González, José M. 2013. The Epic Rhapsode and His Craft: Homeric Performance in a Diachronic Perspective. Hellenic Studies Series 47. Washington, DC: Center for Hellenic Studies. http://nrs.harvard.edu/urn-3:hul.ebook:CHS_GonzalezJ.The_Epic_Rhapsode_and_his_Craft.2013.
4. The Euboian Connection
4.1 The Cultural Argument
This book argues for the central role of Athens in the performance-driven fixation of the Homeric text. Because my argument turns on the dominance of Athens, what Nagy 2001 has called the Panathenaic “bottleneck,” I must take some time to review a competing alternative to Athens that has grown increasingly popular during the last thirty years. This alternative goes back to Wathelet 1981, who proposed Euboia as the location in Greece where the Homeric poems were composed and possibly recorded in writing. His theory was motivated by the desire to find an alternative explanation to what scholars dismissively called the “Attic coloring” of the Homeric poems. This label comprises all the dialectal forms deemed Attic as a last analytical resort, that is, because they cannot be ascribed to any other of the dialects long recognized as major constituents of the Kunstsprache. Since most scholars assume that the Iliad and the Odyssey had received its final shape, usually with the help of writing, well before Athens became the conduit of their transmission to posterity, it is an embarrassment for them to find the text somehow affected by Athens’ dialect. In the past, two main strategies were adopted to explain away these data. Whenever these forms could be readily emended to earlier dialectal layers, scholars assumed that they were trivial (and mostly unintentional) scribal changes that reflected contemporary linguistic practice. This notion was helped by the fact that often the emended, hypothetical older form cured what scholars deemed prosodic blemishes. When an expedient substitute could not be devised, a second strategy was ready at hand: the passage must be a late, inorganic interpolation, either an attempt to doctor the text or an instance of what is contemptuously labeled a “rhapsodic elaboration.” In the search for the ‘true Homer,’ such suppositious insertions could be cleanly excised and the authentic text saved from the bungling experiments of latter-day poetasters. [1] The paramount motivation driving these strategies was the offense caused by what appeared to be evidence of late Homeric composition, however limited in scope (for this coloring was regularly pronounced “trivial” and “superficial”).
But what if one could suggest a source other than Athens for all or most of these data, a source that was old enough to remove the taint of embarrassment? Scholars who embrace an early-archaic recording of Homeric poetry in writing could then allow these forms to stand without the urge to emend or excise them. This is precisely what Wathelet 1981 made possible by substituting Euboia for Athens. The Euboian dialect belongs to the West Ionic variety, and it shares with the Attic some of the traits that had until then been interpreted as Athenian coloring. The theory of a Euboian origin of Homeric poetry had several attractive advantages. The excavations at Lefkandi proved that Euboia was a Greek center of power as early as the ninth century. Thus, the chronology was open for as early a written fixation as one could plausibly defend. The testimony in Hesiod’s Works and Days 650–657 about the funeral games for Amphidamas was further marshaled to make the case for a vigorous tradition of epic performance on the island in the early archaic period. That Janko had dated Hesiod to the early seventh century, a date widely accepted and roughly in agreement with West’s own, lent further credence to the scheme. [2] The cultural importance of Euboia was boosted by the converse contention that during the same period (ninth to the end of the seventh century) Athens had been a cultural backwater. I will show below that this is not true, but it was only too readily accepted by most. Another fact that seemed to buttress the Euboian theory was this island’s early role in trade and colonization, which made it a likely and convenient instrument for the diffusion of Homeric poetry. Dramatic corroboration was provided by the discovery in Pithekoussai of an inscribed skyphos from the late eighth century with two lines of hexameter that, for many, made clear reference to Pylian Nestor’s cup. [3] The regular size and spacing of the characters, together with the use of colons as interpunction between words and perhaps even metrical clauses, prompted Immerwahr to suggest that the inscription reflected Euboian bookhand practices. Voilà: late eighth-century evidence of Euboian poetry books! Positing Euboian dictated transcripts of the Iliad and the Odyssey could not be long in coming. To round out the composition, some suggested that the alphabet had been invented by the Euboians, and at least one scholar, that the purpose of this momentous invention was the recording in writing of Homer’s performance. In view of these arguments it is hardly surprising that a number of prominent Homerists would soon espouse a version of Wathelet’s Euboian theory: West, Powell, the early Cassio, and Ruijgh are only the more prominent; Janko is sympathetic. [4] It is fair to say that the Euboian theory, at least in its more restricted form, has now become the consensus view. Athens, never in favor, has finally been displaced.But the narrative is all too clean, and the theory faces insuperable difficulties. I must make clear at the outset that I do not wish to deny that epic poetry—even poetry whose subject matter might reasonably be classed Iliadic or Odyssean—was performed on the island of Euboia during the archaic period, perhaps even during the Dark Ages. And to the extent that Euboian itinerant rhapsodes visited Athens during the sixth and fifth centuries to compete at the Panathenaia, one cannot completely rule out some influence of the West Ionic Euboian dialect, even as one must reckon with island Ionic and East Ionic. The case for a significant Euboian influence is muted by the absence of any indication in the Homeric Vita traditions that Euboia played a role in the diffusion of Homeric epic. [5] Contrast this silence with traditions that tied Homer to Khios, Smyrna, and other places. [6] All the same, it is possible (and I do not deny) some, probably marginal role of Euboia in the diffusion of the Iliad and the Odyssey. [7] What I think does not withstand scrutiny is the notion that Euboia exerted decisive influence upon their text during the Dark Age and the archaic period. An exaggerated view of the possible Euboian influence is usually attended by a corresponding devaluation of the cultural importance of Athens at the beginning of the archaic period. [8] Cassio (1998:19) cites with approval the comment by Murray (1993:185) that “[t]he city of Athens remained a centre of occupation throughout the Dark Age, and from about 900 was the most prosperous and advanced community in Greece.” To this cultural preeminence witness the fine examples of geometric art by the Dipylon Master and his followers; the abundant eastern imports starting in the 850s; the presence of Attic pottery in the lowest levels of Al Mina; and the presence of eastern goldsmiths in the city in the early eighth century. [9] Only after 730 did the urban center decline until the end of the seventh century. But this decline was accompanied by a resurgence of the Attic countryside, [10] and we cannot therefore make definitive pronouncements about the cultural decline, unless we assume without proof that the aristocratic gentry responsible for the rich burials on the countryside could not have supported communal festivals, poetic performance, and the arts. Even during the seventh century, “the city itself remained large, and … economically advanced in its differentiation into specialized trades and activities, and its attitude to wealth” (Murray 1993:186). [11] Not without reason does Cassio (1998:19) pair the Dipylon oinokhoē and the Pithekoussan cup as two of the earliest and most important attestations of metrical composition. At any rate, those who believe in Athens’ formative influence on the text of Homer during the sixth and fifth centuries do not feel an urgency to oppose to Euboia a culturally and economically thriving Athens from the early ninth to the late eighth century.
4.2 The Linguistic Argument
4.2.1 The third CL
Perhaps the best known proponent of Euboian influence in its strongest form is M. L. West. Ruijgh thinks that West has overstated his position, and himself chooses a weaker form of the theory: Homer visited Euboia for a series of performances and, while on the island, acquired for his Kunstsprache metrically convenient Euboian forms that are therefore sparsely represented in the text of the poems. [12] But what are these forms? Are they likely to be of Euboian provenience? We have to do here with the consonantal clusters *-rw-, *-lw-, *-nw-, and *-sw-, which Ionic simplified in the so-called third compensatory lengthening (henceforth CL) with the corresponding lengthening of the preceding vowel, whereas Euboian, like Attic, simplified them without any CL. Thus *xenwos had the outcome ξεῖνος in Ionic [13] and ξένος in Attic and Euboian. Principally responsible for popularizing the Euboian origin of Homeric poetry was Wathelet 1981. [14] But does his argument withstand scrutiny? We must remember that this theory attempts to substitute an early-archaic (late eighth- or early seventh-century) Euboian formative influence for the impact of Athens on the Homeric text during the archaic [15] and classical periods. The assumption, then, is that the linguistic forces responsible for the Euboian short forms (i.e. the forms that did not suffer CL) must have completed their work by the time these entered the (dictated?) text of the poems. [16] In other words, CL had already happened in Euboia, and the attendant pronunciation affected the composition, performance, and transmission of Homeric poetry there.
But this theory meets with two insurmountable objections. The first is that, as I will demonstrate below, at the early time relevant to the theory clusters with post-consonantal -ϝ- will not have been simplified, and neither alternative outcome (with or without CL) will have been in use. Therefore, the word will have been ξένϝος, not ξένος or ξεῖνος. A second chronological obstacle concerns the process responsible for the alternatives. The simplification with CL follows from a syllabification in speech that assigns the first consonant to the earlier, and the *-w- to the later, syllable: *xen.wos. The simplification that does not result in CL follows from the syllabification *xe.nwos. But I will show below that the inherited Indo-European syllabification is the one exemplified by *-n.w-, in relation to which *-.nw- is an innovation. And this innovation cannot safely be assumed to date from so early at time. If these two objections are sound, a late eighth-century or early seventh-century dictated text would have included the full forms ἔρϝιον (for εἴριον), μόνϝος (for μοῦνος), κενϝός (for κεινός), ξένϝος (for ξεῖνος), etc. [17] After the third CL took place, rhapsodes with sufficient mastery of the Attic dialect could compose verses with short forms like ξένος. [18] Alternatively, attested, metrically secure short forms may reflect selective dialectal updating. [19] Wathelet (1981:831) thinks that he has proof in ἄνοιτο (Σ 473) of the non-Attic character of these forms: although it has not suffered CL, if it were an Attic form, he asserts, one would expect *ἅνοιτο, with the spiritus asper. But matters are not this straightforward, despite the witness of ancient grammarians for this Attic dialectal practice. [20] While exhibiting a preference for the spiritus asper, examples of the spiritus lenis in Attic texts are not hard to come by. [21] Among them: Aristophanes Clouds 506 (ἀνύσας), Frogs 606 (ἀνύετον), [22] and fr. 2 K-A (ἄνυσον); Pherekrates fr. 44 K-A (ἄνυσον); Plato Sophist 230a9 (ἀνύτειν), Sophist 261b6 (ἀνύτων), Republic 486c5 (ἀνύτων), Laws 650a6 (ἀνύσειεν), etc.; Aristotle Rhetoric 1409b4 (ἀνύειν); Aiskhylos fr. 161.2 R (ἄνοις), Agamemnōn 935 (ἤνυσεν); [23] Euripides Bakkhai 1100 (ἤνυτον), Andromakhē 1132 (ἤνεν/ἤνυεν), [24] and Phoinissai 453 (ἀνύουσι). [25] The number and variety of manuscript attestations of the spiritus lenis hardly suggests scribal alterations. In my opinion, they can only be explained if this pronunciation was allowable and current in Athens during the classical period. [26] This deprives Wathelet’s argument of strength, and we are left to consider on other grounds the merit of his suggestion of a Euboian origin for the short forms.
