Poem Division, Paired Poems, and Amores 2.9 and 3.11

University of Pennsylvania ScholarlyCommons

Departmental Papers (Classical Studies) Classical Studies at Penn

1990

Cynthia Damon University of Pennsylvania, [email protected]

Follow this and additional works at: https://repository.upenn.edu/classics_papers

Part of the Classics Commons

Recommended Citation Damon, C. (1990). Poem Division, Paired Poems, and Amores 2.9 and 3.11. Transactions of the American Philological Association, 120 269-290. Retrieved from https://repository.upenn.edu/classics_papers/47

This paper is posted at ScholarlyCommons. https://repository.upenn.edu/classics_papers/47 For more information, please contact [email protected]. Poem Division, Paired Poems, and Amores 2.9 and 3.11

Disciplines Arts and Humanities | Classics

This journal article is available at ScholarlyCommons: https://repository.upenn.edu/classics_papers/47 Transactionsof the American Philological Association 120 (1990) 269-290

POEM DIVISION, PAIRED POEMS, AND AMORES 2.9 AND 3.11*

CYNTIA DAMON StanfordUniversity

The mostrecent editor of theAmores, J. C. McKeown,prints Am. 2.9 and 3.11 as single poems, thoughhe concedes a doubt as to whetherunity is "capable of definitiveproof or refutation";'the editor of theOxford Classical Text,E. J.Kenney, prints two separate poems under each number:2.9a-b and 3.11a-b.2 Rival icebergtips, topping a disputewhich has goneon fornearly a centuryand a half.A bewilderingvariety of proofsboth for and againstunity has beenadvanced, many of thembased on aestheticgrounds; the present paper makesan argumentfrom form. For once it is realizedthat the consecutive pair of poemswas a legitimatecompositional unit in Ovid's day,the arguments of theunifiers have to be usheredout of court. The processbegins with an examinationof material evidence, indications of poemdivision in themanuscript traditions of workswith a formalresemblance to theAmores, books, that is, builtup of numerousrelatively short poems.

Section1: How theproblem arose Lucian Mullerbegan his discussionof themerits of dividingAmores 2.9 and 3.11 withthis reminder: Primumid monebo,quod nemo ignorat,elegias cum non sicutapud nos titulisaut numerisseiungerentur, sed sola litteramaiore, facile potuisse fieri,ut aut iungerenturperperam aut etiam secernerentur carmina. Sola, however,is notentirely accurate, even for the medieval MSS of whichhe was thinkinghere, and it is probablynot at all trueof ancientmanuscripts. Some of theevidence for ancient methods of indicatingpoem division in a con- textcomparable to thatof theAmores is direct;the earliest is providedby the 1stcentury B.C. papyrusfragment of Gallus,in whichpoem divisions are indi- catednot only by a largeinitial letter at thebeginning of thenew poem but also by a relativelylarge space between the last line of one poemand thefirst line of thenext (approximately 3 times the average space betweenlines within a poem)

In preparingthis paper I benefittedgreatly from the advice and criticismof E. Courtney,J. Solodow and W. Batstone,to all of whomI cheerfullyrender thanks. 1 Ovid: Amores.Volume 1, Text and Prolegomena(Liverpool 1987) 92. 2 p. Ovidi Nasonis Amores,Medicamina Faciei Femineae,Ars Amatoria,Reme- dia Amoris(Oxford 1961). 3 L. Muller,"De Ovidii Amorumlibris," Philologus 11 (1856) 89. 270 CynthiaDamon and signswithin that space.4 The 4thand 5thcentury vellum codices of Virgil's Eclogues representa differentera, a differentmedium, a differentgenre and a differentauthor, and show, not surprisingly,an entirelydifferent method of indicatingpoem division.In MP (4th cent.) and R (5th cent.) most of the Ecloguesare preceded by a listof speakerswhich is placedwithin the column of textand writtenwith the same scriptand linespacing as thatof thetext, but at timesin red inkinstead of thetext's black.5 These listsresemble nothing so muchas thelists of speakersat scenebreaks in thecontemporary Bembine MS of Terence.For twoof theEclogues, however, the inscriptiones (8, DAMONIS ET ALPHESIBOEICERTAMEN, M; DAMONISET ALPHESIBOEICERTATIO, P; 10,CONQUESTIO CUM GALLO POETA DE AGRIS,M; CONQUAESTIODEAGRIS CUM GALLOCORNELIO, P) are morein thenature of poemtitles, though they still list the participants.The headingsof the fourthand sixthEclogues, in whichthere is no majorspeaker but thepoet, contain titles based on subject matter:4, SAECULINOVI INTERPRETATIO,R; 6, FAUNORUMSATYRORUM SILENORUMDILECTATIO, PR. It is worthnoting that M and P have titlesfor everypoem of whichthey contain the beginning, including the first poem in thebook (P), whileR has titlesfor all butthe first.6 There is also directand indirectevidence for intra-columnar tituli in thearchetype of Books 1-12 of the

4 P. J. Parsons in R. D. Andersonet al., "Elegiacs by Gallus fromQasr Ibrim" JRS 69 (1979) 129. 5 In V (5th cent.) thered inscriptioof theone Eclogue-beginning(vi) contained in thatfragment is illegible. None of the otherearly Virgil MSS is available for theEclogues. 6 The earliestsurviving bit of Ovid (thefragmentum Guelferbytanum of the5th century,containing bits of ex Ponto 4) does not preserveany poem junctions, but 0. Korn (P. Ovidii Nasonis ex Ponto libri quattuor[Leipzig 1868] x) believes thatit had no intra-columntitles. He reconstructsan archetypewithout titles, too (xxxii). I do not know the basis of R. J. Tarrant's claim (in L. D. Reynolds, Texts and Transmission:a Surveyof the Latin Classics [Oxford1983] 203) that this archetypecontained the poem titlesfound in laterMSS, but thereare no ele- mentsin these tituli thatcould not be derivedfrom the poem (or froma nearby poem, as, e.g., the names Pompeius at 4.15 fromthe introductorypoem 4.1, and Tuticanus at 4.14 from4.12), in almost everycase withvery littletrouble. (The one exception is the informationthat Cotys, the addressee of 2.9, was king of Thrace. This is preservedin Kom's r, a "codex sine nomineab Heinsio notatus." T's one otherunique bit of information,the praenomen Quintusat 1.2 is, in fact, erroneous.)There are, moreover,a numberof differencesin the tituliwhich sug- gest thatat some pointin the traditionsof A, 3, B, E and P (again, using Korn's sigla) scribes were creatingtituli off their own bat: at 1.9 P's titulusis MAXIMO (fromline 32 of the poem), B's is AD CELSUM (fromline 1, bothreferring to the same person). Similarly,variation between the termssodalis and amicus in theti- tuli for 3.6 suggestsindependent creation rather than copying. Clearly P's titulus AD CONIUGEM at 3.1.31 is an independentcoinage designed to complementthe false poem beginningmade here. G. Luck's inference(Untersuchungen zur TextgeschichteOvids [Heidelberg1969] 77; hereafter,"Luck") thatthe titleswere located in the marginis not a necessaryresult of theirabsence in the columnof text,nor even a likely suppositionin view of the nonattestationof marginalti- tles beforethe 9th century. PoemDivision, Paired Poems 271

Epigramsof Martial, which was in existence before A.D. 401 whenGennadius Torquatuscompleted his revision of it.7 Forthe 6th and 7th centuries we haveevidence from the Bobbio fragment of Juvenal,which does not mark the beginning of Satire 15 witha title,8and fromthe earliest manuscripts ofPrudentius (A andB), inwhich titles were left to therubricator tobe completed(A) ornot (B). Likethe inscriptiones in the earlierVirgil MSS, thoseof A lie withinthe column of text and are present at thebeginning of every poem in the collections of short hymns (Cathemerinon, Peristephanon). The red inscriptionesin the8th century Codex Salmasianusof theLatin Anthologyare generally longer than those found earlier, providing a title for eachpoem or excerpt, and occasionally other information such as theauthor's name,source of the excerpt, summary of what preceded an excerptedpassage, etc.The first letters of the line following an inscriptio are similarly red, and oc- casionallyenlarged or ornamented.9 We finda litteramaior at poembeginnings again in an early9th century MS which,among many other items, contains selections from Martial (H).10

