Milman Parry 2

8/12/2019 Milman Parry 2

1/51

Department of the Classics, Harvard University

Studies in the Epic Technique of Oral Verse-Making: II. The Homeric Language as the Languageof an Oral PoetryAuthor(s): Milman ParrySource: Harvard Studies in Classical Philology, Vol. 43 (1932), pp. 1-50Published by: Department of the Classics, Harvard UniversityStable URL: http://www.jstor.org/stable/310666.

Accessed: 21/04/2011 03:32

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at.http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at.http://www.jstor.org/action/showPublisher?publisherCode=dchu..

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

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 [email protected].

Department of the Classics, Harvard Universityis collaborating with JSTOR to digitize, preserve and extend

access toHarvard Studies in Classical Philology.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=dchuhttp://www.jstor.org/stable/310666?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=dchuhttp://www.jstor.org/action/showPublisher?publisherCode=dchuhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/310666?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=dchu


2/51

STUDIES IN THE EPIC TECHNIQUE OF ORALVERSE-MAKINGII. THE HOMERIC LANGUAGE AS THE LANGUAGE OF ANORAL POETRY

BY MILMANPARRYI. The Homeric Language and the Homeric Diction: Older Theories of the Ho-meric Language (p. i); the Homeric Language as a Poetic Language (p. 4); theHomeric Language as an Oral Poetic Language (p. 5).- 2. The Traditional PoeticLanguage of Oral Poetry: the Formula (p. 6); the Archaic Element (p. 9); the Art ofTraditional Poetry (p. 12); the Foreign Element (p. -7); the Artificial Element(p. 19). - 3. The Study of a Traditional Poetic Language (p. 20o).- 4. The Ho-meric Language as a Traditional and Oral Poetic Language: the Ionic Recording(p. 23); Arcado-Cyprian, Aeolic, and Ionic (p. 25); the Arcado-Cyprian Element(p. 26); the Aeolic Element (p. 27); the Traditional Language of Lesbian LyricPoetry (p. 29); the Artificial Element (p. 33); Equivalent Aeolic Forms (p. 35);the Ionic Element (p. 37).- 5. Conclusions: the Theory of an Aeolic Homer Re-jected (p. 40); the Theory of an Aeolic Diction Accepted (p. 43); the History ofthe Greek Heroic Style (p. 46).

I. THE HOMERIC LANGUAGE AND THE HOMERIC DICTIONITHIN the last twenty yearsHomericscholarshave shownthatthe languageof the Iliad and Odysseys a poetic languagemadeto suit the needs of the verse, and they have therebydoneawaywith awhole numberof hypotheses whichwereno longerneeded. A briefac-count of these earlier theories of Homer's language will serve to set

forth the subjectof the presentpages. The readershouldbearin mindthat we are speakinghere at the beginningabout language, and notabout diction or style. All three have to do with the sum of words,word-forms,and word-groupingsused by a man. As language,how-ever, we look at them as usedby a certainpeople,at a certaintime,andin a certain place; as diction,as the material by which thought is ex-pressed;and as style,as the form of thought.OlderTheoriesof theHomericLanguageThe commonview of Homer'slanguage n antiquitywas that which,while it seems the simplest, is likewise the farthest from the truth:


3/51

2 Milman ParryHomer himself chose various formsand words from the dialects whichhe had heard in his travels about Greece.' Sucha view couldof coursebe held only in the lack of any carefulstudy of the Greekdialectsandof Homer's language, for we now know that some of the formsin theIliad and Odysseyare much older than others,while some could neverhave been a part of the everyday speechof any Greek. Also Homer'suse of the forms and words of differentdialects followsa fixed ruleandno varying chance of memory.2 The gravest fault of such a theory,however,is that it supposesthat one man couldall by himselfcreate apoetic language. Sucha thing has been seen nowhere. No single poetcouldeverhave suchpowers;and a poetic language, t is clear,is poeticonly by a convention sharedby the poet and his hearers,so that thegrowthof a poetic languagemust be gradual.The ancients,sincethey had no rigoroushistoricalmethodof literarycriticism,may be excusedfor such a mistake,but not so the authors ofa recent theorywho hold that Homerwas the native of a city whereina mixed populationof Aeolians and lonians had come to speak a lan-guage having the same variety of forms as that found in the Homericpoems.3 Such a view altogetheroverlooks the nature of Greekpoeticdiction as it is to be seen everywherein the poetry of the historicalperiod; by the same reasoningthe population of Attica was partlyDorian. Nor has anyone anywherefound a spoken languagewhichshows even faintly such a variety of formscurrentside by side.A third theory, which found many believers at the end of the last

1 [Plutarch], Life and Poetry of Homer 2, 8; cf. Dio Chrysostom, Orations ii, 23.2 See pp. 4-5.S The latest critic to hold this idea is T. W. Allen in his Homer: the Origins andTransmission (Oxford, 1924), PP. 98-1o9, where he claims to be developing theviews of P. Giles (cf. Proceedings of the Cambridge[England] Philological Society,1916, pp. 7-9), whose very sensible view, however, he has failed to understand. Hedid not know, it would seem, that the theory had already been set forth by Wila-mowitz (SitzungsberichtederAkademie der Wissenschaften,Berlin, 190o6, pp. 52-75,and Die Ilias und Homer, Berlin, 1906, pp. 356 ff.), and straightway disproved(cf. E. Drerup, Das Homerproblemn der Gegenwart,Wtirzburg, 1921, p. iio). Forthe views of a linguist on such a theory see A. Meillet, Apergu d'une histoire de lalangue grecque3, . 171. There is scarcely any need of giving further warning againstthe theory, but one finds even so good a critic as C. M. Bowra falling into the error

in a milder way. In his Tradition and Design in the Iliad (Oxford, I93O),pp. 139-140o,he compares Homer's Ionic and Aeolic with Chaucer's English and French -


4/51

TheEpic Technique f OralVerse-Making 3century,held that Homer'slanguagewas altogetherIonic and that thevariety of formswas due to a simple literaryconservatismwhichkeptthe older formsfromage to age forpurelystylistic reasons.' This view,like the next which will be mentioned,is much better than the firsttwo, since it has a partof the truth in it. Its authors, however,had in-sufficient inguisticknowledgewhen they held that all of the Aeolismsin Homer had at one time been used in earlierIonic, since many ofthe formsin question are the creation of a later period than that ofcommon Greek. Yet their greatest faults were those of giving nothe comparison shows that the author is thinking of Homer in terms of writtenliterature: "Chaucer wrote for a class who knew both English and French, and forwhom his mixed language was intelligible. But it was essentially his own creation.His predecessors wrote in the Anglo-Saxon tradition, but he created a new languagefor English verse. If we press the analogy, it would follow that Homer lived in aworld where different dialects, though existing separately, impinged on each otherand were mutually intelligible. Out of this situation Homer or his predecessorscreated a poetical speech." Bowra, however, somewhat misses the nature ofChaucer's language, as one may judge by the following statement of A. W. Pollard(Chaucerin the Encyclopadia Britannica, 14th ed.): "The part played by Chaucerin the development of the English language has often been overrated. He neithercorrupted it, as used to be said, by introducing French words which it wouldotherwise have avoided, nor boreany such part in fixing it as was afterwardsplayedby the translators of the Bible. The practical identity of Chaucer's language withthat of Gower shows that both merely used the best English of their day with thecare and slightly conservative tendency which befitted poets." Moreover theFrench part of Chaucer's language, as of English, is of a very certain kind, namelyabstract words without which the pattern of European, as opposed to Anglo-Saxon,thought could not be kept, and names of objects brought in by French culture. Itcould be held only in the rarest cases that the Aeolic element in Homer thus repre-sented any contribution of thought or culture foreign to Ionic.As this article goes to press I find that Allen's unlucky theory is accepted byB. F. C. Atkinson (The GreekLanguage, London, 1931, p. 201): "We shall not befar wrong in regarding it [i. e. the Homeric language] as in the hands of the poet ofthe epics a living language against whose everyday use in the island of Chios earlierthan the ninth century we know no valid reason." It is criticism enough to havequoted the statement.

1 The theory was first set forth by K. Sittl (Die Aeolismen der homerischenSprache in Philologus, XLIII, 1884, pp. I-3I), and answered by G. Hinrichs (HerrDr Karl Sittl und die homerischenAeolismen, Berlin, 1884); cf. his De Homericaeelocutionisvestigiis Aeolicis (Berolini, 1875). It was developed in English chiefly byD. B. Monro (Journal of Philology, IX, i88o, pp. 252-265; XI, 1882, pp. 56-6o;Homeric Grammar2,Oxford, 1891, pp. 386-396). See below p. 27.


5/51


6/51

The Epic Technique of Oral Verse-Making 5the Homeric poems were composedin a poetic languagewhereinoldand foreignformshad beenkept andnew formsbrought n by reasonofthe help they gave the epic poets in making their hexameters. Thesepoets ever sought a languagewhich was easier to handle, and for thatreason evermade use of the fact that the olderorforeignform of a wordwas to a Greek,as Aristotle tells us, morepoetic than the form used ineveryday speech.'

TheHomericLanguageas an OralPoeticLanguageIn one way, however, the theory of Witte, even with the furtherwork done on it by Meister, is unfinished:they have logically provedthat the languageof Homeris the work of the Homericverse,but theyhave not at all shown how the verse in this case couldhave suchpower.It did not have it in the later Greek epic, nor in Roman hexameterverse, nor in short do we find elsewhere n ancientor modernliterature(with the very notable exception,however,of the early poetry of thenations) any but the slightest traces of the verse-formacting on thelanguageof the poetry. Clearlya special languagefor the hexameter

could come into being only when poetry was of a very differentsortfrom that which we ourselveswrite, and whichwe know to have beenwrittenthroughoutthe historyof European iterature. To say that theHomeric languagewas the work of the Homeric verse thus implies apoetry whichis, at least to our way of thinking,of a very specialkind,so that while the theory may be provedit cannot reallybe understooduntil we know just what this poetry was.It is my own view, as those who have read my studieson Homericstyle know, that the nature of Homeric poetry can be grasped onlywhen one has seen that it is composed n a dictionwhichis oral,and soformulaic,and so traditional.2So it is for the languageof the Iliad andOdyssey: f we know what an oral diction is we shall have the larger

I Rhetoric1404bIO.2 L'gpithite traditionnelledans Homere(Paris, 1928); Les formules et la metriqued'Home're(Paris, 1928); The Homeric Gloss in Transactions of the American Philo-logical Association, LIX (I928), pp. 233-247; Enjambement n Homeric Verse, ib.LX (1929), pp. 200-22o; Homer and Homeric Style in HarvardStudies in ClassicalPhilology, XLI (1930), pp. 73-147; The Traditional Metaphorin Homerin ClassicalPhilology, XXVIII (1933), PP. 30-43.


7/51

6 Milman Parrybackgroundwhich the theory of a languagemade to fit the hexametercalls for. At the same time the Homericlanguagewhenthusexplainedby the diction will in turn give us the history of that diction.