As noted above, the third CL cannot have happened early enough for a hypothetical Homer to use the short (or long!) forms in performance. A dictated transcript would have had the unsimplified consonant clusters (e.g. *xenwos). It is well-known that the loss of post-consonantal -ϝ- that gave rise to the third CL took place after Attic reversion of [æː] to [aː] [27] following rho. [28] This explains Attic κόρη (< *κόρϝη < *κόρϝᾱ), which otherwise would have reverted to κόρα. Attic reversion, in turn, must postdate the second CL, which recreated the sound [aː] in phonological opposition to [æː]. [29] The second CL was caused by the simplification of secondary *-ns- clusters whose *-s- came from *-ky-, *-ty-, apical stop + σ, or tau before iota. [30] It is unlikely that the reversion alone [ræː] > [raː] would have sufficed to recreate the phoneme /aː/. Using the principle of economy developed by Martinet 1955, Ruipérez 1989a offered a compelling explanation for the evolution of the Greek long vowels. He pointed out that, after the first CL had created the new phonemes /eː/ and /oː/, [31] the overloading of the back series sought articulatory relief by pushing in the direction of maximum aperture (65–66). This resulted in the fronting of original /aː/ > /æː/ in all phonic environments. After /r/, the groups /reː/ /rɛː/ and /ræː/ will have realized the corresponding vowel sounds with a comparatively greater aperture (68). Thus, /æː/ after /r/ will have been a slightly different, more open sound than [æː], what using the IPA I denote [æ̞ː]. [32] This will have been an allophone of /æː/, i.e. a combinatorial variant of [æː] after /r/. It must not, however, be represented as [raː], as if [æ̞ː] = [aː]. This equation is false, for otherwise we would predict ×παράη < × paraæː < × paraː(w)æː rather than παρέᾱ < *pareæː < *paræ̞ː(w)æː. [33] That /ræː/ appears under the combinatorial variant [ræ̞ː] does not change the fact that only after the recreation of the phoneme /aː/ by the second CL could /ræː/ > /raː/. For if we assume that post-consonantal -ϝ- in -ρϝ- was absorbed before the second CL, we would have (in Attic) *korwæː > × koræ̞ː, and this form in turn would have resulted in κόρα after the recreation of /aː/. [34] To avoid reversion, we need the [æː] in *korwæː to collapse with [ɛː] before the loss of *-w-; and this can only have happened after the second CL. [35]
4.2.2 Dating the second CL
The conclusion seems inescapable that the second CL must have preceded the Attic reversion of /ræː/ to /raː/. And this reversion, in turn, must have preceded the third CL, which included the loss of -ϝ- in -νϝ-. This suggests a rather late dating for the third CL, because it is almost certain that the second had not yet happened when the Persian Māda (for Mede) was borrowed into Greek as Μῆδος [36] or when Kαρχηδών was adopted for Carthage. [37] It is true, as Gusmani 1976 demonstrates, that these loanwords cannot date the fronting [aː] > [æː], since e.g. Māda could have been borrowed as Μῆδος before the change and undergone the fronting later; or it may have been borrowed after the change, with immediate replacement of [aː] by [æː], because no Greek sound matched the original Persian [aː]. [38] But if the second CL had already taken place and [aː] was current in the Greek phonic repertory, we must assume that Μῆδος would have been the Greek reflex unless we appeal to the conjectural substratum influence of non-Greek languages spoken in Asia Minor as the source of the shift away from Persian [aː]. [39]
Kretschmer 1907 is perhaps the best known advocate of a substratum influence. [40] Prompted by Herodotos 1.146 he proposed “daß das ionische η auf karischer Aussprache des griechischen ᾱ beruhe” (32). Unfortunately, even after pathbreaking work in the last twenty years [41] we still know very little about Carian today, and Kretschmer’s theory remains largely conjectural. [42] To give it some support, Kretschmer resorted to Lycian, of which a good deal more is known, and tried to make his analysis of it applicable to Carian by pointing out the correspondence of Carian e to Lycian e (on which see immediately below). Kretschmer was not interested in dating the Ionic change [aː] > [æː], but he discussed the borrowing of Persian Māda in the course of his review of Lycian. There is not much that can be said about Kretschmer’s Carian theory, but his approach enables those who argue more generally for the indirect borrowing of Mῆδος not immediately from Persian but from one of the languages spoken in Asia Minor. [43] For this reason, I will consider here the possibility that Ionic Mῆδος might be a Lycian loanword in the light of Kretschmer’s review of the Lycian evidence. [44]
Kretschmer assumes that Lycian Mede was borrowed from Persian Māda and not from the Greek Mῆδος (an alternative that cannot be ruled out and, I would argue, is more probable [45] ). If so, its first e would correspond to the Persian ā. To evaluate the plausibility of this assumption, we need to review his argument in full. Kretschmer claims that the sign for Lycian /e/ at times represents Greek ε, a times Greek α (1907:32–33). Bilingual inscriptions seem to support this assertion. From this fact, he surmises that it must have been “ein offenes e” (32), i.e. [ɛ] between [e] and [a]. But, as I will presently show, I do not believe that his inference, though right, is relevant to the case of Mede. Hajnal (1995:11) agrees with Kretschmer’s inference that Lycian /e/ was open (though not with his further assertion that it was variously rendered by Greek ε and α): “Lykische Vokale sind (zumindest im Vergleich zu den entsprechenden griechischen Phonemen) offensichtlich recht hell (d.h. offen) gesprochen. Dies ergibt sich aus dem Umstand, dass lykische Phoneme in griechischer Transkription durch hellere Laute wiedergegeben werden.” Kretschmer is unaware of what Melchert (1992:44) describes as a “very important synchronic phonological rule of Lycian which has grave consequences for [its] historical phonology”: [46] any low vowel (e or a) assimilates in terms of backness to the vowel of the following syllable. This means that, ordinarily, two successive syllables with a or e in the first only exhibit the sequences e–i, a–u, a–a, and e–e. This Lycian rule largely explains the vacillating choice between Greek ε and α to represent Lycian /e/.
Thus, Kretschmer’s examples fail to prove that Lycian /e/ might correspond to Greek (and also Persian) [aː] (and, therefore, that Lycian Mede may be reasonably expected for Persian Māda). For example, Greek Σιδάριος > Lycian Siderija (TL 1.117) [47] cannot prove that Greek α > Lycian /e/, for the i in -ija would umlaut a hypothetically preceding Lycian a to e. Hajnal (1995:11) correctly ascribes the sound [ɛ] to Lycian /e/ on the basis of Greek renderings of Lycian vowels (where obviously the umlaut rule does not apply). He cites Πυβιάληι for (dat.) Pubieleje (TL 117); and Πυριμάτιος for (gen.) Purihimetehe (TL 6.1–2). Examples like Eρμενηννις < Erm̃menẽni, where Lycian e is now rendered by ε, further support the intermediate character of Lycian /e/. And note the mixed outcome in Esedeplẽmi > Aσ̣ε̣δ̣επλεμ̣ι̣ς. [48] But even if we grant that, as regards its openness, Lycian /e/ stood between Greek /e/ and /a/, why should we assume its sound to be [ɛ] rather than [æ]? In fact, [æ] is preferable because it is implausible that Greek speakers, whose archaic alphabetic scripts often used the same grapheme for [e], [eː], and [ɛː], but never for [a]/[aː] and any of the previous three ([e], [eː], or [ɛː]) should have vacillated between ε/[e] and α/[a] to render [ɛ]. (In other words, hearing the sound [ɛ], which had no place in their phonology, they would have readily associated it with [ɛː], which did. And /ɛː/ was unquestionably an ‘e-sound’ phoneme.) On the other hand, faced with a sound [æ] that could not be readily made to correspond to a current phoneme, [49] it is reasonable to assume that Greek speakers would have rendered it by the phoneme that most nearly approximated [æ]. But the choice of this optimal phonemic match was not obvious. Since [æ] was truly intermediate between α and ε, some vacillation between them is intelligible.
Any wavering between ε and α to render Lycian e is less conceivable if this Lycian phoneme was realized as [æ] and Greek speakers still had a phoneme whose sound was [æː]. Thus, under the transitional circumstances of a fronted *æː < *aː, if speakers of Ionic Greek had initially used the same grapheme for */aː/ and /a/ and where still using it for /æː/ < */aː/, we would expect them to render Lycian e reliably as α. [50] All of this, however, is mere speculation, for the language community that eventually became what we today call Ionic speakers cannot have practiced any writing before the first CL. [51] And where the use of a common grapheme for /a/ and */aː/ was not an established practice, one cannot assume that later /æː/ and /a/ would have been represented by one and the same script sign. A distinction was made at Naxos (with Amorgos and Keos), [52] where the merger of /æː/ and /ɛː/ happens rather late: the epsilon-sign was used for /e/ and original /ɛː/, the eta-sign for /æː/ < */aː/, and the alpha-sign for /a/ and /aː/. [53] It is possible that, absent the pressure of a recreated /aː/, the same grapheme might have been used for /æː/ and /a/, but this, too, is speculation. Elsewhere among the speakers of Attic-Ionic, the same grapheme was used from the earliest times both where we can reconstruct original */aː/ and where words exhibit original */ɛː/. Most assume that this indicates a complete, early merger, although some scholars have dissented. Any consideration of transitional forms that may have evidenced /æː/ is unnecessary, however, because, as Bryce (1986:45) notes, “[t]he earliest clearly attested evidence for the Lycian script dates to c. 500 B.C., the date assigned to the fragment of an Olpe bearing the name Pinike in Lycian characters (N 313a)” (his emphasis). [54] He further writes that “the period covered by the datable epichoric inscriptions extended from the last decades of the 5th century down to the last decades of the 4th century” (1986:50). [55] This suggests that the Lycians practiced the inscribing on stone only many decades after the first surviving attestation of their script, and that otherwise undatable inscriptions are likely to belong to this later period. Therefore, we must firmly reckon with Ionians whose phonemic repertory had no /æː/, only /ɛː/ and /aː/.
I have spent a fair amount of time on Greek renderings of Lycian to establish the open character of Lycian e and to elucidate some of the complexities that must be puzzled out before the written form of a loanword can be made to speak to the phonological issues of interest. But the argument at hand is the alleged borrowing of Persian Māda into Lycian Mede, and the further conjecture that this Mede may have been borrowed into Ionian Greek as Mῆδος. Māda > Mede is puzzling in light of the numerous examples of Persian names with long or short a that were borrowed into Lycian with Lycian /a/: Arssãma for R̥šāma, Arttum̃para for *R̥tambara, Humrxxa for *Humarga, Kizzaprñna and Zisaprñna for *Čiçafarnā, [56] Miθrapata and Mizrppata for *Miθrapāta, Wat[aprdd]ata for *Vātafradāta, Widrñna for Vidr̥na, and Wizttasppa for Vištāspa. [57] Note, moreover, the obviously relevant Parz(z)a (=‘Persian’) for Pārsa. [58] An alleged Mede < Māda is baffling on the assumption that Lycian /e/ corresponds to [æ], but it is utterly implausible if /e/ was [ɛ] instead. [59] The puzzle vanishes, however, if the Lycians borrowed Mede from the Ionians. Schmitt (1982:375) rejects this theory with the remark, “noch verschiedene Fälle von e statt erwartetem a!” But the only incontrovertible case seems to be Ertaxssiraza, [60] and this otherwise inexplicable exception must be balanced against all the other cases that reliably borrow Persian /a/ and /aː/ as Lycian /a/. The proposal of a Greek-to-Lycian loan also dispenses with the perplexing length of Mη-: why should the Greeks not have borrowed it simply as ×Mέδοι? A review of the few unquestionable Lycian-to-Greek loans establishes this expectation. The only exceptions are the occasional long suffixes inevitably used by Greek speakers to incorporate Lycian loanwords into their own morphology (e.g. the masc. ending -της in Πυριβάτης); and Zrppedun- (in a Milyan context), which, if we accept its equivalence with Σαρπηδών, is arguably a prosodic remodeling of the unmetrical ×Σαρπεδών. [61]
There is yet another, even greater obstacle to reading in the Lycian inscriptions a word Mede equivalent to our Mede (< Pers. Māda). Regarding the superficially promising Mede in TL 37.3–4, Melchert writes: “[It] is clearly a personal name. It could be the ethnicon used as a personal name, but since this text doesn’t even give a patronymic, there is no particular reason to assume that this person has such a name, and the personal name Medemudi (TL 110.1) leaves open the possibility that Mede is a perfectly good Lycian name (etymology unknown)” (his emphasis). [62] As to the ablative/instrumental medezedi in TL 44a.37, Anthony Keen suggested that it might be an adjective ‘Median’ with the usual Lycian suffix -ze/i- for ethnica. [63] Melchert writes: “Possible, but the context, as usual in the Xanthos Stele, is far from clear”; and, in any case, the umlaut rule is relevant “and the underlying base could easily have been *mada-.” [64] As to medese in TL 29.7, it is best to quote Melchert again: “We cannot parse this text with any confidence, but the one thing that is clear is that, despite my and others’ attempts to segment the -se as the Lycian conjunction ‘and’, this will not work! Lycian se ‘and’ is always proclitic, never enclitic. So [that] even if medese is also some kind of ethnicon modifying Arttum̃para (though the case of the resulting phrase would remain unclear), here too the umlaut rule means that we cannot determine the Lycian vocalism of the base. This could still be *mada-” (his emphasis). [65] With Melchert, then, we must conclude that “the true Lycian representation of Mede is unknown and that claims of its serving as an intermediary for the Greek, while possible, have no positive foundation” (per litteras).