7 To give only one example. See W. M. Lindsay,Ancient Editions of Martial (Oxford 1903) 34-55. It is unlikelythat the originalsof the tituli preservedin Lindsay's AA and CA familieswere by Martial,but the presenceof a few bits of informationnot derivablefrom the poems themselvesled Lindsay (54) to propose for them,very tentatively,a date "not long afterDomitian's reign."The case is verysimilar for the tituliof Statius'Silvae. These are hardlyStatian, but theydo preservesome detailsof nomenclaturenot given in thepoems and are dated by K. Coleman to "withindecades of Statius' death" (Statius,Silvae IV [Oxford 1988] xxviii-xxxii.To her list [xxviii] add the nomen Vettiusin the titulus for 5.2). They are in any case likely to have been presentin the textfamiliar to Sidonius Apollinarisin the mid-Sthcentury. He uses "titles" to referto fourpoems from the collection (Carm. 22 epist. 6, referringto 1.5, 2.3, 3.1, 3.4); these are not the titulifound in M, but ratherad hoc reminiscencescombining elements from the tituli and the firstlines of the poems. There is no information,however, about the positionon the page of the tituli in this early edition. 8 U. Knoche,D. Iunius JuvenalisSaturae (Miinchen1950) ad loc. Both R and UFI (Knoche's sigla), too, regularlyomit inscriptiones.A librarycatalogue dated by B. Ullman to the late eighthcentury ("A List of Classical Manuscripts[in an Eighth-centuryCodex] perhaps fromCorbie," Scriptorium8 [1954] 24-37, a ref- erence I owe to F. Newton) does referto books 2 and 3 of the Satires by tituli appropriateto the firstpoem in each book. Yet the varianttitles offered by the traditionfor Satires 3 (QUARE UMBRICIUS URBEM DESERAT [theP reading]vs. DE URBIS INCOMMODIS ET DE DIGRESSU [vel EGRESSU] UMBRICII [MSS of the( group]),4 (DE PISCIS MAGNITUDINE [PI VS. SATIRA DE ROMBO PISCE [vel sim.]or CATALOGUS AMICORUM DOMlITANI SATYRA DE RHOMBO PISCE [velsim.] [0]), and9 (QUAERELA NAEVOLI DE REGE IMPUDICO [P] vs. LOQUITUR AD PARASITUM QUENDAM QUI SERVIERAT REGIBUS [vel sim.] [0]) are so discrepantthat in thesecases, at least,no archetypalinscriptio can be assumed. 9 A. Riese, Anthologia latina sive poesis latinae supplementum,pars prior: carmina in codicibus scripta,2 vols. (Leipzig 1894) 1:xiii. 10 The littera maior had, however,been used for the flrstletter of a page or column,and sometimes(e.g. in the Bobbio Juvenal)also for the last letterof a page. (For a discussion of this characteristicof ancientMSS, see E. A. Lowe, Palaeographical papers 1907-1965, 2 vols. [Oxford1972] 1: 196). In a 5th cen- turyMS of Cicero's in Verrem(Vatic.Regin.Lat. 2077=CLA i.115) the enlarged 272 CynthiaDamon

This is a nearcontemporary of thearchetype of RPSY of theAmores (=a). Poem titlesin theMartial manuscript are stillwithin the column of text,but are writtenin a largerscript in black.Large initialletters are also foundin the florilegiumThuaneum of the9th or 10thcentury, again combinedwith intra- columntitles."1 And thisbrings us up to manuscriptsin whichwe have direct evidencefor poem division in theAmores (P, 9/10thcent., SY 1Ith cent.). In PSY poembeginnings are regularlyindicated by an enlargedinitial letter.Additional notice is frequentlygiven in P by capitalletter titles in black in theright or leftmargin (only omitted at 1.6; 2.13, 18; 3.7, 10, 12 and at 2.19 whereml does notindicate the beginning of a poem). Similartitles (in small letters)are slightlyless frequentin S (omittedat 1.4, 5, 6; 2.2, 8, 16, 18; 3.2, 6, 7, 9, and whereno poemdivision is indicated:1.1 [carryingon fromthe in- troductoryepigram] 1. 2, 3; 2.13, 19) and infrequentin Y (presentat 2.8, 9, 11 12).12The agreementof suchtitles as are or are notpresent in PSY is one bitof evidencethat Kenney and Munariuse in determiningtheir relationship as de- scendantsof a, and Kenneybelieves that they were in themargin of a's exem- plar.3 The irregularitywith which titles are presentis in itselfinteresting. In themanuscripts of Virgiland Prudentius discussed earlier we saw thatevery ex- tantpoem beginning was markedby an inscriptio.It seemsthat titles relegated to themargin were more liable to damageand omissionthan the intra-column tituliof ancienttexts.14 That titles were useful in preservingpoem divisionsis shown by the fact thatY, withonly 4 titlesand 3 incipits,has 12 run-on poems (notcounting 2.9a-b and 3.1la-b as run-onsfor the moment), while P,

initial is used instead at the beginningof sections in a prose work, and in the 7th centuryCodex Taurinensisof Sedulius the initialletter of the page is a little larger thanusual, but the initial letterof a section is about 4 times larger than the average letterin the text (see K. Zangemeister,Exempla codicum latinorum litterismaiusculis scriptorum [Heidelberg 1876] no. 16). 11 At least, theyare foundtowards the beginningof the codex. The poems on f. 20 have large initialletters, but on f. 51 the last few epigramsof Martial in this codex do not, althoughthe scribe dignifiesthe beginningof Catullus 62 (which follows the Martial) with a large initial. Lindsay remarksin the preface to his OCT editionof Martialthat the codex is "negligenterdescriptus." 12 I am includingin the titlecategory (for the moment)the incipitsthat begin each book. Because there are no poem titles in the principalMSS for the first poems in the books (where the incipitindicates poem division),it seems reason- able to inferthat at least partof a title's functionwas to serve as divider.An in- terestingconfirmation of thisis foundin P: the titleSUASORIUM AD SE (not in SY) is placed beside the beginningof 2.2, for whichit is not appropriate.Merkel (iv) thoughtit belonged to 2.1; it was apparentlydisplaced by the incipitof Book 2 and transferredto the next possible location. 13 F. Munari,II codice Hamilton 471 di Ovidio (Rome 1965) 58 (hereafter, "Munari,Codice"); E. J. Kenney,"'The ManuscriptTradition of Ovid's Amores, Ars Amatoria, and Remedia Amoris,"CQ 12 (1962) 7n. 3. See also Luck 44, althoughhis hypothesis,that in a the poems were writtenwithout intervening spaces and withnumbers or marksin the marginwhich were eitherunreadable or misunderstoodby copyists,fails to account for the agreementof titles (where available) in PS and Y. 14 Though problemsin poem division occurredin the lattercase, too. On the textof Martial,for example, see Lindsay (above, note 7) 40 withnote h. Poem Division,Paired Poems 273

with34 titlesand 2 incipitshas only 1. S, with22 titlesand 2 incipits,has 5.15 RUN-ONPOEMS 1.Ep.-1-2-3 RSY=a'6 1.8-9-10-11 Y 1.13-14-15 Y 2.1-2-3-4 Y 2.12-13 S17 2.18-19 PSY=a So thatwhile Muller's sola was too pessimistic,it remainstrue that to varying degreesin thevarious codices a somewhatlarger and more ornate initial letter is themost reliable indication of poem division.Just how unreliableit is, however,can be seen fromthe example of Y, where,as I havesaid, with the exceptionof 4 titlesand 3 incipits,it is theonly indication of poem division.In 6 places (apartfrom the normal poem beginnings) the initial letter of theline is largerthan usual, "quasi che il copistavolesse indicare l'inizio d'una nuovaele- gia."18P containsone sucherror,19 and at 2.19.37 thescribe of S has leftspace fora largeinitial letter and addedthe title AD AMiCAM.With the exception of thislast, these errors are not recorded in othermanuscripts. FALSE POEM DIVISIONS 1.2.15 Asper Y 1.4.13 Ante Y 1.6.9 At Y 1.6.17 Aspice Y 1.6.27 Ferreus Y 2.3.15 Fallere Y 2.19.37 At S 3.7.19 A, pudet P

15 Cf. Luck 92, wherehe relatesloss of titlesto errorsof poem division in the archetypeof Catullus. 16 Since R ends at 1.2.50, only SY show the run-on continuinginto 1.3. Scholars are inclinedto thinkthat the missingpart of R (1.2.51-end = R') was the exemplarof P, because the bottomof the last folio of R (which contained5 lines on the recto 1.2.20-24, and 4 or 5 on the verso 1.2.51-3.1 or 2) has been tom off and 1.2 51 is foundat the beginningof a gatheringin P, whichdoes not contain the run-onbetween 1.2 and 3. See Kenney,(above, note 13) 6-7; G. P. Goold, "Amatoria Critica," HSCP 69 (1965) 4; F. Munari,P. Ovidi Nasonis Amores: testo, introduzionee note (Florence 1951) xix-xx; S. Tafel, Die Uber- lieferungsgeschichtevon Ovids Carmina amatoria [diss. Tiubingen1910] 26-31. Contra, D. S. McKie, CQ 36 (1986) 219-38. 17 Note thatPY have no titlehere, so perhapsthe absence of a titlein a led to the run-onin S. 18 Munari,Codice 17. He continues"Forse notevoli,ma per me inspiegabili, sono qui tre fatti:1) il fenomenoe limitatoall'inizio del libro I (fattaeccezione per l'ultimo esempio); 2) tuttii versi comincianocon A o F; 3) i primidue casi si trovanonella primariga della pagina." Observe thatthe errorsof this sortin P and S similarlyinvolve A's. 19 Muller (above, note 3) 89; Munari,Codice 60. 274 CynthiaDamon

Errorsof poemdivision, then, occur not infrequently in theantiquiores.20 In orderto correcterrors already present in a (e.g. therun-ons 1.Ep. 1-2-321 and 2.18-19) one maylook at poemdivisions in manuscriptsof the,3-strain and/or apply non-traditionarycriteria to evaluatethe appropriateness of transmitted poemboundaries. In theabsence of published collations of therecentiores it has been impossibleto assemblea list of poem divisionsshown by O3'sdescen- dants22or to assess theirinfluence on thedivisions inserted by thecorrectors of P and Y at the"proper"23 places in all therun-ons in theirrespective texts. The need fornon-traditionary criteria by whichto judge thesedivisions (and by extensionMuller's divisions in 2.9 and 3.11, whichare nowhereattested) remains,however, and an attemptto establishsuch will occupythe rest of this paper.