2. THE TRADITIONALPOETIC LANGUAGE OF ORAL POETRYThe Formula

In a society where there is no readingand writing, the poet, as weknowfrom the study of suchpeoplesin our owntime,alwaysmakes hisverse out of formulas. He can do it in no other way. Not having thedevice of pen and paper which,as he composed,would hold his partlyformed thought in safe-keepingwhile his unhamperedmind rangedwhere it would after other ideas and other words,he makeshis versesby choosingfroma vast numberof fixedphraseswhich he has heardinthe poems of other poets.' Each one of these phrasesdoes this: it ex-pressesa given idea in wordswhich fit into a given length of the verse.Each one of these fixedphrases,or formulas,is an extraordinarycrea-tion in itself.2 It gives the wordswhich are best suited for the expres-sion of the idea, and is made up of just those parts of speechwhich,inthe place which it is to fill in the verse, will accordwith the formulaswhich go before and after to make the sentence and the verse. Eachformula s thusmade in viewof the otherformulaswith which it is to bejoined;and the formulastakenall togethermake up a dictionwhichisthe material for a completely unified technique of verse-making.3Finally, the formulasof an oralpoetryare not each oneof them without

I Cf. A. van Gennep on the Serbianepic (La questiond'Homere,Paris, 1909, p. 52):"Les poesies des guslars sont une juxtaposition de clich6s, relativement peu nom-breux et qu'il suffit de posseder. Le d'veloppement de chacun de ces clich6s se faitautomatiquement, suivant des regles fixes. Seul leur ordre peut varier. Un bonguslar est celui qui joue de ses clich6s comme nous avec des cartes, qui les ordonnediversement suivant le parti qu'il en veut tirer." Cf. also F. S. Krauss, SlavischeVolksforschungen Leipzig, 90o8), pp. I83-184, and John Meier, Werdenund Lebendes Volksepos(Halle, 1909), pp. 17-19.

2 For a fuller definition of the formula see L'dpithItetraditionnelle, pp. 15-17;Homer and HomericStyle, pp. 80-84.3 For the technique of composition by formulas see L 'pithete traditionnelle,pp. 8-19, 45-145; Les formules et la mttriqued'Homdre,pp. 10-13, 17-23, 48-52;Homer and Homeric Style, pp. 84-89, 140-147-


8/51

TheEpic Technique f OralVerse-Making 7likeness to any other; in that case the techniquewould be far too un-wieldy. They fall into smallergroupsof phraseswhich have betweenthem a likeness of idea and words,and these in turn fall into groupswhich have a larger pattern in common, until the whole diction isschematized in such a way that the poet, habituated to the scheme,hits without effort,as he composes, upon the type of formula and theparticular ormulawhich,at any point in hispoem,he needs to carryonhis verse and his sentence.'A singleman or even a wholegroupof men who set out in the mostcarefulway could not make even a beginningat such an oral diction.It must be the workof many poets over manygenerations. Whenonesinger (for such is the name these oral poets most often give them-selves)2has hit upon a phrasewhich is pleasingand easily used,othersingerswill hear it, and then, whenfacedat the same point in the linewith the need of expressingthe same idea, they will recallit and use it.If the phraseis so good poeticallyand so useful metricallythat it be-comes in time the one best way to expressa certain idea in a givenlength of the verse, and as such is passed on from one generationofpoets to another, it has won a place for itself in the oral diction as aformula. But if it does not suit in every way, or if a better way of fit-ting the idea to the verse and the sentence is found, it is straightwayforgotten,or lives only for a short time, since with each newpoet and

I For the schematization f the formulaicdictionsee L'pithditeraditionnelle,pp. 19-24, 85-94; Homerand HomericStyle, pp. 84-89, 140o-47. W. Radloff(Probender Volkslitteraturern6rdlichenTiirkischentdmme. V. DerDialectderKara-Kirgisen,. xvii) gives the words n whichan oralpoet tells of the easewithwhichhe composes:"Ich kanntiberhauptedesLied singen,denn Gott hat mirdieseGesangesgabens Herzgepflanzt. ErgiebtmirdasWort auf dieZunge,ohnedass ich zu suchenhabe, ich habe keines meiner Liedererlernt,alles entquelltmeinem Innern, aus mir heraus." This is a commentaryon two passagesinHomer:0 44 TrLtap a Oeb0'ip ,6WKEI)OL67jripraE&5irir7t Ovds rorpbiv'ytzv'aeie6et,X 347 abrobl'6aKTOSV ' e.l, GeTsT Jot& V /peaiv oltas7rav-rolas5,


9/51

8 MilmanParrywith each new generationof poets it must undergothe twofoldtest ofbeing found pleasingand useful. In time the needednumberof suchphrasesis made up: each idea to be expressedin the poetry has itsformulafor each metrical need, and the poet, who would not think oftrying to express ideas outside the traditional field of thoughtof thepoetry, can make his verses easily by means of a diction which timehas proved to be the best.Actually,of course,this birthof a dictionis beyond observation,andunless it can really be shown that a people revertingfrom written tooral poetry created anew a formulaicdiction we must suppose that ittook place in a very distant past, since the poetry of an unletteredracehas as muchclaim to age as have any of its other institutions. Butif the birth of a formulaicdiction is only to be describedtheoretically,we can see in living oralpoetrieshow such a dictionis passedon fromone age to another,and how it gradually changes.The youngpoet learnsfrom someoldersingernot simplythe generalstyle of the poetry, but the whole formulaic diction. This he does byhearingand rememberingmany poems, until the diction has becomefor him the habitual mode of poetic thought.' He knows no other

1 Cf. Mathias Murko, La podsie populaireepiqueen Yougoslavieau d&but u XXesiecle (Paris, 1929), P. i2: "Les chanteurs commencent A apprendre A jouer desgusle et a recueillir la tradition 6pique des leur tendre enfance, sur les genoux d'unpare ou d'un aleul, ou d'autres parents, ou de familiers, puis dans le public, laplupart du temps entre dix et douze ans, mais toujours en g6neral jeunes, 'alorsqu'ils ne pensent encore ' rien,' jusque vers l'Age d'environ vingt-cinq ans. Il leursuffit d'ordinaire d'entendre chanter un chant une seule fois, et, quand ils sont plusAges, plusieurs fois." Cf. also the same work, p. 42, paragraph 18, and D. Com-paretti, Traditional Poetry of the Finns (English translation, London, 1898), p. 20.In countries where the art of the singer is a paying profession there is a more formalapprenticeship; cf. James Darmesteter, Chants populaires des Afghans (Paris,1888-I890), p. cxcii: "Le deumnovice va aupres d'un djum 6l6brequi est devenuemattre, ustdd;il devient son shagird,son disciple. Le maitre lui enseigne ses propreschansons, puis les chansons des grands chanteurs pass6s ou presents, et les chansonsles plus populaires de KhushhMlKhan. Il l'emmene ' la hujra, oft l'on se r6unittous les soirs pour causer des nouvelles du jour et 6couter quiconque a un conte aconter ou une chanson a chanter. ... Quand le shagird commence a se sentir assezfort pour voler de ses propres ailes, il quitte son maitre, compose en son propre nomet devient ustad a son tour." Cf. also H. Basset, Essai sur la littmratureesBerberes(Algiers, 1920), p. 331.


10/51

TheEpic Technique f OralVerse-Making 9style, andhe is everkept fromquittingthe traditionaldictionandusingphrasesof his own makebecausehe couldnot findany as pleasingor asuseful as the old ones, and moreover,since he is composingby word ofmouth, he must go on without stoppingfrom one phrase to the next.Since his poetry has being only in the course of his singing,and is notfixedon paperwhere it can show itself to him verse by verse, he neverthinks of it criticallyphraseby phrase, but only faces the problemofits style whenhe is actuallyunder the stress of singing. Thus whateverchange the single poet makes in the traditional diction is slight, per-haps the changeof an old formula,or the makingof a new one on thepattern of an old,or the fusingof old formulas,or a new way of puttingthem together.' An oral style is thus highly conservative;2 yet thecauses for changeare there,and sooner or later must come into play.These causes for change have nothing to do with any wish on thepart of the singlepoet for what is new or strikingin style. They existabove the poets, and are two: the never-ceasing change in all spokenlanguage, and the association between peoples of a single languagebut of differentdialects.

TheArchaicElementAs the spoken language changes, the traditionaldiction of an oralpoetry likewisechangesso long as there is no need of giving up any ofthe formulas. For example,a change in the sound of a vowel or con-sonant which calls for no changein the metricalvalue of a word soon1 Cf. Homer and HomericStyle, pp. 134-147.2 A. Dozon, L'popee Serbe (Paris, 1888), pp. lxxiii f.: "L'Age des pesmas n'est

pas une question facile . resoudre. En presence de l'uniformit6 de style et de languequi les caractdrise, on n'a pour guide, afin de constater du moins leur anciennet6 re-lative, qu'un reste de couleur plus antique ou plus barbare ... pour ces sujets m&mesqu'une c616brit6exceptionnelle maintient dans la tradition vivante et qui peuventtenter quelque chanteur, on y trouvera ' la v6rit6 certains anachronismes: la com-position, le style et l'esprit de la pesma ne varieront pas. Pour s'en convaincre,on n'a qu'. lire, par exemple, la piece des Adieux de Karageorge,qui date de 1813,et la comparer avec les plus anciennes. Rien, sinon l'incident qui en forme lefond, ne vous avertit qu'il y a entre elles un espace de plusieurs si&cles." It shouldbe added, however, that this uniformity of style is due as well to the fact that thelanguage of the older poems changes along with the language of the diction as awhole; cf. below, pp. 11-12.


11/51

IO Milman Parrymakesits way into thepoetic language:the singernaturallypronouncesthe wordas he usuallydoes,and there is not the least thing to keephimfrom doing so. But when a change in the form of a word must alsochange its metrical value it is far otherwise,for the poet, if he thenwished to keepup with the spoken anguage,wouldhave to put upwitha phrasewhichwas metricallyfalse, or give it up altogetherand makehimself a new one. But neither of these two choicesis at all pleasing.The rhythm must be kept fairly regular,' and the oral verse-makingmakes it very hardforhim to find newwords;it is evendoubtfulifwithall the good will and time in the world he could do so in any greatnumber of cases. Each formula, as it was said above, is the long-proven choice of a long line of singers, and it is not possible that aphrasewhich is useful in oral compositioncould be made in any otherway than by a singerwho, makinghis verses throughhis sense of thescheme of the formulaicdiction,created,in the stress of the moment,anewphrasemore orlesslike an older one. For otherwisethe newphrasewould not fit into the scheme of the diction,and since it could be usedonly with an effortit wouldnot be used at all.2 Finally the changeinthe spoken languagewouldvery likely be such that a phraseto expressthe sameideain wordsof the samemetricalpatternwouldbe out of thequestion. The new phrasemust be shorteror longer,or begin or enddifferently. Then the formulas o whichit wouldbe joinedmust also be

1 It often happens, however, that oral poets will change a formula under theinfluence of the current language and yet keep it despite the false verse which is theresult. Kaarle Krohn has noticed this in Esthonian oral poetry (KalevalastudienI.Einleitung in F[olklore] F[ellow] Communications,XVI (1924), PP. 56 f.: "Bei derfeststellung der urform eines altestnischen liedes durch vergleichung der verschie-denen varianten kann somit die forderung aufgestellt werden, dass sie sowohl derilteren sprachform als den metrischen gesetzen der rune entsprechen muss. Fiirdie beurteilung der in den varianten vorkommenden verse und ihrer variierendenformen ist diese doppelte forderung ein ausgezeichnetes kriterium. Ein scheinbarfehlerhafter vers kann, wenn er in die altere sprachform zuriickgedacht metrischregelrecht wird, der urform angeh6rt haben. Als spitere interpolation muss dagegenein scheinbarfehlerfreiervers angesehen werden,der in die altere sprachformzurtick-gedacht eine iiberzahl von silben aufweist." For the same thing in Greek epicpoetry see Les formules et la metriqued'Homdre,pp. 43-58; and see below, pp. 33n. 1, 36, 44 n. i. Such cases show how the usefulness of the formula overtops allelse in oral verse-making.2 Cf. Homer and Homeric Style, p. 147.


