargumentagainst all accent regressionby
Bertil Axelson, "Der Mechanismus des ovidischen Pentameterschlusses,"Ovidiana, ed.
N. I. Herescu (Paris 1958) 130, n. 2; but Axelson appears to overlook the fact that
Ennius' verse shows a coincidence of ictus and syntactical accent.
aoThough the theoryof a Latin secondary word-accent was challenged by William
Beare, Latin Verse and European Song (London 1957) 175, the majority of recent
scholars uphold the existenceof a secondary stress: see Wilkinson,GoldenLatin Artistry
(above, n. 12) 121 ff.;Lienard (above, n. 10) 559; Allen, Accentand Rhythm190-191.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

134

PHOENIX
TABLE II.

ENCLITIC RHYTHMSIN VERGIL: BASE-WORD IS A MOLOSSUS,CHORIAMB,OR DISPONDEE (697)*

A. WithoutElision(524):

1. aeternumque (343) 1 -- 2- - (273) 2--3-

I--I--

2. auxilioque

I- __1_-

- (46) 4-- 5- - (24)

(164) 1--- 2- - (122) 2- -"3 - - (20) 4- "- 5- - (13) 5 - - 6 - "(9)

- 4-- 5 - - (17)

3. tempestatumque(17)

-I--I

B. WithElisionin thesi(173):
1. aeternumqu(e)
(121) 1--2-t
I--I-t

2. auxilioqu(e)

I- --t

(51) 1- --2-t

(84)
(33)

2--3-t
2- --3-t

3. tempestatumqu(e)(1)

(36) 3--4-t
(16) 3- --4-t
-4--5-t

- I--I-t

(1)
(2)
(1)

*Not included in Table II are 12 occurrencesof threeless typical polysyllabicrhythms:

Agamemnoniaeque, - 4- -" 5 (4 occurrences unelided, Geo. 3.550, Aen. 6.489,
Aen. 6.838, den. 10.123; once elided at den. 3.328); gubernatorque, 2- - 3 - - (twice

unelided,Geo.3.345 and den. 1.426) and 4--5and soporiferumque,-4-

--5 -

- (once unelidedat den. 3.269);

(3 occurrences unelided, Aen. 4.486, Aen. 7.711,

Aen. 8.725; onceelidedat den. 10.749).

Limitedsupportfora secondaryaccentmaybe providedby thenine

versesin whichVergilintroduces
the typeauxilioqueat theend of the

hexameterline (5 - -- 6 - -). These patternsare stylisticanomalies,of

course, but there is no reason to believe that they were rhythmically

offensive
in violatingtheestablished
cadence.Five ofthenineare quasiGreekhexameters,
whichconveythesing-song
ofa chantedcataeffect

logue:31

Geo.4.336
Xanth6que
Lig(aquePhllodoc.que
Drymoque
Nisa4eSpioqueThaltaqueCymodoc.que
den. 5.826
jngimut,Glauc.mqueMed6ntaque
mqueAen.6.483
Th?rsiloch.
hjcmdctqt
Lad5na Pher.taque
Demodocumque den. 10.413
den. 12.363
Chl6reaque
Sybarimque
DaretaqueTh.rsilochumque
The exoticand un-Latinqualityof theselinesis underscored
by their
in contrastto thesubtleinterplay
exaggerated
rhythmic
of
harmonies,
clashand harmonyat whichVergilnormally
aimed.However,thevery
factthattheyareexoticmustgreatlyreducetheirvalueas clearphonetic
"'We may note that these five verses account for all occurrencesof the rhythmical
pattern cruentaque (Table III.A.4) in the fourthfoot of the hexameter.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC

135

RHYTHMS

evidence; and the other fourcadences of this type (Aen. 5.589, 6.393,
6.601, and 8.416) includeonly one Latin word.32
A somewhatstrongerargumentin support of secondaryword-accent
is the fact that Vergilcan begin 395 hexameters(on average, 30.6/1000
verses) with encliticcompositesthat have a molossus (273) or choriamb
(122) as the base-word.In view of his establishedpreferenceforfirst-foot
harmony-probably in excess of 70%33-I think it likely that we are
intendedto hear a strongrhythmicalcoincidencein these commonpatternsof the type aeternumqueor auxilioque. The presumedaccent shift
caused by enclisiscan be illustratedby these contrastingpairs of verses:
aeternumlatransexsanguisterreatumbras
habebit
aeternumquel6cus Palinuri ndomen
auxilio tUtosdimittamopibusqueiuua'bo
palmas
auxilioque uocaredeos et te'ndere

Aen. 6.401
Aen. 6.381
Aen. 1.571
Aen. 5.686

The rhythmicimportanceof patternA.1 is apparent fromits use in the

two majestic lines that formthe climax of the Aeneid's epic statement:
deosLatio; genusundeLatinurn
inferretque
Romae. Aen. 1.6-7
Alban.'quepatresatque altae mroenia
If these compositesare to receivea secondaryaccent when theybegin
a verse,we must surelygive them a similarintonationwhen they begin
at the second or fourtharsis,or (in the case of groupA.3) at the penthemimeralcaesura. This assumptionis strengthenedwhenwe findpatterns
A.1 and A.3 linkedeffectively
in fourversesof identicaltexture:
canes importunaeque
obscenae'que
uolh.cres Geo. 1.470
facis tempestatumque
potentem Aen. 1.8034
n.mborumque
deos obtestanturque
Latinum
Aen. 7.576
implor4ntque
d?siectiqueduicesdesolatiquemanipli
Aen. 11.870
One reason forthe much lower frequencyof types A.1 and A.2 at the
second arsis may be the fact that this arrangementalmost invariably
produces threesuccessiveharmoniesin the firstthreefeet:
lilia uerbenasquepremensuescumquepapauer Geo. 4.131
restitit,
Eurydicenquesu'amiani luce sub ipsa Geo. 4.490
32Theone Latin word used in this position (ancipitemque)occurs in Aen. 5.589, where
the labyrinthimage is thoughtto account forthe metricalanomaly; cf. R. D. Williams
ad loc. It is significantthat Horace concluded three sapphic stanzas with one-word
adonic verses that seem to confirmthe secondary accent: Fabriciumque (C. 1.12.40),
M?rcuriisque (C. 1.30.8), and militialque(C. 2.6.8).
33Cf.n. 24 above. Ott's statistics forfirst-foot
harmonyin Aentid 1 and 6 are 68.53%
and 70.97%; the inclusion of secondary accents would raise these figuresby about 6%.
harmoniesof 82.0%, 70.2%, and 66.2% for
Greenberg(above, n. 7) 33 shows first-foot
the Eclogues, Aeneid 4, and Aeneid 12 respectively.
'4Cf. Geo. 1.27, Aen. 1.53, 1.255,Z3.528.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

136

PHOENIX
TABLE III

ENCLITIC

RHYTHMS IN VERGIL:

A. Without Elision (237):

1. armaque

I- -(102)

2. armentaque - | -

(94)

BASE-WORD

1- --(69)

3. Tiberinaque-I- -(34)
4. cruentaque [I--- (7)
B. WithElisionin thesi(13):
1. armaqu(e) I--t
(11) l--t

2.armentaqu(e)- - -t

(2)

IS A TROCHEE

- 4-

OR HAS A TROCHAIC ENDING (250):

(34)

.-4(14)
-4--- (5)
(8)

-4-

-t (1)

5- --(33)

- 5-

- (60)

-5- --(20)
-5--- (2)
5--t

-5-

(3)

-t (1)

