Part I: Generalities

I§1. The heading etymology / vocabulary, just mentioned, is of and by itself an indication of recent trends in the study of Greek.

I§2. I postpone the details until we reach that heading. ⊛I begin here by making a general statement about the study of Greek words, which requires both synchronic and diachronic perspectives.⊛ (In using these terms synchronic and diachronic, I have in mind the definition of synchrony and diachrony by Ferdinand de Saussure: “… synchronie et diachronie désigneront respectivement un état de langage et une phase d’évolution.” [1] )

I§3.⊛From an overview of recent works on ancient Greek vocabulary, we can see that most of these works focus on the Indo-European etymologies of Greek words. This is to be expected, since a basic fact about the Greek language is that it stems from a prototypical language-group known to linguists as “Indo-European.” Such works tend to prioritize the diachronic perspective at the expense of the synchronic. In view of this tendency, it is important to stress that the diachronic perspective, which is basic to etymology, needs the support of a synchronic perspective.⊛

I§4. One of the foremost masters of diachronic analysis in general and of Indo-European linguistics in particular, Antoine Meillet, advocated synchronic analysis as the point of departure for diachronic analysis (a shining example is his Méthode comparative, 1925).

I§5. Meillet shows his own all-pervasively synchronic perspective in this most telling definition of language: “Une langue constitue un système complexe de moyens d’expression, système où tout se tient.” [2] The line of thinking represented by Meillet has been described this way by Ernst Risch: “Charakteristisch ist … das Streben nach systematischer Klarheit und die Tendenz, möglichst viele Erscheinungen unter einem einzigen gemeinsamen Nenner zu vereinigen.” [3]

I§6. That said, it is relevant to note that perhaps the greatest achievement in Greek-language studies in, say, the last hundred years—and the first to be mentioned here—was made possible precisely by way of synchronic analysis. This achievement was the decipherment of Linear В by Michael Ventris in 1952. As Ventris proved, the script known as Linear B was a writing system used by scribes in the second millennium BCE to record the earliest attested form of Greek, conventionally designated as Mycenaean Greek.

I§7. This decipherment, as noted in a foundational book co-authored by Ventris and his collaborator, John Chadwick (the first edition of this book by Ventris and Chadwick was published in 1956), made the study of Mycenaean Greek a vast new sub-section of classical philology. [4] Despite its prestige, however, “Mycenaean philology” is still far from being integrated into the central discipline known as classical philology. ◊What I said in 1972 about the marginalization of “Mycenaean philology” still applies in 2008, perhaps even more so.◊

I§8. Initially, such marginalization was caused in part by an attitude of indifference or even hostility on the part of {17|18} some classical philologists. [5] Chadwick offers a vivid account in his introductory book on the decipherment of Linear B. [6] Another cause of marginalization, however, is the ongoing self-isolation of Mycenaean philology from classical philology. ⊛There are signs of this self-isolation even in the introductory book of Chadwick. I made that point more fully in a review of the introductory book by Chadwick (Nagy 1969).⊛

I§9. In general, the most important working rule in Mycenaean studies is a strict adherence to the internal evidence of context (Chadwick 1960); a prime illustration is an analysis, presented by Ventris and Chadwick themselves, of the “horse-tablet” Ca 895 from the archives of Knossos. [7] Instead of paraphrasing the account that is offered by the two authors, I give here the actual text of the tablet, followed by the text of a later account published by Chadwick only after the untimely death of Ventris. In this later account, we see Chadwick’s lively personal reminiscence of a decisive moment in the discovery procedure that led to the decipherment of Linear B: (The capitalized lettering here indicates ideographic symbols; the italic lettering transliterates the Linear В syllabary; the arabic numerals represent the Linear В digital system; the superscripts for ‘male’ and ‘female’ represent two different ligatures used by the scribe; the sign ⦚ shows the breaks in the tablet—as described by Chadwick in what I quote below.)

I§10. Without internal analysis, made possible by applying a synchronic approach, it is difficult to situate Linear В forms diachronically by reconstructing either forward or backward in time–forward from the second millennium BCE into the first and backward from the second millennium into the third and beyond, back to a time when the Greek language was not yet differentiated from other Indo-European languages. In fact, the actual decipherment of Linear B would have been impossible without a synchronic approach.