12/51

TheEpic Technique f OralVerse-Making IIchanged,and so on. Thus by no wilfulchoice,but by the constraintofhis techniqueof verse-making,the singer keeps the formulathoughitslanguagehas become archaic.'As it happens, this archaic languagedoes not at all displeasehim.His style is thus lifted above the commonplaceof daily speech andmade distant and wondrous. But though the old wordsand forms arethus desirable,they are neverwilfullysoughtafter. When the formulacan be changedit sooneror later will be, and the cleavagebetween theold and the new in the style dependson whether it is easy or hard tochange the formula.2 An oral diction may thus in time become veryarchaic, since even though a word has been lost altogether from thespoken language its context in the poetry will teach the poet and hispublic its meaning. In the case of words which are not a neededpartof the thought, such as the ornamentalepithets, the meaning of thewordmay even be lost altogether.3In time, however,a point must bereached in the case of each formulawhere its meaning,needed for thethought, is lost, and here an even heavier constraint than ease of

1 For this strong conservatism of the formulaic technique cf. Les formules et lamrtriqued'Homere, pp. 43-65.2 It should be added here, however, that a form or word easily changed maynevertheless be kept for a long time because it is bound with the words which gobefore or after into a larger word-group which the singers feel as a single whole;but such survivals are not apt to be common. Cf. below, pp. 35-37 ff.3 Cf. The Homeric Gloss, pp. 243-244; A. Rambaud, La Russie tpique (Paris,1876), pp. 18-19: "Les chanteurs ne comprennent pas toujours ce qu'ils chantent:la langue a vieilli et plus d'un vers s'est alter6. Si on leur demande compte d'uneexpression singuliere ou d'un passage obscur, ils repondent invariablement: 'Celase chante ainsi,' ou bien: 'Les anciens chantaient ainsi; nous ne savons ce que celaveut dire' ... Ce qui prouve la tenacit6 de la m6moire populaire, c'est que lepaysan de l'Onega continue achanter les 'chenes robustes,' et 'la stipe de la prairie'et 'la plantureuse campagne,' bien que ces traits de la nature kievienne ne re-pondent en rien a la nature qu'il a sous les yeux, et que de sa vie il n'ait vu unchene. 11parle de casques, de carquois et de massues d'acier, bien qu'il n'ait memepas une idee de ces sortes d'armes, de 'l'aurochs au poil brun' et du 'lion rugissant,'bien que ces animaux qui ont pu exister dans l'ancienne Scythie, lui soient aussiinconnus que les quadrupedes australiens.... Ces scrupules n'ont pas empech6qu'il ne se glissAt parfois dans les bylines des d6tails etrangement modernes. ....C'est ainsi qu'on voit des h6ros &crire ur du papier timbre, ou encore, sur le pointd'attaquer un dragon ou un g6ant, braquer sur lui une lunette d'approche." Cf.also, Basset, op. cit., p. 319.


13/51

12 Milman Parryverse-makingcomes into play: the formulamust be given up cost whatit may, and the singersmust do the best they can to findanotherone totake its place. Thus the languageof oral poetry changesas a wholeneither faster nor slowerthan the spokenlanguage,but in its parts itchangesreadilywhen no loss of formulas s called for, belatedly whenthere must be sucha loss, so that the traditionaldiction has in it wordsand formsof everydayuse sideby side with others that belongto earlierstages of the language.' The number of new words and old wordsvaries, of course,from one oral poetry to another as differentfactorshave force: a complexverse-form,a fondness for tales of an heroicpastrather than of the present,and the practiceof poetryby a classof pro-fessional singers all tend toward a greater conservatism,whereas ashort verse without enjambement,a changein the way of living of apeople,and the lack of a classwhosegainit is to keepthe bestpoetryofthe past all allow a quickerchange. But the principleof change andconservatismof languageis the same in all cases.

TheArt of TraditionalPoetryI have written so far, in telling of how the languageof oral poetrycomes to be archaic,as if the formulawerethe unit of diction,and suchit is in the end. But in practice the oral poet by no means limits hisborrowingto the single formula;ratherhe uses whole passages whichhe has heard. This is, indeed, his whole art: to make a poem like thepoemshe has heard.2 I know only too well that this is sure to suggest1 For numerous examples of the conservatism of the oral poetic diction seeO. Bdckel, Psychologieder Volksdichtung Leipzig, 1913), PP. 59-63. B6ckel himself

altogether misses the nature of the poetic language; witness his use of the termSchriftsprache.2 Cf. W. Radloff, op. cit.: "Man glaube nun nicht, dass dieses Improvisiren einjedesmaliges Neudichten ist. Es geht dem improvisirenden Sanger gerade, so wiedem Improvisator auf dem Klavier. Wie der letztere verschiedene ihm bekannteLaufe, Ueberginge, Motive nach der Eingebung des Augenblicks in ein Stimmungs-bild zusammenftigt und so das Neue aus dem ihm gelaufigen Alten zusammenstellt,so auch der Sdnger epischer Lieder. Er hat durch eine ausgedehnte Uebung imVortrage, ganze Reihen von Vortragstheilen, wenn ich mich so ausdrticken darf,in Bereitschaft, die er dem Gange der Erzahlung nach in passender Weise zusam-

menfiigt. Solche Vortragstheile sind die Schilderungen gewisser Vorfille undSituationen, wie die Geburt eines Helden, das Aufwachsen eines Helden, Preis der


14/51

TheEpic Technique f OralVerse-Making 13the thoughtof plagiarismto those not familiarwith oralpoetry, but itmust be understoodabove all that plagiarism s not possiblein tradi-tional literature. One oral poet is better than another not becausehehas by himselffounda morestrikingway of expressinghis own thoughtbut becausehe has been better able to make use of the tradition. Hestrives not to create a new ideal in poetry but to achieve that whicheveryone knows to be the best. This is true even of the poetry whichmay tell of happeningsof the singer'sown day: the event may be new,but it willbe told in the traditionalway on the patternof passagesfromotherpoems,andin more or less the samephrasesas wereused in thosepassages,so that the only differencebetween the poemmadeabout thepresentand that whichtells of the past is that the formerwill be madefromthe memoryof a largernumberof differentpoems.' For if the taleWaffen, Vorbereitung zum Kampf, das Get6se des Kampfes, Unterredung derHelden vor dem Kampfe, die Schilderung von Pers6nlichkeiten und Pferden, dasCharakteristische der bekannten Helden, Preis der Schbnheit der Braut, Beschreib-ung des Wohnsitzes, der Jurte, eines Gastmahles, Aufforderung zum Mahle, Todeines Helden, Todtenklage, Schilderung eines Landschaftsbildes, des Einbrechensder Nacht und des Anbruchs des Tages und viele Andere. Die Kunst des Sdingersbesteht nur darin, alle diese fertigen Bildtheilchen so aneinander zu reihen, wiedies der Lauf der Begebenheiten fordert und sie durch neu gedichtete Verse zuverbinden." Cf. also Murko, op. cit., p. 18; Basset, op. cit., p. 307.1 Cf. G. Gesemann's account of the composition in 1914, in the military hospitalof Kragujevac, of a poem on the death of the son of one of the hospital surgeons(Studien zur siidslavischen Volksepik,Reichenberg, 1926, p. 66): "Der Singer sangsofort drauf los, das erstemal bezeichnenderweise mit Einkleidung des Ganzen ineinem der hdufigsten traditionellen Kompositionsschemata. Nattirlich, er war janicht dabei gewesen, als der junge Mann fiel. So stilisierte er das Ereignis in einerWeise, die es ihm ermiglichte, etwa hundert Verse herzusingen und seiner Aufgabezu gentigen, ohne sich auf reale Einzelheiten einlassen zu miissen: Da liessen sichzwei Raben auf dem Dache der Kaserne nieder mit blutigen Fliigeln und blutigemSchnabel, da fragt sie der Oberst, von wo sie kommen. Sie kommen aus der Macva,wo grosse Kampfe sind. Sie werden gefragt, ob die Serben gesiegt haben, ob Sabacnoch in Feindes Hand ist usw., nein die Stadt ist befreit, die Serben haben gesiegt.Ob der Sohn nun bald auf Urlaub kommt, mit einem Orden geziert? Einen Ordentrigt er, aber heim kommt er nicht mehr. - Ein paar Tage spiter h6rte ich densel-ben Sanger in einer anderen Krankenstube dasselbe Lied singen, und siehe da, erhatte nicht nur das eben angeffihrte Kompositionsschema aufgegeben, indem er esnur noch als Einleitung benutzte, dann aber gleich nach der ersten Frage an dieRaben zur Schilderung einer Schlacht tiberging, die sich durch ziemlich viel rea-listische Zilge auszeichnete und auch das Bild des Gefallenen irgendwie persbnlicher


15/51

14 Milman Parryis old, and, as is usually the case, regardedas more or less true, thesingermay tell it just about as he heardit.Yet no graver mistake could be made than to think the art of thesinger calls only for memory. Those who have sought to record oralpoetry in lands where it still lives have straightway found that thesame poem, that is to say, a poem on the same subject, could be sungbadly or well, and that the people carefullyset apart the poor singersfrom the good.' Still the fameof sucha singercomes not fromquittingthe tradition but fromputting it to the best use. The poorersingerwillrepeata poemwith the loss of its most pleasinglines or its most drama-tic moments,but the good singerwill keep what is striking,and evenadd, on the pattern of other poems, lines which he knows will please,and new incidents, or give a fuller tale with many such borrowings.He may even have heard the same tale told by a singer living at a dis-tance who inheritedfrom a differenttradition; then he will fuse thepoems, using the best in each. Thus the highest sort of oral verse-makingachieves the new by the best andmostvariedand perhapsthefullest use of the old. This is the meaningof what Telemachussays:zu zeichnen versuchte. Er hatte offenbar die Fiihrung durch ein festes Schema derErzaihlungnicht mehr n6tig, er hatte wahrscheinlich auch von Kameraden irgend-welche Einzelheiten inzwischen geh6rt, die er jetzt verwandte. Aber eins blieberhalten: die episch-heroische, feudale Stilisierung der Einzelziige und des Gesamtge-halts." Cf. also M. Murko, Neues iber siidslavische Volksepik in Neue Jahrbiicherfiir das klassische Altertum,XLIII (1919), p. 294; John Meier, Werdenund Lebendes Volksepos,pp. 1i-1i7.1 Cf. Murko, op. cit., p. 21: "Un bon chanteur peut faire d'un poeme mediocreun poeme remarquable, et un mauvais chanteur giter le meilleur poeme. Ce n'estpas Atort que, souvent Vuk Karad'i6 cherchait un chanteur de qualit6 pour se fairedicter tel chant qui ne lui avait plu. Les auditeurs apprecient, eux aussi, cet art duchanteur. Un bey m'exprima un jour son admiration en ces termes: 'Moi, je nesaurais meme pas faire une composition de trois mots.' En Herzegovine, on m'aparle de paysans qui auraient donn6 le meilleur boeuf de leur 4table pour savoirchanter un seul chant."Les chanteurs sont des artistes, le fait qu'ils se montrent extremement jalouxl'un de l'autre le prouve encore. Un jour, A Sarajevo, aprbs avoir recueilli desphonogrammes de trois chanteurs, je donnaih tous trois la meme r'compense. L'und'entre eux refusa de l'accepter. Je flairai aussit6t que je l'avais froisse de quelquemanidre. Les personnes pr6sentes me previnrent en effet qu'il se consid6rait commeun bien meilleur chanteur que les deux autres."