III

Table III displays the group of 250 compositesin which the encliticis
attached to a trocheeor a wordendingin a trochee.The questionof their
accentuationhas been a matterof continuingcontroversy.In two provocative studiespublishedin the late 1960's, Jean Soubiran ralliedto the
defenceof the imperialgrammarians,backed by carefulscholarshipand
impressiveingenuity.35
Only the accentuationsarmdqueand armentdque,
he argued, could explain the relative infrequencyof these types in the
fifthfootof Latin hexameters.I remainunconvinced,and my scepticism
is apparentlyshared by Lienard and Allen.36Combined factorsof syntax
and style-admirably analyzed by Soubiran himself-seem far more
compellingthat the alleged clash of ictus and accent. A fifth-foot
rhythm
thatoccursas oftenas 115 times(8.9/1000verses)can hardlybe described
as a strikinganomaly,and it seemsunlikelythat Vergilwould have been
so ready to violate his normalhexametercadence. In any event, elision
of these patternsis veryunusual (13/250 = 5.2%), and the question of
accent regressionis not an issue.
IV
Table IV displaysall cases of Vergilianencliticelisionin arsi, exclusiveof
compositeswithmonosyllabicand pyrrhicbase-words.It is only forthese
eight rhythmicpatterns that a theoryof enclitic accent regressionbecomes phoneticallyplausible.
Elision permittedthe introductioninto hexameterpoetry of enclitic
compositeswith a dactylic base. To be sure, therewere other phonetic
expedientsavailable: like Cicero,Vergilcould resortto synizesis(Lauinjaque uenit,Aen. 1.2) or syncope (suppostaquefurto,Aen. 6.24; exposta""Sur les mots de type armaque dans l'hexametrelatin," Pallas 14 (1967) 39-58, and
"Sur les mots de type armentaquedans l'hexametre latin," Pallas 15 (1968) 57-101.
"6LiEnard(above, n. 10) 554; Allen, Accentand Rhythm159.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

137

ENCLITIC RHYTHMS
TABLE IV
ENCLITIC

RHYTHMS

IN VERGIL:

ALL OCCURRENCES

OF ELISION

in arsi (170)

and pyrrhicbase-words)
(excludingmonosyllabic

A. DactylicFeet (65):
1. corporaqu(e)

(46) 1 - - l a (41) (1)*

5- -I a (4)

2. Saturniaqu(e)

(13)

-4-

l a (7) ) -5- -- a (6)

---I a

-I- --I a

3. Cerealiaqu(e)
--l- --i a
4. exsultantiaqu(e)

I--I--- a

(4)

44-

a (1)

--5- --a

(3)

4--5---I a (2)

(2)

B. SpondaicFeet (105):
1. caelumqu(e)
(24) 1 - a (19) (1)?
I--la

6--lh

2. aeternumqu(e) (58)

-4--

a(58)

3. coloremqu(e)

(16)

-4--

a (1)

(7)

-4--

a (7)

TOTALS: (170)

(60) (2)*?

(74)

-I--la

I--l a

4. perituraequ(e)

----l a

(4)

-6--I h(15)

(15)

(19)

*2- -- I a (one occurrence

only: Aen. 11.634)
?2-- I a (one occurrence
only: Aen. 1.78)

