1. SOLVING THE HUMAN PROBLEM IN GREECE AND MESOPOTAMIA

One famous case for Mesopotamian influence on Greek poetry rests on the resemblance of the Greek Cypria to the Babylonian Atra-ḫasīs.Footnote ² Both are extended narrative poems about a catastrophe that ends the age of myth and begins our own epoch. Each presents this catastrophe as the chief god’s plan to cull the burgeoning human population. In Atra-ḫasīs, the catastrophe is a series of plagues and famines, followed by a flood. In the Cypria, it is the Trojan War.

In Atra-ḫasīs (earliest manuscript seventeenth century b.c.e.), humans multiply. As their noise keeps the chief god, Enlil, from sleeping, he decides to wipe them out (Atr. I.352–60):

Fortunately, Enlil does not succeed in eradicating humankind. Thanks to the god Enki and his favourite mortal, Atra-ḫasīs, humans survive not only the plague but drought and a series of famines. Finally, Enlil sends a great flood. But Enki warns Atra-ḫasīs to build a boat, and he again escapes destruction.

The Cypria too began with divine population control (fr. 1):

We have this from a scholium on Il. 1.5, explaining ‘Zeus’s plan’ to relieve the earth by a cull of mortals (Schol. [D] Il. 1.5, van Thiel):

Some of this information is not apparent in the Cypria fragment. First, Earth made an embassy to Zeus: she complained of the weight of mortals, and asked him to lighten her burden. Second, Zeus’s actions were justified not only by practical necessity but also by the impiety of human beings.Footnote ⁵ Third, the plan of Zeus was tempered by Momus: the scholiast goes on to explain that Momus advised Zeus to marry Thetis to a mortal, and to father Helen. The birth of Helen and Achilles brought on the Trojan War, which relieved the earth of her burden by mass death (ἀφ᾿ οὗ συνέβη κουφισθῆναι τὴν γῆν πολλῶν ἀναιρεθέντων)—but not by the complete eradication of humanity.Footnote ⁶

Momus is not otherwise attested as Zeus’s advisor, but in name and function he matches Apsu’s vizier Mummu (Enūma eliš 1.30–72).Footnote ⁷ Shortly before this episode in Atra-ḫasīs, three gods cast lots for dominion of the sky, the earth and the underworld sea (1.11–18), and this matches an Iliadic scene in which Poseidon tells how he, Hades and Zeus cast lots for their domains of sea, underworld and sky (15.184–93).Footnote ⁸ Finally, there is a straightforward route of transmission. Atra-ḫasīs circulated widely in the Neo-Assyrian empire during the eighth and seventh centuries (Lambert and Millard, 31–6); precisely ‘that epoch when Cyprus, though rich and powerful, was still formally under Assyrian domination … the commemorative steles of Assyrian kings were erected in the cities of Cyprus’. No wonder, then, that the motif found its place in the Cypria.Footnote ⁹

A formidable case, but the matter is not as simple as it appears. Cypria fr. 1 has long played a role in another comparative project: Indo-Europeanists have noted its resonances with a very different poem.

2. SOLVING THE HUMAN PROBLEM IN GREECE AND INDIA

The Mahābhārata, greatest of Indian epics, took form between the middle of the first millennium b.c.e. and the middle of the first millennium c.e. The core of the poem is a succession dispute in the Kingdom of Kuru.Footnote ¹⁰ Two branches of the ruling dynasty, the Pāṇḍavas and the Kauravas, claim the throne, leading to a great battle and the ultimate victory of the Pāṇḍavas. But, as in the Trojan cycle, the battle has a double cause: war between mortals is the gods’ device to cull the population. Earth is crawling with human avatars of demons (Dānavas), so numerous that she is sinking gradually beneath the sea (12.202.10):

Desperate, she makes an embassy to the chief god, Brahmā. He decides that the gods will descend to earth in human form, and slaughter the Dānavas in a great war between Pāṇḍavas and Kauravas (Mbh. 1.58.1–52).

The theme is repeated many times throughout the Mahābhārata.Footnote ¹² In another version Earth begs the gods to relieve her great burden, and Viṣṇu assures her that they will initiate a war involving the King of Kuru (11.8.25–6a):

Thus the elements that pointed to a relationship between the Cypria and Atra-ḫasīs are also present in the Mahābhārata.

The Greek poem shares more with its Indic than with its Mesopotamian counterpart. The Cypria and the Mahābhārata agree that the weight of humankind is the cause of the cull, while in Atra-ḫasīs their noise disturbs Enlil. In Atra-ḫasīs, the chief god acts from personal annoyance; in the Mahābhārata and the scholium on Il. 1.5, because of Earth’s embassy. And while Enlil visits plague, famine and flood on the human race, Zeus and Brahmā prefer a great war. The Cypria and the Mahābhārata also evince a richer structural correspondence than any we can draw between the Cypria and Atra-ḫasīs. Both are long-form epics, covering the full course of the final war of the heroic era preceding our own.Footnote ¹³ In both, a council of gods sets the war in motion, though there is a proximate cause in the world of mortals (the abduction of Helen and Yudhiṣṭhira’s dice-game with Duryodhana). Both narrate the adventures of the heroes prior to the war, the build-up to the conflict, the gathering of the troops, and the battle itself.