16/51

TheEpic Technique f OralVerse-Making 1577V yap dlOLa77VaXXoM 7rLKXOUGVivOpcrot21 TLs &KOVOVThUL P4EWTaT7I) ayCL7tre'XfTaL. 1

It is the same in all thrivingoral poetries. The good singerwins hisfame by his ease and versatility in handling a tradition which heknows more thoroughly than anyone else and of which his talentshows him the highest use, but his poetry remains throughout thesum of longerand shorterpassageswhich he has heard.2The formulathus is by no means the unit of the singer'spoetry, butit neverthelessever tends to becomeso, forno singerever tells the sametale twice in the same words. His poem will always follow the samegeneralpattern, but this verse or that will be left out, or replaced byanother verse or part of a verse, and he will leave out and add wholepassagesas the time and the mood of his hearers calls for a fulleror abriefertelling of a tale or of a given part of a tale. Thus the oralpoemeven in the mouthof the samesinger s ever in a state of change;anditis the samewhenhis poetryis sungby others.3His greatnameand the

1 a 351-352.2 Cf. Gesemann, op. cit., p. 68: "Ein Improvisator muss improvisieren k6nnen.Er muss nicht nur ein Dichter sein, um unter Umstanden ein neues Lied singen zukinnen - das war besonders bei Viinjid der Fall - er muss als Haupterfordernisseines Dichterberufes nicht nur einem alten, UiberliefertenLiede sein 'adiquate'Form geben, d. h. die h6chsten kiinstlerischen Miglichkeiten, die ein Stoff in sichtrigt, herausarbeiten K6nnen - er muss also nicht nur im Rahmen traditionellenfiuhlensund mit traditionellen Stilmitteln eine von der eigenen kiinstlerischen Per-sbnlichkeit durchwarmteLeistung hervorbringen kbnnen, sondern er muss, um allesdieses zu kdnnen, eine vollkommene Beherrschung Uiberdie Formenelemente seinesKunststils fertig mitbringen, wenn ihm eine Improvisation oder ein teilweise im-provisierter Vortrag gelingen soll."3 The researches of M. Murko on this point will long remain a model of methodfor students of oral poetry (op. cit., pp. 16-I7): "On a longtemps cru, et l'on croitencore, que les chanteurs ne modifient pas les po&mes. .. J'ai deja dit qu'au con-traire il peut a volonte raccourcir ou allonger ses chants et que le meme poemepeut etre tres different, quant au fond, dans les versions de divers chanteurs. 11estbien certain que dans de telles conditions un texte ne peut demeurer immuable.Deux fois, j'ai emporte avec moi le phonographe perfectionne de l'Academie deVienne. Je n'ai pu enregistrer avec cet appareil les longs chants 6piques, mais ilm'a suffi de fragments de moins de 30 vers pour constater quelque chose d'inat-tendu. Comme il 6tait prescrit de noter chaque texte avant l'enregistrementphonographique, je demandais au chanteur de s'exercer, au prealable, une fois


17/51

16 MilmanParryfame of his verse may urgethose who have learned from him to a morecarefuland more faithful use of memory than that which they wouldshow for the poetry of a lesser singer. But the memoryof the hearerdepends after all on his being habituated to the diction as a whole,rather than on the learningof the poem wordby word, so that he toomust changeand add and leave out verses and partsof verses,and thisprocesswillgo on until all that is left of the poemare its singleformulasand shorterpassageswhich are the finalunits in the traditionaldiction.It should be added here that an oral poetry practiced by guilds ofsingers with masters and apprenticeswould tend to a more faithfuldevant le pavilion, tandis qu'un stinographe notait le texte. J'avais ainsi a la foistrois textes, et j'en ai meme eu quatre dans un cas. La comparaison a montr6 quece ne sont pas seulement des mots isol6s ou l'ordre des mots, mais des vers entiersqui apparaissent sous une forme entierement nouvelle ou disparaissent, si bien quesur 15 vers dictes, par exemple, il n'en reste plus que 8 chantes. Un bon chanteurmusulman du nord-ouest de la Bosnie modifiait a chaque fois le premier vers lui-meme."Il dicta une premiere fois:

Beg Osman beg rano podranio (figure 6tymologique)"Le bey Osman bey s'est lev6 de bon matin";puis en s'exergant:Beg Osman beg na bedem izidje"Le bey Osman bey est mont6 sur les remparts";

et puis il chanta:Beg Osman beg niz Posavlje gleda"Le bey Osman bey regarde la plaine de la Save."

. . Dans le monast&reorthodoxe de Duii pres de Trebinje, en Herzegovine, nousavons entendu les chants d'un paysan attache au monast&re. .. Un des moineset l'instituteur avaient 6crit le commencement du chant sous sa dict6e. Je les priaide noter les variantes au cours du chant, mais ils furent contraints d'y renoncer desle second vers. ... 11 est desormais bien clair pour moi que les chants que nousposs6dons aujourd'hui imprim6s n'ont tous et6 qu'une seule fois chant6s, ou plusexactement dictis, et cela, lors de leur mise par 6crit. C'est pourquoi aussi toutesles tentatives faites pour reconstituer un chant dans sa forme originelle sont vaines.La comparaison des differentes variantes ne peut nous permettre de determinerque le contenu primitif ou encore des parties ou des vers." The bibliography ofthese phonographic studies is given op. cit., p. 7, and the author has given a sum-mary of them, Neues iibersiidslavische Volksepik, in Neue Jahrbiicherfiir das klas-sische Altertum, XLIII (I919), pp. 273-296. Cf. also Radloff, op. cit., pp. xvi-xxviii; Basset, op. cit., pp. 306-307.


18/51

The Epic Technique of Oral Verse-Making 17keepingof poems which had won fame, and that one singermight winsuch a name that his discipleswould find their profit in keeping hispoetry as nearly without changeas they could; but then they are nolonger singersbut rhapsodes,their task is not that of creationbut onlyof memory, and they are merely keeping from age to age the versewhichwas first composedby a singerwho made his poetry, in the waythat we have seen, by an ever varying use of what he had sung andheard others sing.

The ForeignElementWhenpoems thus pass fromone singerto anotherin the sameregionthe languageof the poetry undergoesno changeother than that whichtime may set upon it. But when the poets of one locality hear thepoetry of a singerwho speaks another dialect of their language theirown traditional poetic language may undergo a much more rapidchange. One must suppose that the two dialects are enoughalike fortheir speakersto understandeach otherfairlywell, and that the poemsfrom abroadare such as to please. The fame of some singermay havespreaduntil other singerscame from afar to hear him; or the way oflife in one region may have broughtabout a great liking for poetry, sothat it was practicedmore intensely and carriedto a higher point; orthe singersmay have made theirliving by carryingtheirsongsabroad.1In someway, then, the foreignpoemsareheardby the local singersandrepeatedmore or less as they have been heard, and just as they havebrought into their poetic language new words and forms of their

1 For the way in which the poetry is spread cf. A. Hanoteau, Poesies populairesde la Kabylie (Paris, 1867), p. iv: "Ces poesies sont repandues parmi le peuple pardes chanteurs de profession qui parcourent les villages et vivent des offrandes dupublic. Cette profession est ordinairement hereditaire et se transmet de pare enfils, souvent pendant plusieurs gendrations. ... Quelques-uns neanmoins...vivent retires dans leur villages. Leurs vers ne restent pas dans l'oubli pour cela.Dis qu'ils ont acquis une certaine reputation, les chanteurs qui n'ont pas requ ledon poetique viennent, souvent de fort loin, enrichir aupres d'eux leur r6pertoire.Moyennant une retribution assez lgere, mais toujours proportionnelle aux succesdeji obtenus par l'auteur, celui-ci leur repete ses chansons jusqu'a ce qu'ellessoient grav6es dans leur memoire. Ils vont alors les repandre dans le public et lesapprennent, par le meme proced6, A eurs colligues." Cf. Hymn to Apollo, vv.173-176.


19/51

18 Milman Parryspoken language,so do they make the foreign poetry fit their spokenlanguagein so far as they can do so without any too great loss. Thenew poemsthus take on straightwaya local color,but they keep thoseforeignformswhich cannot be changedwithout harm to the verse, aswell as wordswhose meaningmay be knownonly fromthe context orwhich may be meaningless.' In time these poems, by the unendingprocessof changewhich has beentold of above, becomefusedwith thelocal poetry, yet even when they have been lost as poems they leavetheir markupon the poetic language. Comingfroma traditionwhichhas developed separately,the foreignpoemshave in them many pleas-