Section2: Argumentsfor poem division By the 12thcentury, Y, at least,had thepoems divided in such a way as satisfiedreaders and critics until Scaliger and then Bentley objected to thesepara- tionof 2.2 and 3.24 The unanimousconsensus of the manuscriptsas to the unityof 2.9 and 3.11 remainedunchallenged much longer, until in 1856 L. Mullerinsisted, upon aesthetic grounds, on theirdivision.25 With this he set in

20 S, in fact,demonstrates one source of this fallibility.Although the firstline (6 times) or two (28 times) of each poem were indentedto provide space for a large initialletter, it was never put in. While in some cases the appropriateletter (or letters-in 2.12 and 2.17 the firstletter of the second line is omittedas well) is noted in the margin,the manuscriptoffered subsequent copyists plenty of opportunityfor error,particularly since the indentationitself as well as the marginalhelp-letter was absentat 1.4, 2.2, and 2.5. 21 1.2-3 are joined in SY, hence presumablyin a. Since the beginningof 1.3 is markedby a large initial letterand marginaltitle in P, however,and since S and Y both demonstratean inclinationto createoriginal run-ons, there remains a possiblitythat the divisionwas presentin a and thatS and Y made simultaneous errorshere. But cf. C. E. Murgia,"The Date of Ovid's Ars 3," AJP 107 (1986) 90n. 26, arguingthat "the divisions of poems found in the MSS of not only Propertius,but Ovid's Amores and Tibullus, reflectmedieval editorialdecision ratherthan transmitted evidence." 22 I have been able to glean only the followingscraps of information:accord- ing to Munari's edition 1.Ep and 1.1 are separatedby Politianus,Marius and the Codex Arundelianus(now Edinburgh,Bibl. Nat. 18.2.9. See M. D. Reeve RhM 117 [1974]: 142 and 138). 1.1 and 2 are separatedin EX2G2(Kenney's Ea, Vb second hand, Ab second hand). 2.18 and 19 are separateand have titlesin B and H (Kenney's Va and H). The false divisionof S at 2.19.37 is also presentin BAJ (B=Kenney's Va, othersnot in Kenney). 23 The correctionsto P are noted in Munari's collation of P (SIFC 23 [1948] 113-52), but not in his editionof the Amores. I am inferringfrom the phrase "nullumintervallum" in Lenz's collationof S (RIL ser.2, 69 [1936] 633-57) that the properdivisions were never indicatedin S (by "proper"or "correct"poem divisions I mean simplydivisions that have been acceptedby editorsand criticsto date). 24 For thisthey had the supportof the 15thcentury Codex Hafniensis. 25 "Haec in uno carmineconiuncta fuisse nunquammihi persuadebo.Ita mutati animiexemplum in eodem carminenusquam inveni, si exceperisde quo infradice- tur III 11; nec potest tale existere.Etenim non puerilemodo, sed plane est inep- PoemDivision, Paired Poems 275 motiona debatewhich continues even today.26 The unifiers have been more diligentin seekingout arguments. Jacoby argues that 3.11 is a singlepoem sinceit is basedon a singlemodel (Catullus 85).27 Jager sees both 2.9 and3.1 1 as poemsof "seelische Entwicklung" and "innere Kampf" and of 2.9 concludes that"Die Einheitdes ganzen Gedichts besteht wiederum darin, daB ein in sich zusammenhiingenderseelischer Vorgang unmittelbar dramatisch dargestellt ist."28Cairns provides generic arguments for unity: since 2.9 and3.11 are re- nuntiationesamoris (and in thisgenre "change of mind"frequently occurs) the contradictioninthe second halves of the poems of all thatwas said in the first halvesis somethingone might expect to fmd, not an aberrationtobe dealtwith surgically.29These arguments30 share a fundamentalweakness: they assume that tumita subitoconverti animum poetae, ut quod non uno alteroveversu sed per tot disticha omnibus precibus devovit, id iam omnibus precibus expostuletet ef- flagitet. Hoc autem evenit, nisi putarismecum a v. inde 25 novum carmen incipere" (90). The possibilityof dividinga poem withinthe Amores had been consideredand rejected 18 years earlier by Otto Gruppe (Die romische Elegie [Leipzig 1838] 375-79) who, however,was not concernedwith consistencyin poetic units,but was looking for arithmeticalniceties in the collection. 26 Kenney (above, note 2) notes with approval (x) that by removing3.5 and dividing 2.9 and 3.11 he gets books with 15, 20 and 15 poems respectively (which, he says, could hardly have come about by chance). J. C. McKeown (above, note 1), afterreviewing the numbersof poems in books of Augustanpo- etrv (91-92), printsthe poems as single units in his text. 2 F. Jacoby,"Zur Entstehungder romischeElegie," RhM 60 (1905) 86-87. That is, fromthe unityof the model, the epigram,he infersthe unityof the elegy. But why, in view of the clear referenceto Catullus8 in 3.1 la (perferet obdura, 7 cf. Cat. 8.11 sed obstinata menteperfer, obdura) and greatersituational similarityof these two poems (i.e. makinga resolution),should the slightlyless directechoes of Catullus 85 in 1lb (hac amor hac odium,34 cf. Cat. 85.1 odi et amo) lead us to posit unityof model? Moreover,Weinreich (Die Distichen des Catull [Darmstadt 1964] 72-76) has shown that Catullus 85 is not the only treatmentof simultaneoushate and love, in otherwords thatboth poets were uti- lizing a topos, the earlier fashioningfrom it a 2-line epigram,the later finding in it matterfor 25 couplets. Indeed, I thinkit is this, thatOvid took a demon- strablysuccessful epigram theme and expandedit into a full-blownelegiac treatment, that is Jacoby's point, and when critics like Weinreich (1004n.41) and Jager(Zweigliedrige Gedichte und Gedichtpaarebei Properz und in Ovids Anores [diss. Tilbingen,1967] 142; hereafter,"Jager") say thathe has made a strongcase for the unityof 3.11 theyare undulystressing an almostincidental remark. Wein- reich himselfcorrectly perceives (72-76) thatthe "unityof model" argumentdoes not apply here and claims that "Die innere Einheit des Gedichts kann darin erblicktwerden, daB der ersteTeil den Grundfiur das Hassen vorfuhrt,wihrend der zweite den Zwiespalt ausmalt." This analysis of the poem is unexceptionable enough,but is it proofof unity? 28 Jager144-53. His views are largelyechoed by G. Lorcher(Der Aufbau der drei Bucher von Ovids Anores [Amsterdam1975] 15-23). The additionalargu- ments she adduces (17, 22-23)-the equivalence of the corona of 3.11a.29 and the votum of l1b.40, and a 3-partstructure for 2.9-are unconvincing.Jager had criticized(149) Mtiller's argumentfor being "grundsatzlichsubjektiv" and seems to feel thathis defense of unityis somehow less subjective because the poems have been transmittedundivided. As we have seen, however,the transmissionof poem divisions is not entirelyreliable, and the archetype'sdivisions must be submittedto the same examinatio as its texL 276 CynthiaDamon the only compositionalunit available to Ovid (or Propertius)in whichar- tisticallyrelated sections sharing subject matter and language3'were possible was the singlepoem.32 A briefreview of two poems,Amores 1.11-12, will makeit clearthat this is notthe case. Thesepoems clearly have the same subject matter, the tabellae going to ask fora rendezvousand returning with the answer "no." The speakerand addressee are thesame in both.Enough time has elapsed betweenpoems, however, for Nape to deliverthe letter and bringback thereply, so thesituation is notquite identical.33As forartistic interaction, it is presentat severallevels. To begin with,it is difficultto understandthe situation at thebeginning of 1.12 without thebackground of 1.11: theaddressee and contentsof thetabellae are unknown, thereference to Nape trippingover the threshold(1.12.4) is obscure(who is Nape and whatdoes she have to do withthe tabellae?), the interlocutorat 1.12.5-6 unnamed.Then 1.12 clearlyreflects the structureand languageof 1.11, thoughI wouldnot go so faras Davis and call it"a point-by-pointcontra- dictionof everythingthe poet said in theprevious elegy."34 He does, however, correctlypoint out that for there to be parallelismat all, one mustoppose Nape 29 This argumentwas initiallyadvanced (in Generic Compositionin Greek and Roman Poetry [Edinburgh1972] 139-40) for 3.11. Note that wheneverCairns uses generic considerationsto defend the unityof a poem, thatpoem is a non- standardmember of its genre: on Am. 3.11, see p. 139, Prop. 1.8, p. 150, 2.28, p. 154. In "Self-imitationwithin a generic framework"(Creative Imitationand Latin Literature,edd. D. West and A. J. Woodman [Cambridge1979] 121-41) Cairns returnedto the fraywith a certainamount of exasperation,insisting that unityhas been "amply demonstrated"or "demonstratedbeyond doubt" for both 2.9 (127) and 3.11 (131). He too (127-32) catalogs "internalcorrespondences" betweenhalves and diagramsstructural relationships (ring structurein 2.9, parallel in 3.11). He also (130) points,Jacoby-like, to a reminiscenceof Prop. 1.9.8 (atque utinamposito dicar amore rudis) in the crucial bridginglines 2.9.22-25 (noting 'posito', dicat 'amore'[25, cf. deposito, 22] and the "witty"recall of Propertius'adjective rudis withthe noun rudis [22]). With the exceptionof this last, the featureshe notices are presentand important,but they cannot prove unityfor reasons to be discussed. 30 The argumentfrom book structure-a clear book structurerequiring either a single poem at the place wheredivision might be made (as, e.g., in Prop. 4.1, on which see E. Courtney,BICS 16 [1969] 73) or two poems whereone is currently found(as, e.g., at Prop. 1.8, on which see the lengthynote by J. T. Davis, Dra- matic Pairings in the Elegies of Propertiusand Ovid [Bern 1977] 27-29n. 2; hereafter,"Davis")-has not been profitablyapplied to the problemof 2.9 and 3.11. The argumentsof G. Wille ("Zum kuinstlerischenAufbau von Ovids Amores," in Navicula Tubingensis: studia in honoremAntonii Tovar, ed. F. J. Oroz Arizcuren[Tilbingen 1984] 389-423) are quite insubstantial. 31 I have borrowedthese criteriafor unityfrom R. E. White,"The Structureof Provertius2.28: DramaticUnity," TAPA 89 (1958) 254. 3 Jager,for instance, says (148) of 3.11 "Leitmotive tragen dazu bei, die Einheitdes Gedichtszu unterstreichen"and goes on to show verbal and metaphor- ical echoes of Ila in lib. 33 Accordingto Davis (80 withnote 18), the lover entrustedthe tabletsto Nape one evening to be delivered to the lady the next morning(mane 1.11.7), but thereis no need to wringextraneous situational details out of the poem. Mane, afterall, mightbe takenwith peraratas to emphasize the lover's eagerness and volubility even at an early hour in the morning.So Lenz and Barsby com- mentariesad loc. Poem Division,Paired Poems 277

