1. Unrecorded traditions

All Greek hexameter poetry was performance poetry. It was not designed for private reading or for analysis in classrooms: it was a thing of live poets and live audiences. We shall look at the circumstances of performance in Chapter IV. For now, it is most urgent to emphasize that the surviving written tradition is only a very poor reflection of the material that was available to poets of the time.

In the very earliest oral hexameter poetry, traditional motifs and generic forms acted as a kind of stockpile that a poet could draw on for improvisatory, or semi-improvisatory, purposes.Footnote ¹ At the same time, since the existence of traditional elements depends on those elements' suitability for repetition, the elements exerted a creative force of their own. Poetry in that environment was a collaboration between bard and tradition: bards selected traditional elements for their poetic purposes, while traditional elements evolved to be suitable for those purposes. Performing a pre-existing poem was not a matter of verbatim copying but of composing the poem anew. This is usually called ‘recomposition’.

As time passed, reiteration became more important than recomposition. The development of the hexameter tradition, from its earliest stages through to the fifth century, is the story of a gradual transition between at least three performance models:Footnote ²

• • ‘Recomposition’: story outlines are traditional, but details and verbal expression may be improvised or innovated using a toolkit of traditional set pieces and motifs.

• • ‘Recaptured’ performance: re-performance of existing poems with stricter constraints on reproduction determined by stricter social conditions, with an emphasis on the social nature of performance, in religious festivals, symposia, and so on.

• • ‘Recited’ performance: replication of an authoritative recorded version, with an insistence on verbatim repetition.

There is naturally disagreement over where to place any specific poem on this spectrum. It is no longer possible to imagine that any Greek poem that has survived in writing belongs solely to the first of these categories. But the underlying point is secure: traditional set pieces such as type scenes, proems, and the hymnic form served both improvisatory and mnemonic functions. As performance shifted towards recapture and repetition, these elements came to be more valuable as mnemonic devices than as improvisatory building blocks, and their role in the composition of new poems became more indirect. For later hexameter poets such as Choerilus and Panyassis, these elements were not compositional aids so much as genre markers for their poetry, echoing and evoking classic models of the past.

Poets never stopped innovating on traditional material, of course. In Chapter II we saw how the Shield poet innovates within the traditional format of the arming scene. But the pool of available material shifted from a stock of traditional devices towards a stock of specific texts. In the Theogony, when the Muses instruct Hesiod in poetry, we should think of it as a traditional motif. But when Parmenides uses the same idea, we should be thinking of allusions to specific works (fr. B 1.11–12, 22–6):

Parmenides' unnamed goddess is not just a stock motif. Her didactic role specifically echoes that of the Hesiodic Muses, who ‘know how to say many false things similar to genuine ones': the allusion frames a central theme in Parmenides' poem, the dichotomy of truth versus belief. The reference to the gates of Night and Day echoes the didactic and oracular role played by Night in the Derveni Theogony (or another similar poem): this frames Parmenides' poem as a mystic initiation into hidden knowledge.Footnote ³ Later hexameter poets echo Parmenides in their turn. Empedocles borrows his imagery of a highway (hamaxiton) as a metaphor for learning, and the proem of Choerilus' Persica borrows the image of a chariot as a metaphor for poetry.Footnote ⁴

Poets' use of source material that is now lost, or was never written down, is a deep-seated problem in studying early hexameter. Even experienced scholars sometimes make the mistake of assuming that a reference to a myth is automatically a reference to a poem: if a vase represents a Trojan War scene, it must be a homage to a hexameter poem; if a legendary figure has some kind of cult at the site of his or her grave, the stories told about him or her must be epic narratives. This goes even for material within the poems themselves: if the Works and Days refers to ‘Strife’, it must supposedly be a response to a specific text, namely the Strife of the Theogony; if the Iliad relates the genealogy of Diomedes and his father, Tydeus, that must supposedly be borrowed from a Theban epic. The nature of the surviving evidence feeds this habit. Our early evidence for myths and their transmission consists solely of poetic narratives and pictorial representations, so it is naturally tempting to think of these two media when we think about myth-making.