It is not plausible to assert that a Mesopotamian story of the flood travelled west into the Aegean, east to the Indus Valley, and underwent identical metamorphoses in both of its new homes.Footnote ¹⁴ Nor is there a plausible route of transmission between India and Greece in the era of the Cypria. For these reasons, the motif of the overburdened earth has long been a canonical parallel for Indo-Europeanists.Footnote ¹⁵ The Greek and Sanskrit languages (along with many other languages of ancient Eurasia) evince correspondences explained by their descent from a common ancestor, Proto-Indo-European (PIE).Footnote ¹⁶ Indo-Europeanists treat the Cypria and the Mahābhārata as divergent developments of a poetic tradition current among prehistoric speakers of PIE. Just as Greek and Sanskrit reflect a proto-language, so do the Cypria and the Mahābhārata reflect a proto-story.

This explanation seems to obviate the theory that the Cypria has borrowed from Atra-ḫasīs. If the features shared between the Cypria and the Mahābhārata are inherited into both poems, they must predate the divergence of Greek and Sanskrit into independent languages. The complex of motifs is older than Assyrian presence on Cyprus, indeed, older than Atra-ḫasīs itself.

3. EASTS AND WESTS

Two incompatible genealogies of the Cypria’s aetiology of the Trojan War are therefore current. On one account, the motif of a divine cull was imported from a Mesopotamian flood narrative around the seventh century, and incorporated into a mature and developed epic tradition about the Trojan War. On the other, the motif is no younger than the third millennium: it was part of the Trojan cycle from its inception, since the stories of the Trojan and Kurukṣetra Wars each evolved from some ancestral story of a cataclysmic battle to relieve the overburdened earth.

The puzzle neatly illustrates a tension in the priorities of these fields: in Indo-European Studies, the organizing principle of research is inheritance; in Ancient Near Eastern Studies, it is diffusion. From the Bronze Age on, the Eastern Mediterranean was a highly networked region in which cult, myth, art and technique travelled widely. For scholars of Greece and the Ancient Near East, diffusion within this environment is the object of research. But for Indo-Europeanists it is an obstacle. To recover material inherited from a common ancestor, external influences on descendants must be identified and excised: Greek material is therefore complicated by its Near Eastern elements.Footnote ¹⁷ Indeed, some scholars object that IE Studies distorts Greek culture, overvaluing ‘inherited’ material and dismissing Eastern Mediterranean influences. ‘The Indo-European abstraction’, in the words of López-Ruiz, ‘has become a method for isolating what is “purely Greek” (another abstraction) and thereby enforcing an artificially sharp dichotomy with the “non-Indo-European” other, the oriental, the Semitic, the East.’Footnote ¹⁸

These suspicions are justified; IE Studies has served racist political projects within and beyond the academy since it began.Footnote ¹⁹ But this contemptible history accompanies a record of concrete discovery. The comparative method in linguistics is sound. Greek is an IE language, and aspects of Greek literature have been illuminated by comparison to other literatures composed in IE languages.Footnote ²⁰ Especially in the last few decades, Greek literature has re-emerged as a core concern of Indo-Europeanists, and anyone interested in the prehistory of Greek literature must integrate this research with the ongoing renaissance of work on Greece and the Near East.Footnote ²¹ The puzzle of the Cypria shows, if nothing else, that scholars in one field cannot ignore results in the other. How then should we bring the two together?

One approach was modelled by West. Learned both in the literatures of the Ancient Near East and in IE Studies, West wrote two compendia of parallels between Greek and other literatures. In The East Face of Helicon (1997), he set Greek material beside texts from ancient Anatolia, Mesopotamia and the Levant, and argued that Near Eastern culture had shaped Greek literature through contact beginning in the mid second millennium b.c.e. In this book, West explained parallels by diffusion from a culturally generative centre to a culture-poor periphery: ‘Culture, like all forms of gas, tends to spread out from where it is densest into adjacent areas where it is less dense’ (1). Ten years later came Indo-European Poetry and Myth (2007), which drew on ancient texts in Greek, Latin, Hittite, Sanskrit and Avestan, medieval texts in Celtic, Slavic and Germanic languages, and nineteenth-century folklore and poetry in Lithuanian, Latvian, Armenian and Ossetic. In this later work, West employed the Indo-Europeanist method of treating features found in more than one of these languages as inherited from the prehistoric speech community ancestral to them all.