(enjoyingpraise 1.11.1-6) in the firstpoem to the tablets(suffering blame 1.12.9-14) in thesecond. This substitutionhe explainsby referenceto Ovid's "self-interestand self-possession,"noting that Nape will be needed iterum (1.12.5).35Verbal echoes make the new focus the clearer: Nape isfida (1.11.6) and in ministeriis...utilis (1.11.3-4), thetablets are fidas ministras(1.11.27) beforethey report the discouraging answer and inutilelignwn after (1.12.13). The militarymetaphor, too, is appliedto bothNape and thetabellae (militiae signa tueretuae 1.11.12;victrices tabellae 1.11.25). The followingchart shows thestructural parallels, admittedly imperfect, between the two poems: 1.11 1.12 (1-6 descriptionof situation) 1-6 rais forNape 7-8 command:ite hinc... 7-8 command:accipe.. 9-14 blaen fortabellae 9-12 whyNape is usefulto 15-20 whytabellae are notfit carrylove letters to containlove letters (13-18 instructions) (21-22 self-reproach) 19-24 what0. hopesthe 23-26 whatthey should be used tabellaewill be usedfor for 27-28 self-reproach 25-28 theirprojected fate: 29-30 theirprojected fate: dedicationto Venus decay Finally,the charm of thepoems lies in theirinteraction. The elatedlover, who was so surethat he wouldget theanswer he wantedthat he had alreadycom- posed an epigramdedicating the tablets to Venus,gets his come-uppancewhen theyreturn with a "no" and is similarlyextravagant in his depression.The "no," afterall, onlyapplies to thatevening (hodie, 1.12.2). The persona of each poem takenby itselfis thelover typical of elegy,the juxtaposition of thetwo creates a good-naturedmockery of that persona. No one, to myknowledge, has ever suggestedthat these two poems be combinedinto a singlepoem, but they clearly constitute a compositionalunit withsimilar subject matter and artisticinteraction between parts. In lightof this,it seemsthat White's criteria for unity will notwork, and if a pairof poems can be considereda structuralunit, the other arguments for the unity of 2.9 and 3.11 are greatlyundermined. In whatfollows I willlook at thepaired poems in theAmores-poems so closelylinked by subjectmatter and situationas to forma compositionalunit-to see if theycan give us any toolswith which to makeeither a confidentdivision or a moresecure internal bond for 2.9 and 3.11.

Section3: Consecutivepaired poems Althoughthe term "paired poems" can be usedto describepoems related in many ways,36I will be consideringhere only thosepairs thatare relatedas 34 Davis 81. 35 Davis 84. 36 Virgil's book of Eclogues, for example, contains thematicpairs (e.g. 1 and 9, both about land confiscations),formal pairs (e.g. 3 and 7, both amoebean 278 CynthiaDamon

2.9a-band 3.1 la-b arerelated if they are pairs, that is, spatially (i.e. thepoems areadjacent) and causally (i.e. the situation inthe second poem presupposes the eventsof the first). There are four such pairs in the Amores: 1.11-12, 2.2-3, 7- 8, 13-14.37These have been labelled "dramatic pairs" by Davis, and the discussionof them that follows is largelybased on his analyses. In Davis'model of the dramatic pair, the temporal relationship between po- emsis specifiedwith some exactitude: the order in whichthey occur is irreversible,there is a pausebetween the two poems in which the action continues, andthe entire three-episode incident is presentedas ifin progess. There is, furthermore,a causal relationship joining the three episodes: events arising out of thesituation inthe first poem occur during the pause, and the second poem con- tainsthe speaker's reaction to those events (and not to the first poem). One such eventis frequentlythe departure of theaddressee of thefirst poem, which changesthe tone of the speaker's remarks about that person. In writinga dramaticpair the poet is confrontedwith the technical problem of signaling to the readerthat the break between poems which he sees in histext (which looks, presumably,like any other poem division) is ratherakin to a scenechange than toa breakbetween unrelated entities. This can only happen if the reader finds unambiguoussituational and verbal references tothe first poem in the first few coupletsof the second. These features, then-a temporal and causal relationship ofthe three episodes contained in the two poems and the pause between them, anda clearindication of theserelationships very early in thesecond poem- characterizethis compositional type.38 This model, however, is a compositeput togetherfrom four Propertian and four Ovidian examples, and for the purposes ofthis paper is willbe importanttosee howthe real pairs in the Amores flesh outthis skeletal model.39 We havealready seen that the pair 1.11-12 is builtout of three episodes- 1.11contains the lengthy preliminaries to the dispatch of the tabellae, during thepause the request and replyare exchanged,in 1.12 thelover reacts to Corinna'sinfelix littera (2). Theorder of these events is clearlyirreversible. At thebeginning of 1.12 (whichat firstreading we expectto be an unrelated poem4O),we hear that some tabellae have returned from somewhere (1) andthat dialogue) and pairs linkedby source (e.g. 2 and 3, both imitationsof Theocritus), and it is out of pairs such as these thatthe structureof the book is built up. In theAmores, on the otherhand, thereare pairs withclose thematiclinks (e.g. 1.4 and 2.5, 2.19 and 3.4) whichserve no obvious structuralpurpose. 37 Jager adds 2.11-12, but the term "pair" is uncomfortablystrained by its applicationto 2.11-12, 2.11 being a propempticon and 2.12 a celebration,not of Corinna's safe return,but of her successfulevasion of vir, custos,and ianua (2.12.3). Davis discusses fourPropertian pairs as well: 1.8a-b 1.11-12, 2.28a-b, 2.29a-b. 38 In Davis' view (21-22), 2.9 and 3.11 do not display all of the characteristics of this model. He does believe they should be divided, however. On this, see Section 4 below. 39 I have preferredDavis' model to Jager'sbecause the latter-compactnessof presentation,dramatic treatment and polar oppositionof poems-is less clearly defimed(35) and consequentlymore difficult to applyusefully. 40 1.11 appears a perfectlyordinary self-contained entity and we are satisfied thatit has come to a close when we see importantwords fromthe beginningof the poem reappearing in the last couplet (fidas sibi ministras,27, cf. in Poem Division,Paired Poems 279