Scholars of the pictorial tradition, by contrast, are keenly aware that this conclusion is nonsense. Many studies in the last three decades have decisively and consistently rejected the notion that pictorial representations are to be interpreted as illustrations of poetic accounts.Footnote ⁵ The most pointed counter-examples are the Gigantomachy and the story of Troilus: both legends were colossally popular in pictorial arts but, so far as we can tell, no hexameter poems ever gave them anything more than a passing mention.Footnote ⁶

The Gigantomachy is the battle of the Olympian gods against the Giants, where Heracles saves the day. In the classical period the most authoritative ‘accounts’ of the Gigantomachy were pictorial: the decorations of the Alcmaeonid temple of Apollo at Delphi, and the metopes on the east side of the Parthenon at Athens. Homer and Hesiod mention the Giants only in passing; Pindar gives slightly more information; but a full report of the story does not appear in any surviving textual source until pseudo-Apollodorus, in the first century bce.Footnote ⁷ The absence of the Gigantomachy from the poetic record is so glaring that ancient writers often confused it with the Titanomachy,Footnote ⁸ and Clay has wondered whether an older form of the Theogony might once have had a Gigantomachy where the Prometheus episode now is.Footnote ⁹

The story of the Trojan prince Troilus, ambushed and pursued by Achilles and murdered at the altar of Thymbraean Apollo, accounts for a staggering twenty per cent of all Greek depictions of Achilles. The pictorial record shows a remarkably uniform storyline, with numerous shared details and only occasional inconsistencies.Footnote ¹⁰ But not a single contemporary poem gave a detailed account of the story, so far as we can tell. The Iliad mentions Troilus' name once (24.257). The Cypria at least reported the story (arg. 11), but the sheer quantity of material assigned to that eleven-book epic means that it must have been a relatively brief report: it is certainly very difficult to imagine it as the inspiration for one-fifth of all depictions of Achilles.Footnote ¹¹ The Cypria is hardly likely to have been more striking than a comparable episode in the Iliad, Achilles' encounter with another Trojan prince, Lycaon (Il. 21.34–135), which provides the occasion for some of the most powerful dialogue in all epic, but of which not a single pictorial treatment survives from the Archaic period.Footnote ¹² Other poetic treatments are too late, and too marginal, to be responsible for the story's towering popularity throughout the Archaic era.Footnote ¹³

This is not to say that pictorial representations of myths are totally unconnected to poetic traditions; that would be swinging the pendulum too far the other way. The current consensus is that artists could creatively engage with poems when they wanted to, though this is best understood as beginning only in the late sixth century.Footnote ¹⁴ There are even a few depictions that appear to have been modelled on a specific poetic source: West cites three vases from 540–490 bce which imitate the binary structure of Arctinus' Aethiopis, one side depicting the ‘Amazonis’, the other the ‘Memnonis’.Footnote ¹⁵ But that kind of privileging of a poem is extremely rare.

So to account for the potential range of media for the transmission of myths, we need a broader perspective. A reference to a myth could be alluding to any of the following:

1. Specific hexameter poems: the Homeric epics, the poems discussed in this book, and other poems that circulated in writing but which no longer survive.

2. Oral epos legend: in other words, purely oral epic. This category refers to narratives traditionally related in hexameter, but which never made it into the written record. The existence of such poems may be reasonably inferred if a narrative shows definite signs of being shaped by traditional hexameter features such as those outlined in Chapter II. A good candidate is the story of Jason and the Argonauts: the story had a heavy impact on the Odyssey,Footnote ¹⁶ but, since we have no attestation of any dedicated epic on the subject, there is a presumption that no specific written poem was known in the classical period.Footnote ¹⁷

3. Other poetic genres: mythological elements appear frequently in the surviving remnants of other poetic genres, and lyric poets sometimes used different material from what we see in hexameter poems. For example, the story of Hermes' cattle raid against Apollo in the Hymn to Hermes had previously been used in a lost poem by Alcaeus. Alcaeus had Hermes also steal Apollo's bow and arrows, but in the Hymn this is only mentioned as a possibility.Footnote ¹⁸ However, it is relatively uncommon for a lyric poem to be devoted specifically to a mythical narrative; Stesichorus' lost poems are the main examples.Footnote ¹⁹ There may also have been purely oral exemplars but, since lyric poetry is not so dominated by traditional set pieces as hexameter, we have no way of detecting their existence.