1 Cf. K. Krohn, Kalevalastudien I in F[olklore] F[ellows] Communications,XVI(1924), Pp. 76-77: "Dass noch in der neuesten zeit estnische lieder uiberdie Narovanach Ingermanland gewandert sind, beweist am klirsten ein auslaufer des est-nischen liedes von der freierei der himmelslichter, der auf der insel Lavansaari inder nahe der westingermanldndischen kiiste aufgezeichnet worden ist. In diesemfinden wir nicht nur einzelne wijrter, die in ihrer estnischen form und bedeutungbeibehalten, wie z. b. opunen (fi. hevonen)'pferd,' soittamaan = saitamaie 'fahren'(fi. bedeutung 'spielen'), oder durch ein iihnlich lautendes finnisches wort ver-schiedener bedeutung ersetzt worden sind, wie z. b. poikinensa 'mit seinen sihnen'< poisikene 'sthnchen,' lassa 'als kind' < las 'lasse,' sdngyn 'des bettes' < sdrgi'des hemdes,' vilu 'kilte' < Viru' Wierland.' Auch ganze sttze sind bis zur unver-standlichkeit und zu reinem unsinn verdreht worden. Der estnischen aufforderungzum tanze, bis eine mark aus dem boden, ein ferding zwischen den zehen, aus derdrehung des schuhabsatzes hervorspringt (marko maasta, veerik varvastevahelta,kinga kanna kierimisti), entspricht in dem finnischen abklatsch: marka (statt

mnarkka)maasta, verikorvatei vaella 'die blutigen ohren wandern nicht,' kimmikammi kieremidstdunverstandlich). Weiter wird im estnischen geschildert, wie dieschlafende jungfrau; Hebemestakeitas kielti 'aus dem federbette warf die zunge,'kineles kivikojasta 'sprach aus dem steingebaude'; diese zeilen sind im finnischenohne riicksicht auf den gedanken lautlich nachgeahmt worden: hedelmdst heithkielld wbrtlich: 'von der frucht sie verbiete,' kojota (?) kivi 'stein' kovasta aus demharten.' Ahnliche sinnlose tibertragungen aus dem finnischen finden wir in denzauberliedernNordostestiens. Fi. maito 'milch' ist einfach als maidu tibernommen,obgleich ihm im estnischen piim entspricht. Das in Ingermanland in Tyrndn koskiverdrehte Tyrjdn koski 'wasserfall von T.' ( = sttirmischer Tiberiassee) ist mit denworten tiirna kaske wiedergegeben, die im estnischen 'zwergbirke' und 'birke'bedeuten. Der finnischen zeile: suonia sovittamahan 'um die adern zusammenzu-passen,' entspricht in einer estnischen handschrift: Sohvia ei soovita Maie 'Sophiaempfiehlt nicht Marie,' ein unsinn, der nicht nur vom singer, sondern auch vomaufschreiber herrtihrt." Cf. also Meier, op. cit., p. I8; Bi~ckel, op. cit., pp. 59-63;Basset, op. cit., pp. 314-31 5.


20/51

The Epic Technique of Oral Verse-Making 19ing and useful formulaswhich are kept even after the poems whichbrought them to the new land have been lost, and in these formulaslive the formsand wordsof the foreigndialect. Then on the patternofthese formulasotherswillbe madewhich,while they areforeign n theirlanguage,are neverthelessnative work. Even as the poets kept archaicformulas and made new formulas with archaisms, so do they keepforeign formulas and make formulas with foreign forms, so that aforeign word or form may show that a passage in a given poem wasmade abroad, or may prove no more than a contact at one time be-tween the poetryof two regions. The poet andhis hearers, t shouldbenoted, in no way think of these words and forms as the words of a cer-tain locality: like the archaicelements, they simply serve to carry thestyle above the commonplaceof everyday speech.'

TheArtificialElementFinally, to the archaic, the new, and the foreignelementsmust beadded a fourth and last: the artificial. Since the languageof an oralpoet is already far removed from daily speech, his public will not

wonderat him if he shoulduse a formwhichhas neverreallybeenusedanywhere. A whole new word no poet couldmake, since no one couldunderstandhim if he did, but he may make a form like another. Thatis to say, he may make the artificialby analogywith the real. The rea-son for such a creationis of coursethe same whichleads the singerstokeep the old and foreignforms,namely the need of a formula of a cer-tain length which can be gotten only by this means. One poet, drivenby this need, and makinghis verses underthe sense of analogy whichbinds togetherthe whole techniqueof his diction, will hit upon such aphrase,anotherwill take it up, and it too will win its own place in thetraditionalpoetic language.2Another kind of artificialformis due tothe only partial adaptation of old or foreign forms. In certain casesthe meter will allow part of a word to be modernizedbut force thesingers to keep the rest of the wordunchanged.31 Cf. above, p. 5 n. I.2 For the part played by analogy in an oral style cf. L'gpithite traditionnelle,pp. 85-94; Homer and Homeric Style, pp. 145-147.3 The present article discusses only the second of these two kinds of artificialforms (pp. 32-34). For the first kind see K. Meister, Die Homerische Kunst-


21/51

20 Milman Parry3. THE STUDY OF A TRADITIONAL POETIC LANGUAGE

Such is the making of the language of an oral poetry. That theHomeric poems were oral is shown by their diction, which, beingformulaic,can only be traditional and oral. Putting the two sets offacts together,we see that the variety of words and forms which solong puzzledHomeric scholars is the natural and necessaryconditionof the Homeric diction. Being oral it must be traditional,and beingtraditional it must have in it old wordsand forms, and it could bewithout foreignwords and forms only if the people among whom itwas developed had been cut off from the rest of Hellas. Until verylately scholarshave started with the study of the forms and wordsinthe Iliad andOdyssey,and have sought in them an answer which couldnot be truebecause,thoughthey were not awareof it, they werebasingtheir searchon their belief that they were written in just the samewaythey themselves would write poetry. But the sounderway, it is nowclear, is first to learn the oral nature of the Homeric poems- this isshown us by the diction quite apartfromthe language- and then toturn to otherpoetriesof the same kind, where we learn,besidesmanyother things, that the language of Greek heroic poetry could nothave been any other than it is.Even more, the knowledgethat the Homeric diction is traditionalgives us the method we need for the study of the Homericlanguage,and shows us what we may hope to learn from it.

I. - Thespokendialectof the authorof an oralpoem s shownby hispoetic language,whichwill tendto be the same as his spokenlanguagewherever ehas no metrical easontousean olderorforeignwordor ormor construction.Many scholars,whenforced to showwhy the languageof an oral poem follows a given dialect wherever the meter willallow, have supposedthe text to be due to a scribewho, in copying apoem from another dialect, changedit to his own languagewhere hecould. To the bookish mind such a process may seem quite natural;yet it is hard to seewhy a scribeshouldhave wanted to do sucha thing.sprache(Leipzig, 1921); such forms, to give a few examples from many, are nvoxijesbeside the nominative 7Ovoxos,poa"crami as a plural to rp6bcorov, the subjunctivein -,wt, SC^, iv, Oiatva, all the forms with lengthening of a normally short vowel,and so on.


22/51

The Epic Technique of Oral Verse-Making 21If he was merely copying the poem for other readers, why should hethink that they would find it harderto read the originalthan he foundit? If it was in orderthat the poemmightbe read to localhearers,whyshould he change only single forms or small groups of words? Theforeignwordsand forms where the metrical value was differentfromthat of his own dialect would be quite as puzzlingto his hearers,andsince he had writing materials to aid him at his task he should notfind it muchharderto changethe languageof the poemfrombeginningto end. When one does away with the factor of oral verse-makingandits ever-presentcheck on changethere is nothing left to show why thechange of language in a poem should be only partial. Further, thechangeof dialect whichone thus findsin an oralpoetry is so regular nits smallestpoints that one would have to supposethat such a scribekept card indexes. But all such theories, after one has grasped thenotionof traditionaloralpoetry, are seen to be forced.1

II. - On theotherhand an oralpoet, composingn a dictionwhichol-lows his own languagewhere t can,may beusing phrasesandpassageswhich are neitherhis own worknorthatof otherpoetsof the samedialect,whether f his ownorof an earlier ime,butborrowingsromthepoetryofanotherdialect. Thus one cannotsay that a given phraseorpassagein apoem is the work of the author, or of another poet of the author'slinguistic group, just because it has forms of the author's language.Such a phrase or passage may have been taken from another dialectand changedonly where the formshad the same metricalvalue. Thusthe proof that a given phrase or passage is the originalwork of theauthor's dialect calls for the same demonstrationby metrical value asthe proofthat it is the workof a foreigndialect.21I. - A givenword, orm, or groupof wordscan beproved o be theoriginalworkof poets speakinga givendialectonlywhen t can be shownthatno otherdialectwhichhad hada partin thehistoryof thepoetryhad,in either ts spokenor its poetic anguage, hesame wordor orm orgroupof wordswith the same metricalvalue. That the poetic as well as thespoken language of another dialect must be taken into account is afact which is usually overlooked. But it is clear that the poetic lan-guage of one region is as subject to change under the poetry of its

I Cf. above, pp. 9-12, 17-19 and below, pp. 24-25. 2 Cf. pp. 43-47.


23/51

22 Milman Parryneighborsas is that of anotherregion,and just as likely to show thesame variety of forms. Thus the formwhichmay seem to be the workof the author'sdialect may instead be taken fromabroad,whereit wasa foreignform taken fromstill a third dialect.'

IV. - Conversely, wordor ormorgroupof wordswhich s metricallyfalse, or ails to makesense,mustbetheworkof a dialectwhosewordsandformswhenusedwouldmake heverse orrect, rgive t meaning. In mak-ing useof this principle,however,the criticmust be quitesurehe has todo in a given case with a word or phrasewhich is reallymetricallyfalseand meaningless.2V (exceptionto I). - Aforeignorolderormmaybekept n thepoeticlanguageevenwhen hepoet'sown anguagehasaformwhichcould ake tsplace,butsuch a keeping,apartfrommetricalreasons,will be due totheregularuse of the orm alongwith otherwordswhicharealwaysusedas agroupand whichthe poet eels as such,or to the poeticcharacter f theword,or to some othersuchspecial reason. This is most apt to be sowhen the words or formsused with it are themselvesforeignor espe-cially poetic, thus makinga largerword-groupwhich the singerfeelsas a whole, so that he changesnone of its parts.3

VI (exception to IV). - The workingof a formulaic dictionmayitself be thecauseof metricalaults. These will be of two sorts: thosewhich are due to the joining of formulaswhich do not fit, and those1 In the paragraphromKrohnquotedabove(p. i8) mentionwasmadeof thepassageof poemsfrom Esthoniato Finlandand from Finland o Esthonia. Onpage61 of the samework he authorgivesa comprehensiveiagramof thepoeticinfluencesbetween he different egionsof Finlandand Esthonia:West- -- Sa-(Osterb.) -> HinterkarelienFinn- -> vo- -- Finn.-Ost-u. Nordk.land --> lax -- Finn.-StidkarelienItEs-


24/51

The Epic Technique of Oral Verse-Making 23which come from changing a correct formula to fit the needs of asentence.'

VII (exceptionto III). - A form whichseemsold orforeign may bea creationby analogy rom forms whichare really so. The form, how-ever, still stands to show that the poetry was at one time influencedby anotherdialect or that the tradition of the languageis old.VIII. - A word, orm, or groupof wordswhich is old orforeign isnotin itselfproofthat the verseorpassage n which t is foundis thework

of an olderorforeign singer. One must ever be ready to admit that agiven poem may be made by putting together anything from singletraditional words or phrases to whole traditionalpassages.24. THE HOMERICANGUAGES A TRADITIONALNDORALPOETIC ANGUAGE3

Any attempt to localize the traits of Homeric language must belargely balked by the conditions of the search: the lateness of the in-scriptions, their small number which allowsus to know only a part ofthe wordsand forms of any one dialect, and our complete, or almostcomplete, lack of them for many regions. The evidence quoted bythe ancient grammariansis simply by itself untrustworthybecausethey had no sound linguistic or textual method, and so must be leftaside unless it happens to agree with the evidence of the inscriptions.The manuscriptsof the Ionic prosewriters likewisegive us little help,since they have sufferedfrom the ignorance,and even more from themistaken linguisticnotions, of their editors and copiers. The remainsof Ionic and Aeolic verse are more helpful, though they too have suf-

1 Onthispointcf. Les ormuleset la mntrique'Homore, p. 10-42. HomerandHomeric tyle,pp. I38-140. Thesetypesof metrical aultarenot discussed n thepresentarticle.2 Cf. above, pp. 12-17,andbelow,pp. 43-47.3 Savewhereothersourcesarenamed, hedialectalevidenceused nthefollowingpages is taken fromone of the followingworks:C. D. Buck, Introductiono theStudy of the GreekDialects 2 (Boston, 1928); F. Bechtel, Die GriechischenDialekte

(Berlin, I921-1924); O. Hoffmann,Die Griechischen ialekte(Gbttingen,I89I-1898).