que ponto,Aen. 10.694); but these experimentswere rare indeed.37Less

extraordinaryare the versesin whichVergilused elisionat the sixtharsis
to accommodate a dactylic composite in the fifthfoot. There are 15
examplesin all (A.1 through4), usually identifiedby the representative
pattern Saturniaquearua (Aen. 1.569).38 Because they have been well
analyzed by Eskuche, Norden, and Soubiran, thereis no need to discuss
them at length.For our purposes,the importantpoint to notice is that
theyseem beyondreasonabledoubt to maintaintheaccustomedharmony
of ictus and accent in the fifthfoot; viz., agminaquearmat(Aen. 7.648),
Titdniaque astra (Aen. 6.725), Ceredliaque arma (Aen. 1.177), and
haurit(Geo. 3.105 and Aen. 5.137). Whetherwe regardthe
xsulta'ntiaque
word-accentas a regressionor as the normalsyntacticalpronunciationof
the compositewill depend on our attitudetowardsthe liminaque/liminaque alternativesmentionedabove (note 8). If it is a regression,it is surely
caused by the combinedeffectof elision and verse-ictus.
When we turnto the beginningof the line, we see that the composite
type corporaqu(e)acquires a frequency(41 instances) that may allow us
"aForCicero,cf.Shipley(above,n. 6); see also Nordenon Aen. 6.24.
38SeeEskuche(above,n. 12) 386; Norden,AeneisVI, AnhangXI.1.9,456 (to Norden's
listof 14, add Geo.2.464); Soubiran(above,n. 14) 464-466.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

138

PHOENIX

to considerit as unexceptional.Again,I would conjecture,theremustbe a

harmonyof ictus and accent, regressionor no regression.This first-foot
rhythmicpatternhas attractedless scholarlyattentionthat its counterpart in the fifthfoot.Stylistically,it can be seen to have particularvalue
in self-contained"golden" lines or otherwisesymmetricalverses,because
it allows the placementof a neuterplural adjective or noun in the first
position:39
impiaque aeternamtimuerunt
sae(culanoctem Geo. 1.468
grandiaqueeffossismir4biturossa sepulcris Geo. 1.497
corporaqueagrestinu'dantpraedurapalaestra Geo. 2.531
Because such verses are rarer in the Aeneid than in the Georgics,the
encliticpatternis proportionately
less common;thestatisticsare Eclogues,
1 (3.46); Georgics,19; and Aeneid, 21. In his use of the pattern,Vergil
was probably influencedby Catullus, who is notoriouslyfond of such
:40 fivebalanced versesin Catullus 64 (235, 264, 316, 345, 351)
symmetries
beginwitha compositeof thisshape. The elided first-foot
dactyl is found
also in Propertiusand Ovid.41
It is not characteristicofVergilto elide dactyliccompositesat the third
arsis.There is one exampleonly (Aen. 11.634), a unique Vergilianrhythm
that probablyshows total coincidenceof ictus and accent:42
4rmaquecorporaqueetpermixticaede uir6rum
semianimesuoluunturequi . . .
The phenomenonof accent regressionhas been shown above to be a
possible but not inevitable explanation for the dactylic composites in
elision. When we considerVergil'suse of spondaic and bacchiac composites like caelumqu(e) and coloremqu(e) (groups B.1 and B.3), this
phoneticprincipleassumesparamountimportance.
As one mightexpect, the topic of encliticaccent regressionhas been
discussedprimarily-almostexclusively,in fact-in the contextofhypermetricverses,wherethe hexametercadence virtuallyforcesa regressive
shift.There is basic agreementon the question. Even Soubiran, who
presentsan eloquent case against regressionas a generalconsequenceof
"The femininesingularadjective or noun is an obvious alternative; forotherexamples,
see Geo. 1.9, 1.222, 2.157, 2.396, 3.366, 4.24, 4.470; Aen. 3.280, 7.615.
40See Kenneth Quinn on 64.7 in Catullus, The Poems' (London 1973) 301.
41Propertiusbegins 9 hexameters with elided dactylic composites: 1.3.25, 1.6.17,
2.1.77, 2.13.29, 2.15.33, 3.6.17, 3.10.25, 3.11.3, and 4.1.19. Ovid has 6 examples in Metamorphoses1: 39, 62, 501, 528, 598, 742.
42The accentuation of monosyllabicconjuctions like et and aut is always problematic;
see E. D. Kollmann, " 'Et' in arsi afterelidable syllables in the Vergilian Hexameter,"
Studii Clasice 14 (1972) 67-84. Statius Theb.10.275 is cited as a parallel rhythmto Aen.
11.634 by R. D. Williams (above, n. 26) 421.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC RHYTHMS