If I am right that Lycian borrowed Mede from Ionian Mῆδος, it does not matter whether they so did before or after the second CL, so long as the Persian Māda had entered Ionic before that terminus. [66] And the Einreihung of masc. Greek -ος as Lycian -e is possible as the example of Λύσανδρος > Lusñtre shows. [67] This substratum theory, then, has many problems, and it raises as many questions as it seeks to answer. Fortunately, we can dispose of it, for, as Ruipérez (1989a:71) notes, his explanation renders superfluous “l’hypothèse qui prétendait expliquer ce changement par l’action du substrat, (ce qui revient, en dernière analyse, à substituer une inconnue à une autre inconnue).”
On the other hand, no recourse to a substratum is possible for Kαρχηδών, which must have been borrowed from the West into Ionic from a *Kαρχᾱδών current in a Sicilian Doric dialectal milieu. [68] Scherer 1975 has questioned the validity of appealing to this word, on the grounds that it may reflect a reformation with a productive -ηδων. [69] While this alternative cannot be excluded, the independent existence of -ηδων as a productive suffix for foreign names is poorly supported by the data and its alleged application to the Phoenician loanword lacks motivation. Far from clear is what precisely the Ionians are supposed to have reformed. To speak to the alleged suffix first, the analysis of Greek noun formation shows that the productive ending was -δων, not -ηδων (Chantraine 1933:360–362 §293). If from a morphological standpoint the suffix -δων appears as -ηδων (and -εδων), this is because “l’η se retrouve généralement dans le système verbal … ou dans d’autres formes nominales.” [70] The argument is not transferable to Qart-ḥadaštī, which commends the attested Carthada [71] and the suggested *Karthādōn [72] or *Karkhādōn. [73] It is also possible that the reformation of -da into -δων reflects the incorporation of Qart-ḥada- into Greek as a noun derivative with the suffix -ων. Chantraine (1933:160 §119) notes that words with the suffix -ων/-ονος do not share the common semantic outlines of a focused group of derivatives. The majority of these nouns are old Indo-European survivals. This underlines its productive nature at an early stage. It is therefore not unreasonable to hold to an early reformation of the Phoenician for Carthage that resorted to this suffix. There is also some evidence of later derivatives. [74] Scherer (1975:143) adds the proposal that Kαλχᾱδών (Chalcedon) in its Ionian form (Kαλχηδών) might have served as the model for Kαρχηδών. [75] But the Megarian foundation of Chalcedon is given as 676 or 685 (cf. CAH 3.32 160). Carthage must have come into Greek long before Chalcedon could have influenced the form of its adoption. Moreover, it is odd to suggest that Kαλχᾱδών in its Ionic version would have inspired the Umformung (with an alleged suffix -ηδων) of a Phoenician name mediated by a Doric dialectal milieu, rather than its Einreihung with the Doric suffix -ᾱδων which more readily fitted its original form. [76] Even the existence of an Ionic Kαλχηδών for the Megarian Kαλχᾱδών argues for the Ionians’ adoption of this name too at a time when /æː/ had not merged with /ɛː/. [77] Critics for whom the change from ᾱ to η in Carthage was not driven by an early-date ökonomische Lautsubstitution or Einreihung, cannot explain why the name would not have been adopted, even by the Ionians, as Kαρχᾱδών. [78] The considerations above render the attempts to neutralize the evidence of Kαρχηδών unconvincing.
One final attempt to blunt the implications of these loanwords must be mentioned. Gusmani (1976:82) introduces the notion of “analogische Lautsubstitution”: foreign /aː/ would have been automatically replaced by /ɛː/ because borrowing happened in a linguistic milieu that had embraced this correspondence as an established practice. [79] The only example he adduces (from Schwyzer GG I.187n1) [80] is Mιθρήνης in Diodoros of Sicily, from *Mιθρᾶνης < Miþrā. But Fischer’s Teubner edition of Diodoros prints Mιθρίνῃ at 17.64.6 [81] and Mιθρίνους at 17.21.7. [82] Schmitt (1978:401) s.v. MIΘΡ-INA remarks concerning the variants Mιθρίνης/Mιθρήνης: “Mag [Mιθρίνης] auch ‘nur’ eine Handschriftenvariante bei Arrian und Diodor sein, so findet sie sich bei Arrian 3, 16, 5 aber in dem Archetypos der erhaltenen Handschriften und bei Diodor in dem einzig zuverlässigen Überlieferungszweig.” [83] Schmitt also directs pointed criticism at Schwyzer and Scherer for accepting Mιθρήνης. [84]
4.2.3 Dipylon oinokhoē
Arguments like Bartoněk’s, which turn on a written text of Homer as evidence to prove that the second CL had already happened, are without merit because they beg the question. [85] We must conclude that the third CL can hardly have happened as early as the proponents of the Euboian theory assume. Crespo (1999:166) offers a terminus post quem of 800 BC (the earliest possible time for the loanword Kαρχηδών), but this is not a good signpost since Mῆδοι brings its down to the turn of the seventh century BC. This date would seem, however, to bring me in conflict with the terminus ante quem that Crespo (1999:170–171) offers: the Dipylon oinokhoē, dated to the second half of the eighth century. Jeffery’s date, ca. 725, is a prudent compromise of the various proposals (Jeffery 1990:76 no. 1; cf. CEG 1.239 no. 432). Crespo’s rationale is that the word already shows the vowel contraction [æː(h)ɔː], i.e. ὀρχεστο̄ν < *ὀρχεστεων < *ὀρχεστǣων < *ὀρχεστᾱων; to which he adds, without comment, that vowels that come from [ɛːha] > [eæː] had probably contracted by the same terminus. This does not necessarily challenge my chronology, for the contraction [ɛːha] > [eæː] > [æː] can be placed before the separation of Attic from Ionic, as Peters (1980:303) does, and hence well before the second CL (see no. 7 in Peters’s chronology, printed for convenience immediately below). But Peters also makes the contraction [eɔː] > [ɔː] contemporaneous with [eā̆] > [æː]. Since after rho the latter outcome did not revert (cf., for example, Attic πλήρη < *plēre(s)a), its corresponding contraction must postdate Attic reversion, which in turn postdates the second CL. The combined effect of Crespo’s and Peters’s chronologies—Crespo cites and, to some extent, depends on Peters—commends the view that the second CL must have happened already by the date of the Dipylon oinokhoē. In other words, making the contractions /e/ + /ā̆/ and /e/ + /ɔː/ concurrent would place the second CL before 725 BC.
The assumed concurrence, however, is hardly necessary, and since it conflicts with the dates I established above other existing alternatives must be considered preferable. [86] In fact, as I will now show, contractions like that of ὀρχεστο̄ν must have been current in ordinary speech even before Attic grew dialectally distinct from Ionic. This inference follows from adding to the testimony of the Dipylon oinokhoē the implications of the form γεννῆται and instances of contraction in Homeric poetry, where especially conducive phonic contexts favored ordinary linguistic practice over the notoriously conservative tendency of traditional epic.
In dating the second CL Crespo seems to follow a sequence like the one proposed by Peters (1980:303), which I reproduce here for convenience:
Common to Attic-Ionic [87]
1. h → ø/Vi ___Vi
2. EiːEi → EiEiː
3. /Vi / + /Vi / → /Viː/
4. [aː] → [æː]
5. [æː] → [ɛː]/___Cæː
6. QM
7. /e/ + /æː/ → /æː/, /e/ + /ɛː/ → /ɛː/
2. EiːEi → EiEiː
3. /Vi / + /Vi / → /Viː/
4. [aː] → [æː]
5. [æː] → [ɛː]/___Cæː
6. QM
7. /e/ + /æː/ → /æː/, /e/ + /ɛː/ → /ɛː/
Attic
8. [æː] → [aː] /r___
9. /e/ + ⎡/o/, /oː/ ⎤ →⎡ /oː/ ⎤, /e/ + ⎡ /a/ ⎤→ /æː/
⎣ /ɔː/ ⎦ ⎣ /ɔː/ ⎦ ⎣ /aː/ ⎦
10. w → ø/V___V
11. EiːEi → EiEiː (reordered) [88]
12. /Vi / + /Vi / → /Viː / (reordered)
13. QM (reordered)
14. [æː] → [aː]/{i, e}___
15. Contraction of EiEj (Ei ≠ e)
9. /e/ + ⎡/o/, /oː/ ⎤ →⎡ /oː/ ⎤, /e/ + ⎡ /a/ ⎤→ /æː/
⎣ /ɔː/ ⎦ ⎣ /ɔː/ ⎦ ⎣ /aː/ ⎦
10. w → ø/V___V
11. EiːEi → EiEiː (reordered) [88]
12. /Vi / + /Vi / → /Viː / (reordered)
13. QM (reordered)
14. [æː] → [aː]/{i, e}___
15. Contraction of EiEj (Ei ≠ e)
To these, Peters also appends: [89]
16. h → ø/Vi ___Vj after 1 and before 6;
17. Shortening of long diphthong, Eː → E/___yyVː, after 10 but no longer after 15
18. Osthoff, still after 9 but no longer after 12
17. Shortening of long diphthong, Eː → E/___yyVː, after 10 but no longer after 15
18. Osthoff, still after 9 but no longer after 12
The motivation to place the contraction of /e/ with /ɔː/ in a sequence prompted by a study of Attic reversion was to explain uncontracted forms like βασιλέως. If one must account for such forms not through analogical processes but through sound changes, one would have to place contraction of /e/ and /ɔː/ before the loss of intervocalic -w- (since Attic εο/εω from *εϝο/εϝω did not contract). [90] And if we believe that the contractions of all remaining heterovocalic sequences starting with /e/ were roughly contemporaneous, we would have to assign /e/ + /ɔː/ and /e/ + /a/ to the same stage. This accounts for 9 and for Crespo’s choice of the Dipylon jug’s ὀρχεστο̄ν as the terminus ante quem for the second CL, which must be placed between 7 and 8. The third CL would follow 10. Only 10 is not a helpful terminus for absolute chronology, [91] whereas, if the chronological signposts defended above are right, a post 700 BC date for 8 would provide a good upper boundary for the third CL. But notice my emphasis on the tentative nature of both conditions. Neither is necessary. In fact, the masc. gen. pl. ᾱ-stem contraction in evidence in the Dipylon ὀρχεστο̄ν must have started late during the common Attic-Ionic period and must have been accomplished in Attic well in advance of 8 (i.e. of Attic reversion). A full demonstration of this claim requires that we reevaluate the phenomenon of Q[uantitative] M[etathesis].