4. Oral non-poetic legend: there can be no doubt that material in this category existed in great quantity, but we have no access to it, since the surviving evidence is poetic and pictorial. Oral non-poetic legend presumably overlapped with the poetic and pictorial traditions to some extent, but we cannot know how great the overlap was.

5. Pictorial tradition: we must assume that pictorial artists could engage with any of the above categories if they wished, and could even – though only rarely, and late – evoke specific poems. But the pictorial tradition had its own set of well-defined themes and tropes; in some cases it may well have had its own stories too.

The Gigantomachy and Troilus are good candidates for categories 4 and 5. Their popularity lies squarely in the pictorial tradition and not at all in the poetic tradition. If they had an earlier origin, that origin must lie in oral non-poetic legend.

There is an enormous quantity and diversity of material to which we have no access. Because of that, it is an extravagance ever to imagine that a story, or a character, or even a scene, must point to a specific poem. Oral epos legend is nearly always going to be a more parsimonious interpretation than imagining a reference to a specific poem; oral non-poetic legend will be more parsimonious still. An allusion never proves a poem.

2. Hexameter and the formula

Some aspects of the prehistory of the epic tradition can be deduced from language and metre. The most striking characteristic of early hexameter poetry is its extensive and sophisticated system of metrical formulae, and it is widely thought that these formulae were a key part of the toolset that allowed the very earliest, pre-Homeric, epic poets to improvise.

Dactylic hexameter consists of a strict quantitative rhythm of twelve beats, or hemipedes:Footnote ²⁰

(Here ‖ indicates the start or end of a line.) These symbols represent long and short notes, so in musical terms this notation equates to

except that in most cases ♫ can be replaced by ♩.Footnote ²¹ Each line has a tendency to divide into two or more cola, or ‘prosodic phrases’: partly syntactical, partly rhythmic entities. The most common cola are 5 + 7 hemipedes and 5½ + 6½.Footnote ²² There are other, lesser divisions after hemipedes 2, 3, 7, and 8, which can be combined with each other and with the major divisions in regular ways to produce a line of three or more cola: thus some common divisions are 2 + 6 + 4, 3 + 2½ + 6½, and 5 + 2 + 5.

The formulaic system consists of many traditional phrases – formulae – that can be slotted into each of these cola, in such a way that the rhythm of the line as a whole emerges perfectly. In this way, it is thought, a pre-Homeric improvising bard could hypothetically reel off hexameters ad lib. Some formulae are used strictly, others less so. Here are some formulae corresponding to the 5 + 2 + 5 division:

Mixing and matching these will always produce a perfectly formed hexameter. Incidentally, it would be a mistake to think of this system of formulae as a restriction on a poet; rather it is one of many tools that our hypothetical improviser has at his disposal.Footnote ²³

It is usually understood that, as time passed and oral improvisation gave way to literary poetics, the usefulness of formulae as compositional tools gave way to their usefulness as mnemonic devices. As with larger-scale generic elements, metrical formulae moved from being aids in improvisation towards being markers of genre. Certainly there are only slender traces of the formulaic system in fifth-century hexameter poetry, such as Parmenides’ poem or the shorter Hymn to Dionysus (but formulae play an important role in later parodies and imitations, such as the Hellenistic Battle of the Frogs and Mice).

But Homer did not have a monopoly on the system's versatility. Hoekstra has analysed a set of Homeric formulae that developed to fit both the 5 + 7 and the 5½ + 6½ colometry (that is, both penthemimeral and trochaic caesuras).Footnote ²⁴ Some of these formulae gain the extra short syllable through grammatical inflection:

Other pairs are equivalent in both sense and grammatical case:

Hoekstra sees this kind of formula as a very archaic type, since none of his exemplars shows very recent linguistic developments.