25/51

24 Milman Parryfered from copiersand mistaken theories of language, and their evi-dence, as we shall see, bears on the poetic rather than on the spokenlanguage. Thus the study of Homericlanguagemust be based aboveall on the inscriptions.

The Ionic RecordingThe languageof the Iliad and Odysseyhas at least one very commonsound and one very common form which, as the prose inscriptionsshow us, were used only in Attic and Ionic speech, but which might

have been replaced without harm to the meter by the sound andform of the otherdialects:rtfor originala, and the thirdsingularof theimperfect iv. Originala is found in inscriptionsof all the other dia-lects, and is is found where that form of the verb occurs outside ofAttic and Ionic, namely in West Greek,Boeotian,Lesbian, Arcadian,and Cyprian.' Neither-q or a nor'v couldbe the archaismof anotherdialect.2 These traits of Attic-Ionic, though only two, play such apart in the languagethat they are more than enoughto show,in viewof what was said above,3 that the singer (or singers), or rhapsodes,who composed,or gave final form to, the Iliad and Odyssey,spokeeitherIonic orAttic. These traits of languagedo not, however,neces-sarily show whether the recording was due to Ionic singers,whoseverse-makingwas a constant creation, or to Ionic rhapsodes,whowere mere reciters, although all the evidence of contemporaryoralpoetry which I know points to the singer,and none to the rhapsode.Nor do they show in what measurethe diction of the poem- words,phrases,verses,or passages- was the originalcreationof Ionic ratherthan foreignsingers.That the spoken language in question was Ionic and not Attic isshown by the followingsounds and forms: n for originala even afterE,, p, where Attic wouldhave -a; -a--, 7v (El + aiv), lVJLKa,L~-T1iorAttic -rr-, av, ?VEYKa, Tria. Of these Ionic traits at least three - 71 Since there is no evidence that the language of the Homeric poems has anyother elements than those of Ionic, Aeolic, and Arcado-Cyprian, the other dialectsare referred to under the general term of West Greek.

2is s for *fl', cf. Vedic Sanskrit dh.3 Pp. 9-10.


26/51

The Epic Technique of Oral Verse-Making 25after E,t, p, -uu-, and ii - could not be Attic archaisms.' The vari-ous Ionic dialects are not well enough known to allow a more exactlocalization.2

Arcado-Cyprian,Aeolic, and IonicThe various elements of the Homericlanguageare drawnfrom threedialects- Ionic, Aeolic, and Arcado-Cyprian. As we have just seen,the language was last affected by Ionic, and we have the followingevidence to showus that the Arcado-Cyprian lementwas not broughtinto the languageby any direct contact betweenArcado-Cyprianand

Ionic poetry, but came in alongwith the Aeolic element. The Homericpoems have Aeolic at for Ionic Elwhen the next word is KE, al KEbeingfelt more or less as one word becauseof the foreignKE. The Arcado-Cyprian form of the phrase, however, has El (Cyprian ' KE, ArcadianElK iv), and had this phrase ever been known to the Ionic poets itmust, because of the greaterlikeness to the Ionic form, have straight-way taken the place of al KE.3 Likewise, the poems have Aeolic apers,iyles, etc., whereas the knowledgeof the Arcado-Cyprian orms /is,ViMs,tc., would have broughtabout their use because of their greaterlikeness to Ionic i~Ets, iEtds, etc., s doubtless becoming *7~i. A6-Etvat and Opp,where Arcado-Cyprian has oFEivatnd O'p, point thesame way; otherwise one must be willing to grant the bare possibilitythat 66/LEYaLmight have taken the place of o60Fvatby analogy withother non-thematic infinitives in -Ervat, and that the legend of theBeasts (cf.below,p. 37) mightnot have beenknownin Arcado-Cyprian

1 Even though -uu- and -TT- (cf. Buck, op. cit., p. 65) be only different writingsfor the same sound, -TT-nevertheless proves an original Ionic recordingof the poems.The relation of (VELKao 'VE'yKa, and of lar-Ti o fori-a, is disputed. On ~ after e, L,pcf. W. Ridgeway, Origin of Tragedy(Cambridge, England, 19go), p. 3.2 West Ionic -TT- and -pp-would show that the recording was not Euboean. Theuse of 67rot, 7roTE, etc., is attested by inscriptions of Amorgos, Thasos, and Ceos,while the single form in K so far brought to light is 6Ko-a from Erythrae. Thismight show that the last singer (or singers) or rhapsodes spoke island Ionic, butthe evidence is slight (cf. Bechtel, op. cit., III, pp. 87-88; H. W. Smyth, Soundsand Inflections of theGreekDialects. I. Ionic (Oxford, 1894), pp. 289-293). The useagain of such forms as 6KC0S,eKo/at,etc., in Ionic prose would show that literaryprose was developed by a different linguistic group (Miletus?) from that to whichthe recording was due. More cannot be said on the grounds of language.3 Cf. below, p. 35.


27/51

26 Milman Parrypoetry. The history of the Homeric language is thus seen to followthe order Arcado-Cyprian,Aeolic, Ionic, whatever may have beenthe influence back from Aeolic to Arcado-Cyprianand from Ionic toAeolic.

TheArcado-CyprianElementThe followingHomeric forms are found, as far as the evidence ofthe inscriptionsgoes, neither in Ionic nor in Aeolic, but in Arcado-Cyprian; if they are foundelsewhere t is in West Greek:1 the infini-tive of contract verbs in --vat (Arcadian Karv4povpvat,Cyprian KVUIE-

pjvat, Homeric Oop^vat, d~vat, and also, therefore, Pwevat); the de-clension of iFr-stemsin -rs, -77P Cyprian ljEp's, ypaq/ns,Arcadian htEP~V,Homeric ?avv, "Apyv,M~yrnv);he suffix -rEpos n the sense of one ofa pair of things (Arcadian rCjppivrTEov ybvos,Homeric O7XbrEpat,ECWTE-pat, d&yp6rEpos,tc.); Arcadian 5W4la, ad&Xto,cXE"0ov,mia-ra;Arcadianand Cyprian wr6mXt Homeric wr6Xls, Wrr6X-/os, cf. Eustathius 6 bwr6l.XE.osKvrpwYv).2The followingHomeric wordsare found in neither Ionic nor Aeolicinscriptions,but in Arcado-Cyprian;they occur in the Greeklitera-ture we know only as poetic words:In Arcadianand Cyprianatoa, Eprwy, erXWLY, oLFos.In CyprianFaiva, avryw, aprTw, aiT?rp, O(os6, ir6ts, i,aos, XpaV'w.In Arcadian a7rw, aKqO'S, GmPapaL,EXEVUOS,X6aw.3If we considerhow small a part of Arcado-Cyprian ocabularyit iswhich we know from the inscriptions,this number of poetic Homeric

I On Arcado-Cyprian in Homer see H. W. Smyth, The Arcado-CyprianDialectin Transactionsof theAmerican Philological Association, XVIII (1887), pp. 59-133;C. M. Bowra, Homeric Words in Arcadian Inscriptions in Classical Quarterly,XX(1926), pp. 168-176; Hoffmann, op. cit., I, pp. 276-283.

2 It is hard to see what sure conclusions can be drawn from the Arcadian nameof a festival 'EKor6vota. Arcado-Cyprian b36Xopa&s also found at Oropus andEretria. Of the forms given above the following are found in West Greek: AXos;a/iara (in Aetolian); the pairing -repos (in Elean).3 "Eprw and aprbw occur also in West Greek. Nv is found in Arcadian, inCyprian, and in Boeotian, but its use in no one of these three places, is that foundin Homer. Hesychius glosses obvov.. Kbrptot 6pb6ov, which is some reason totake iptobvtos,the epithet of Hermes, as Arcado-Cyprian. If E. Forrer's transla-

tion of the Boghaz-Keui tablets (OrientalischeLiteraturzeitung,XXVII, 1924, PP.114-118) is correct they show Kolpavosto be an "Achaean" word.


28/51


29/51

28 Milman ParryThessalian r'Ewrn,w7rEELpdKovTES,Homeric riocvp~s, lre4ErW/3oXa, -pLr--Xb6Evos, EIrXE,'ErXEro, rXo/atL, etc., lrekXp,7rWpLtos,b p, and the vari-ant reading XL4E7-raLor OXi4EraLin p 221); declension of rlF-stems in-os, -7L (Lesbian pactlqos, Thessalian paot~Xos, Boeotian f3aokXE-);a/wEs,UpaEs,etc. (Lesbianand Thessalian,no evidence for Boeotian);the apocopatedforms of the prepositions;the dual (Thessalian &E-aEb, A[&?]4rotL,oeotian 6brofLarod v,AwaOrav, Lesbiancv]5pe, 7rErtLff7a7a).'(2) In Lesbian and Thessalian the developmentof a followedby aliquid or nasal into double liquidsor double nasals respectively (Les-bian 4'Etyvat, Zovvbcow,"AXXqKT0S,Thessalian ~qyEv, Atovvabrot,HomericEflEpEva, /pE/V, tLXO4tlu77~S,tEp4opE,pE3SEvv6s,Ap'yEVVS, dYtdvvto,vveov,aXX7KTroS,MhXXapE,ppeov, EVppooS, KarappEw); the change of *Tr nd *06into au (Lesbian and Thessalian O6icos,etc., Lesbian IJqcqcos,Homeric6cTao, ~&Tacos, tc.); KaXXoSnstead of Ionic KaX6rin compounds, and inthe comparativeand superlative(Lesbian KanXXiKXtL,tc., ThessalianKaXXL4pobvrELos, tc., cf. Boeotian KaXFOs, omeric KaXXtkvaaaa, KaX-Xtov, KaXXLyvbvaLKa,etc.); KE instead of iv.