Méndez Dosuna 1993 has convincingly shown that the traditional understanding of QM as the phonetic exchange of vocalic quantities cannot be sustained by typological parallels. It also fails to explain an array of data in the literary texts and inscriptions. Méndez Dosuna’s theory of synizesis with CL is superior to the alternatives and should be adopted. My abbreviated exposition of it will use without distinction vocalic sequences that developed from the loss of intervocalic -s-, -y-, and -w- (the phonetic process was the same in all cases). But the reader should bear in mind that -s- and -y- > -h- > -ø- much earlier, and they provide the contexts that gave rise to vocalic contraction (or a preliminary stage thereof) before the Ionic migration. Note also that I intend to apply this model to the masc. gen. pl. ᾱ-stems. [92] I will not consider the further question whether and how this particular case might be extended to other cases, such as *ηου in *νηοῦ > νεώ (cf. Méndez Dosuna 1993:98).
From the restricted set of vowels in contact to which QM applies, the Spanish scholar concludes that QM does not reflect a generic phonotactic restriction on vowel sequences /VːV/. Neither are there non-Greek typological parallels for the process /VːV/ > /VVː/ (1993:99). At this point Méndez Dosuna turns to the Homeric text. Although long forms like βασιλῆος predominate there, as one would expect in a very conservative medium, there are numerous instances of QM, the overwhelming majority of which are in synizesis (ε͜ω and ε͜ᾱ). Monosyllabic scansion is the ordinary outcome of first-declension masculine εω (ἀγκυλομήτεω). For the gen. pl. of first-declension feminine nouns (ἀγορέων), [93] the subjunctive (στέωμεν), and a few other categories, the scansion ⏑ – is rare. [94] Such synizesis is also attested in archaic inscriptions, of which the most famous example is the Nikandre inscription (CEG 1.221–222 no. 403 or Del 3 758). There we meet with Δεινοδίκη͜o … ἀλή̆͜ο̄ν, where the Naxian sign here transcribed as η stands for /æː/ < */aː/. [95] At this point, Méndez Dosuna asserts the central thesis that QM is not an independent phonetic process but one that follows from synizesis. In /VːV/ > /VVː/ the lengthening of the second vowel does not result from the transfer of the first vowel’s quantity but from the loss of its syllabic quality and its transformation into a glide: ηο > ε̯ω. Glide development and CL are coordinate processes. The outcome is monosyllabic, what had long been considered the metrical license of synizesis. [96] Since the prevocalic shortening *ǣω > *æω in the sequence -ο̄ν < *-εων < *-ǣων < *-ᾱων shares its phonotactic dynamics with QM, it is proper to study the contraction ὀρχεστο̄ν under that heading.
Strictly speaking, of course, since the period of time I am considering precedes Attic reversion, I have in view */-aː(h)ɔːn/ > */-æː(h)ɔːn/ > */-æ̯ɔːn/ ~ -ε̯ων (i.e. -ε͜ων) > -ῶν. The -h- corresponds to intervocalic *-s- and *-y-, but not to *-w-. That mid vowels like ‘e’ might lose syllabicity and develop into glides might seem surprising to classical philologists, who only consider -w- and -y- in the phonemic repertory. But this fact has been conclusively documented for Romanian and Spanish. Chitoran 2002 demonstrates that the Romanian glide-vowel sequence [ya] and the diphthong [ea] (i.e. monosyllabic [ea]) are phonologically different (but there is no phonological difference between [wa] and [oa], which seem to be phonetically neutralized) (220–221). [ya] and [ea] are produced differently: in both the first element is a glide, but they differ in height (210). [97] Bowen and Stockwell (1955:237) cite the following Spanish examples when uttered in rapid conversation: [e̯a] /beatitud/ (‘beatitude’), [a̯e] /maestrita/ (‘little teacher’), [98] [e̯o] /leonés/ (‘of the city of León’), [o̯e] /poetisa/ (‘poetess’), [a̯o] /ahorita/ (‘right away’), [o̯a] /toallita/ (‘little towel’). [99] Against the common presumption that loss of syllabicity implies the optimal (or prototypical) glides -y- and -w-, Méndez Dosuna remarks: “The implication is excessive. One thing is to observe that the loss of syllabicity of the e ([eV] → [e̯V]) may—and, in fact, usually does—lead to the closing of the semi-vowel ([e̯V] → [i̯V]) … . Another thing entirely is to establish a necessary relation between both processes; or (what amounts to the same thing) to fuse these two successive and independent stages into one process [eV] → [i̯V].” [100]
For a study of the constraints that allow one language to glide both high and mid vowels, and another, high vowels only, see Casali 1995. [101] Languages that are intolerant of heterosyllabicity resort to the following strategies to resolve hiatus: diphthongization, glide formation, vowel contraction, consonantal epenthesis, and elision. [102] In this case we have to do with glide formation in Ionic, followed by coalescence in Attic. Méndez Dosuna’s analysis reveals that monosyllabic (synizetic) ε̯ω is actually the expected norm, to which disyllabic ε.ω from QM is secondary and typical of cadences with rallentando (101n11). [103] At all events, QM has penetrated the conservative medium of Homeric epic only to a rather limited extent. Although current in everyday speech, in Homer both -ηο- (without QM) and primary -εω prevail over ε̯ω, whatever its source. [104] Synizesis of /eo/ in Ionic was not attended by CL because /o/, not /e/, was the vowel that suffered the partial loss of syllabicity and diphthongized, [eo̯] > [eu̯]. In the language of moraic phonology, /e/ and /o/ syllabified as a tautosyllabic nucleus and their moras combined into the bimoraic diphthong. [105] For contractions like γένους in Attic, /e/ will have developed into a glide and joined the moraless onset (-νε̯-); its stranded mora will have been filled by spreading the -ο- leftward, yielding /oː/; [106] finally, the onset will have been simplified by glide-deletion, with no effect on the quality of the -ν- (i.e. without palatalization) because of the mid-height of [e̯]. [107] A clear case of Attic synizesis is στερρός < στερε̯ός. Sihler (1995:223 §233) observes that words like γεννῆται ultimately grew from *geneā- via *γενι̯ᾱ- < *γενε̯ᾱ-. [108] This proves that the development in question (loss of syllabicity of ε and closure into ι̯) must have predated not only Attic reversion but also rule 7 in Peters’s sequence above. This crucial observation supports the early date of the contraction ὀρχεστο̄ν. If so, the Dipylon inscription does not contradict my chronology, for in Attic /eo/ and /eɔː/ coalesced before Attic reversion.
Homeric instances of -ω- < -αω-/-αο- support the claim that the contraction which ὀρχεστο̄ν illustrates predated the individuation of Attic. Leumann (1950:223n20) had already complained that “über die Entwicklung von (āo) ηο zu att. ω statt εω findet man keine Auskunft” in Schwyzer, Lejeune, or Chatzidakis. Thumb and Scherer (1959:255 §311.9a) remind us that the Ionic feminine gen. pl. article was τῶν (in Homer, τάων ×21, twenty of these in the first foot, τῶν ×8). The IE pronominal declension of the feminine gen. pl. article (originally a demonstrative pronoun) was *teh 2 sō̆m. For feminine and masculine ᾱ-stem nouns, the gen. pl. was *-eh 2 ō̆m. With the early loss of intervocalic sigma, however, the difference between the near identical *-āhōn and *-āōn (*-aōn, remodeled) was obliterated, and Mycenaean Greek offers -ᾱων. [109] The contracted pronominal form must have prompted the corresponding nominal contraction in favorable prosodic and phonic contexts. [110] It is easy to understand, for example, why -έης and -ίης would contract the gen. sg. to -έω and -ίω. [111] Desyllabized -ε̯- following an accented ε or ι would soon be lost or coalesce with it (as a homorganic or near homorganic glide): [112] e.g. *-éaːyo > *-éæːho > *-éæːo > *-éæ̯ɔː > /-éɔː/ for masc. ᾱ-stems. Thus we find ἐϋμμελίω to ἐϋμμελίης (Δ 47 165 Ζ 449); Bορέω to Bορέης (Ξ 395 Ψ 692 ξ 533); Ἑρμείω (Ο 214; v.l. -είαο); Aἰνείω (Ε 534; v.l. -είεω); Ἀσίω (Β 461; v.l. Ἀσίῳ). [113]
Synizesis naturally displaces the accent away from the desyllabized glide onto an adjacent element. [114] This displacement can be seen in some Homeric Greek futures. Originally desideratives in *-(h 1 )se-, [115] the corresponding form *-ese- that was regular after resonants was extended to verbs in -ίζω. After the loss of the intervocalic sigma, the first epsilon was desyllabized and the accent of the first-person singular displaced to the first mora of the omega, *-έσω > *-ε̯ῶ. [116] This made the sequence rife for contraction after iota by the same principle stated above. Hence, *-ιε̯ῶ > -ιῶ rather early, as we see in κτεριῶ (Σ 334), κομιῶ (ο 546), and ἀεικιῶ (Χ 256). [117]
I hope to have established that at a very early time, when Attic-Ionic was still a single linguistic community, the gen. pl. of ᾱ-stems had already experienced synizesis and coalescence. This must have been pervasive in ordinary speech, although it only made limited inroads into the high register of Homeric speech. But its presence there is unmistakable, and early instances of contraction where the phonic context was especially conducive are also incontrovertible. This renders plausible the proposal that ὀρχεστο̄ν reflects a contraction that predates Attic reversion. Attic developed great tolerance to contraction and commensurate intolerance to hiatus. One may reasonably expect Attic to have finished early the process that Attic-Ionic had already significantly advanced by the time of the migrations.
4.2.4 Syllabification of -nw- clusters
Having disposed of the potential objection raised by the Dipylon oinokhoē, [118] I can return to the date of the third CL and advance another reason why this must be placed rather late—in all probability later than the time of Euboian cultural efflorescence during which, according to some scholars, the Homeric poems were committed to writing. The outcome of the third CL assumes for a word like ξένϝος the West Ionic syllabification ξέ.νϝος and the East Ionic ξέν.ϝος. [119] Indeed, Wetzels (1986:310) follows Steriade (1982:120) and shows with CV phonology that the CL in ξεῖνος cannot be explained by reconstructing a stage in which -w- was brought into coda-position by metathesis. [120] Deletion of -w- in an onset -.νϝ- would invoke the “empty node convention” of Ingria (1980:471). [121] Because the resulting West Ionic syllabification ξέ.νος is well formed, there would have been no resyllabification, and hence no impact on the first syllable /kse-/ (i.e. no CL). Matters are different for East Ionic syllabification ξέν.ϝος. Here, deletion of onset -.ϝ- would result in ×ξέν.ος, which is not well formed; the corresponding resyllabification would produce what Hayes named the “double flop”: [122] the -ν.- would shift into the melodic slot vacated by the -.ϝ-, and the new empty position is filled in turn by spreading the preceding vowel segment /e/ > /eː/.