However, this technique for building versatility into the formulaic system continued to see innovations in post-Homeric poetry. Here are several more pairs that exist in 5 + 7 and 5½ + 6½ variants:

These formula pairs are of mixed age, to judge from their linguistic character. The first pair relies on having one variant with an Aeolic form περιπ(ε)λ-, the other with Ionic περιτελλ-; the second pair relies on Διός and the late Ionic innovation Ζηνός.Footnote ²⁹ The third pair relies on a fossilized rhythmic irregularity caused by a lost w sound, or digamma (hiatus before *ϝέργα ϝιδυίαι); but the fourth and fifth rely on digamma not being observed (short -ον before εἶδος, elision before Ἑσπερίδες; in Homer, digamma is always observed on *ϝέσπερ-). For all pairs other than the first we have to turn to non-Homeric poetry to see both variants put into practice.

The continuing development of such versatile variants after the formation of the Homeric epics points to continuing development in the tradition. To be clear, it is not proof that the surviving poems were themselves oral productions, or subject to oral improvisation to any particular extent. But it does show that, at the time when the Catalogue of Women and some of the major Hymns were being produced, the epic style was not just a matter of imitating older models. It was a living, evolving tradition with continuity that may well have reached back to the late Bronze Age.

3. Prehistory of the epic tradition

Occasionally it is possible to dig deeper into old linguistic elements of the epic tradition and detect strands of earlier traditions. So to some extent, it is possible to speak about a prehistory of the epic tradition. The most abundant evidence for this prehistory comes from Homer. It is conjectural in many respects, and there is ongoing debate over how to interpret the evidence.

Extant hexameter poetry is primarily in the Ionic dialect, with a sprinkling of other dialectal forms. These dialects are conventionally imagined as influencing the epic tradition as follows:

• • Arcado-Cypriot dialect: earliest stratum; few traces, but common

• • Aeolic: also prehistoric; very many metrical formulae rely on Aeolisms

• • Ionic: core dialect of extant hexameter

• • Doric: uncertain date; very rare in Homer, more common in the Theogony Footnote ³⁰

• • Attic: latest stratum; traces only, perhaps introduced by Athenian editing and adaptation of older Ionic texts

Other than the core Ionic dialect, the most important are the Arcado-Cypriot and Aeolic elements. Arcado-Cypriotisms are few in number but account for some exceptionally common words in the epic dialect, such as αὐτάρ (‘then’), δῶμα (‘house’), ἴερος (‘sacred’), and κέλευθος (‘path’): these all appear frequently in non-Homeric hexameter too. Linguists sometimes conflate Arcado-Cypriot elements with the Mycenaean dialect, since its closest known relative is Arcado-Cypriot, and so they interpret these elements as a supposed ‘Achaian’ phase derived from Mycenaean.Footnote ³¹

Aeolisms are more common because they are deeply embedded in many systems of metrical formulae. The general impression is that Ionian poets incorporated Aeolic formulae into their own poetry and would typically ‘translate’ them into Ionic where possible. But sometimes they preserved the Aeolic originals, if their rhythms were different from those of the Ionic equivalents: poets had more versatility if they kept both forms around. For example, a poet could choose between Ionic πρός and Aeolic πότι (and, in Homer, also πρότι) for the idea ‘towards’, depending on whether he wanted a one- or a two-syllable word; for ‘to be’ he could choose between Ionic εἶναι and Aeolic ἔμμεναι, ἔμεν, and other variants; and we saw above how he could choose between Ionic περιτελλομένων or Aeolic περιπλομένων. There are very many more examples.Footnote ³²

Some cases are less clear-cut. In Classical Ionic and Attic, o-stem (second declension) nouns took the ending -ou in the genitive case. In hexameter we find -ou alongside a more archaic form -oio, which is probably Aeolic in origin;Footnote ³³ but often -oio is a false archaism, used in imitation of genuine Aeolic formulae, as in Ζηνὸς ἐριγδούποιο (‘of Zeus the loud-booming’, 2× non-Homeric), where Ζηνός is a late Ionic form.