I§11. ⊛That said, I turn to a point I made in my review of the introductory book of Chadwick (Nagy 1969). In this review, I argued that much further work is needed in situating the language of Linear B diachronically. In particular, there is a need to make the study of Mycenaean Greek less isolated and more relevant to the study of alphabetic Greek.⊛

I§12. ⊛Here and hereafter, I use the term alphabetic Greek with reference to Greek texts written in the era of the Greek alphabet, from the eighth century BCE onward.⊛

I§13. ⊛In the same review of Chadwick’s introductory book (Nagy 1969), I stressed the importance of finding contextual matches in alphabetic Greek that correspond to contextual matches in the Linear B texts.⊛

I§14. ◊As a specific example of such matches, I highlighted in this work (Nagy 1969) the collocation of the words τελεστα and δαμος in an inscription from Elis (DGE 413.8–9), to be compared with the collocation of the Mycenaean words te-re-ta = telestās / telestă and da-mo = dāmos in the Linear B land-tenure tablets from Pylos.◊

I§15. ◊As another specific example, I highlighted in this same work (Nagy 1969) the collocation in Odyssey xix (188) of the Cretan place-name Amnīsos with the divine name Eileithuia, which I compared to the collocation of the Cretan place-name a-mi-ni-so (Amnīsos) with the divine name e-re-u-ti-ja = Eleuthia in a Linear B tablet from Knossos (Gg 705). Both Eleuthia and Eileithuia are variations on the same sacred name, and Eleuthia is actually attested in the alphabetic era, as we see in the evidence from Laconia and Messenia. [9] Further, Strabo (10.4.8 C476) describes Amnīsos as the seaport of King Minos of Knossos, and he notes in this context the existence of a hieron ‘sacred space’ of Eileithuia there. In the same Linear B tablet from Knossos that I mentioned earlier (Gg 705), Amnīsos is designated as the site where the goddess Eleuthia is to receive an offering of honey. In the alphabetic era, well over a millennium later, we find a comparable mention of an offering of honey to Eileithuia. According to Pausanias (6.20.2), who flourished in the second century CE, there is a hieron ‘sacred space’ of Eileithuia at Olympia in Elis: within that space, a local daimōn (ἐπιχώριος δαίμων) is tended by the priestess of Eileithuia, who prepares for the daimōn offerings of barley-cakes kneaded in honey (μάζας … μεμαγμένας μέλιτι).◊

***

I§16. I now turn to another important body of attested Greek textual evidence that requires a combination of synchronic and diachronic approaches. The case in point is ancient Greek epic, as primarily represented by the Homeric Iliad and Odyssey. ⊛From here on, the term epic will be used to refer short-hand to the Iliad and Odyssey.⊛

I§17. The language of epic is an artificial language. In other words, it is a poetic meta-language or Dichtersprache. The better we understand the genre of epic, the clearer it is how its Dichtersprache differs from the natural language that gave rise to it. ◊When I say natural here and hereafter, I mean simply not artificial. And when I say the genre of epic, I am speaking about a historical (and prehistorical) contingency, not about a universal category of literature (Nagy 2005 §§33–40).◊

I§18. In this context, it is essential to cite first and foremost the work of Milman Parry (especially 1928a and 1928b), who produced the most authoritative formulations of such key concepts as oral poetry and formula. It was Parry who showed that ancient Greek epic evolved out of an oral poetic tradition, and that the Dichtersprache of epic was basically a formulaic language. Parry also showed that, with the passage of time, the mechanisms of oral epic poetry (whether recited or sung) become independent of the natural or non-artificial language that engendered these mechanism. Thus a given mechanism may atrophy or even become extinct in the natural language while in the oral poetic language of epic it may not only survive but even become extended—overextended, from the standpoint of the original natural language that afforded the initial precedent. ◊Here and hereafter, I mean original merely in the sense of earlier—with reference to the earliest recoverable phases of reconstruction.◊

I§19. Whatever grammatical rules we may devise for epic, then, may often accommodate processes independent of the natural language, bearing witness to erosion of the original hierarchy of constraints. What is more, such rules may turn out to be limited, that is, not universally applicable.

I§20. Here is an example. We know from the Indo-European cognates of Greek νιφ- ‘snow’, such as English snow, that we must reconstruct νιφ- with an initial *sn-; now some of the formulaic collocations in epic must go back to a prehistoric time when word-initial *sn- was still extant in Greek: hence the making of position by initial ν- in the metrical scheme of, say, ὄρεῑ νιφόεντι in Iliad XIII 754, scanned ⏑ ⏑ – ⏑ ⏑ – ⏑. (Here and hereafter, the ordinal numbers of the scrolls or “books” of the Iliad and Odyssey will be cited by way of upper- and lower-case roman numerals respectively.) Sporadically, however, the factor of precedent extends such a license beyond its etymological confines: a case in point is the making of position by initial ν- in the expression ἅμα δὲ νέφος at Iliad IV 274, scanned ⏑ ⏑ – ⏑ ⏑, even though this ν- does not go back to *sn- (as we see from the Sanskrit cognate nábhaḥ).