139

elision,seems grudginglyprepared to admit that the hypermetrichexametersmay be legitimateexceptions.43

Vergilhas 19 such versesthat end
in -que,with4 spondaic base-wordsand 15 bacchiac.44The two patterns
of hypermetricrhythmcan be illustratedas follows:
B. 1 sternitur
infelixalieno uulnere,caelumque
Argos. Aen. 10.781-782
aspicit et dulcis moriensreminiscitur
B. 3 omnia Mercuriosimilis,uocemquecoloremque
et crinisflauoset membradecoraiuuenta. Aen. 4.558-559
In readingtheseverses,whydo we feela naturalinclinationto change the
accent of theelided composites?The phoneticexplanationis quite simple:
we are merelyshiftingfromthe syntacticalaccent of the compositeto the
customaryisolate accentof the base-word,underthedynamicinfluenceof
the verse-ictus.The enclitic becomes virtually a detached phoneme,
linked by synaphea to the followingverse.
If the sixth-footaccentuation of 6 - - {h is cailumque (or possibly
cae'limque),should not a similarrhythmbe heard whensuch wordsoccur
in the firstfoot,withelisionat thesecond arsis?Table IV. B. 1 showsthat
thereare 19 patternsof this type in Vergil:45
arces
circumdedit
septemqueuna sibi mturo

Geo. 2.535
(cf. den. 6.783)
Geo.
4.442
igneemque
framfluuiumque liquentem
h4rribil.mque
et
ueldmen
Aen.
1.711
ac4ntho
croceQ
paliamque pictum
et
non
8.625
t.extum
Aen.
enarrabile
h4stamque clipei
Although this proposal can be nothingmore than conjecture,I believe
that the ictus and isolate accent again combine to produce a first-foot
harmony,parallel to that of the corporaqu(e)type in A.1.46
"Soubiran(above,n. 14) 466-468;see also Alice H. Carpenter,"Hypermetric
Lines
and InterlinearHiatus in Latin HexameterVerse,"P2 9 (1930) 351-362. Carpenter
subscribesto the Germancompromisetheoryof a "hovering"accent (schwebende
sixthfoot(catlilmque); thispositionhad
Betonung)on bothsyllablesofthehypermetric
been developedearlierby AlbertGrangerHarknessin "The Word-Group
Accentin
Latin Hexameter,"CP 3 (1908) 42. For otherreferences,
see A. S. Pease on den. 4.558.
44Spondaicbase-words:Geo.3.377, den. 1.448,8.228, 10.781; bacchiac:Geo.2.344,
2.443, 3.242, den. 1.332,2.745, 3.684, 4.558, 4.629, 5.422, 5.753,6.602, 7.470,9.650,
textual
10.895, 11.609. Soubiran,who lists only 17, presumablyfollowsa different
variants.
readingin Geo. 2.344 and Aen. 3.684, whereMynorsadopts hypermetrical
45Thelistof 19: Geo.1.279,1.406,2.535,3.451,4.341,4.442; den. 1.711,3.445,5.371,
6.280 (ferreique
withsynizesis),6.650,6.783,6.839,7.326, 8.291,8.625,9.344, 11.255,
12.336.Threeof theseare wordsin -cumque(Geo. 1.406,Aen. 3.445, 11.255); because
suchas *quodcum
or *quaecum),
theyhave no isolateaccent(i.e., thereis no base-word
theseindefinite
compoundsmay be special cases. The uniqueexampleof a spondaic
in Aen. 1.78.
compositeelidedat thethirdarsisis quodcumque
46Asimilarconjecturefor20 cases in Lucretiuswas made by WilliamA. Merrill,
"Lucubrationes
Lucretianae,"CPCP 7 (1924) 239.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