Steriade (1982:125) argues that Attic has preserved an archaic version of Common Greek syllabification to which the Ionic syllabification is an innovation. [123] It is true that at times Attic is demonstrably more archaic than Ionic (e.g. in retaining the dual). [124] But if Steriade is right, it is puzzling that epic formulas overwhelmingly reflect the supposed innovative syllabification, and that in turn the allegedly archaic one has not left a deeper mark on Homeric prosody. [125] The proposal that the scansion of Vedic Sanskrit and Homeric Greek reflects innovative syllabification is untenable. [126] Poetic prosody, especially at an advanced stage of development, is doubtless conventional and may depart markedly from ordinary speech. Yet the rules of versification are not set by committee; [127] they develop as stylizations of rhythmic patterns of language and music in performance that gain over time the approval of the audience and gradually acquire traditional authority. One must therefore regard the syllabification entailed by the archaic hexameter as a stylization of the prosody of natural language dominant during the time of its forging. [128] A careful reading of Clements 1990 reveals that syllabification according to the sonority hierarchy is not only dependent on a possibly historically contingent scale of relative sonority (cf., for example, the different relative sonority of -w- and -r- in Attic and Ionic), [129] but also on language specific rules that Miller (1994:8–9) calls “P[articular] G[rammar] stipulations,” “PG exceptions,” and “PG restrictions.” [130] It is well and good to call the archaic Greek syllabification of Homeric epic typologically “marked” [131] so long as one does not thereby suggest that an artificial metrical convention was operative ab origine and that the syllabification of the archaic hexameter never matched the natural speech rhythms of proto-Greek. [132]
Pulgram and his followers might argue that the poetic scansion λεπ.τός was conventional and did not reflect the λε.πτός of everyday speech. [133] But the lengthening in prose of the last vowel of σοφός to form σοφώτερος and σοφώτατος, when contrasted with the unchanged κουφότερος/κουφότατος, supports the view (pace Guion 1996:76–77) that prose and poetry shared one notion of light and heavy syllables. [134] Similarly, the accent of trisyllabic neuters in -ιον falls on the stem if the syllable is light but on the ending if heavy: θύριον but πυκτίον and χρῡσίον. [135] Original IE syllabification supports λεπ.τός. Even in Attic, the forms κενότερος and στενότερος point to original forms κενϝός and στενϝός and strongly suggest that IE practice was followed by original Attic syllabification too. [136] What, then, are we to make of Horrocks’s claim that Mycenaean ke-se-ni-wi-jo points to /kse.nwios/? [137]
The first thing to be said about ke-se-ni-wi-jo is that it has two spelling variants, ke-se-nu-wi-jo and ke-se-ne-wi-ja. To call the last “exceptional” (so Morpurgo Davies 1987:98, followed by Miller 1994:21) is to beg the question whether the small number of surviving attestations may be fairly thought to render very unlikely the Bronze-Age syllabification /ksen.wios/ (for this syllabification we should expect ×ke-se-wi-jo). [138] At any rate, we should note the agreement of Viredaz (1983:133–134) and Morpurgo Davies (1987:92) that Mycenaean reflected original IE syllabification, and that where its spelling seems to show otherwise, an explanation other than an innovative syllabification must be found. Morpurgo Davies’s proposal is that Mycenaean scribes exhibit a stance vis-à-vis the graphic representation of consonantal clusters that is strikingly similar to the one espoused by late grammarians. Indeed, she shows that a Linear B scribe could make a compound explicit by separating the words with an internal divider (this is not done with derivational suffixes) or by marking a graphic hiatus; and that he could treat as a unit a sequence of accented and unaccented words (proclitic or enclitic). This implies a working notion of the word as a unit of utterance (97). From this, she goes on to claim a link between the writing practice of Linear B, Cyprian, and alphabetic Greek. The syllabification commended by late grammarians was neither entirely based upon, nor wholly divorced from, the real syllabic division of the spoken language. It reflected both the phonetic reality and the psychological perception of syllabification (101). For Herodian, if a cluster started a word it could also start a word-internal syllable (e.g. κτίζω vs. τέ.κτων). Allowable word-final consonants may also have influenced his analysis. For Morpurgo Davies, perceptions of syllabic division in Mycenaean Greek were also affected by a combination of real syllabic division and the model provided by word beginnings and ends. To return to /ksenwios/: the -nu-wi- of ke-se-nu-wi-jo would naturally follow from the quick utterance of -nwV-, so that we cannot infer solely from it the existence of words that began with nw-. But the attestation of a complex sign ‘nwa’ [139] gives it some support. Thus, we can hardly infer from ke-se-ni-wi-jo the syllabification /kse.nwios/. [140]
If Mycenaean, then, does not support the claim that Attic preserved the archaic syllabification ξέ.νϝος, and neither do the forms στενότερος and κενότερος, we must accept the fact that the further we go back in time, the more we are unlikely to meet with the innovative syllabification responsible for the lack of CL in West Ionic after the loss of postconsonantal -w-. Cassio (1998:17–18) is right to affirm that at the time of the migration all Attic-Ionic speakers must have syllabified ξέν.ϝος. It is surely out of the question that a tenth-century rhapsode in Lefkandi could have syllabified ξέ.νϝιος, nor can Attic correption be explained by reference to a Euboian phase during the Geometric Period. The cumulative import of the linguistic arguments reviewed here controverts the now popular view that what scholars before Wathelet had interpreted as Atticisms are best explained as Euboian forms preserved by a hypothetical early-archaic written text of Homer. [141]
Although the main thrust of my criticism has been aimed at the theory of ninth- or eighth-century dictated scripts of the Iliad and the Odyssey, the same arguments challenge the soundness of West’s belief in mid seventh-century texts. Indeed, if, as West (1988:166) claims, “Euboea [was] the area in which the epic language acquired its definitive and normative form,” starting with possible performances for the “anonymous king … at Lefkandi” in the tenth century BC, [142] any instance of a word that would not have suffered CL in West Ionic but is in fact lengthened in our texts is eo ipso problematic. For, if West is right, much of the formulaic language would have come into being while the syllabification -C.w- was Panionic. In all words with such clusters, the syllable before the -w- would have scanned long. But if the definitive and normative shape of the texts came into being during 300–350 years of Euboian performances, rhapsodes would have had to adapt the formulaic language to the innovative syllabification as it developed. Hence, most (if not all) of the attested compensatorily lengthened words should have been reduced to the short Euboian forms. In other words, we should not expect ἶσος, ξεῖνος, κοῦρος, κᾱλός, etc., but ἴσος, ξένος, κόρος, κᾰλός, and so on. Not the latter West Ionic forms, but the former East Ionic, would be in need of justification. This must be so, unless we assume such a strong conservative tendency in the system and so little linguistic renovation that, after the loss of postconsonantal -w-, CL was artificially applied under prosodic compulsion against the vernacular West Ionic in order to save the existing formulas. Had this in fact been the case, one can hardly call the centuries of the alleged Euboian phase “definitive and normative”; one might better consider the Euboian bards as a minor, unimportant link in the transmission of texts that might as well, so far as the linguistic and formulaic shape goes, have remained in the hands of East Ionian bards.
None of these difficulties would afflict the Athenian transmission: since the definitive Athenian period in the evolutionary model covered the sixth, fifth, and fourth centuries BC, the poetic language would have come under the influence of the Attic audience and dialect as a fully formed, largely Ionic formulaic system, including East Ionic CL. I am not denying, of course, that performances of ‘Homeric’ traditions of epic took place in geometric and early-archaic Athens (as they probably did in Euboia and elsewhere). Nor do I wish to imply that the ever-changing poetic language failed to be affected at all by eighth- and seventh-century Attic (although the effect, if it became part of the paradosis, would probably be undetectable now, covered over by later Attic influence). More will be said below against a definitive archaic Euboian phase of Homeric epic in connection with the famous Ischian skyphos.
Footnotes
[ back ] 1. Cf. Cassio 1998:16–17.
[ back ] 2. See Janko 1982:200 for a convenient diagram, with his reasoning at 228–231. For West’s dating of Hesiod’s floruit to ca. 700 BC, see West 1966:45. West thinks Hesiod to be older than Homer (cf. West 1995:208–209).
[ back ] 3. But see below, §5.2.
[ back ] 4. See West 1988:165–172; Powell 1991 passim; Cassio 1991–1993:200 (but cf. Cassio 1994:66 and n. 78); and Ruijgh 1995:47–50. Cf. Janko 1992:18n33. Cassio 1998 provides a commendable reexamination of this question that broadly supports my argument here.
[ back ] 5. One must remember that these Vita traditions reflect in biographic form the competition for Panhellenic acceptance by local versions of the Homeric and cyclic poems. By attempting to tie Homer’s birth or an episode of his life, including his activity as a composer and performer, to a particular individual or location (city, island, etc.), various parts of Greece sought to appropriate for their local versions the symbolic capital of Homer as the Panhellenic poet per excellence. This involved not only authorizing—as to ‘authorship’ and ‘authority’—epichoric multiforms of the Iliad and the Odyssey, but also articulating a relationship between cyclic (and other) poetry and the Iliad and the Odyssey. For lives of poets generally, see Nagy 1999a:279–288, 296–316 and Nagy 1990c:75–76. For the lives of Homer, see Nagy 2010a:31–47 and Graziosi 2002 passim (see “General Index” under “Lives of Homer”).
[ back ] 6. Homer’s descent from Mousaios (so Vita 6.9 and Gorgias apud Proklos’ Khrēstomatheia, p. 100 line 5 in Allen’s edition) arguably regards Athens.
[ back ] 7. Euboia’s place in the poems is no more conspicuous than Athens’. In particular, its ten-line section in the Catalogue of Ships is comparable to the immediately following eleven lines that regard Athens. Both locations appear in the initial group headed by the Boiotians, and, as far as strength of numbers, Athens’ fifty ships compare favorably with Euboia’s forty. (On Athens’ and Menestheus’ place in the Catalogue, see Kirk 1985:179–180 and 206–207.) As Cassio (1998:14) remarks, the Euboian’s leader Elephenor is killed early in action at Δ 463–471 and is otherwise absent from the poem (cf. Kirk 1985:387). Athens’ Menestheus, also a relatively minor figure, appears considerably more often, with interventions at Δ 327 Μ 331 373 Ν 195 690 Ο 331 (with the Ἀθηναῖοι at Δ 328 Ν 196 689 Ο 337). Athens’ historical (and celebrated) connection with Salamis, the birthplace of the prominent hero Ajax, who immediately follows the city in the Catalogue, might well have sufficed to obviate the need to gratify an Athenian audience with a greater role for Menestheus. Instead, Ajax could be embraced as Athens’ own preeminent Iliadic hero (cf. Shapiro 1989:154–157). And whatever may be said for Euboia as a point of reference in the Odyssey (in my opinion, far too much has been made of it; cf. Cassio 1998:14), one cannot read Cook 1995:128–170 without feeling the greater potential debt of the poem to Athens.
[ back ] 8. E.g. Wathelet 1981:831–833 and Ruijgh 1995:47.
[ back ] 9. All in Murray 1993:185.
[ back ] 10. Morris (2000:241–242) observes that grave goods began to decline in Athens before 800, but in the Attic countryside the reverse was true.
[ back ] 11. Thomas and Conant (1999:84) offer a traditional view of Athenian decline: “[B]y the eighth century its preeminence had disappeared.” But they too note that Attic graves surpassed the Athenian in their richness and (citing Whitley) that there may have been an affirmation of local interests vis-à-vis those of the urban center. Their inference of decline from the poverty of Athenian burial goods is challenged by Morris (2000:287–306), who sees in Athens’ return to cremation for adults during the seventh century a deliberate elite reaction to the “middling” ideology of citizenship that was in the ascendant elsewhere (288–290). Like Athens’ seeming lack of involvement in the Panhellenic games, hoplite warfare, and colonization, the poverty of its burials responded to the attempt by the Athenian elite to “cut themselves off from the growth of polis institutions” and look back “to the simpler religious practices and social relationships of the Dark Age” (294, 302). If Morris is right, one must wonder whether this reaction presupposes a cultural climate that would fail to support the performance of epic poetry or, rather, the precise opposite.
[ back ] 12. Ruijgh 1995:15–16 and 47–48.
[ back ] 13. The -ει- stands for the lengthened -ε-, which gave rise to a long, closed /eː/ that contrasts with the inherited, more open /εː/ represented by the grapheme -η- (or the corresponding sign in an archaic Ionic script).
[ back ] 14. Previously in Wathelet 1966:154–169 and Wathelet 1970:153–157, and ascribed there tentatively to Aiolic (following Schwyzer GG I.228, with bibliography) or else to West Ionic or Attic. See now his recent summation in Wathelet 2008. Wathelet’s ascription to West Ionic (Euboian) of forms previously considered Attic was well received by West 1988:166 (“epic-Ionic … is not East Ionic, it is Central or West Ionic”). Wathelet 1970:155n150 realizes the chronological difficulties of making archaic Euboian the source of these traits: “La date du troisième allongement en Ionie (début du second quart du premier millénaire) semble trop récente, si l’on situe Homère vers 800.” He also thought desirable to find a common explanation both for these forms without CL and for instances of the so-called correptio attica (1970:155n150). Wackernagel 1970:120–122 had already noted the Aiolic and Euboian options. Of the Aiolic he pointedly noted: “was geht das Lesbische der Sappho den Homer an? Das Äolische, das in Homers Sprache steckt, ist viel älter als jenes.” He rejected in turn the Euboic source for reasons of economy: since Attic was unquestionably responsible for some forms, ceteris paribus it also seemed reasonable to him to ascribe these other to Attic. (Both statements are in Wackernagel 1970:121.)