I§21. Upon further examination, it becomes clear that the combinations of word-initial *s- with any given sonorant (R), such as n-, created a precedent after the phonological disappearance of *s- in these combinations. So now a word-initial *R- could make position in the epic language—just as the original word-initial *sR- could make position. And without the comparative lexical evidence of cognates in other Indo-European languages, from which we discern an original contrast between *sR- and *R-, the diachronic locus of diffusion for making position with R- would be nearly impossible to determine by way of applying only a synchronic approach.

I§22. Even before Parry, there were studies of the language of Greek epic that showed a keen awareness of the factor of Dichtersprache as {19|20} opposed to natural language. It should be noted, however, that these studies tended to emphasize the artificiality of the Dichtersprache, not the internal dynamics producing it.

I§23. One of the most intuitive of such studies emphasizing la puissance créatrice du mètre was an article by Kurt Witte (1913) in the Pauly-Wissowa classical encyclopedia: his explanation of epic εὐρέα πόντον—by contrast with εὐρὺν πόντον in the natural language—as an artificial creation on the model of εὐρέι πόντῳ has become a Paradestück for illustrating the idiosyncrasies of Homeric Greek. [10]

I§24. The prime concern for Witte was the question of dialectal layers in Homer (on which more in Part II, dialectology), and this interest was pursued further in the book Die homerische Kunstsprache, by Karl Meister (1921). ◊As I will elaborate later, I distance myself from various theories of dialect “layers” in Homer.◊ In any case, the question of dialectal differences is not germane to the present discussion.

I§25. It is Parry’s concept of the formula, and the dynamism of jeux des formules, that has lead to a more profound understanding of Dichtersprache, with its self-sustained equilibrium and momentum partially detached from the natural language but constantly affected by it and originally even united with it.

I§26. Parry extended his work on the epic language to develop a definition of oral poetic composition (1930, 1932), and, after his untimely death in 1936, his work was successfully continued and enhanced by Albert Lord (especially in The Singer of Tales, 1960; ◊second edition 2000◊). The formulations of Parry and Lord concerning oral composition extended not only to Greek epic [11] but also especially to South Slavic heroic poetry (which they studied by applying rigorous fieldwork procedures). [12]

I§27. Parry’s approach to ancient Greek Dichtersprache will be relevant to each of the subsections in Part II: phonology (on the distribution of “digamma,” written as ϝ), morphology (on the construct ἱερὸν μένος plus genitive of name), syntax (on the anomalous combination Ζεῦ πάτερ … ᾿Ηέλιός τε), etymology / vocabulary (on Ἥρη, etc.), dialectology (on the pronouns ὔμμες, etc.).

I§28. The interplay of formulas in ancient Greek epic needs to be studied diachronically as well as synchronically. A diachronic approach helps understand the conservatism inherent in the linguistic heritage of formulas. Configurations {20|21} that otherwise would have long ago become extinct remain embedded in this or that expression preserved by the formulaic system. It is to ancient Greek epic that we owe the perpetuation of the most archaic words in the Greek repertory, often coexisting side-by-side in the same verse with the most recent (on which more in Part II, etymology / vocabulary).

I§29. When it comes to the coexistence of archaisms and innovations in epic, the archaisms generally outnumber the innovations. For an example, I highlight 143 Homeric occurrences of noun + epithet combinations referring to the sea:

I§30. Vocabulary is not the only level on which archaisms outnumber innovations in epic. For another example of this phenomenon, let us look now on the level of morphology: besides 50-odd Homeric instances of old athematic-stem ὦρτο ‘arose’, there are only three instances of the innovative thematic-stem ὤρετο ‘arose’ (XII 279, XIV 397, XXII 102). The evidence of statistics is most telling here.

I§31. Because of such relationships between archaisms to innovations, statistics may also be used to calibrate diachronically the relationships between relatively older and newer grammatical categories. For example, let us test the conclusion, reached by way of applying the methodology of comparative grammar, that the verbal class in -όω is derived or extended from (and originally restricted to) {21|22} the aorist system, while the verbal classes in -άω and -έω are derived or extended from the present system. The statistical evidence of the Homeric Iliad and Odyssey serves as an indication:

I§32. Here the evidence of the Homeric corpus shows that the number of present / aorist verbs in -όω is proportionately inverse to the number of present / aorist verbs in -άω and -έω; thus the epic here statistically reflects the evolution from (a) the prevalence of the original constraints on these three verbal classes to (b) their subsequent eventual breakdown.

I§33. So much for attempts at a diachronic evaluation of grammatical categories in epic. There have also been attempts to extend such inquiry into determining on grammatical grounds the relative earliness or lateness of particular passages. For example, G. P. Shipp (1953) concludes from the data on innovations assembled in Chantraine’s Grammaire Homérique (1953, 1958) that newer configurations are more frequent in Homeric similes than in the rest of Homeric narrative. [15] Shipp goes on to argue that the similes themselves may be a relatively recent accretion in Homeric narrative. An objective critique of this theory is offered by Ruijgh, who bases his counterarguments primarily on the factor of genre-conditioning in formal language. [16] A given Homeric simile may well be as old as or even older than the narrative surrounding it, but its genre may be more recent. That would mean that there could be a higher proportion of grammatical innovations in the similes. A similar argument may be extended to the themes of the Iliad and Odyssey . The question is, how would these themes affect the grammatical texture?