Nape (who appears only in thispair) is again involved(4). The speakeris thinkingof somethingthat happened just recently(modo, 3) and whichpor- tendedhis present unhappy state (omina, 3). All thissuggests a connectionwith theprevious poem, in whichNape was entrustedwith the delivery of some tabellae of whichhe had greatexpectations. The termsof thereply (hodie posse negat,2) hintthat perhaps the tabellae contained a requestfor an appointment, which accords too well with the situationof 1.11 (where the lover wants Corinnato thinkhe is longingfor an eveningwith her, spe noctisvivere dices 1.11.13) forthere to be no connectionbetween the two poems. Then, the gaps in thereader's understanding of thesituation of 1.12 (to whomwas theletter sent?was a rendezvousreally the object? who is Nape andwhat does shehave to do withthe tabellae? who is beingaddressed in lines 5-6?) are filledin very nicelyif he assumesthat the situation sketched in 1.11 is retainedin 1.12. The pointof cera refertain 1.12.8,too, only becomes clear against the background of 1.11.19-24 wherethe lover speculates about just how fullhe wouldlike the cera to be in Corinna'sanswer (and finallydecides that a simpleveni would be sufficient).Verbal echoes do notplay a particularlyimportant role in alerting thereader to theconnection between 1.11 and 1.12,though tabellae (1.12.1) and Nape (4) do recallthe situation of 1.11 by repeatingthe words for important elementsof thatsituation.41 The Cypassispoems, 2.7-8, serveas Davis' paradeigmaticpair.42 They show thecharacteristic irreversible three-part temporal and causal sequence:in 2.7 the lover hotlydefends himself against the charge of infidelity,during a pause of uncertainbut not momentary duration Corinna passes judgmentand departsand thelover and Cypassiscontrive a meeting(or are lefttogether by Corinna,which seems less thanprobable), then in 2.8 thelover tries to recover Cypassis'goodwill, which probably withered somewhat under the scorn she had heardheaped upon the idea of an affairwith a slave (2.7.19-22, 25-26, cf. 2.8.9-10). The firstpoem is broughtto a neatclose by theecho of thefirst line (reus in crimina)in thelast (criminis esse reum).But thismakes it all themore delightfulwhen the situationsketched in the firsttwo lines of the second poem-thata hairdresseris involvedagain (2.8.1, cf.2.7.17, 23-24) and thatit is thevery Cypassis mentioned in theprevious poem (and nowhereelse)-de- mandsthat the two poems be readas a pair.A particularlynice featureof this pair is that2.8 greatlyimproves our appreciationof 2.7 by showingthat Corinna'saccusation was, in fact,justified (which we mayhave suspectedall ministeriis utilis,3 andfida reperta mihi, 6). For a general discussion of ring compositionin the Amores see F. Bertini,"La Ringkompositionnegli Amores Ovidiani e I'autenticitadell'Elegia 11.5," RCCM 18 (1976) 151-60. In thiscase, the verbal echoes serve two purposes-in the contextof the poem theygive a sense of closure,in the contextof the pair theyease the shiftof focus fromNape to the tabellae. The dedicatoryepigram, too, is frequently,though by no means always, encounteredat the ends of poems (cf. Am. 2.6, Her. 2, 7, Prop. 4.3; Tib. 1.9, [31.2). 41 The use of the epithetinutile (1.12.13) for the disappointingtabellae is certainlya reflectionof the utilis used to complimentNape at 1.11.4, but occurs too late to informthe reader thatthe poems are related,though it may confirm his impressionthat they are. 42 Davis 19-21. 280 CynthiaDamon along). The discoverythat Nape was presentduring that interview (2.8.15-16) also servesto makeour picture of thesituation clearer and explainswhy the sequel was necessaryat all. At theend of thesecond poem we again finda refer- ence back to itsfirst line (quotque quibusque modis, 28, cf.in millemodos, 1), and thepair as a wholeis drawntogether and tiedoff, so to speak,by the recurrenceof index(2.8.25), whichrecalls indicio (2.7.26) fromthe end of thefirst poemand index (2.8.5) nearthe beginning of the second.43 The lastof theundisputed pairs is 2.13-14. In thefirst poem Corinna lies seriouslyill fromthe effectsof an attemptedabortion.44 The eventsof the pause betweenthe poems are neverexplicitly described,45 and we can onlyinfer thatCorinna's condition has improvedby assumingthat, if in 2.14 thespeaker gives way to his ira (whichwas formerlyrestrained by his metus,2.13.4, cf. 27), somethingmust have happenedduring the pause to removehis fear.The firstpoem of thepair is broughtneatly to a close-a dedicationconcludes the prayersection of thepoem, and thehesitancy with which the final reproach is uttered(si tamenin tantofas est monuissetimore) recalls the metus (4) of the beginning.At thestart of 2.14, however,the reader is leftgroping for a context in whichthe exasperated but unspecific complaint of lines 1-4 makessense. The discoverythat it is an abortion(teneros convellerefetus, 5) thathas provokedthe speaker's anger and the continuation of themilitary metaphor from thelast line of thepreceding poem incline him to connectthe two poems, and whenhe needs an identityfor the addressee of 2.14 (at tuae, 8) thegirl who attemptedan abortionin 2.13 is themost likely candidate. These situational linksare thestronger in thatthe subject of abortionis nottreated elsewhere in theAmores, so thatthe relative scarcity of verbalechoes need notpreclude a connection.46 The threeepisodes of thepair 2.2-3 are,in outlineat least,quite similar to thoseof 1.11-12. In thefirst poem of each pair,the lover tries to persuadea slave fromhis mistress'household to do hima service,during the pause he re- ceivesa negativeresponse (in thecase of 1.11-12,a "no" fromthe recipient of thetabellae, not from Nape), and in thesecond poem he ventshis angeron the conveyorof thatresponse (in 2.3, on Bagoas himself,in 1.12, on thetabellae).47The durationof thepause between 2.2 and 3, however,need only be long enoughfor the custos to denythe lover's request for entrance, whereas between 1.11 and 12,Nape has deliveredthe tabellae, presumably dressed her mistress'

43 The end of the conciliatorysection of 2.8 (1-22), too, is signalizedby ring composition:concubitus..-tuos (22) recalls the same phrase fromline 6. 44 Davis argues (111-12) fromthe presenttense of iacet (2.13.2) and the prayer addressed to the goddess who presides over childbirth,Ilithyia (19-26), thatthe actual miscarriagehas not yet takenplace during2.13. 45 Of course if the events of the pause were describedin the same way as the eventsof the poems are described(i.e. as if in progress),they would not happen duringa pause, but it would be possible to narratewhat happened during the pause and still preservethe distinction. 46 Such words as do recur (e.g. ventre, gravida, temerasset2.14.15, 17, cf. ventris,gravidi, temeraria 2.13.1) are bound up withthe subjectof abortion. 47 I shall use the eunuch's name throughoutin the form to which Kenney emendedit ("Notes on Ovid," CQ 8 [1958] 59-60). Poem Division,Paired Poems 281

hair,48and foundan opportunityto returnwith the infelix littera. The orderof eventsin 2.2-3 is irreversible.The feelingof closureat theend of the first poem is notparticularly strong: there is no ringcomposition and thesummary precibus in the last line merelyindicates that one attemptat persuasionhas come to an end and a responseis expected.49When thereader starts on 2.3 thinkingthat he has a new poem in hand,he meetswith an exclamationof distress(ei mihi,cf.flete meos casus, 1.12.1) whichis "explained"by thefact thatthe addressee, a custos,is againa eunuch(quod dominamnec virnecfemina servas /mutua nec Venerisgaudia nossepotes, 1-2).50 As was thecase with Nape (1.11-12) and Cypassis(2.7-8), neitherBagoas norany other eunuch in the role of a custos appears elsewherein theAmores. The verbalecho of dominam...servas(1, cf.quem penes est dominam servandi cura, 2.2.1) provides a furtherlink to theprevious poem, and theform of servas showsthat the ad- dresseeis thesame in bothpoems.51 Now, accordingto Davis, thisis unusual in thecontext of a dramaticpair. With the partial exception of 1.11-12 (where Nape is theaddressee in thefirst poem and forat leasttwo lines, 5-6, nearthe beginningof thesecond),52 dramatic pairs in Propertius and Ovid showdifferent addresseesin thetwo poems, with the result very often that "the person spoken to in thefirst poem is spokenabout in thesecond as thoughthey are notwithin earshot."153Several things then-the continued presence of Bagoas, thebrevity of thepause (and perhapsalso of thesecond poem), the similarity of situation between2.2-3 and the singlepoem 1.6 and therelatively open endingof

48 Davis 80. 49 Pace Lenz ("Ovidio: Amores II 2 e 3, una sola poesia?" Maia 17 (1965) 121; hereafter,"Lenz"): "Queste parole segnanoinequivocabilmente la finedel discorso al guardiano."Comparison with 1.6 is instructive,for in that address the lover pauses threetimes (afterlines 20, 26, 40) to see if the line of argumenthe is then followingis having any effect.The comparisonof 1.6 and 2.2-3 is also rewardingfor its juxtaposition(incidental to our purpose,but none the less amus- ing) of the ianitor of 1.6, whose uncooperativenessis hypotheticallyascribed to his being in bed with his own amica (45-47) and the uncooperativecustos of 2.2-3, whose deficiencesin thatarena are dwelt on at some length(2.3.1-6). 50 The citationsprovided by Brandtand Nemethy(especially PlinyN.H. 13.41 in hortoBagou: ita vocantspadones, qui apud eos [Persas] etiamregnavere) suffi- ciently demonstratethe implications of the name Bagoas. Cf. L. Alfonsi (Latomus 28 [1969] 208): "Questa tradizionecospicua, filosoficae scolastica, formatasiin epoca ellenistica,ha fattodi Bagoo il tipo dello spado per eccel- lenza." See also his earliernote, Latomus 23 (1964) 349. For a detailed discussion of a famousBagoas, see E. Badian, "The EunuchBagoas," CQ 8 (1958) 144- 57. 51 This echo at the beginningsof the poems is balanced by one at the ends (precibus, 2.2.66, 2.3.17) which serves to round off the pair very neatly. The same ringcomposition over the pair as a whole is foundin 2.7-8 and 13-14. 52 Though it has been arguedthat, in view of the 3rd personreference to Nape in line 4, lines 5-6 are an apostrophe.See Davis 79. Otherwise,of course, one assumes thatthe second couplet is an aside, not addressedto Nape thoughshe is present. 53 Davis 21. 54 And this is so thoroughas to show the identical exclamation of disap- pointment,ei mihi (2.3.1, 1.6.52), at the same strategicpoint-the transition fromentreaty to threat-in the lover's discourse. 282 CynthiaDamon

2.2-might induce one to wonderwhether the poems ought,in fact,to be united.Scaliger's and Bentley's reasons for suggestingprecisely this are unknown,and thesupport of the 15thcentury Codex Hafniensisis virtually worthless,yet the arguments for retaining the division have notbeen entirely convincing.Because thispair alone provides a precedentfor the pause of onlya moment'sduration between poems of a pair,it seems worthwhile to examinein somedetail the arguments for the division which the best manuscripts attest.55 In an articledevoted to thequestion, F. Lenz makesthe following main points56:1) theending of 2.2 wouldbe spoiledby theimmediate addition of 2.3, in whichthe request is madeagain (lines 17-18); 2) thereis an important changein situation-in2.2 thespeaker has onlyjust seen thepuella (line 3) whilein 2.3 thereexist "rapporti intimi"57 between the two and he can already speakfor her (fallere te potuit...non caret effectu quod voluereduo, 15-16); and 3) therehas beena concomitantchange of tone-aftera well-calculatedprogres- sion fromcontempt58 to amiabilityin 2.2, in 2.3 thespeaker rudely refers to theeunuch's deficiencies.59 Jager notes the change of toneand adds twopoints:

55 Pauses of a moment'sduration are of course frequentlyfound within single poems, generallyin poems in which the action is presenteddramatically rather than narrated.There are pauses before, e.g., 1.14.51, 2.8.23, 2.14.41, 3.2.19, 3.3.41. 56 See above, note 49. I have omittedsome of his argumentsin this summary. He findsthe combinedlength of the poems (84 lines or, if one omitsthe disputed verses, 74 lines) much longer than average for its book. (Average length for Book 2 [not including2.9] is 42.1 lines, for Book 1, 51.1 lines and for Book 3 [not including3. 5 or 11] 59.2 lines.) But this proves nothing,for both Book 1 and Book 3 also containpoems whose lengthis much longer than average: 1.8 (114 lines) and 3.6 (106 lines). One mightpoint to the placementof 2.2-3 near the beginningof its book (the otherlong poems are more centrallylocated), but the argumentwould be tenuousat best. I cannotassess the value of Lenz's second point,that the markseparating 2.2-3 in Y differsfrom that which separates2.3- 4 or 2.4-5, since I have not seen a photographof the poem division and the dif- ferenceis not notedby Munari,but I do note thatLenz is mistakenin sayingthat 2.4-5 are run togetherin Y (2.1-2-3-4 are run on, but 2.5 is apparentlysepa- rate). He also findsconfirmation of his argumentin Ovid's use of the Cporroq e{p?,vrin theme(2.3.3-4), which,he says, marksthe begiing of a poem. The topos of thenp(-roq e{p?5rijq does not seem to have a necessaryaffinity for any one positionin a poem. It occurs at or near the beginningof Am. 2.11 and 2.14 and Tib. 1.10, mid-poem in Prop. 1.17.13-14, 4.3.19; Tib. 1.4.59-60; Hor. Odes 1.3.9-12 (to name only a few). 57 Lenz 122. 58 Lenz thinksthat the parodyof the too-solemntone of the beginningof 2.2 implies"profundo disprezzo" (123). Yet the parodyseems more for theperception of thereader than for Bagoas' ears, and I would say thatthe movement of thesua- soria (as opposed to thatof the poem) is fromsomewhat overdone flattery to the aforementionedamiability. 59 Lenz finds the tone of 2.3.1-6 "piiu compassionevolee sdegnato che bef- fardo" (123), but it seems to me that,in view of ei mihi (1) and the Ovidian lover's well-knownselfishness, the indignationof lines 3-4 is at least as much for his own sufferings(i.e. because what he is sufferingat the momentis the indirectresult of thatfirst castrator's action) as for those of the eunuch. On the tone of 2.3 see also J. Booth, "Double-entendresin Ovid, Amores 2.3," LCM 8 (1983) 101-2. Poem Division,Paired Poems 283

4) therhetorical status changes (in an unspecifiedmanner) at thebeginning of 2.3; and 5) because thereader has to deducewhat happened during the pause, 2.2-3 is a dramaticpair (in whichcompositional type this indirectnessis common)rather than a two-partpoem (in whichthe events of thepause are ex- plicitlydescribed).60 Davis observesfurther that the structure of 2.3 reflectsthat of 2.2 and considersthe case forunifying 2.2 and 3 "all butclosed."61 But is it? The amiabletone of thefinal request in 2.2, whichLenz considerssuitable onlyfor the end of a poem,is in realityequally (or more)suitable for the end of a suasoria,which need not coincide with the end of a poem.62That the request is repeatedafter a differentkind of persuasion has beenattempted need not violate theintegrity of thefirst requesL As to thesecond point, if one could provethat a lengthof timesufficient forthe development of "rapportiintimi" had elapsed between poems, one could withconfidence enroll 2.2-3 amongthe other dramatic pairs. But theevidence forintimacy is thinindeed. Lenz's distinctionbetween the meaning of dominam (sc. tuam) at 2.2.1 and thatat 2.3.1 (sc. tuam or meam), in despiteof the obvious verbalconnection between the lines, is simplyarbitrary.63 Given the presenceof theverb servare, which can mean"keep an eyeon" or"observe," but is especiallyused forkeeping something safe for someone, the surface meaning in bothplaces is likelyto be dominamtuam, but the ambiguityarising from thefact that the reader assimilates the word dominam in an elegiaccontext be- forefitting it to servareadds a touchof piquancy-also in bothplaces. As addi- tionalproof of thenew intimacyLenz citesthe non-gnomic perfect tense of potuit(2.3.15) and voluere(16), whichhe thinksmust be based on some discussion,at least,between the lover and thepuella. But in theimmediately preced- ing lines thespeaker has givenone justificationfor his presumptionin speak- ing on behalfof thegirl: love is whatis herbeauty and age are fitfor (apti 13) and it wouldbe a shameto wastethem (14), thereforeshe wouldhave wanted whathe wantedas soon as theopportunity presented itself (in theform of his note,2.2.5), hencethe perfect tense. We can see fromher replies-she saysthat therendezvous is notpossible (not that she is notwilling; non licet, 2.2.6) and thatit is notpossible because there is a custosin theway (notbecause she has scruplesof herown, 8)-that indeedthe lover is justifiedin his voluere.No greaterdegree of understandingbetween the "lovers" should be postulatedfor the backgroundof 2.3.6f

60 Jager30. 61 Davis 97 and 86n. 1. 62 Of the poems containingsuasoriae (using the termloosely for any mono- logue in which the speaker is primarilyconcerned with giving general advice, e.g. 1.8.23-108, or persuading the interlocutorto do somethingspecific, e.g. 1.6) in theAmores, 5 also describethe aftermathor responseto the speech (1.6, 8, 13; 2.8, 3.6) and 8 do not (1.3, 4; 2.3, 11, 18, 19; 3.4, 14). In the case of the apologia of 2.7, the responseis given in the second poem of the pair. 3.2 is not exactlya suasoria, since much of the persuasionis non-verbal,but the girl's renly(also non-verbal)is describedin the second to last line of the poem (83). 3SLenz 122. 64 Davis (86n. 1) rightlydissents from Lenz on this point, thoughhe accepts the restof the argument. 284 CynthiaDamon

Neitherchange of tonenor change of argument demands a newpoem to ac- commodateit. In 2.8 thechanges in thelover's tone are parallelto (and more extremethan) those in 2.2-3; he movesfrom the conciliatory flattery (1-4) and fondwheedling (21-22) whichprecede the proposition to reproach(ingrata, 23) and seriousthreats (25-28)65 after Cypassis' "no," yetthese changes are com- fortablyand effectivelycontained within the limits of a singlepoem. Different stylesof argumentation,too, are foundin singlepoems. In 1.6, forexample, thealternative tactics of entreaty(3-39)66 and threat(57-60) are thetwo halves of thesingle attempted subornation of theianitor (as, indeed,the lover himself says: omniaconsumpsi nec te precibusqueminisque movimus, 61-62) and as suchcan (whichis notto say must)be presentedin a singlepoem. Jager'sassertion that 2.2-3 belongto thecategory of dramatic pairs because of theway in whichthe reader learns what happened during the pause deserves carefulconsideration. He findsthat in individualpoems which contain a pause and situationchange of some sortthe reader is alwaysinformed as to exactly whatevent in thepause promptedthe change of situation.67His examplesare 2.8.23, wherethe change of tonefrom wheedling to threateningis explained by quid renuisfingisquenovos, ingrata, timores?; 3.2.19, where the speaker is star- tledout of his enthusiasticimaginary race at line 18 and we learnfrom quid frustrarefugis? cogit nos linea iungithat he was all thewhile inching closer to thegirl on hisright (in accordancewith the precepts at Ars 1.139-42) whileshe triedto moveaway; and 1.14.51-52,where the lover's increasingly thoughtless diatribeis broughtto an abruptconclusion by thegirl's flood of tearsand blush of shame(me miserum,lacrumas male continetoraque dextraI protegitingen- uas picta ruboregenas).68 Specific details of thiskind are lacking,he says,at thesituation change between 2.2 and 3-the readermust infer from the con- trafactualcondition in 2.3.5-6 thatBagoas has notbeen helpful. The beginning of 2.8, wherewe findthe lover busily patching up his relationshipwith Cypas- sis and mustinfer that Corinna has departed,is similarlyoblique. The readeris leftwondering what her verdict was, althoughin view of thespeaker's lack of concernabout the state of hisaffair with Corinna in thispoem, his rathercocky at quanto....praesentior ipse /per Venerisfecinumina magnafidem (17-18) and especiallyhis claimto havedone Cypassisa service(21) one maysuspect that he had been forgiven.At thebeginning of 2.14, onlythe license given to the lover's ira (formerlyrestrained by his metus,2.13.4) informsthe reader that