140

PHOENIX

We are leftonlywiththe veryinterestinggroupof fourth-foot

patterns
in Table IV. Is it not reasonable to surmisethat encliticelision at the
fiftharsis produces ictus-accentharmonyin the precedingfoot? As in
the case of the hypermetricverses, the pulse of the ictus coincideswith
the isolate accent of the base-word,so as to cause the reader to shift
naturallyand unconsciouslyto that familiarisolate accent.This would be
a logical explanation for enclitic accent regression.Of all patterns of
Vergilianencliticelision,none but the rhythmsin Table IV meet these
necessarymetricalconditions.
In the corpus of extant Latin poetry, Lucretius is the only one of
Vergil's predecessorsto have made significantuse of this fourth-foot
rhythm,and his 16 exampleslack subtletyor variety.47
Vergil,in contrast,
the
into
an
artistic
developed
pattern
rhythmicaldevice,used 74 timesin
total: thoughit is absent fromthe Eclogues,it is foundin every book of
the Georgicsand the Aeneid,occurringon average once in 163 lines (6.1
times/1000verses) withinthese two poems. Vergil'sencliticconjunction
is -que in everycase except Aen. 11.457 (piscosoveamne Padusae), unless
one also reads -vein Aen. 2.37 (subiectisqueurereflammis).
The eightexamplesof therhythmin dactylicfourthfeet(A. 2 and A. 3)
may or may not involve accent regression,as we have seen with other
dactylicrhythmsabove. They are illustratedby these two verses:
A. 2 res Agamemnoniasuictriciaquea'rmasecutus Aen. 3.5448
A. 3 Curetums6nituscrepitantiaque
aera secutae Geo. 4.151
The 66 remainingexamples must show accent regression,if our hypothesis is valid; in all these cases, a spondaic fourthfoot is followedby
encliticelisionat the5th arsis:
B. 2 an memorem
p6rtusLucrinoqueaddita claustra
disiecjtquerateseuertitque
a.equorau.entis
silentarrectisqueauribus astant
conspexere,
B. 3 antiquiLaur.entisopacaque ilice te'ctum
B. 4 corticibusque
cauis uitiosaequeilicis aluo
.t sedethdcanimoperitu'raeque
addereTr6iae

Geo. 2.16149
Aen. 1.43
Aen. 1.152
Aen. 11.851
Geo. 2.45350
Aen. 2.660

47Lucretius1.431, 501, 677, 681, 686; 2.47, 149, 246; 3.780; 4.347, 555, 951, 1187;
5.448; 6.20, 686. The earliest use of the rhythmis an elegant example in Ennius (Sat.
inde loci liquidas pilatasque aetherisoras. There is no example in the
3-4): contemplor/
poems of Cicero or in Catullus 64.
48Theothersix verses of the type A.2 are Aen.3.174,6.490, 8.535, 11.6, 11.729, 11.909.
49Thereare 58 versesof the type B.2, 14 in theGeorgicsand 44 in the Aeneid: Geo. 1.240,
2.33, 2.161, 2.507, 3.27, 3.35, 3.57, 3.369, 3.423, 3.513, 4.4, 4.44, 4.296, 4.521; Aen. 1.43,
1.69, 1.152, 2.1, 2.37, 2.80, 2.195, 2.301, 2.681, 3.227, 3.386, 3.579, 4.579, 5.47, 5.53,
5.210, 5.490, 5.584, 5.865, 6.260, 7.127, 7.386, 7.575. 8.304, 8.448, 8.520, 8.627, 10.142,
10.207, 10.645, 10.776, 10.801, 10.896, 11.186, 11.398, 11.457, 11.826, 11.835, 12.158,
12.551, 12.618, 12.700, 12.830, 12.848.
56The other five verses of the type B.4 are Aen. 8.669, 9.24, 9.113, 11.190, 11.887.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC

RHYTHMS

141

If one examines the group of 74 verses as a whole, several stylistic

featuresbecome immediatelyapparent. There is a strongtendencyfor
therhythmto createa self-contained
syntacticalunit betweenthepenthemimeralcaesura and theend oftheline; thoughmanyofthesehexameters
may not show finalpunctuation,only seldom does the syntax impose a
necessaryenjambementwiththe followingverse.In 50 ofthe74 cases, the
introductorybase-word is an adjective or participle,usually balanced
with a noun at the end of the verse (32/50) or less commonlyin the 5th
foot(14/50). (Of the remaining24 base-words,18 are substantives,4 are
verbs,and 2 are adverbs.) The Vergilianpatternsare richin assonance,a
quality that may be enhanced by the accent regression:
clariscuntsonitusarmo.rumque
Hen. 2.301
ingruitho.rror
ora parentum Aen. 2.681
namquemanustntermaestorumque
ora parentum Hen. 11.887
exclusia.nteoculoslacrima.ntumque
The harmoniccadence can providea powerfulrelease to heavy,spondaic
verses:
haud uatumignarusuenturique.nscius ae.ui Aen. 8.627
inclusus mu'rishostiliqueaggeresa.eptus
Aen. 11.398
All 74 examples are memorableverses, showingconsistentartistryand
patternedsonority.That the rhythmicpatternimpressedVergil'sRoman
readers may be judged by its common appearance in Ovid and later
poets.51
v
This enquiryhas suggestedthat accent regressionis likelyto occur only
when an encliticis elided at the end of a metricalfoot,underthe circumstances definedin Table IV. As we see fromTable V, however,thereis
one special situationwhereVergilcould impose regressioneven without
elision. In 17 verses,we findthe device of the -que/-quecoordinationin
which the first-que is prolongedin arsi (a type of diastole). Vergil was
clearly attracted to this rhythmicallicence, as his poems account for
exactly half of the examples to be foundin extant Latin hexameters.52
Though the techniquewas a Homeric imitationthat departed fromthe
51Therhythmoccurs in every book of Ovid's Metamorphosesfora total of 50 cases, all
-que (on average, once in 240 lines or 4.2 times/1000verses). All 50 examples begin at the
penthemimeralcaesura, with 22 dactylic fourthfeet and 28 spondaic. Of the 50 introductory base-words,47 are adjectives or participles,with 44 nouns in balance (19 in the
5th foot and 25 at the end of the line). See, forexample, Met. 1.476, 538, 730; 2.81, 273,
324, 679. Book 1 of Statius' Thebaid has two examples: lines 6 and 535.
52Thereare 34 extant examples; in addition to Vergil's 17 and Ovid's 12, the device is
foundalso in Accius Annales 1 (= Festus 130.15), Grattius Cynegeticon130, Germanicus
Aratea 1.262, Ilias Latina 168, and Silius Italicus Punica 7.618.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

PHOENIX

142

TABLE V
ENCLITIC

RHYTHMS

1. liminaque

2. terrasque
3. Noemonaque

IN VERGIL:

SPECIAL

(4) 1-

1--2I--I-(12)
-4J--5(1)(1)

PATTERN:

ENCLITIC

PROLONGED

in

arsi (17)

2- (4)
(12)