I§34. Grammatical criteria can be used for establishing different traditions, if not different authorship. For example, in his book Homerische Wörter, Manu Leumann (1950) points out that the adverbial use of ἀπριάτην ‘without purchase’ in Odyssey xiv 317 is a “false” extension from the “correct” adjectival usage of ἀπριάτην, as in Iliad I 99; on this account the possibility is raised that the “author” of the verse at xiv 317 is different from the author of the verse at I 99. [17] An important modification must be added to this approach, however: we cannot assume that the source for the verse at xiv 317 was necessarily the verse at I 99: such an assumption “presupposes what we certainly do not know and have no reason whatever to believe—that the Iliad’s phrase ἀπριάτην ἀνάποινον (or the like) could not have been known tο the Odyssean poet from the traditional stock of phrases common to all poets, and existed nowhere in the world but in that one line of the Iliad.” [18] The mutual exclusion illustrated by this example establishes different traditions only—and not necessarily different authorship. {22|23}

I§35. As for the Homeric Hymns, it is again by way of grammatical criteria that we can see how they are not directly derivative from the Iliad or the Odyssey and how their background may even have stemmed from an earlier phase of epic. This earlier phase may have dated back to a time before the establishment of the Iliad and the Odyssey as they have survived (for the notion of establishing fixed Homeric texts, I cite Lord 1960, especially chapter 6). [19] In other words, we may be dealing with a phase when the Dichtersprache of epic was not yet moribund (that is, before the onset of fixed texts). And such a phase may have given rise to elements in the Homeric Hymns that are clearly independent of the Iliad and Odyssey. Such independence is demonstrable wherever the Hymns preserve a grammatical archaism corresponding to an innovation in the Iliad and Odyssey. Granted, the situation is more often the reverse, and that is why it is assumed by many that the Hymns are in all respects more recent. A case in point is the innovative thematic-stem πολυπιδάκου ‘rich in springs’ (Ἴδης) in the Hymn to Aphrodite (54) as opposed to the older athematic-stem πολυπίδακος (Ἴδης) in the Iliad (XIV 157, etc.). But there are counterexamples, however rare. [20] I list here three such counterexamples, which are sufficient to show that while the Hymns may be in some ways newer than the Iliad and Odyssey, they are nevertheless at least partially independent survivals of an unattested stage of epic Dichtersprache that gave rise to the Iliad, the Odyssey, and the Homeric Hymns:

1. older athematic-stem χρυσάορα ‘with sword of gold’ in the Hymn to Apollo (123) vs. innovative thematic-stem χρυσάορον in the Iliad (XV 256)
2. older κατάκειαι ‘you lie down’ (with intervocalic σ phonologically lost) in the Hymn to Hermes (254) vs. innovative κεῖσαι (with intervocalic σ morphologically restored) in the Iliad (XIX 319).
3. older verse-final (τό σε φράζεσθαι) ἄνωγμεν ‘we bid (you consider this)’ in the Hymn to Apollo (528) vs. innovative *ἀνώγαμεν, with paradigmatic extension of α; the similarly verse-final but 1st singular (τά σε φράζεσθαι) ἄνωγα of the Odyssey (xx 43, etc.) could not have formulaically generated the older 1st plural ἄνωγμεν, while the predictable *ἀνώγαμεν could not fit metrically.

I§36. The factor of Dichtersprache, of course, extends beyond the language of Homer and the Homeric Hymns; not only is it equally relevant in non-Homeric “epic” such as Hesiodic poetry but also in the genre (or, more accurately, genres) of lyric.

I§37. ◊I include iambic and elegiac poetry in this category of “lyric” and restrict the category of “epic” to poetry composed in the meter known as the dactylic hexameter.◊

I§38. In the works of Dover, Page, and Scherer (all 1964) on the poetry of Archilochus, we can see various attempts at describing the conditioning of poetic language {23|24} in lyric as distinct from epic. Whereas Page argues that the technique of composition in elegiac poetry is essentially the same as in epic, Dover points out that the elegiac language of an early poet like Tyrtaeus already employs non-Homeric phraseology in metrical environments where Homeric phraseology would have fit just as well. [21]

I§39. A case in point is a passage in Tyrtaeus F 7 (verses 21 and following), a text contextually parallel to a passage in the Homeric Iliad XXII (verses 66 and following), where we see Priam musing about his own fate:

I§40. Dover remarks: “All awkwardness could have been avoided if [Tyrtaeus] had availed himself fully of epic diction and said νέῳ δέ τε; but, like all the early elegists and the composers of verse inscriptions, he eschewed those combinations of particles which are characteristic of epic and distinguish it from drama and prose.” [22] Or again, in the preceding verse in the same passage from Tyrtaeus (verse 26), we read:

I§41. The syntactical disagreement here between singular and plural neuter could have been obviated by replacement of νεμεσητόν with νεμεσητά—which implies an earlier observance of initial ϝ- latent in (ϝ)ἰδεῖν: such observance is a lexically-conditioned metrical precedent regularly operative in the epic (I emphasize that the overtness is not phonological but metrical: see Part II, phonology). And yet, the wording of Tyrtaeus regularly avoids this metrical precedent. “In this respect [Tyrtaeus’] principle is that of the Ionian elegiac poets, but it is conspicuously at variance with that of all the archaic verse inscriptions, which, whatever their metre and whatever the degree of epic phraseology they adopt, observe digamma [ϝ] in regions where the vernacular observed it and omit it in regions where the vernacular omitted it.” [23]

I§42. In short, Dover has shown that “the language of [Tyrtaeus] is derived not primarily or directly from epic.” [24]

I§43. Also relevant is the argument of Page concerning the poetry attributed to Archilochus, the iambic and trochaic meters of which can be traced back to a preliterate period of poetic tradition. [25] Page compares an archaic poem attributed by some to Homer, called the Margites, which was composed in dactylic hexameters interspersed with occasional iambic trimeters. Nevertheless, Page thinks that the attested verses of Archilochus—not only the elegiac verses but also the iambic and trochaic ones—derive most if not all their traditional elements directly from the epic. [26] As he concludes, “The formula-element comes almost exclusively from the epic, and the new style is formed by more or less extensive adaptation of traditional phrases combined with components, generally in moderate measure, of premeditated word-selection.” [27] This line of thinking allows for the separateness {24|25} of lyric from epic only in terms of innovative borrowings from epic, not in terms of idiosyncratic archaisms in lyric. There is a parallel line of thinking in the analysis of Tyrtaeus by Dover (1964), as well as in Page’s Alcman (1951), or again in his Sappho and Alcaeus (1955a). Another parallel is the approach of Zumbach to the Homeric Hymns. [28]

I§44. I argue against this line of thinking. I start by questioning Page’s theories on the poetry native to the island of Lesbos. Although he concedes that the poetic language of a figure like Sappho, unlike that of Archilochus, generally resists the influence of epic, [29] he invokes epic influence wherever he sees in the language of Sappho any deviation from what he thinks are the regularities of the non-poetic language of Lesbos. [30] The possibility of at least residual non-epic native Lesbian archaisms is not seriously considered by Page. It is as if the only traditional elements in Lesbian poetry were those that are traceable to epic.

I§45. A work by Harvey (1957) on epithets in lyric adopts a stance similar to that of Page: contextually skewed usages reflecting ornamental (and therefore inherited) epithets are explained as archaisms only in terms of stylized cross-references to the epic. [31]

I§46. Here is a basic objection: to deny the possibility of archaisms generated by lyric independently of epic is to ignore the evidence of a primary external feature of Greek poetry and song, meter. From a metrical point of view, lyric is independent of epic. As Ulrich von Wilamowitz-Moellendorff has shown (1921), the dactylic hexameter of epic is highly complex and derivative, whereas the meters of lyric as we find them in the poetry of Sappho and Alcaeus are by comparison more simple and basic. [32] Moreover, Meillet (1923) has shown that there is a structural correspondence between the meters of the poets of Lesbos and the triṣṭubh / jagatī meters of Vedic Sanskrit. In terms of the theory propounded by Meillet, the comparative method allows the meters of the Lesbian poets to be derived directly from reconstructed Indo-European prototypes. [33] On the other hand, Meillet thinks that the complex metrical structure of the dactylic hexameter resists any such direct derivation. [34]

I§47. It is a given that epic, as represented by the Homeric Iliad and Odyssey, ultimately prevailed as the prime medium of pan-Hellenic song culture. So the likelihood of its influence on other forms like Lesbian lyric is also a given. Nevertheless, the comparative evidence afforded by the metrical structure of lyric forms shows that in some ways lyric was independent of epic and preserved elements that were more archaic than the elements we find in epic. Besides the available evidence on the level of form, there is also evidence on the level of meaning. A prime example, as Marcello Durante has demonstrated (1960, 1962), is the phrase κλέος ἄφθιτον ‘imperishable glory’ in a poem of Ibycus (F 282.47): the context of this phrase is more specific—and more accurate etymologically—than the corresponding context of the same phrase in the Iliad (IX 413). [35] What I mean here when I speak of etymological accuracy has to do with the cognate context of the cognate phrase śrávas … | … ákṣitam ‘imperishable glory’ in the Rig-Veda (1.9.7). The significance of these cognate phrases κλέος ἄφθιτον and śrávas … | … ákṣitam is enhanced by their Indo-European heritage.