65 The threatin 2.3.18 is mild by comparisonwith these. 66 The lover triesone argumentafter another: it is only a small thingI ask (3- 4), pity me, I am at your mercy(5-16), you owe me a favor (19-23), it is for yourown good (25-26), whatare you afraidof anyway?(27-39). 67 Jager30n. 48; also p. 37. Afterall, if the eventsof a mid-poempause were neverreferred to, it would be almostimpossible for the readerto know thatthere had been a pause. In paired poems, however,a pause is inevitablythe resultof the physicalbreak between poems and the poet can affordto be less explicit. 68 There is a counter-exampleof sortsat 2.14.41-the angrylover has gone too far with his merito (40) and retractshis prognostications,but there is no indicationas to whetherit was Corinna's reactionto the sketchof 39-40 (i.e. a situation parallel to 1.14.51-52) or his own realization of what he had said (parallel to 2.9b, on which see below) thatbrought about the change of mind. Poem Division,Paired Poems 285

Corinna'scondition has improved.The mainevent of thepause between1.11 and 12, however,is fairlyevident from the first couplet of 1.12-Corinna has said "no." Of coursethe pause also includedNape's goingand coming, which is onlyhinted at by thereference to hertripping over the threshold (1.12.4)69 so thatthe reader still has to makesome deductionson his own. Withthe partial exceptionof 1.11-12, then, Jaiger's perception of a differencein thetreatment of situationalchanges in dramaticpairs and individualpoems seems accurate. It is thusthe only one of themany arguments for leaving 2.2-3 as a pairthat stands up to muchscrutiny, for Davis' observationthat the structure of 2.3 reflects thatof 2.2 is rendereduseless (for the purpose of demonstratinga pair) by his own analysisof 2.2, accordingto whichthe two halves of thesingle poem are likewisereflections of one another.70Indeed this technique of conveyingin- formationindirectly, necessitating as itdoes theparticipation (hence enjoyment) of thereader, may be one factorcontributing to thespecial kind of poeticsuc- cess achievedby the dramatic pairs. Our modeldramatic pair, then, has thefollowing characteristics. In two ad- jacent poems threeepisodes of a singleincident are presented.Some of the eventsof thesecond episode (which occurred during the pause) must be deduced fromthe tone or underlyingassumptions of thesecond poem. The pause may requirea moment'stime (2.2-3) or muchlonger (1.11-12, 2.7-8, 2.13-14). The connectionof thesecond poem to thefirst is indicatedby a strongsitua- tionalsimilarity which becomes evident very early in thesecond poem, and the situationallink is generallyreinforced by verbalechoes throughout the second poem.The addresseemay or maynot be differentin thetwo poems. 2.7 is ad- dressedentirely to Corinnaand 2.8 entirelyto Cypassis,but in 1.11-12 and 2.13-14 theaddressee at thejunction (which is thecrucial point for a readertry- ing to decideif he has a pairin handor not)is thesame in bothpoems, while in 2.2-3 theaddressee is Bagoas throughout.The firstpoem of a pairgenerally comes to a convincingclose, so thatthe addition of anotherpoem comes as a surprise,while the conclusion of thepair is oftenindicated by thereappearance of termswhich were prominent in boththe first poem and theearlier part of the second.It remainsto see how 2.9 and 3.11 comparewith this composite dramaticpair.

Section4: 2.9 and 3.11 In thefirst 24 linesof 2.9 thelover utters an elaboratecomplaint about the unfairnessof Cupid's latestattack on him,Love's own soldier.He makesan oblique bid foran immediatecure withthe story of Achillesand Telephusin lines 7-8 (cf. Tr. 5.2.15), thensuggests that Cupid seek a kindof preymore productiveof gloryfor its captor (13-18, cf. 6). He offersparallels which jus- tifya morepermanent solution, retirement from love's service(19-24), with the image of the lover as a soldierreemerging as the cap of the list and

69 Davis notes (79) that this is "the only directreference to an event which h4ppenedbetween two pairedpoems" in any of thepairs he discusses. luDavis 94-97. 286 CynthiaDamon providinga sense of closure (23-24,71latent since line 4).72 In line 25, the speakersuddenly distances himself from the advice vive posito amore (i.e., the sortof thinghe was thinkinghe wouldlike to have done in thefirst 24 lines) by supposingthat it comes fromoutside himself (si quis). Because his hypo- theticalinterlocutor is a god, his rejectionof therejection of love is themore emphatic.We haveseen howthis radical change of mindwithin the boundaries of a singlepoem bothered Muller, so let us supposefor a momentthat a new poembegins at line25. A readerthinking he was enteringupon a newand unre- latedpoem here would first of all be struckby theprominence of vive(25), re- memberingthat the previous poem endedplacide viveretempus erat (24). He would soon discoverthat the subjectmatter-renunciation of love-was the same as thatof thepreceding poem, and thatthe point of view takenwas the oppositeof thatexpressed before. But not diametrically opposite, for the speaker is noteager to fallin love,but rather an unwitting(nescioquo, 28) and unwill- ing (miserae mentis,28) victimof theincerta Cupidinis aura (33), whichis likenedto theunopposable forces of an unrulyhorse and a strongwind at sea (29-32). Giventhe tense and moodof pertaesum est and relanguit (27), thecum is a generalizing"whenever," and lines27-34 refernot to thepresent situation butto past renewalsof love (cf.saepe, 33 and vixillis prae me notapharetra sua est,38). The readerwill remember that the lover of 9a, too,had experienced frequentfallings into and out of love (totiens,23).73 In lines 34-37, we find Amortaking up his weaponsagain, recalling the beginning of 9a (5, 13-14, especiallyfige, 34 andfigit, 5). This image may be too commonto be an unambiguouslink between poems, but the fact that the lover is in each poem nudus (13, 35), thoughadmittedly in differentsenses, is moresuggestive of connection.74Lines 39-42, withtheir discussion of thepeaceful nights of the non-lover,carry the reader back to thelast line of 9a, whenthe speaker says he deservedto retireand enjoya lifeof peace (defunctumplacide vivere,24). The vigorousstulte, quid est somnusgelidae nisi mortisimago? (41) exploitsthe oxymoronlatent in defunctumvivere.75 At the end of 9b indesertameo pectore regnagere (52) showsan attitudevery different from the complaint o in corde meo desidiosepuer (2) withwhich the lover began; we have alreadyseen this

71 For merere used absolutelyto signifymilitary service cf. TLL s.v. mereo, 803.44-57. Sub is used witha militarymerere at, e.g., Fasti 4.381, but not in any of the passages wheremerere used absolutelymeans "to deserve" (cf. TLL 807.80-808.3) or "to serve" (cf. TLL 809.78-810.3). 72 Even Jager,who views 2.9 as a two-partpoem, admits (150) the strong endingof 9a: "Die ganze Argumentationist von Anfangan auf diesen SchluB aus- gerichtet,der das Gebet durch diese zielstrebig gebaute Reihe zu einem geschlossenenGanzen abrundet." 73 In fact,that is the pointof writingtotiens as opposed to longum, or a phrase like qui dominaemerui sub amore totannos. 74 Cf. 1.2.29-31, for example, where the lover has been wounded by Love's weapons,but is not nudus. Also, theexempla of horse and ship are repeatedin 9b (29-30, 31-32), althoughthe old soldierand gladiatordo not reappear. 75 Cf. 1.8.108 ut mea defunctaemolliter ossa cubent,where Ovid uses defunctae in the sense of "dead." PoemDivision, Paired Poems 287

sortof ring composition over the pair as a wholein 2.2-3, 7-8 and13-14.76 It is reinforcedby the recurrence ofpopulis in the last line of the pair: in its first appearance,in 9a, Love's peoplewere suffering from his attacks(nos tua sentimus,populus tibi deditus, arma, 11) butby theend of theepisode Love is invitedto enjoy the reverence of men and women alike (ambobus populis sic veneranduseris, 54). Inother words, if one supposes a pairof poems here, one findsthe same kind of situational and verbal links as we foundin Davis' dramaticpairs. If it is a pair,what happened during the pause? In 9a thelover feels heought to have fmished his tour of duty as love'ssoldier and deserves a rest; in9b heclaims that whenever he gets to feeling tired of love something always stirshim up again. During the pause he musthave been thinking over past oc- casionswhen he hadthought his love had cooled, but was revived through no effortof his own. His perceptionof the inevitably cyclical nature of hisloves changeshis earlier petulance into a somewhatresigned acceptance of the new affair. We havethen the features of a dramaticpair, namely one self-sufficient poemfollowed by another poem with such surprising situational, verbal and structuralreferences tothe previous one that the reader soon suspects he is deal- ingwith a diptychrather than two unrelated canvases. As in 2.7-8, thead- dresseesof the two poems are different, at least at thejuncture. 2.9a is,as the MSS titulitell us, an addressad Amorem,but Amor is aboutthe only being to whomthe first 10 linesof 9b couldnot be addressed.The inferences one can makeabout what went on duringthe pause in lightof 9b.25-28explain the changeof attitude which is troublesomeifone reads 2.9 as a singleelegy. And finally,one of the greatest benefits of dividing the lines between two poems is that,as in 1.11-12and 2.7-8, the pair amounts to somethingbetter and more Ovidianthan the sum of its parts. 9a containsa standard elegiac complaint-in hiscommentary Brandt produces many Greek and Latin parallels. It is a fairly straight-facedpresentation ofthe serious lover; the military imagery with which itbegins must be intendedto recall 1.9, in whichthe lover has as seriousand strenuousan occupationas thesoldier. In 9b,on the other hand, the emphasis is on thefickleness of love (incertaCupidinis aura, 33; quod dubiusMars est,per te,privigne Cupido, est, 47, cf. 49-50) and thelover is fullycognizant of the deceptions(43) andself-deceptions (44) involved.This, too, has its parallels in theelegiac corpus,77but the combination of twostandard but elsewhere mutu- allyexclusive types of lover in a singlepersona reveals clearly the preposterous amalgamthat elegiac love is. If,on theother hand, one tries to read 2.9 as a singleentity, one is faced witha loverwho in line 25 sayshe would refuse to do whathe has been asking tobe allowedto do forthe previous 24 lines,and there would be no indication ofa pausein which he could have reconsidered his decision. We haveseen an exceptionto Jilger'sclaim that the events of a pausein a singlepoem are