Masculine a-stem (first declension) nouns also have dialectal variants in the genitive. These include all patronymics ending in -ēs (Atreïdēs [‘Atreus’ son’], Laertiadēs [‘Laertes’ son’]). Usually they retain the two-syllable ending -āo, but sometimes they take -eō pronounced as a single syllable. Conventionally, these forms are understood as Aeolic and Ionic, respectively. Where -eō appears before a vowel, it may well be a stand-in for the older form with a syllable elided: for example, Λαερτιάδεω Ὀδυσῆος (‘of Laertes’ son Odysseus’) and Ἀτρεΐδεω Ἀγαμέμνονος (‘of Atreus’ son Agamemnon’) might be rewritten forms of hypothetical older Aeolic formulae, *Λαϝερτιάδᾱ’ Ὀδυσῆϝος and *Ἀτρεϝίδᾱ’ Ἀγαμέμνονος, with only cosmetic changes to make them look like Ionic. Several considerations support the intuition that these are often archaisms in disguise.Footnote ³⁴

The principle that older formulae, including Aeolic formulae, were often adapted simply by modernizing or Ionizing their pronunciation, is a well-accepted one among linguists. In this way poets were able to draw on older formulae and on the Aeolic formulaic system while also acquiring the extra versatility offered by contemporary Ionic.

What is debated is the relationship between the two traditions: whether we should think of a diffusion model, where the Aeolic and Ionic traditions coexisted at least partially, and Ionian poets adapted Aeolic formulae while both traditions flourished; or a phase model, where the Aeolic tradition preceded the Ionic tradition, and Ionian poets adopted the Aeolic tradition wholesale to create a new Ionic tradition from scratch.Footnote ³⁵ In the phase model, archaisms and Aeolisms are one and the same thing.

Excursus: a-stem genitives and the phase model

This debate is esoteric, but we shall look at one problem here, as a taster for readers interested in going down the rabbit hole. Readers whose eyes are already glazing over may want to jump ahead to page 71.

One key piece of evidence that is frequently cited in favour of the phase model is the treatment of a-stem genitive endings, mentioned above. The original proto-Greek ending was *-ā(h)o, and Aeolic -āo is close to this. In early Ionic, however, *-ā(h)o came to be pronounced *-ǣo; over time ǣ merged with ē, producing the ending *-ēo; and the vowel lengths were transposed – a process called quantitative metathesis – resulting in the form we see attested, monosyllabic -eō. If an Ionian epic tradition had existed at the time when Ionic still had *-ǣo/*-ēo, then we should expect extant hexameter to display some remnants of that period in the form of formulae with the ending written *-ēo (*-ηο). But only -āo and the post-metathesis Ionic form -eō appear. This implies the existence of a period when there was no epic tradition in Ionic, only in Aeolic: the Aeolic phase.

This argument has been challenged in various ways. Horrocks and Jones point out that some old pre-metathesis Ionic forms do appear in Homer in other contexts: we see στείομεν (< *στᾱ́ομεν) and νηός (< *νᾱϝός) alongside the metathesized forms στέωμεν and νεός/-νεως.Footnote ³⁶ Jones also demonstrates inconsistencies in the treatment of other Aeolic ā vowels: sometimes they are left as α, as in the ending on νεϕεληγερέταο ‘cloud-gatherer’; at other times they are updated to η, as in νεϕελη- in the same word, or in some Aeolic verb forms such as συλήτην.Footnote ³⁷

Jones's argument raises the worrying prospect that the whole business might be more a matter of orthographic conventions than of dialects. The letters η (eta) and ω (omega) did not exist in Archaic and early Classical Athens: prior to 404 bce, Athenian copies of the Iliad would have begun

In this predicted orthography there is no visible difference between the vowels in the a-stem ending on Πελειαδεο and the e-stem ending on Αχιλεος. There is also no difference between a metathesized Ionic ending and a modernized Aeolism: *Πελειαδεο could represent either Πηληϊάδεω (the text as transmitted) or the older form *Πηληϊάδηο.Footnote ³⁸ Orthodox opinion has it that the εω/ηο distinction existed in the ‘original’ Ionic texts of Homer. But it is sometimes suspected that our text of the Homeric epics – and, perhaps, other hexameter poems – is the product of a systematic transliteration from the most prestigious copies from the classical period, the Athenian texts.Footnote ³⁹ This hypothesis is known as the ‘metagrammatism’ or ‘metacharacterism’ theory. It is conjectural in Homer's case, but we can guarantee that this process did take place with Attic authors such as Aeschylus and Sophocles. It is possible that the εω/ηο distinction came about as a result of such a transliteration; and that the distinction was not introduced with the aim of distinguishing between different dialects but between different syllabifications.