[ back ] 15. Starting at least as early as the sixth century BC.
[ back ] 16. Alternatively, ordinary Euboian speech during Homer’s visits to the island must have exhibited the results of this process, and *xenwos must have been pronounced ξένος.
[ back ] 17. For these and similar forms, see Chantraine GH I.159–162. He admits, for example, that ξένϝος and καλϝός may have stood in old formulas, but thinks metrical considerations the primary cause for the existence of short and long doublets (ξένος vs. ξεῖνος).
[ back ] 18. For a convenient list of attested short forms, see Wathelet 1970:154.
[ back ] 19. This is perhaps the case of ἄ̆νοιτο (Σ 473), from ἄ̄νω < * ἄ̆νϝω (a thematic form of ἄνυμι). LIV 2 532–533 considers ἄνυμι a reformed *hanāmi (root *senh 2 –). As Schulze (1967:107–108) observed, however, ἄνοιτο may hide a metrically equivalent, older ἄνῡτο < *ἄνυιτο.
[ back ] 20. Cf. Dindorf 1870:46 and Schanz 1881:v.
[ back ] 21. A strong tendency towards uniform spelling among editors is responsible for printing the asper for the lenis even in the absence of manuscript support.
[ back ] 22. Cf. Frogs 649 (ἀνύσεις). The Budé edition without ms. support silently adds the spiritus asper to these forms.
[ back ] 23. West prints ἥνυσεν without ms. support.
[ back ] 24. Corrected by Diggle to ἤνον, after Borthwick.
[ back ] 25. Diggle prints ἀνύτουσιν. These loci are simply illustrative of the extent of the practice. This is by no means a comprehensive list.
[ back ] 26. At the very least, its pervasive use in tragic poetry could have influenced rhapsodes, who, cognizant that Ionic psilosis was the Homeric norm, might readily adopt it.
[ back ] 27. Both, realizations of original /aː/.
[ back ] 28. See, for example, Schwyzer GG I.187–189 and Sihler 1995:50–51. Sihler argues convincingly for reversion (51). For the chronology of reversion after rho vis-à-vis reversion after epsilon and iota, see Brugmann 1898. In his study of Greek words ending in -αιρα, Peters (1980:250–305) reviews Attic reversion and probes alternative relative orders of the relevant sound changes. He does not attempt an absolute chronology nor does he consider the second or third CLs.
[ back ] 29. That is, it reintroduced the phoneme /aː/.
[ back ] 30. See Crespo 1999 and Sihler 1995:217.
[ back ] 31. I.e. half-close long ‘e’ and ‘o’ that contrasted with the half-open original long ‘e’ and ‘o’, here represented as /εː/ and /ɔː/.
[ back ] 32. For basic guidance on phonetics and the I(nternational) P(honetic) A(lphabet), I recommend Catford 2001.
[ back ] 33. See Sihler 1995:51. For bibliography on this word, see DMic Ⅱ.84–85 s.v. “pa-ra-wa-jo.” See also Peters 1980:295–304.
[ back ] 34. In other words, the long vowel that after /w/ was pronounced [æː] would have been realized as a more open allophone [æ̞ː] following the loss of /w/. Only after the second CL would [ræ̞ː] > [raː] and the outcome would be reassigned to the recreated phoneme /aː/. Under this scenario, its final form would have been just the same as if we had started with × koraː before [aː] was fronted to [æː].
[ back ] 35. Confused by the fact that /æː/ must have existed as a more open combinatorial variant after /r/, some scholars have suggested that this allophone by itself might have recreated the phoneme /aː/ (cf. Bartoněk 1966:103, López Eire 1970:18, and Crespo 1999:182). The confusion is twofold. /æː/ after /r/ cannot have been pronounced [aː]. Otherwise, there would have been no fronting in this environment nor any need for reversion. But we have seen that only reversion predicts the correct Attic παρέα. And, as Ruipérez (1989a:67) remarks, one can hardly admit that in /raː/ the vowel was always realized as [aː], since the /roː/ that arose from the first CL must have exerted the same structural pressure on /raː/ that displaced [aː] > [æː] in other phonic environments. Moreover, this combinatorial variant is not a development of ×/ræː/, that is, it is not a sound change that followed the hypothetically preexisting pronunciation ×[ræː]. /æː/ was always realized after /r/ as the allophone [æ̞ː]. This more open realization is a Greek phonetic fact that held true at all times for any vowel that followed /r/ (unless the vowel was already maximally open). The dialect of Naxos serves to illustrate this: different graphemes were used for /aː/, /æː/ < *aː, and /εː/ < *eː. (See, for example, Schmitt 1977:101, Ⅳ.15.M2, and Schwyzer Del 3 758–759 together with Jeffery 1990, pl. 55, nos. 2, 11.) There was never a ×[ræː] (i.e. a /ræː/ whose long vowel was pronounced just like the long vowel of /rwæː/) that developed into [ræ̞ː] under a sound change that then ceased, with the result that its outcome could be distinguished from a ×[ræː] that arose later from [rwæː] by the loss of /w/ (which, under the assumed circumstances, would not have experienced a subsequent opening to [ræ̞ː]). Only these false premises could produce two groups, [ræ̞ː] and [ræː], that would merge respectively with /raː/ and /rɛː/ after the second CL. Without such a scenario we are left without a sound that can be phonemically opposed to [æː] before the second CL (here, hypothetically—and falsely—/æ̞ː/ and /æː/ after /r/). This lengthening is therefore prerequisite for the recreation of the phoneme /aː/.
[ back ] 36. With a terminus post quem of the early seventh century BC (Gusmani, 1976:79).
[ back ] 37. With a terminus post quem of the eighth century (Gusmani 1976:79).
[ back ] 38. The latter is Gusmani’s “ökonomische Lautsubstitution” (Gusmani 1976:81); the former, his “Einreihung” (1976:77). Cf. Szemerényi (1968:145), who pronounces valid the argument that takes the second CL as a terminus ante quem, but turns to Πέρσης and Ξέρξης to cast doubt “whether the whole problem [of these loanwords] concerns Greek phonological history at all.” Lejeune (1972:235 §249n2) voices similar reservations, but Gusmani (1976:80) has convincingly disposed of these difficulties.
[ back ] 39. In other words, the Greeks who borrowed Māda would have articulated Persian /aː/ under the influence of some non-Greek language that fronted [aː] to [æː]. If so, this phenomenon would be independent of the sound change [aː] > [æː] that followed the first CL and affected all Attic-Ionic speakers. Hence, it could not be integrated into a chronological sequence of Greek sound changes. This subgroup of the Ionic linguistic community would have uttered Mā- as [mæː-]. As the loanword spread to the rest (presumably, after [æː] had merged with [εː]), other speakers of Attic-Ionic replaced [mæː-] with [mεː] (ökonomische Lautsubstitution). Hence Mῆδος.
[ back ] 40. In an earlier work, Kretschmer (1892:286) had not advanced a specific view of the circumstances in which the Ionians allegedly borrowed Māda as Mῆδος. He only stated that it must have involved the change of ᾱ to η. Cf. Schwyzer GG I.187. I gratefully acknowledge the generous help of H. Craig Melchert with various aspects of my analysis below.
[ back ] 41. For a comprehensive introduction, see Adiego 2007.
[ back ] 42. Schwyzer GG I.187: “Einleuchtend, aber nicht näher erweisbar ist die Annahme, der Wandel von ᾱ zu η sei vom Karischen aus zunächst ins Ostionische gekommen.” Schmitt 1982:375 (under 1.c Mede) seems to approve.
[ back ] 43. Gusmani 1976:82.
[ back ] 44. It bears emphasizing that Kretschmer does not propose this. He merely suggests the influence of Carian vocal articulation (1907:33): “[D]ie Annahme, das ionische η für ᾱ beruhe auf karischer Artikulation, [entbehrt nicht] eines gewissen Anhaltes in den Tatsachen.” Lasso de la Vega (1956:287–289) also favors the substratum theory. But he grounds his support on Kretschmer’s doubtful analysis and on Sundwall’s wrong ascription of the vocalic sound [æː] to the Carian sign Kοππα (cf. Adiego 2007:169). Concerning Carian vowels, /a/ is used for Greek α, whereas /e/ is used for Greek η (Adiego 2007:236). Thus, e.g. ada > Αδα; and, conversely, Λυσικράτης > lysikrata-. (The final a for η in lysikrata- is a Kaunian accommodation to the lack of e in their vocalic system. Kaunian was forced to render the close::open opposition of ε::η by i::a.) For e, consider Oὐλιάδης > uliade; and, conversely, mane > Mανης. Carian regularly syncopates or reduces unaccented vowels (Adiego 2007:240–241), so that e virtually always represents an accented vowel which may also have been phonetically long (if not contrastively). This fact helps to explain the consistent rendering of e by η and of o by ω. Melchert tells me (per litteras): “Even if the Carian vowels were relatively open, I know of nothing to suggest that the Carian e was as low as the Lycian. That is, I know of no instances of Carian e represented by Greek alpha.” Thus, I do not find any support in Adiego 2007 for Kretschmer’s notion of a Carian articulation.
[ back ] 45. See below, §4.2.2.
[ back ] 46. See Melchert 1992:44–45 for a full exposition of this umlaut rule.
[ back ] 47. TL stands for Kalinka 1901. These texts are also in Friedrich 1932. For an introduction to Lycian see Melchert 2004b.
[ back ] 48. Wörrle 1995:410.
[ back ] 49. That is, for Ionic Greek, after transitional [æː] had merged with [εː].
[ back ] 50. Only after the recreation of /aː/ was [æː] pushed toward [εː] and /æː/ eventually merged with /εː/. Only then would the grapheme for /εː/ have been used for any merged vowel sounds that had come from */aː/. Without the structural pressure on [æː] and the shift to [εː] that ensued, as to openness [æː] must have been felt closer to [a] than to [e].
[ back ] 51. The Greek alphabet, in all likelihood, was not even in existence.
[ back ] 52. Schmitt 1977:101.
[ back ] 53. See Jeffery 1990:304 no. 11.
[ back ] 54. Neumann assigned two graffiti to an earlier time: N 300a to 580–550 BC, and N 330b to the second half of the seventh century. But the provenance of these is Rhodian, and their Lycian identification uncertain (cf. Bryce 1986:45). A number of coins bearing the legend KVB and using Greek script could be dated as early as 520 BC.
[ back ] 55. E.g. TL 117, the Σιδάριος inscription where Lycian Pubieleje is rendered into Greek by Πυβιάληι, dates from the end of the fifth century BC.
[ back ] 56. This is the form in Schmitt 1982:380 no. 2.g; Melchert 2004a gives it as *Çiçafarnā.
[ back ] 57. All these are found in Schmitt 1982 and Melchert 2004a. Schmitt adds four Lycian names that seem to exhibit Lycian /e/ for Persian /a/ (cf. Schmitt 1982:377 and 383). Two depend on doubtful reconstructions of the Persian: Erbbina < *Arbina (?), whose E- at any rate could be due to umlaut; and Wexssere/Waxssere < *Hvaxšara (?) (Schmitt 1982:378 no. 2.c and 382 no. 2.l). Melchert tells me per litteras, “I see no compelling reason to believe [Schmitt’s] derivations.” In the other two, Ñtarijeuse/i < *Dārayauš and Erijamãna < *Ariyamanā, the apparent Lycian /e/ for Persian /a/ may be due to the umlaut rule. This leaves only Ertaxssiraza < R̥taxšaça. (Ertẽmi < Ἄρτεμις, cited in Schmitt 1982:383n65, follows readily from the umlaut rule.)