76 Indeserta makes its only appearancein classical Latin here; the litotesem- phasizes the extentof the reversal.It is not accidentalthat we fimda lover em- broiledwith two puellae in the subsequentpoem (2.10). 77 For Cynthia'sperfidy, e.g., Prop. 1.8, 11, 12, 15, 16, 18; also Tibullus 1.5, 1.6 and [Tib.] 3.16. 288 CynthiaDamon

alwaysclearly described, but nothing comparable to thisunexplained and drastic reversal. Jager'scomparison of 2.9 and 2.10 (both,in his analysis,two-part palin- odes) onlyserves to makethe difference between a pairof poems and a two-part poem theclearer. For him,the contrast between the complaints addressed to Venus in 2.10.11-14 and the increasinglyenthusiastic acceptance of love in lines 15-38 is comparableto thechange of mindexperienced in 2.9a-b. Jager's assumptionthat a wish to be freedfrom love is implicitin thecomplaints,78 however,receives no supportfrom the text-at mostthe speaker hints that he mightlike to be relievedof thesupernumerary girl (12). He certainlydoes not suggestthat trees be denudedof theirleaves, or thesky of its stars,or thesea drainedof itswater (13-14). Furthermore,there is reallyno changeof mindat all involvedin thetransition sed tamenhoc melius(15). The lover may still preferto be troubledby only one girl,but, for all hiscomplaining (tamen), real- izes thattwo are betterthan none. He is simplyfollowing the rational course of tryingout a differentperspective on a painfulsituation, and he makesthe proce- dureclear witha seriesof connectiveparticles. And thecompositional "cues" thatalert the reader to thepresence of a sequelare naturally absent from 2.10. The changeof mindthat takes place at 3.11.33 has been less troublesome to criticsthat that of 2.9.25, and consequently fewer have argued for splitting up thepoem. In 3.11 thelover's rejectionof his mistressis by no meansas un- hesitatingas the renunciationof love was in 2.9a-he has to gird himself (perferet obdura,7) forthe painful process of applying remedies (sucus amarus, 8, cf.Rem. 299-308) in orderto cure himselfof his affliction.As in Catullus 8, thelover calls to mindthe unpleasant episodes of therelationship (3.11.9- 15, 21-26) and again like Catullus makes the "'entscheidendentaktischen Fehler"79of recallingits pleasantaspects as well (blanditias,verba potentia quondam,31). The loveruses Propertius'metaphor for the end of a love affair (the ship safe in harborafter an arduouspassage, Prop. 3.24.15-17, cf. Am. 3.11.29-30) to alignhimself with a successfulrenuntiatio amoris, but Ovid's readerhas alreadyseen theinvoluntary renewal of love likenedto a shipbeing sweptback intomid-ocean (2.9.31-32). All of thismakes the lover's relapse less surprising.The orderof eventsat the beginningof 1lb also makes the changeof mindhere less awkwardthan that at thebeginning of 2.9b. In 3.11 thelover progresses from a denialof love to troubleduncertainty (luctantur pec- tusque leve in contrariatendunt / hac amor, hac odium,33-34) to an in- creasinglyconfident reaffirmation of his love (sed, puto,vincit amor, 34, cf. quidquideris mea sempereris, 49), whereasin 2.9 thespeaker who expressed his desireto be retiredfrom love's servicewithout any misgivingsin lines 1-24 declinesthe favor in line25, givingno reasonsbut those implicit in thegeneral remarksof 27-34. Even if comparisonwith 2.9 makesthe change of mindat 3.11.33 seem less abruptthan it mightbe, however,the emergence of doubt (33-34) is stillcompletely without explicit motivation. Something happened beforeline 33 to revivelove, and thatsomething must have happenedduring a

78 Jager154. 79 Jager 146. Poem Division,Paired Poems 289 pause. The readercan onlyassume that, just as one cannotsay "notA" without callingA to mind,so thelover's defiantdesine blanditiaset verbapotentia quondamperdere (31-32) brought those able pleadersto lifein hismind.Y* The pause may be of onlya moment'sduration, or moretime may have slipped away whilethe lover's mindwas occupiedwith such pleasing visions.81 This lack of informationabout the eventsof the pause is, as we have seen, an importantcharacteristic of the dramaticpair, so let us supposehere too that thereis a new poem beginningat line 33 and see if the pair resemblesthe compositepair in otherfeatures as well. Nothingin 1la promptsthe reader to expecta sequel (whichis notto say that,as in the case of 2.7-8, thepresence of a sequel does not confirmthe reader'ssuspicions that the speaker is protestingtoo much),indeed the lover's finalnon ego sumstultus ut antefui (32) embracesboth his desireto be freed fromlove (1-8, 28-30) and his shameat havingbehaved as he did in love (9- 16, 21-26) and so roundsoff the poem neatly. The closureis reinforcedby the reappearanceof ferendis and duraviin line27 (cf.ferre, tulisse, 4; obdura,6). Earlyin thesecond member of thepair we fmdthe expected situational and verbal linksto thepreceding poem. In thefirst couplet it becomesevident that the situationinvolves the struggle between love and hate,which is whatprovoked wordslike perfer et obdura,dolor (7) andsucus amarus (8) in 3.1la. The verbal connectionbetween poems is strong-vincit(34) cf. vicimus(5); vitiis(44) cf. vitiis(1)-and theinsulted gods of 1la (periuratosin mea damna deos, 22) reappearin llb (fallendos...saepe deos, 46). Finally,the recurrenceof the ship/lovermetaphor at theend of thepair (51) clearlyrecalls the use of that metaphorat theend of thefirst poem (29-30). Ourassumption, then, that 3.1 la-b is a pair,has theadvantages of explain- ing whythe change of mindat line 33 is apparentlyunmotivated, of allowing thewell-orchestrated conclusion of 3.1la itsfull effect and of producingboth surprise(at thepresence of a sequel)and gratification(from the confirmation of his previoussuspicions) for the reader. Nothing in thepair sets it at variance withthe model dramatic pair and it is onlythe internal nature of theevents of thepause whichdifferentiates this pair from Davis' dramaticpairs. Here, too, I thinkwe can see Ovid usingthe pair formatto makeplayful fun of elegiac

80 The remembranceof Lesbia's kisses in Catullus8 (a poem whichthe reader of 3.11 has repeatedlybeen invitedto recall) had a similareffect on Catullus. 81 Because of the multiplicityof addressees(amor 1-2; self 3-8; audience9-10, 21-26; girl 11-20, 27-32) and the inconsistencyin the formused to address the girl (1st person noster, 20 [cf. vicimus, calcamus in 5 where nos=the speaker alone]; 2nd person tu 11, tuus 18 [cf. tibi=speaker,7]; 3rd person erat, 25, 26) the addressof lla would seem to take place entirelyin the speaker's imagination. In 1lb, thereis only one 3rd personreference to the girl (valet illa, 44), who is otherwisethe addressee.It would be possible-reading 44 as an aside-to assume thatthe girl arrivedduring the break betweenpoems and thatthis is what caused the freshoutbreak of love in 33, but since the poem ends withthe speakerurging a course of action upon himself(dem, utar, 51), I am inclinedto see thispoem, too, as taking place in the speaker's imagination.Cf. Lenz (commentary, ad loc.): "In dem Augenblick,da er wahnt,seine Leidenschafthabe den tiefsten Punkterreicht und er sei jetzt sicher(29f.), trittihr Bild vor seine Seele, und er hortihre Worte." 290 CynthiaDamon conventions.The situationof 1la is thatof theangry lover who wantsto puta haltto an affair.It is also foundat, for example, Catullus 8 and Propertius2.5 and 3.24. In thecase of thesepoems we are nevertold about the aftermath- whetheror notthe lover perseveres-although their position in theirrespective corporamight incline one to thinkthat the renuntiatio amoris of Prop. 3.24 was successfuland to have strongdoubts about the efficacy of Catullus8 and Prop.2.5. Ovid takesup thisimplicit change of heartand exposes it in all its minutiae to thelight of day,again, I think,subtly mocking the elegiac lover. The germof theidea can be foundin Prop.2.5 itself,where the angry lover urgeshimself to end therelationship quickly, before he changeshis mind(9- 14). The gentlenessof therest of thatelegy rather belies his intent,and gives thecue fora displayof Ovidianwit.82 So muchfor the benefits of division.On theother hand, there is no advantageto be gainedfrom reading 3.11 as a single poem (exceptfidelity to themanuscript tradition, which doesn't really deserve it), and the unexplainedchange of mindat line 33 is at least disturbing.If anythingmore is neededto inclinethe balance toward division of thepoem, its similarityin themeand treatmentto 2.9a-b (wherethe reasons for division are stronger)should do thetrick.

82 But cf. G. Williams,Tradition and Originalityin Roman Poetry (Oxford 1968) 508.