On this hypothesis, where older Athenian texts had *εο, later editors systematized the orthography depending on the poetic rhythm. On bisyllabic ⏖, εο was retained. On monosyllabic —, the orthography was changed to εω by analogy with monosyllabic –εω in contemporary Ionic.Footnote ⁴⁰ And on bisyllabic — ⏑, -αο was adopted, using Aeolic orthography where possible because *-ηο no longer existed in Ionic.Footnote ⁴¹

Yet –ηο did not die out everywhere. The Central Ionic dialect of Naxos preserved a distinction between ǣ and ē up until the late sixth century bce, without metathesis: the Naxians used η for ǣ, and ε for both ē and ĕ. One famous verse inscription, the ‘Nikandre inscription’, has as its second line (CEG 403, c.600 bce)

This gives two instances of monosyllabic -ηο-, exactly the missing link that ought not to exist in the phase model. In ‘standard’ epic orthography the line would run

Another verse inscription, CEG 405, shows that this orthography was retained into the late 500s. So a Naxian text of the Iliad would have written the first line

Iliad 21.86 is a good candidate for this standardization theory:

Conventionally the -εω of Ἄλτεω is regarded as a metathesized Ionism: it cannot represent an elided *-α’ since it is followed by a trace consonant (*yός). But the manuscripts show variation between Ἄλτεω, Ἄλταο, Ἄλταω (!), and Ἄλτεο;Footnote ⁴² line 85 has Ἄλταο (— — ⏑); and 86 has a cluster of other Aeolisms (-εσσι, πτολέμο-). This is an ideal place for conjecturing an Athenian orthography of -εο, representing bisyllabic *-ǣo in 85, monosyllabic *-ǣo in 86, and where confusion over the most appropriate orthography led to inconsistency in Hellenistic editions.

Note also the following cases, all with difficult syllabification for Hellenistic editors: Il. 5.60 (Ἁρμονίδεω; -αο in one manuscript), 5.534 (Αἰνείεω vulgate; frequent variants -είω, -είαο; -εία’ Leaf), 8.16 (Ἀΐδεω; -αο Plutarch and one manuscript), 20.484 (Πείρεω; -εως Zenodotus and several manuscripts; -εος in two manuscripts), Od. 3.181 (Τυδεΐδεω; -αο, -εως in one manuscript each), all difficult because followed by trace consonants such as w; and Il. 16.554 (Μενοιτιάδεω; -αο vulgate), difficult because of the ambiguous metrical value of the following syllable.Footnote ⁴³ These variants are distributed across multiple critical editions.Footnote ⁴⁴

A conjecture is not proof, of course, and the treatment of a-stem genitives is not the only pillar supporting the phase model (though it is an important one). The theory that poems in Ionic were transliterated from the Athenian alphabet is disputed. West strongly rejects the idea that orthographic standardization in the Hellenistic Period was ‘a hazardous throw in which any ε or ο was liable to be mistaken for η or ω’;Footnote ⁴⁵ still, the orthographic alterations suggested here are not haphazard, but an extremely systematic standardization.Footnote ⁴⁶ Horrocks regards the Naxian treatment of η in CEG 403 as evidence of conservative orthography, not of conservative pronunciation;Footnote ⁴⁷ yet it is hard to imagine conservatism in orthography at such an early date. Either way, however much we trust or doubt the extant spelling conventions, there is continuing debate on the question of the phase model versus the diffusion model.

Linguists believe that in a few phrases it is possible to see glimpses of formulae even older than the Aeolic and Arcado-Cypriot layers. The phrase kleos aphthiton (‘unfading glory’) was mentioned in Chapter II. That could possibly go back to Indo-European origins: if so, it is a piece of poetry from the Neolithic Age. That is a very exceptional case.

More proximate, though still very ancient, is a group of phrases which are thought to date to a time when an early form of the Greek language still retained a syllabic r̥ sound, that is, r treated as a vowel. This phoneme existed in Proto-Indo-European, but died out in most Indo-European languages except for Sanskrit and the Slavic languages. The Homeric phrases in question are either unmetrical or metrical only by an extraordinary analysis, such as Ἐνυαλίῳ ἀνδρειϕόντῃ ‘(like) man-slaughtering Enyalios’ (5× Iliad); or else ‘corrected’ versions of phrases that would be unmetrical if pronounced normally, such as νὺξ ἀβρότη ‘immortal night’ (1× Iliad) for the more standard, but unmetrical, pronunciation *νὺξ ἀμβρότη. In their original forms these would have been perfectly metrical:

with r̥ serving as a short vowel in both phrases. These reconstructions are guaranteed by other considerations: for example, the -mr̥t- of *amr̥tā ‘immortal’ is cognate with Latin mort- (‘death’).