[ back ] 58. Schmitt 1982:376 no. 1.e and Melchert 2004a s.v.
[ back ] 59. Pace Hajnal 1995: “So ist lyk. <↑> /e/ eine Variante von griech. , da lyk. /e/ vermutlich als offenes [ę] gesprochen wird und somit griech. [a] sehr nahe steht” (8). Better, Melchert (per litteras): “[T]he [Lycian e] vowel must have truly been intermediate (something of the sort of an English [æ] in ‘man’.” Cf. Catford 2001:131 fig. 40 with 148 fig. 43 and the discussion in 151.
[ back ] 60. As noted above, §4.2.2 n. 57.
[ back ] 61. Cf. Kretschmer 1907:32 and Kamptz 1982:312–313.
[ back ] 62. Per litteras.
[ back ] 63. Apud Melchert 2004a:38. Already in Neumann 1969:380 §19.1b.
[ back ] 64. Per litteras.
[ back ] 65. Per litteras. For attempts to read -se as an enclitic or else segment mede-se (which can hardly be right, considering the interpunction that follows medese), cf. Keen 1998:149–150 and Schmitt 1982:375–376.
[ back ] 66. Otherwise, it would have been borrowed into Ionic as Mᾶδος, and thence into Lycian as Mad-. Lycian Med- could reflect [mæːd-] or [mεːd-].
[ back ] 67. Other twists are possible, such as borrowing Mῆδος as *Meda (cf. Exeteija < Ἑκαταῖος or Milasãñtra < Mελήσανδρος), which would subsequently umlaut to *Mada-!
[ back ] 68. Cf. Friedrich 1921:104 and Laroche 1972:88.
[ back ] 69. Cf. Gusmani 1973:45, which explains ‘Lepanto’ and ‘Scarpanto’ as modifications of the original ‘Náupaktos’ and ‘Kárpathos’ after toponyms like ‘Ofanto,’ ‘Levanto,’ and ‘Taranto.’
[ back ] 70. Chantraine 1933:362.
[ back ] 71. The Latin Carthāgō results from progressive dissimilation.
[ back ] 72. Friedrich 1921:103.
[ back ] 73. Laroche 1972:88.
[ back ] 74. Chantraine 1933: ἀρηγών, τρυγών, κατηφών (160). Cf. Schwyzer GG I.486–487.
[ back ] 75. A recourse to ἀνθρηδών, τενθρηδών, and πεμφρηδών is a shot in the dark. That these could have motivated the shape of the Greek name for Carthage is a peculiar proposal.
[ back ] 76. For the history of Kalkhedon, see Merle 1916 (for its Doric dialect see p. 9). That Karkhedon only survives in its Ionic form in the literary sources should not surprise us: Kalkhedon provides a good parallel. Only inscriptions preserve its original Doric form. But one does not expect the name Karkhādon of a non-Greek settlement to have been common in inscriptions. Cf. IG Ⅶ 2407.6; FD Ⅲ 1, 497.17; and SEG 30.1117.7, 30.1118.8 (on these two, see further 32.914), and 52.536.
[ back ] 77. Establishing this equivalence at a later time would require a transparent morphological or etymological motivation that does not seem to exist. For a parallel, consider Mῆδος, which despite its original Pers. Māda only exists in its Ionic form, even in Doric contexts. Only in Cyprus, with its own independent traditional contacts, do we meet ma-to-i = Mᾶδοι (Del 3 679.1). Cf. Forssman 1966:141.
[ back ] 78. Scherer presumably believes that the second CL had happened by the time of its borrowing, when /aː/ was a working phoneme! Under this scenario, the Ionians would have acquired the name of the settlement from speakers of Doric and they would have had no motivation to replace the Doric ᾱ by the Ionic η.
[ back ] 79. This automatic substitution would have produced what might be considered parallels to later hyper-Ionic forms.
[ back ] 80. Also in Scherer 1975:143. Gusmani 1976:80 shows that Ξέρξης is not an example of this alleged practice; nor is Πέρσαι, unless we argue that the borrowing happened after the second CL. But even Gusmani speaks of “der wohl ungefähr um die gleiche Zeit [als Meder] entlehnte Name der Perser (iran. Pārsa).” If so, it too would have undergone the same process as Māda, with the additional shortening of η > ε after /æː/ merged with /εː/, either because of the fanciful tie with Περσεύς or owing to Osthoff’s law. (Peters 1980:318, 333 makes additional suggestions that are independent from Osthoff.) Even if we assumed that it had been borrowed into Greek after the second CL, the opposition Māda::Mῆδοι might have motivated the corresponding Pārsa::*Πηρσαι (before shortening). This seems at first an example of the “analogische Lautsubstitution” I am opposing here. What makes this one motivated and a permissible theory is the close association of Medes and Persians. This would allow speakers of Ionic, who knew that [aː] in the native Māda corresponded to /εː/ in their own dialect, productively to apply this mapping to the closely related Pārsa. A fanciful relation to Perseus as their eponym could only help: he was thought to have ruled the Argolid, where original /aː/ was preserved (cf. CAH 3.32 31.). This is the kind of motivation that is missing for Māda, which would lack an applicable precedent.
[ back ] 81. This is the reading of RX; F has Mιθρήνῃ.
[ back ] 82. The reading of X. R has Mιθρίννους, and F Mιθρήνους.
[ back ] 83. Mιθρίνην is indeed the sole reading for Arrian’s Anabasis 3.16.5, which Roos emended to Mιθρήνην after 1.17.3–4 (see his apparatus ad loc.).
[ back ] 84. Schmitt 1978:401n24: “Gleichwohl geht Eduard Schwyzer … von der Form Mιθρήνης bei Diodor aus, um den Lehrsatz zu begründen, dass ‘auch in spätern Transkriptionen ion. ρη für fremdes rā’ stehe! Hierauf beruht auch Anton Scherer … .” For all the attestations, with textual notes, see Schmitt 1978:429.
[ back ] 85. Bartoněk 1966:100. Accepted by Szemerényi 1968:146–147.
[ back ] 86. Note the similar chronological disjunction between the contraction of /eaː/ and /ea/, more striking—if either of them is—than the one between /eɔː/ and /ea/ owing to the greater likeness of the sounds involved. The former (i.e. contraction of /eaː/) corresponds to no. 7 in Peters 1980:303 ([eaː] > [eæː] > [æː]), the latter (i.e. contraction of /ea/), to no. 9. As Peters 1980:274 remarks, “a priori hätte man nun jedoch eine Gleichzeitigkeit der Kontraktionen von /e/ + /ā/ und /e/ + /ă/ erwartet.” Regarding no. 9, note that [eaː] > [aː] in forms like the fem. acc. pl. χαλκᾶς < *-eaːs < *-eans; but [eaː] > [æː] in the corresponding nom. sg. (so, χαλκῆ < *-eaː).
[ back ] 87. C = [-syll] except for y. E = a e o. QM (=Quantitative metathesis) involves EiːEj → EiEjː with Eiː = /æː/, /εː/, or /eː/ and Ej = /a/ or /o/. For an explanation of these restrictions, see Méndez Dosuna 1993:99 and 109–111.
[ back ] 88. On the notion of reordering, see the bibliography in Miller 1976:140.
[ back ] 89. V = [+syll, -cons].
[ back ] 90. Crespo 1977:193.
[ back ] 91. We do not know how much time may have transpired between 9 and 10. Besides, even if we knew that 8 had been completed by 725 BC, this fact would not determine how much earlier it might have taken place. Thus, we could not rule out a third CL that had already happened by the time of a late eighth-century ‘Homer.’ But one need hardly regard Crespo’s terminus post quem of 800, not because it is false, but because it is not stringent enough. Crespo has conveniently ignored the evidence of Mῆδοι (which is ca. or post 700), perhaps to accommodate his terminus ante quem without contradiction.
[ back ] 92. On the genitive singular of masculine ᾱ-stems, see Szemerényi 1956.
[ back ] 93. Where the meter does not enforce -ᾱων. I hope to explore in a future work the origin of these alleged Aiolisms.
[ back ] 94. Cf. Chantraine GH I.64 for ᾱ-stems. Primary sequences like θεῶν are hardly ever subjected to synizesis.
[ back ] 95. As Schwyzer GG I.245 observes, here η must represent not length but the quality (timbre) of the vowel. For a somewhat different interpretation, see Méndez Dosuna 1993:100n10.
[ back ] 96. This argument explains what has often been thought a violation of regular accentual patterns in worlds like πόλεων. Following Méndez Dosuna (1993:100n9), I will use more or less interchangeably the terms ‘synizesis,’ ‘glide development,’ and ‘loss of syllabicity.’
[ back ] 97. Cf. Chitoran 2002: “The shorter total duration and transition duration of [ea] are consistent with the representation of the diphthong as a simple segment contained in a syllable nucleus. The longer duration of [ja] supports its representation as a sequence of two segments, filling an onset and a nucleus” (214). Restating her conclusion in the terms used at 209–210, one could say that although both represent glide sequences, -e̯a- and -ya-, their disparate phonetic character subjects them to different morpho-phonetic constraints.
[ back ] 98. For convenience, here and in the other Spanish examples I use the graphemes to denote the glides without implying that their precise phonetic values are those of the corresponding IPA signs. Hence, ‘a̯’ should not be taken to imply that this glide has the open quality of a central, maximally open ‘a’. As Méndez Dosuna (1993:100 and 100n28) observes, by its own nature a glide cannot be maximally open.
[ back ] 99. As a native speaker of Spanish I disagree that “[e]ach vowel in the cluster makes a separate syllable” (Bowen and Stockwell 1955:237). This claim accounts for their representation [ĕa] rather than [e̯a] and so on. It is true that the faster and more pronounced the gliding (i.e. the greater the loss of syllabicity), the more /e/ becomes /y/, /o/ becomes /w/, and /a/ becomes /ø/. But even at a high degree of gliding the phonetic realization of /toa-/ in /toallita/ is clearly distinguishable from /tua-/ in /Tuareg/. Cf. Real Academia Española 1973:60 §1.4.15, which regards the Spanish sequences /eo/ and /oe/: “con tendencia al hiato en determinadas voces … y en general al diptongo en la conversación rápida”; and Real Academia Española 2009–2011:3.339 §8.9i, a recent, authoritative restatement of the same view, now generalized to “secuencias vocálicas formadas por vocales que presentan el rasgo [-alto] (/e/, /o/ y /a/).” Ulivi’s formulation in reference to Romanian (apud Chitoran 2002:209) offers a more adequate description: compared to [e] and [o], based on the larger amount of frication visible on the spectrogram [j] and [w] are “consonantal.” All are glides, but the former two are vocalic (semi-vowels), while the latter are consonantal (semi-consonants).
[ back ] 100. “La implicación es excesiva. Una cosa es comprobar que la pérdida de silabicidad de e ([eV] → [e̯V]) puede—y, de hecho, suele—conducir al cierre de la semivocal ([e̯V] → [i̯V]). … Otra cosa muy distinta es establecer un determinismo absoluto entre ambos procesos o, lo que viene a ser lo mismo, fusionar estas dos etapas, sucesivas e independientes, en un único proceso [eV] → [i̯V]” (Méndez Dosuna 1993:108n25).
[ back ] 101. Cf. Rosenthall 1994:56–57.
[ back ] 102. Casali 1997:497 and 1996:1–2. Casali’s dissertation focuses on elision and contraction (which the author calls ‘coalescence’).
[ back ] 103. That is, dieresis is used metri gratia.