1. liminaque(4)
Geo.1.164
trahea(que
el infquopSnderer4stri
:rfbulaqui
d,en.3.91
mouii
Ifminaqu'laur4squed4i,tot45que
carfnis
r65stra
7.186
spfculaqu(clipefque
erfptaque
12.363
Sybarfmque
DaritaqueThUrsiloc/hmque
Chl6reaquf
2. terrasque
(12)
mdriscael4mque
Ed. 4.51
ftrrasqul
traclkrque
prof4ndum
Geo.1.153
I4ppaequf:ribolfque,
intirquenitintiacglta
1.352
u#ntos
abltusquipluui4squeetagfn:isfrfgora
t6nqtd6mus,6mniapltnis
1.371
E4riquiZeph.yrfque
l4ppaequitribol&que
3.385
dbsjnt;
f4gep4bulalafta
mdriscaelUmque
4.222
trrasquitract,4que
prof#ndum
4.336
LigfaquePhAllodocfque
Xanth4que
Dr$imoquf
d4gathfrsi den. 4.146
Cr-':esquf
pictfque
Dryopfsquefr6munt
Br6ntesqui
8.425
cen4dusmm&bra
Pyr4gmon
Steropfsque
cSrnuacrfstae
(nsemqu(clipe4mque
12.89
el r4hbrae
12.181
u6c9,quaequeaitheris4lli
f$ntisquifluuiisque
12.443
reflctis
4ntheusquiMnest:hesSque
r4unt,6mnfsque
3. Noemonaque
(1)
A4lcandra4mque
HalUimqueNolmonaquf
PrytanimqueJen. 9.767 (cf.Iliad 5.678)
Ovid Metamorphoses.
Type 1 (5): 5.484,7.265,10.262,10.308,11.36
Type 2 (5): 1.193,3.530,4.10, 8.527,11.290
Type 3 (2): 13.257,13.258(cf.Iliad 5.677-678)

normalrulesofLatinprosody,
it wasusedonlyundercarefully
controlled

phoneticand metricalrestrictions.63

In thecontextofthepresentstudy,it is important
to observethatthe

device is limitedto the 2nd or 5th arsis of the hexameter,thus affecting

the accentuationof base-wordsthatoccupy the 1stor 4th footrespectively-the same verse-positionscommonly affectedby accent regression
see Norden,deneis VI, AnhangX, 450-452;
"'For analysesof theserestrictions,
S. E. Winbolt,LatinHexameter
Verse(London1903)200-201;J. P. Postgate,Prosodia
Latina(Oxford1923)32-34; A. E. Housman,"Prosodyand Method,"Ct 21 (1927) 12;
Hugo Pipping,"Ein Beitragzur ramischenMetrik,"SocietasScientiarum
Fennica,
Commentationes
Humanarum
Litterarum
13.1 (1942) 1-12; R. G. Austinon den. 4.146;
N. E. Collinge,Collectanea
Linguistica(7anua Linguarum21; The Hague 1970) 200.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions

ENCLITIC RHYTHMS

143

beforeelision in Table IV. One can easily see the rhythmicalkinship

betweenthe followingpatterns:
V. 2
Cr.tesqueDryopesquefremunt
pict4queAgathyrsi Aen. 4.146
IV. B. 1 Coeumque.apetumquecrqetsaeuumqueTyph.dea Geo. 1.27954
Vergil's fiveexamples of dactylic composites(Table V.1 and 3) can be
interpretedas supportforthe normalunelided accentuationliminaqueor
liminaque.55The evidenceis not conclusive,however,forthe rhythmcan
be equally well explained as an accent regressioncaused by the verseictus,like the65 cases in Table IV.A. There is good reasonto believe that
Vergil's twelveexamples of spondaic compositesin Table V.2 show the
same pattern of accent regressionthat we have conjectured for their
elidedcounterparts.
Once again, therefore,
we can note the effectof the hexameter'srhythmic pulse-beaton the accentuationof encliticcomposites.Here, I submit,
is furtherevidence that the ictus may be the dynamic cause of accent
regression.
UNIVERSITY

OF VICTORIA

4The similaritycaused Christensen(above, n. 25, 182) to explain Geo. 1.279 as a case

oflengthened-que. This was an understandableerror;but lapetus is always scanned with
a vocalic I in Latin poetry(cf. Horace C. 1.3.27, Ovid Met. 1.82, etc.).
"6Cf.n. 8 above, and see Wagener (above, n. 5) 509-511.

This content downloaded from 132.66.11.211 on Mon, 02 Nov 2015 13:12:17 UTC
All use subject to JSTOR Terms and Conditions