I§48. There is a wide-ranging survey of this and other such examples of indogermanische Dichtersprache in a book by Schmitt (1967), which features prominently the comparative evidence from ancient Greek, including the case of κλέος ἄφθιτον. There are also further studies of further Greek evidence, centering on particular words or phrases: examples include Thieme (1938, 1952abc) on ἀριδείκετος, ἐρικυδής, νέκταρ / ἀμβροσία, Ἀίδης; also Durante (1958, 1960, 1962) on ἔπεα πτερόεντα, ὕμνον πλέκειν, ὠκέες ἵπποι.

I§49. ◊At this point in the 1972 version of the book, I announced my intention to publish what eventually became an entire monograph centering on the Greek phrase κλέος ἄφθιτον ‘imperishable fame’ and the cognate Indic phrase śrávas … | …ákṣitam (Nagy 1974). In the monograph, I concentrate not only on the cognate phraseology but also on the cognate metrical frameworks of this phraseology.◊ {25|26}

I§50. The meter as well as the meaning of such cognate phrases as Greek κλέος ἄφθιτον and Indic śrávas … | … ákṣitam needs to be studied in greater breadth and depth. Even though metrical studies have proven to be a most effective tool for resolving broader questions of genre, a great deal remains undiscovered about the structure and dynamics of Greek meter, on both the diachronic and the synchronic levels. In the case of diachronic studies, two most productive works available as points of departure for further investigation have already been mentioned: Wilamowitz (1921) and Meillet (1923). [36] In the basic area of metrical typology, I cite the work of Lotz (1960). [37]

I§51. On the synchronic level, a groundbreaking inquiry into Greek meter has been W. S. Allen’s correlation of verse-ictus with a theory of Greek lexical stress (1966). In terms of this theory, the factor of stress is independent of intonation, the patterns of which are morphologically as well as phonologically conditioned except in finite forms of the verb (where the conditioning is solely phonological). Moreover, the phonological conditioning for stress is different from the phonological conditioning for lexical intonation familiar from the most elementary grammars (I cite, for example, the “law” that says that a proparoxytone becomes paroxytone when the ultimate vowel is long).

I§52. Allen’s theory forces a reformulation of the overall history of Greek accentuation: from the standpoint of the phonological evolution from ancient to modern Greek, the lexical heritage preserves the patterns of intonation and loses those of stress, but at the same time it replaces the phonological dynamics of intonation with those of stress. In other words, the evolution into the modern Greek stress-system, with patterns inherited from an old intonation-system, reflects a sort of chiastic compensation. As for the actual phonological conditioning of stress in ancient Greek, Allen’s formula is as follows:

I§53. Here are two examples, where the highlighting of the vowels of syllables indicates stress:

I§54. The testing of this formulation on further samples of dactylic hexameter and iambic trimeter leads to this conclusion: Greek metrical ictus essentially coincides with Greek lexical stress. (The verse-final syllable ⏓ counts as latent ⏑ or – when the preceding verse-rhythm is …– ⏑ … or … ⏑ – … respectively. This “law of indifference” has been explained by Allen. [40] )

I§55. From a diachronic point of view, Greek lexical stress can be motivated as the underlying factor of meter itself. Applicable again is this dictum: the dynamics of poetic language, whether or not they are still dependent on the natural language, nevertheless originate from the actual grammatical rules of the natural language. [41] Appropriate too is the following remark of Wilamowitz: “Das geschichtlich wichtige Ergebnis ist, dass die ausgebildete Metrik in dem was ihr gemeinsam ist und auch in ihren Anomalien auf einen Zustand weist, in dem sich alles vertrug.” [42] What is more, Allen’s theory “gives an immediate and simple explanation of a number of the ‘laws’, ‘canons’, ‘bridges’, etc., regarding the positions at which heavy word-finals may or may not occur; all reduce simply to the avoidance of word-division where this would produce conflict between stress and ictus—more particularly in the coda section of a metrical structure.” [43]

I§56. Especially productive is the application of Allen’s formula—which he calls a hypothesis—to “Porson’s Law” in iambic trimeter. [44] And it is this very applicability, in the specific instance just mentioned and in others, that best refutes any charge of circularity. As Allen himself argues, “the detailed testing of the hypothesis involved certain specific phenomena, which had not been considered as such in establishing the hypothesis, and which in some cases concerned quite different metres—and was nevertheless found to have considerable explanatory power in relation to them.” [45] What is more, the hypothesis itself was derived from empirical observations about the nature of the two basic classical Greek metrical rhythms, dactylic and iambic, “and the fact that any single set of correlations was traceable, that a hypothesis based on them produced a high proportion of agreement in both types, and that the distribution of agreement and disagreement in the line showed clear and intelligible patterns, seems likely to be significant.” [46]

I§57. Further, Allen’s hypothesis allows new insights into such grammatical phenomena as enclisis and word-juncture. [47] Even further, it allows new insights into stylistic devices achieved with meter.