(3) In Lesbian 67rrws, brroL,OTL,, etc.; the infinitive of non-thematicforms in -/eat (Lesbian LEvaL, Ef5pEva, O/LEVaL,5bECLt, Homeric Ey-pEvaL, ELEVCaL,EBi~va, 6OLEVaL,etc.); ap3pb76ro(Homeric ip43porov,Opor0b, aiyPp6oEos); a- from t&a- (cf. Lesbian Zovvbcow,Sappho `d'5' tX4tapav, Homeric 6c8Eos, acrpeo's, etc.).(4) In Lesbian and Boeotian the aorist in -ov-.(5) In Thessalian and Boeotian the thematic and non-thematicinfinitive in -yuev(Thessalian riyyev, O6ev, 8 eEv, Boeotian 6,UEV,etc.,Thessalian KperVLEV,7rpaCI~trv,

Boeotian 0eppeyv, etc., Homeric Etaev,OEev, 56~er',tc., d,yvleu, epE'ye, etc., and the variant readings roXE-/z'A~LgEvII 834, d0KovU'iev 79, etc.); the genitive in -oLo (Thessalian ofPelasgiotis and Perrhaebia HIavaovvedoo, roXiMoo, tc.).(6) In Boeotian the genitive in -io (Boeotian 'ApLroTao,tc.); thegenitive in --wov Boeotian SXpx'wv, etc., cf. Thessalian KOLVIoVV, tc.);rol, rat instead of ol, at.2

1 Cf. A. Cuny, Le nombreduel en grec (Paris, 19o6), pp. 454-466, 487-505.2The vocative in -a is attested only in Lesbian verse, which also has -a, so thatthere is nothing to show us which was the common form. The evidence for -qt asAeolic is too slight to be given much weight (cf. Bechtel, op. cit., I, p. 269).


30/51

The Epic Technique of Oral Verse-Making 29Of all these Aeolic traits the only ones which are found in otherdialects are the following: in CyprianKE;in Arcado-Cyprian he de-clensionin -"Fos, "FL,tc.; in West Greekthe non-thematicinfinitivein -p~E,Jcos, etc., rot, rcd; and in Arcadian and West Greek apocopein the prepositions. 'Ar, ar, and br, however, are Thessalian only.'The only cases of apocopein Ionic inscriptionsare two occurrencesofrap. The Homeric words which have Aeolic (and original) a whereone would look for Ionic 7qre dealt with below (p. 36). The tracesofthe digammain Homer are likewiseAeolic, and allow us to say fromwhich of the three Aeolic groups the poetry passed to the Ionians,

but before dealing with this sound we must understand the natureof the Lesbianpoetic language.The TraditionalLanguageof LesbianLyric Poetry

The same forces which created the poetic epic languageof Homercreated the poetic lyric languageof Sapphoand Alcaeus. The scantremainsof these two poets do not allow us to show, as we can do forHomer, that theirdictionis formulaic,and so oral and traditional. Wedo know,however,that Solonand Theogniswerestill followingan oraltradition of iambic poetry,2and that they lived at that time, alwaysso preciousfor our ownknowledgeof oralpoetriesof the past andpres-ent, when verse-makingwas oral but writing known and used as ameans of recordingand keeping.3All that we know of the use of writ-ing in Greeceat the beginningof the sixth century points to the samething for Sapphoand Alcaeus. Yet while we may still feel some doubtas to the way in which they made their verses, there is not the leastdoubt that their poetic language was drawn from an oral tradition:

1 One might add -aFuin Cyprian as equivalent to Boeotian -Eo.2 Cf.Homer nd Homeric Style, pp. 91-92.3 Cf. Murko, La poesieepique en Yougoslavie"... Mujo Selimotif, paysan, nesait pas lire, chante des poemes qui durent jusqu'& des quatre heures" (p. 46)."... Ilija GagljeviC,riche meunier, qui a dict6 un gros recueil de chants populairesau cure catholique" (p. 34). "Mais le plus grand ennemi du chanteur, c'est l'in-struction moderne. Les recueils ont fait perdre l'interet aux chants populaires (je

n'ai gagne la confiance de nombreux chanteurs qu'en leur assurant que je ne pren-drais pas note de leurs poemes)" (p. 30).


31/51

30 Milman Parryonly in an oral poetry does one ever find such a variety of forms thathave each one its own metrical value.'Thus Sapphoand Alcaeus use the endingsof the spoken language,that is of the inscriptions, -w, -a, -av, -EEa, -oaL, -oVTEs perfect activeparticiple), but they also use -oo, -ao, -awv, -a-, -ocs, -6s; althoughthe forms with -oa- were the current forms, they also have 6aos, raos,brLaw, and first aorists such as AcrKacr', oXEcrav; beside Lesbian 7r6bs,E&,Kav,Ei7XOv,pos, iaXos, pos, they have WrTXts,"3ooav, E"KEOOV,Epos,EaXtos,4pa; besiderap, which we alwaysfind in the inscriptions, heyhave the longer irapa, and where their speech used adrv hey shortento adr; they have the uncontracted v6osbeside vj.The foregoingforms are either archaic or found in other dialectsthan Lesbian; the following are artificial forms: aiL/Eatv beside AeolicpuLpv;the genitives /IO6Ev, LTeOEP,EOEP;Nrp?~Lbsbeside Nqph~Es; thetransferof endingsfrom one declensionor conjugationto another,asin rb6Xos,rTEIJ~p)os,xqto0a; the use of v-movable which, in verb-formsat least, was foreign to Lesbian; artificial forms such as &papv(w,rOLKL-

X6bEpot, ppeO-t,7roXvavaKT7a, &Xopyvav,beside 6pivw, 3pat, 6pEaT,rbXXv.In all these cases there is a differencein the metrical value of theforms; the languageis the work of the verse.It is my own view2 that initial digammahad been altogether lostin Lesbian by the time of Alcaeus and Sappho. Where it still seemsto be calledfor to prevent hiatus or makeposition,it is not the sounditself, but its one-time presence,which is felt, much in the same waythat the French feel the one-time presence of h-aspire. It was likelythat even in everyday speech certain word-groupsthat had had aninitial digammawere long kept without elision,as the combinationofthe unelidedarticle before h-aspirekeeps the traces of that sound inFrench (3'olmay be suchaninstance),but the greaternumberof casesin Lesbianpoetry,as in Homer,must have been due to the keepingofthe poetic formula. The poets and theirhearers,being used to theseformulas, would feel no fault where there was hiatus or failure tomake position, while on the other hand, if they were using newerphrases, they were free to treat the word as if it had never had the

1 For the language of Sappho and Alcaeus cf. E. Lobel, 2drwrovsMXr (Oxford,1925), pp. xxviii-lxxvi; 'AXKaLovM'rXnOxford, 1927), pp. xxviii-xciv.2 On F- in Homer cf. Lesformules et la mntrique'Homdre,pp. 43-56.


32/51

TheEpic Technique f OralVerse-Making 31sound. Thus we find in Sappho yXW^o-J 'da-y, ,alvera o,1 and inAlcaeus irvEzbuova''vwt,bro E"pyovbeside ~air'o.TOS "p'yo),Xi0 ' iTEpWEPy-_ ,_ ); likewise the trace of the initial digamma is seenin Sappho's "EtrE, and in Alcaeus' d]yrvetrl[e and iEvaboe, where themeter kept the unshortenedform beside d1Trovnd".X-[

in SapphoandEd7rEand TXE7-ro n Alcaeus. There are in the two poets some33 placeswhere an initial digammawould spoil the meter.The keepingof fp- in someformwhichis notedin ourmanuscriptsasOp-(Sappho fp660o, p6&vPoS,Ipai a) - how it was sounded is doubtful- was a poetic device to keep for these words the power of length-ening the foregoingsyllable. Had such a treatment ever been a partof the spoken language the poets wouldhave used it when the secondelement of a wordbeganoriginallywith *Fp-,and Herodianwould havequoted from Lesbian poetry *f'pdar'y and *ip?7KTro0 (*(-FP'-Y, *-fpKTroS), and not Ebp'yq and aVipKaros,s he did. These two formsare beyond doubt, becausethey show the treatmentwhich the spokenlanguage gave to the group vowel-digamma-consonant-vowel,as inaVEO *SFOw), ,vaios (*vaFU~o), and EbpvolXaos(*E-Fpvol-XaFo~)of theinscriptions. Nevertheless such forms as evbparynd aivprKTroSouldnever have been a part of the spoken language,since -FP-would thenhave been treated as an initial and not an intervocalic sound-group,and they would have become *pdyryqand *aPpqKro.0S.Evbpa7yr ndaiprlKTro an only be understood as poetic forms made to keep themetrical value of *AFpCaiy nd *aFp77KT~S. he following forms inHomer show thisLesbian treatment of the digamma:6EbbesideIonic3Z; XEV0,~evav,etc., beside Ionic Xea, Exav, tc., robpas (*ar6Fpas),airlppas (*aewEFpas),aXkacptvos*TaXkFpP0os),aXaVpo4(*Ka'kaFpo4/).A like treatment was given in Aeolic to the group vowel-consonant-digamma-vowel. Eia&e from *~aFabE)s cited by Choeroboscus as

1 The reading of a papyrus fragment of Sappho (Lobel E 3, 6) is without value,since antiquity, no less than our own times, had its grammarians who, failing tounderstand the hiatus, wished to restore the digamma, e.g. Apollonius Dyscolus,who quotes MumraLTEP Oievo show that the third personal pronoun began with adigamma. Likewise John the Grammarian(Hoffman, op. cit., II, p. 217) states thatthe Lesbians wrote foWVov,ut Aulus Gellius wrote the quotation from Alcaeuswvv-biova 'ivwo, while others emended rbltova to wvebovas. Balbilla has no moreauthority than the grammarians.


33/51

32 Milman ParryAeolic. Homer has edab&beside 6e, aiblaxoL *e"FlaXot),t aVephovra(*aPFepl'ovTa). Of these words orobpas, a7rlrnpas, TakabpLvos,KaXa^vpol?,evake, abtaXoL,a EpbovTa, could have come only from the poetry of theAeoliansof Asia Minor. They are the proofthat the Aeolicin Homerwas brought into the Ionic epic from the Lesbian epic languageat atime when spokenLesbian had lost the digamma.2 Likewise poeticand Lesbian is the treatment of a short vowel followed by digammawhen the short vowel is followed by two more short syllables. Theglossographers quote as Lesbian KaviXEov, 4avb6opot, etc.; the sametreatment is seen in Homer's aXEv6p~Evo,tc., beside adXEovro,etc. Tothis same source belongs the artificial treatment of intervocalic -3F-as -6- in 'C'aw-as, 6e6We-v,tc.3There is no sure case of a form borrowedby the Lesbian singersfrom some other Aeolic dialect: whereverthe Aeolic form in Homerdiffersfrom that of the Lesbian inscriptionsthe form may be archaicLesbian. Yet the readinesswith which the Ionic singerstook over theAeolic forms would ratherpoint to just such an exchangebetween theAeolic groupsbefore the migration to Asia Minor. The Lesbianlyriclanguage offers the same difficultyof deciding between forms whichare Thessalian or Boeotian but might also be archaicLesbian; nor isthe source of the non-Aeolic forms always altogether sure. 'AbXtos,&Ca, EKEO~OV,"Epos,and rr6XTs, owever, are Arcado-Cyprian and notIonic; moreover, the non-Aeolicendings of the dative -ots and -ats,which are the only ones found in Arcado-Cyprianinscriptions,arelater in Ionic than -owL, -yLaL,which were not altogether supplanteduntil well into the fourth century. N-movable, which is generallyclassed as an Ionic trait, is nevertheless found in the dative plural ofconsonant stems in Thessalianand in verb-forms n some Cyprian n-scriptions. There is thus nothing to show that the foreignelement ofthe Lesbian lyric language was not drawn wholly from Arcado-Cyprian. This is a point to rememberwhen we seekfor forms n Homer

1 That 'azxw began with a double consonant is proved by 23 cases where the wordmakes position after a short final vowel, e. g. oEpcXEpala 'xcwv, 8 times.2 Lesbian is used here, of course, in the sense which it bears as a linguisticterm.3 I shall discuss more fully in another article the traces of the digamma inAeolic and Ionic verse.