These archaic phrases do not appear in early hexameter outside Homer. They are probably pre-Mycenaean in origin: syllabic r̥ had already been lost in the Mycenaean dialect attested in Linear B tablets from the Late Bronze Age.Footnote ⁴⁸ In principle, syllabic r̥ might have survived longer in a proto-Aeolic dialect independent of Mycenaean: though Mycenaean is the only Greek dialect attested from the Bronze Age, it was not the only dialect being spoken. The ancestor dialect(s) of both Aeolic and Ionic must have co-existed with Mycenaean. However, the balance of probabilities favours linking Mycenaean and proto-Aeolic in this case, since we see the transformation *r̥ > or, ro (as in *amr̥t- > *ἀμβρότ-) in both Mycenaean and Aeolic. In Ionic the transformation was *r̥ > ar, ra, so we can guarantee that formulae with those spellings were never used in Mycenaean, and must have come from Ionic or proto-Ionic: for example ἦπαρ ὑπὸ πραπίδων ‘in his liver beneath the midriff’ (3× Iliad; perhaps < *ēk^wr̥ supai k^wr̥pidōm?).

The origin of the hexameter itself, as a verse form, is another matter of ongoing debate. This topic is even more obscure and intractable. One area of disagreement is whether or not the hexameter has its origins in Indo-European metre. Indo-European metres are characteristically isosyllabic: that is, they rely on a particular number of syllables per line. This describes Aeolic poetry nicely. But the hexameter is isochronic, with pervasive use of contraction: that is, the practice of substituting — for ⏖.

West and Nagy have nonetheless suggested isosyllabic origins: West, in a combination of two cola, a hemiepes and a paroemiac (— ⏖ — ⏖ — plus ⏓ — ⏖ — ⏖ — ⏓); Nagy, in a pherecratic with dactylic expansion (⏓ ⏓ — ⏖ — ⏖ — ⏖ — ⏖ — ⏓). Both are based ultimately on the Aeolian glyconic metre. West's model, with its versatile ⏓ after the break, successfully predicts formulae of the kind we saw earlier, adapted to both the 5 + 7 and 5½ + 6½ colometries, but does not predict the other common colometries that we see in hexameter, nor the development of bridges (positions in the line where colon-end, and therefore word-end, is avoided); Nagy's model arguably accommodates many colometries and bridges that we see in Homer, but is not so comfortable in explaining why the same features are not preserved in the lyric metres that he cites as parallels. Neither explains the origin of contraction very elegantly (that is, — for ⏖). Barnes follows Fränkel and Porter in favouring a four-colon model, which is undoubtedly the most general way of describing the pattern of colometries that we see in hexameter, but is so general and versatile that it is purely descriptive: it has no power to explain the origins of those cola or the forces driving their development. Hoekstra advocates abandoning research on the hexameter's origin altogether, and insists that only synchronic analysis makes sense. Ruijgh interprets the intractable nature of the problem as evidence that the hexameter must not originate in isosyllabic metres after all: it must have come to the Mycenaeans from a non-Greek isochronic poetic tradition, perhaps poetry of the Minoan culture of Bronze Age Crete.Footnote ⁴⁹

That is hardly the last word on the subject. The more common view is that the hexameter was originally isosyllabic, though the process of its derivation is disputed. This topic is devilled by the complexity of the formulaic system and the versatility of the colometric system. It is all too easy for a theory to gloss over inconvenient details. Purely statistical or automated analysis of metrical formulae and colometry should be the basis for a genuinely objective discussion; to date, however, only a handful of attempts at automated analysis have been made, most of them very narrow in scope or deeply flawed.Footnote ⁵⁰ For the time being the prehistory of the epic tradition is, if not totally obscure, at least very much open to doubt and debate.