I§58. For an example of such devices, I focus on the utilization of an exceptional space allowing disagreement between stress and ictus: if the penthemimeral caesura of the iambic trimeter divides a spondee, that is, if a word-break in the third foot occurs between two long syllables, then stress clashes with ictus (both stress and ictus are indicated by way of highlighting; “|” is foot-juncture, “‖” is caesura): vs. hypothetical

I§59. Nor is a spondee avoided in this space: according to Allen’s statistics, verses with penthemimeral caesura have a spondee for the 3rd foot 75 per cent of the time. [48] Allen therefore raises the possibility that the resulting tension between stress and ictus here was deliberately induced, as a verse-initial counterbalance to the cadence. [49] That the non-avoidance is deliberate is also suggested by the fact that Sophocles often uses this space for the sake of contrast in repetition, as we see in this example (with stresses highlighted): Here is another example:

***

I§60. Here I turn to the relevant linguistic evidence of lyric poetry. We encounter more difficulties here, because of the generally poor state of the textual tradition. Not only has there been an irretrievable lapse in transmission for the majority of lyric texts but even what few texts remain are highly vulnerable to corruption, given the nature of their fragmentary survival as quotations or paraphrases (Zitatfragmente) and the like.

I§61. The occasional discovery of a relatively complete ancient text containing an early lyric composition, such as the Louvre papyrus (ca. first century CE) containing Alcman’s Partheneion, may reveal the inaccuracy of the later transmission that had previously been the only basis for establishing the text. Here, for example, are (1) lines 64–65 of the Partheneion as attested in the Louvre papyrus, followed by (2) the version derived in quotation-form from the medieval textual tradition: Such examples show the need for sound and systematic grammatical investigation of the lyric fragments. [52]

I§62. The textual tradition of Alcman illustrates still another complication. In the Alexandrian exegetical tradition, a consistent awareness of the contrast between a relatively unlocalized epic dialect and a relatively localized lyric dialect prompted sometimes misdirected efforts at maintaining fidelity to the manifold and varying idiosyncrasies of localisms. For example, the Laconian provenience of Alcman’s poetry gave rise to the regular substitution of σ for ϑ in the Alexandrian edition of the Partheneion: thus πάσον for πάθον in line 35, παρσένος for παρθένος in line 84, and so on. Granted, the process θ > σ is a Laconian phenomenon, attested also in such contexts as the Laconian wordings attested in Thucydides (σύματος for θύματος, 5.77) and in Aristophanes (σιός for θεός, Lysistrata {28|29} 81). Such instances of s (< θ) are also clearly attested in the latter-day descendant of the Doric dialectal family to which Laconian belonged, Tsakonian. Examples include to séri ‘harvesting-time’ (τὸ θέρος) and silikó ‘female’ (θηλικός). [53] Nevertheless, attestations of σ for θ in Laconian are relatively later, and the more archaic inscriptions show θ. [54] As Risch argues, if only θ is possible for the era of Alcman (whom he dates to the seventh century BCE), then it follows that “Dieser typische Lakonismus σ statt θ muss daher… auf nachträglicher Modernisierung beruhen.” [55] In other words, the Alexandrian exegetical tradition may be responsible for dialectal features unwarranted by the previously transmitted lyric text.

I§63. There may have been even more serious Alexandrian modifications: on the grounds of genuine formal convergences between the dialect of Laconia and the dialect of Cyrene, remote and familiar respectively from the Alexandrian standpoint, the editorial tradition may have overextended these convergences by selective application of Cyrenaean forms to the text of Alcman: “bleibt als einziger Ausweg nur noch die Annahme, dass der uns vorliegende Alkmantext in verschiedenen Punkten nicht authentisch ist, sondern nachträglich an den Dialekt von Kyrene angepasst wurde.” [56]