34/51

The Epic Technique of Oral Verse-Making 33which couldonlybe Ionic: if the form is alsoArcado-Cyprian,we mustgrant that it may have been a part of the Lesbian epic language.

The ArtificialElementThere arein the Homeric anguagea numberof artificial ormswhichcan be understood only as Aeolisms which were changed by Ionicsingers to forms nearer those of their spoken language, though theycould not make them altogether Ionic. The change in each case wasbrought about by a purely oral process. Thus the Lesbian poems

which the Ionic singers earnedhad in them a numberof perfectactiveparticiples in -wv, -ovros. Where the forms of the Ionic participle hadthe same metrical value these were put in their place, save in a fewwords where the Aeolic ending and the meaningof the word led theIonic singers to mistake the forms for presents, as in KEKX?'yovaTO,ETpL-'yovras.' Usually, however, the Ionic singers, when the forms of the twodialects were metrically different,were drawnby their habit of usingdifferent endings for the present and perfect to the endings -60ro, -O'T,which the rhythm forced them to lengthen to -Cros, -(ort. Thus we findreOvWyra besider6Ovbro70WWTr7&rTEeside7re7r-bra, KEK/?rJWTabesideKEK/rl06Tas, etc. This same struggle of the Ionic singer between theforeign form and the habit of his daily speech is likewise the sourceof divbave. He had heard on the one hand the Aeolic poetic *E'avave,but all the usage of his speech tended to ijvae, so that, speaking thetwo forms as it were at once, he made 47ipave. This form, it shouldbe noted, shows that the initial digammahad been lost in Ionic.The so-called"distracted" formswere the workof singerswho, tornbetween their desire to keep the metricalvalue of the genuine uncon-tracted formson the one hand, and their habits of daily speechon theother, in which they used the contracted forms, made such artificialforms as /lvo0u/LEvot, bopb, 6pbwvrcs, pWcwait,tc. Thus the poet whohad heard plva-6AEvoLn verse, but said pI/lEvoL in talk, would tend tobegin the word with yuvw-,whereuponthe rhythm would force him to

1 KEKXYOyprT was a variant reading of Aristarchus at 30, and is also found asa variant readingat y 256, M 125, 1 430. It is found in all the manuscripts at P 756and 759. At B 314 Zenodotus read LTvrLovras, which can only be an attempt toIonize rerpiyovras.


35/51

34 MilmanParrykeep without change the latter part of the poetic form and makepowP'pEVoL. hen faced with a verb of the Aeolic -aco conjugation,such as p-wcoa,where he wouldusually say c0380a, e wouldbe drawninto using the first two syllables of the spoken word '3ow-,and thenwhen forced by the rhythm to supply two more syllables would usethe ending of the spokenword-cowa,hus using the o twice and mak-ing '0coraa. When faced with the poetic 6paw beside the spoken6pW,the singerwould be drawnto the spokenform,but the rhythm wouldforce him to shorten the w-syllableto o, which would be felt only asa poetic sound to fill in the verse until the real last syllable could begiven, making6pbo. The samethinghappenedin certain nouns: oaos,where the spoken form was ow's, became 6Wbs; rpaovEs, where thespoken form was rpOvEs,became rpcovps; ados, where the spoken formwasaos,becameabos; tc. Anotherwayin whichtheuncontracted ormswere broughtnearer to the usual contractedformswas by lengtheninga short root-vowel and making r7pworaa-e from rpowra~EoTe,Trr&VTrofrom 1oraovro, etc. That this change was a purely oral process isshownby the fact that when the root had an a there was a tendencyto keep the genuine uncontracted forms, as in aotlbtal (the voicerepeating in -d~c the movement of dot-), Kpa'dwv,vatET6ovL, etc.A likeoralcreationof artificial orms s foundin such formsas 'IXtov,&yptov, uodttov,6ov, etc., found always before a double consonant,which can have come only from *'IXloo, *&~yploo,*06odLoo,*o0, etc. Inanothercase the loss of the ending*-oo ed to the makingof the poeticform Kpvu0ELsn the phrases iir&yrptovKpVb6EVroI 64) and KaKoLpXaVOvOKpvoEOaa77Z 344) for *Er$L7ll/oo KpVOEVrosnd *KaKo~?XYVooKpVEoG-77S.Van Leeuwen is doubtlessright in thinking the poets were guided bythe model of the adjective 6KpC6ECs,'ut of far more weight in eachinstancewas the need of keepingthe formula,and whereverthe formerpresenceof the *-ooendingis found thereis a markedformulaicdevice.Thus the form 'IMov is found only in the phrase 'IMov rpor&poLOe(3times); 6ov s found only in a special type of clause, 6ovKXEOS Oi)vrOT6XE~Tra(B 325, Hymn to Apollo 156), ovUKpTros r7Tp/~oLyTroV (a 70).'OLAOLov,hich is found only in the phrase 6bAodovoX7MOLOor *6~oLoow-roXtoo,s used six times in the Iliad and twice in the Odyssey,and

1 Enchiridium dictionis epicae2, (Lugduni Batavorum, 19o8), p. 169 n. c.


36/51

The Epic Techniqueof Oral Verse-Making 35here the Arcado-Cyprian 7rT6XEUosight show that the ending *-oobelonged to the Arcado-Cyprianpoetic language. But we cannot besure, since a Lesbian poet might have used the foreign 7r6XEpos n anew phrase.' But whether *-oo s the originalof Arcado-Cyprian 0o,or a middle stage between Aeolic -otoand -w, the creationof the arti-ficial forms in question can have been due only to singerswho had tokeep the formulas.

EquivalentAeolic FormsIt is likewiseonly the theory of oral verse-makingwhich can showwhy certain Aeolic forms were kept when an Ionic form might havebeen used. In at least one case it was only because there was no suchform in Ionic: OcEwas kept becauseIonic used 0o6'for both masculineand feminine. But usually the Aeolic form was kept for less simplereasons, which could be only those of a poet who was drawn moretowards the foreignform he found in poetry than towards the formwhich was habitual with him in his daily speech. An equivalentAeolicform might be kept for one or moreof the followingreasons: the formwas used along with other words, so that the poet felt the groupas aunit and sought to change none of its parts; the form was known tohim morefrompoetry than fromspeech,so that the habitof the poeticlanguage was greater than that of his spoken language; the syntaxof the form was foreign,and thus set it apart from the currentform.Aeolic at for Ionic El is regularly found in the phrases ac'KE,al y p,and a'OE. In the phrase aciKE t has been kept because of the foreign

KE with which it was felt more or less as a single word, like iv for El a'v.So soon as a singlewordis put betweenat and KE,however,the motivefor keeping the Aeolic form is lost and we have el 3E KE,EitLV KEV,YyLp KEV,etc. The use of al -yp, and its metrical variant atOE,s foreignto Ionic speech, being found only in Ionic and Attic prose in highlyemotional passages where the author meaninglyassumes the tone ofpoetry.

'EpL-,which seems to have come firstfromArcado-Cyprian,2endedin the Ionic vocalization to become Apt-. The form Ept- was kept in1 A Thessaliannscriptiongivesol TroX&apXoPTes.2 Cf. p. 26 n. 3.


37/51

36 MilmanParrycertain fixed phrases: e.g. EptabYXEvEsrrot (5 times); ipbflpE~Ecrapot(22 times); ply~ovros r6o-ts "HprS(7 times); pLKVUcabar^V75 times);EpLKv~iacJpa (twice); EpLKv6a-TEKVa(once).' It was these formulasthat kept the prefix, but it was kept unchangedfor another reason:where the radical of the word is a noun apt- is used, while apt- wasbroughtin where the radical is a verb or adjective and thus properlycalls for an adverbial prefix. Thus we find EptL/3Xac,plytovros, EPLKU-8?yS, Pi pUvKo,SpLo-OEVS, PLo-TTvXOS, pLTLUrOS, 'LWoS, but aply'Wopos,aptL6LKETOS,ap'i7,XOS,lpt7rpE7r17,ap,ppafbls,a&pLto-aX's.n iptobvtosandptrlpos the meaning of the radical is so vague that the prefixis scarcefelt as such,and in pL6~,qXjst was the poetic 77X-,or the prosaicOaXX-,which kept the whole poetic word unchanged.There is no need, if we would understandwhy Xabswas not changedto *Xy6swhereasvaosbecamemqbjs,o arguethat therewas no wordXEc~sin Ionic. Indeed two names in the Iliad and Odysseyprove an IonicXEwOs:AELcSKPLrTO and AEtc~,0s re artificial forms which have been madefrom *Aa6KpTrosand *Aa~bjys(*AaFoFlrs) after the spoken *Ae-WKPLTroand *Ac&-6s. 'A-yXEcO-n -ros 6' 'A-y'XEwserTEELrEx 131, 247), wherethe Aeolic would be *ro?06' 'AytXaosEL7rE, shows that there must havebeen many names of this type in early Ionic (cf. 'Ava?LXEo,ApXEXEos,OEP0XEW,XE4Ofopov,f the Ionic inscriptions), for the form ELrEwascommonenoughin the Homericpoetic languageand 'A'yEXaoss foundin otherverseswhere the changewould not have been so simple. TheforeignXabswas kept simply because it was more commonin poetrythanXEW&as in speech,so that the singerswere more habituatedto thepoetic than to the prosaic word. Each time a singer met vaos in verse hewould tend to modify it in the direction of the spoken vE&s. But Xa6s,which figured in so many poetic phrases - 7ro~yva Xaf&v56 times),KoLpavXaiav (II times), Xa6v 'Axatcv (19 times), etc., in the epithet ofgods Xao-oroos 6 times), in heroic names - and was moreover a wordwhich by its meaning had a special dignity, won a place for itself farbeyond the reach of Xos. Where on the contrary a word had no suchspecial place in the poetic language, but differed only in form from thecommonprose word, the singerswould be drawn to the currentform.Thus &osand raos became iEws nd ri-ws, contrary to the usual laws of

I Cf. K. Witte in Pauly-Wissowa, s. v. Homeros, col. 2121.


38/51

The Epic Technique of Oral Verse-Making 37the heroic meter.1 It is wrong, however, in the case either of thesewords or of the wordswhich make hiatus througha lost digamma, orin such phrases as 7/?s 3belcavrEs, ,uXa 3tiv, o pEpiaXEaIdXovrPE,tospeak of metricalfaults, for these irregularitiesof the rhythmare con-stant and accepted, and so arerightly to be felt as the correctrhythmicusage.The Aeolic formO-qp- herever t

Milman Parry 2

Documents

Transcript of Milman Parry 2