[ back ] 104. Against the notion that QM is an ‘artificial’ metrical device, see the persuasive arguments of Méndez Dosuna 1993:104.
[ back ] 105. For moraic phonology applied to CL, see Hayes 1989. For a recent review of moraic phonology that probes its limitations and explores alternatives, see Kavitskaya 2002:17–35. Although moraic phonology as defined by Hayes 1989 fails to explain tautosyllabic CL caused by onset deletion in some languages (Kavitskaya 2002:27–28), it is well suited for the description of Greek and Latin phonological processes because the growth of the theory was driven in part by them. A phonetically-based approach like Kavitskaya’s has much to contribute, but its insights can be incorporated into moraic analysis through language-specific rules. Neither traditional moraic CL nor Kavitskaya’s phonologization model are designed to account for the left-to-right CL involved in Greek QM and vowel contraction (Kavitskaya 2002:33 and 187). But the moraic phonology can be readily extended by the addition of language specific rules. Since Greek QM and vowel coalescence do not preserve moras when its bimoraic upper limit is transgressed, it may be best not to think of these processes as involving strict CL, for such CL assumes the conservation of moras. This does not controvert that glide formation exhibits a tendency toward moraic conservation, subject to language-specific limitations. Consider the remarks in Kavitskaya 2002:102 regarding phonetic naturalness and phonologization. The phonologization of a hypothetical trimoraic vowel or vowel-coda that resulted from the phonetic process of glide formation would be subject to the contrastive limit of two moras. Note also how Kavitskaya 2002:99 explains the absence of CL with loss of intervocalic /r/ in Samothraki by reference to a morpheme-structure constraint. An Optimality Theory of CL is of less interest to me because of its inherently non-derivational approach.
[ back ] 106. No association lines are crossed. Cf. Kavitskaya 2002:33.
[ back ] 107. Cf. Méndez Dosuna 1993:111.
[ back ] 108. Schwyzer 1914:195–196 and Chantraine 1999:222 s.v. γίγνομαι. I cannot accept Peters’s substitution of *genewā– for the more plausible *genehā- (1980:254n212; cf. Chantraine 1933:91 §70 and Sihler 1995:52 §56a). Cf. Schwyzer GG I.315 for the proposal of an expressive gemination, to which contrast Chantraine’s comments. Gemination was the original pandialectal form of CL. Some dialects simplified it after the first wave of palatalizations with the ensuing CL of the preceding vowel (García-Ramón 1982:114 and Sihler 1995:195). Gemination was preserved in the palatalizations with this secondary -y- < -e̯-. From the point of view of moraic phonology, one must distinguish the palatalization caused by the height of the newly formed glide from the lengthening of the palatalized consonant. The mora stranded by glide formation was filled by the palatalized consonant through gemination. Then followed glide deletion. See Méndez Dosuna 1993:112 for στερρός.
[ back ] 109. Sihler 1995:272–274 §265.4 and §267.
[ back ] 110. Cf. Peters 1980:268n223 on the influence of πυρῶν upon κριθῶν in πυρῶν ἢ κριθῶν (Λ 69).
[ back ] 111. Thumb and Scherer 1959:260 §311.12.
[ back ] 112. The moraic value of desyllabized glides is zero and their coalescence does not add length. The case of -ύης with gen. sg. -ύω is different. Discounting the effects of analogical leveling, before æ̯ (~y) a long ῡ could split and develop a homorganic glide: -ῡ- > -ῠϝ-. Therefore, -ῠ́ϝι̯ω > -ῠίω (Lejeune 1972:169). But in the case of a short upsilon the intervocalic glide would be lost: -ῠ́æ̯ω > -ύω. Among the names that Thumb and Scherer (1959:260 §311.12) offer exempli gratia are Παγτύω (Miletos), Ἐρασύω (Khios), and Παναμύω (Halikarnassos).
[ back ] 113. Other examples in Meister 1921:184.
[ back ] 114. Cf. Méndez Dosuna 1993:124 and n. 64. Read also his illuminating comments regarding the absence of QM in the Homeric simplices λαός and νηός, which he relates to morpho-phonological considerations of “phonic volume”; and his further thoughts on τέως and ἕως (112–116).
[ back ] 115. Sihler 1995:508, 557.
[ back ] 116. Cf. Sihler 1995:190 §195.a.
[ back ] 117. Meillet and Vendryes 1979:190 §292, 214 §324. For a different explanation that assumes a textually unattested -ίω, see Wackernagel 1893 and Chantraine GH I.451. Glide formation can result in CL to the left or to the right of the glide. Traditional CL compensates to the left, whereas vowel coalescence compensates to the right. Where -εω- > -ε̯ω- we cannot assume that ε̯ was raised to -y- and that the resulting glide palatalized the previous consonant. (This is possible but, since Attic eventually depalatalized such consonants, indeterminable, unless the rule proposed immediately below holds true.) The parallel of Romanian warns us that [e̯a] and [ya] can be phonologically different. Palatograms of Romanian [Ce̯a] and [Cya] show more contact for [Cya] and contact only on the edges of the palate for [Ce̯a] (Chitoran 2002:209); and -y- in [ya] is longer than -e̯- in [e̯a], supporting the notion of a two-segmental /-ya-/ (with non-moraic -y-) and a mono-segmental /e̯a/ (2002:209 and 205, 207). This suggests the following rule for Greek: when the glide is raised to -y-, CL is leftward and issues in a palatalized geminate; otherwise, it is rightward and issues in contraction (subject to a bimoraic syllabic upper limit). If this rule holds true, gemination would prove original palatalization. A Greek example that did not involve leftward compensation was the so-called Attic future, of which there may be an example at Ρ 451, βαλῶ. Otherwise, we would expect an outcome similar to γεννῆται or πολλό-, with (initially palatalized?) gemination (Sihler 1995:223). But we do not find ×μεννῶ to μένω or ×βαλλῶ to βάλλω. In this case we also have to reckon with paradigmatic leveling, since -e̯- in -ε̯ω [e̯ɔː] might be raised to, and approach, -y-, but in -ε̯ει [e̯eː] it would not.
[ back ] 118. For the loss of intervocalic ϝ in *παϝιζει, see §5.
[ back ] 119. For a short summary of the argument, see Cassio 1998:17–18. Unconvincing is the synchronic view in Chantraine GH I.161 that sees both syllabifications as equally available to Ionic rhapsodes and chosen merely for metrical convenience. This begs the question that only a diachronic perspective can hope to answer: how these two forms came to be part of the Homeric system in the first place. Steriade 1982:124–125 follows Chantraine.
[ back ] 120. Steriade (1982:120) notes that Ionic failed to delete preconsonantal -w- in derived and underived environments alike, so that -VwC- cannot be assumed to have developed into -V̄C-.
[ back ] 121. Cf. Steriade 1982:114–115.
[ back ] 122. Hayes 1989:265–266.
[ back ] 123. She prefers this view—which she supports with a questionable interpretation of Mycenaean Greek spelling, on which see immediately below—because it suits her attempt to establish a universally applicable sound law {s,y} → h/σ[___. In order to prevent postconsonantal /y/ > h, she needs to syllabify intervocalic sonorant + -y- clusters as complex onsets. Wetzels (1986:303–309) shows that Steriade’s explanation cannot be maintained.
[ back ] 124. Brugmann-Thumb GG 423 §434.
[ back ] 125. The Attic treatment of muta cum liquida is usually considered the innovation. Cf. Ruijgh 1985:120n57.
[ back ] 126. This view is supported by Steriade 1982; Guion 1996; and Consani 2003, esp. 39–50 (following Pulgram 1981). Cf. Miller 1990:173 nos. 4–5, Miller 1994:17n4, and Baechle 2007:73n24.
[ back ] 127. Pulgram (1981:86) wonders “why Greek metrics needed more metrically long syllables in its versification than the language provided in the shape of phonologically heavy syllables.” His solution: “Could the answer be that Greek metrics was borrowed, in a prehistoric, pre-Homeric period, from a language with another phonological system that had a suitable number of long syllables … ? Or was this the condition of an earlier pre-Greek or proto-Greek phonological system into whose period falls the origin of Greek metrics … ?” (86). A little later in the same article he considers the Rig Veda and conjectures that “an older syllable-counting metrics in a language relatively rich in light syllables was replaced by a quantitative metrics”; and this, in turn, required perhaps an increase in the number of heavy syllables which were readily supplied by the application of specific metrical rules designed to this end (90).
[ back ] 128. Devine and Stephens (1994:37–38) argue that “[t]he metrical evidence cannot be discounted because it conflicts with an a priori theory of syllabification … . It is probably easier to motivate the alternative assumption, namely that the orthographic evidence partially fails to reproduce the syllable divisions of normal speech. … Writing is a slower activity than speaking, and the writer needs to adopt certain specific strategies to slow down his speech to align it with what is being written.”
[ back ] 129. Cf. Woodard 1997a:77 and 246.
[ back ] 130. With reference to Vennemann 1988, Guion 1996:79 notes that Steriade’s view—which she endorses—of the evolution of syllabification from IE to Mycenaean and Attic violates the so-called ‘Coda Law,’ even as it observes the competing ‘Head Law.’ To this, she tellingly adds: “It is unclear which type of change is more natural.”
[ back ] 131. So Miller 1994:9 and 16.
[ back ] 132. Miller (1994:17n4) thinks that “epic meter is more conservative in terms of syllable division (e.g. Hom. pat.rí vs. Attic pa.trí ‘to father’) than Mycenaean.” Woodard 1997a makes the Mycenaean and Cypriot spelling independent of syllable structure (cf., for example, 48), and thus implicitly disputes the orthographic basis for the inference that Mycenaean had already adopted the innovative syllabification.
[ back ] 133. Cf. Morpurgo Davies 1987:93.
[ back ] 134. So, λεπτότερος and λεπτότατος.
[ back ] 135. For the rule and its exceptions see Chandler 1881:101 §343 and 104–107 §§347–352. For these and other examples of prosodic considerations common to poetry and prose, see Morpurgo Davies 1987:93 and Ruijgh 1985:120–121.
[ back ] 136. These forms were noted by Monro 1992:382–383 §405.3 (first published in 1891). Cf. Cassio 1998:17.
[ back ] 137. Horrocks 1987:276–279.
[ back ] 138. The same scribe writes ke-se-ne-wi-ja and ke-se-nu-wi-ja (Morpurgo Davies 1972:107–110). Cf. Viredaz 1983:167 §§29–30. On ke-se-ne-wi-ja see also Consani (2003:94 and 114), who writes that it is evidence of “una sillabazione tipologicamente marcata /ksen.wV/” (114). I agree.
[ back ] 139. Morpurgo Davies 1972:110, used for Ἐνυάλιος and περυσινός (with unexplained -w-).
[ back ] 140. Cf. Wetzels 1986:312 regarding Mycenaean sonorant + -y- clusters. His article convincingly proves against Steriade the heterosyllabicity of these clusters. Note also his statements about the heterosyllabicity of *ekwo- > i-qo = ikk w os > ἵππος (312).
[ back ] 141. I do not understand how Kavitskaya (2002:51) proposes to supersede an analysis grounded on syllabification like the one by Steriade and Wetzels. Her argument seems to be that in the sequence -VCw- the consonant -C- was heard with a labial off-glide that affected the phonetic duration of the vowel -V-. When the consonant was delabialized, the longer vocalic transition that the labialized consonant had called for was reinterpreted as a longer intrinsic duration of the vowel and phonemicized as such. The logic seems reasonable enough. But Kavitskaya fails to address why the -n- in *ksenwos should have been labialized in Ionic (with the corresponding phonemicization of the longer vocalic transition after it was delabialized), whereas in Attic such vocalic transition was not phonetically relevant. Surely the difference must lie in the different syllabification strategies effective before such loss.
[ back ] 142. Cf. West 1988:166n93.