I§64. On methodological grounds, however, we need not necessarily fault the Alexandrians’ use of dialectal material from Cyrene as a linguistic point of reference for establishing the text of Alcman. Cyrene was founded by Thera, which in turn was founded by Laconia. So Cyrene probably preserved linguistic and other social archaisms that stem from Laconia. It is important to add, in support of the validity of this Alexandrian approach, that the application of Cyrenaean forms was selective (as with the textual convention of σ for θ, which is Laconian only but foreign to Cyrene). Risch concludes: “Wenn man bedenkt, dass Kyrene für Alexandrien schliesslich die nächste dorische Stadt war und dass kein geringerer als Kallimachos aus dieser Stadt stammte, wird man die Möglichkeit, dass die alexandrinischen Gelehrten sich bei der Bereinigung des Alkmantextes bis zu einem gewissen Grade nach dem Vorbild des kyrenäischen Dialektes richteten, nicht von vornherein verneinen dürfen.” [57]

I§65. In analyzing the text of Alcman, then, we need to reckon with as many as three possible phases: (1) the archaisms of an inherited Dichtersprache, (2) a Laconian veneer, (3) Alexandrian editorial modifications based on dialectal studies of Cyrenaean and Laconian. [58]

***

I§66. The factor of Dichtersprache is not confined to Greek epic and lyric. [59] It is ever-present also in such genres as Attic tragedy and comedy. [60] For example, in the dialogue of tragedy, forms that feature a borrowed Doric ᾱ are attested in metrical situations where a native Attic equivalent with η is not available. So we find forms like νᾱός νᾱῶν vs. native Attic νεώς νεών, but exclusively νηΐ νῆες rather than the metrical equivalents in Doric, vᾱΐ νᾶες. [61] Meillet has broadened the perspective from the confines of Dichtersprache into a more general notion of Kunstsprache, since even literary prose is distinct from the natural language as such. [62] {29|30} The basic principle is this: “Chaque grand groupe dialectal a tendu à se créer sa langue littéraire propre.” [63] In the more restricted sphere of poetry, Wilamowitz had noticed the same phenomenon when he remarked: “Versmass und Sprache gehören zusammen.” [64]

I§67. Extending this principle from poetry to prose, Meillet offers the following formulation:

I§68. What with the firm establishment of Ionian as the official vehicle of prose, it becomes clear why Athenian prose had such a hard time becoming Attic:

I§69. It is characteristic of Meillet, who is known for “die Betonung, dass die Sprache und jedes Wort ein Glied des sozialen Lebens ist,” [67] that he gives a social motivation for the dialectal repartitions of Greek Kunstsprache:

I§70. After considering the full extent of Kunstsprache in ancient Greek, we come to the realization that what we really lack is a sufficient attested corpus of the natural language:

I§71. Given that the Hellenic institution and way of life subsumed under the term πόλις was pivotal in the maintenance of individual dialects in the earlier part of the first millennium BCE, the emergence of a common Greek language, a Κοινή, became inevitable in the later part of that millennium, with the emergence of powerful leagues and empires that transcended the institutions of the πόλις. But the dialectal components of this Κοινή need not have been inevitable. How, then, do we explain the Attic-Ionic basis of the Κοινή that actually did evolve? Even in the era of the πόλις, there had been a latent tendency toward the ultimate leveling-out of localized idiosyncrasies:

I§72. A prototype of une langue commune, a Κοινή, was the Ionic dialect as spoken in the second half of the first millennium BCE. [71] But the constitution of what actually goes under the name Κοινή was more complicated, resulting from a whole series of dynamic historical processes: (1) the hegemony of the Achaemenid Empire over Ionian cities, (2) the growth of the so-called Athenian Empire, (3) the ascendancy of Macedonia in the Hellenistic world, (4) the conquests of Alexander the Great, and (5) the superimposition of the Roman Empire. The consequent social effects on the evolution of the Greek language have been masterfully outlined by Meillet. [72] The key to the prevalence of Attic through these processes is cultural prestige:

I§73. The mixed Attic-Ionic basis of Κοινή is most clearly explained by Meillet:

I§74. This formulation by Meillet is a tour de force in displaying an awareness that language operates on two levels: internally, it operates by way of grammatical rules (φύσει), while externally it operates by way of social convention (θέσει).

I§75. So, elements of natural language previously repressed in official Attic and Ionic come to the fore in Κοινή, but even here we see the persistence of a dynamic tension between la langue courante and la langue littéraire. [75]

***

I§76. Before we leave the topic of the Κοινή, I mention in passing the hypothetical model of a Κοινή par excellence, the Common Greek that linguists attempt to reconstruct by pushing their reconstructions of attested phases of Greek as far back in time as possible. From the standpoint of nineteenth-century linguistic science, mention of this hypothetical stage of Greek, Common Greek, should have appeared first in this book, since the methodological point of departure for the study of any Indo-European language is deemed to be the prototypical form of that language. My emphasis on current trends in the study of Greek, however, has given me the excuse to avoid such a beginning for this book. Current progress in the study of the Greek language depends primarily on continuing refinements of perspectives on the actual linguistic material available. That, in essence, is the contribution of philologie to grammaire comparée. [76] {32|33}