West’s diptych offers an attractive solution. There are two distinct strata in Greek culture, equally important, but each demanding methods of its own.Footnote ²² There is first the pre-Aegean stratum, ‘a heritage from the past, not a continuing tradition’ (IEPM, v); second, a period of development within the Aegean, during which cultural material was transmitted to the Greeks from neighbouring civilizations. Up to the moment the speakers of Proto-Greek enter the south Balkan peninsula, they belong to Indo-Europeanists, who use other IE-language texts to determine what they brought with them. On arrival, they become property of specialists in the Ancient Near East, who chronicle how Eastern Mediterranean influence gradually transformed them into the Greeks of the alphabetic period. Hellenists need only attend to both fields and consolidate their results, revealing how IE patrimony conditioned the reception of Near Eastern material, or how diffusion from the East altered inheritance.Footnote ²³ Naturally, there are complications: some literatures are linguistically IE but geographically Near Eastern, and it is not always possible to tell a hoary theme of PIE antiquity from a more recent areal development. But it is no objection to a programme of research that it encounters difficult cases, and Hellenists with a hand in each field may even decide such territorial disputes.Footnote ²⁴

This article argues that this approach, however attractive, is fundamentally misguided. There are no separate Indo-European and Near Eastern strata in the prehistory of Greek literature. This is not a matter of our inability to sort the material into appropriate categories. There is nothing to sort it into. The conditions that justify a distinction between IE and non-IE phenomena in linguistics do not apply in the historical study of literature.

My case is as follows. In section 4, I argue that narratives are frequently transmitted across speech communities by multilingual individuals. Consequently, lines of linguistic descent do not predict lines of narrative descent: poems preserved in unrelated languages may share ancestry. In section 5, I argue that current models predict that, throughout the Late Eneolithic and Early Bronze Ages, many speakers learned PIE as a second language, and that the language functioned as a medium for the transmission of cultural material in a sphere centring on the Eurasian Steppe. In section 6, I present evidence that this sphere extended into Mesopotamia, and that some speakers of PIE also spoke Semitic languages. Finally, in section 7, I argue that this gives us resources for a more satisfactory solution to the puzzle of the Cypria. The Mahābhārata and Atra-ḫasīs are not alternative and incompatible sources of the Greek poem. All three share ancestry in a Eurasian-Steppe tradition.

4. SPEECH TRADITIONS AND CULTURAL TRADITIONS

The Indo-Europeanist approach to Greek literature is founded on the comparative method in linguistics. By identifying systematic correspondences among languages, we can prove that they share an ancestor, determine the sequence of changes by which each has diverged and, by reversing these changes, recover features of that ancestor. In linguistics, this method is relatively uncontroversial.Footnote ²⁵ A more difficult question is how far the method extends beyond language. Culture is less predictable than phonology. Greek Ζεὺς πατήρ corresponds phonetically to Sanskrit Dyáuṣ pitṛ́, and the sound changes that produced each are regular and expressible as laws. But when we reconstruct PIE *Dyḗus ph ₂ tḗr ‘Sky Father’, we are dealing with more than sounds: if speakers of PIE passed down this name, a Sky Father must have figured in their pantheon. Yet no law explains why earth appeals to Zeus in the Cypria, but to Brahmā, not Dyauṣ, in the Mahābhārata.

This is not in itself an objection to cultural comparison. Syntactical, lexical and semantic changes are also irregular and unpredictable, but it is legitimate for linguists to compare these phenomena in an Indo-European framework, for their shared ancestry is proved by systematic phonetic correspondence. Just so, two IE-speaking cultures sang of an epoch-marking war, brought on by the chief god to relieve the earth of mortals. Do these then descend from a proto-epic, sung by people who spoke the language ancestral to Greek and Sanskrit? Can we compare extant narratives of IE-speakers to reconstruct a PIE narrative?

An affirmative answer requires a minimum of three inferences. The first two are explicitly articulated in many introductions to the field, most succinctly by Puhvel: ‘Protolanguage implies protocommunity, protocommunity entails protoculture.’Footnote ²⁶ Languages exist in speech, so PIE must have been spoken by a historical group of people—the protocommunity—and, like all such groups, PIE-speakers must have had beliefs, values, rituals, stories, and so on: the protoculture.Footnote ²⁷ But it is not enough that PIE-speakers had a culture—we must be sure that we can access it through our sources. So we need a third inference, less often expressed, but crucial to the project of IE cultural comparison: that we may reconstruct the protoculture by the same methods as the protolanguage—namely, comparison of extant descendants. As noted, Indo-Europeanists select these descendants on the basis of their language of composition: poems in Greek and Sanskrit are fair game; poems in Akkadian are not. This is sound method if, and only if, linguistic ancestry is a good proxy for cultural ancestry.

‘Culture’ is a broad category, and the relation between speech traditions and cultural traditions is unsettled in anthropology.Footnote ²⁸ Some cultural phenomena are closely bound to language, and so resist transmission between speech communities: in the study of IE metre or wordplay, the third inference is sound.Footnote ²⁹ But the case for an IE origin of the Cypria depends wholly on narrative parallels.Footnote ³⁰ Narratives and narrative devices, characters, images and ideas of all sorts are expressible in any natural language, and frequently spread between different speech communities. Here lines of descent do not track linguistic descent, and the third inference is mistaken.

Internal language change occurs when native learners imperfectly replicate some element of the older generations’ speech, but such changes only take hold in the speech community at large if they are common to a whole cohort.Footnote ³¹ Contact-induced change occurs when native speakers of one language import features of their second, or when second-language learners impose features of their native language onto their target—but again such changes only generalize within a significant bilingual population over a long period.Footnote ³² It is, therefore, difficult for outliers to impact the development of a language. Not so for the narrative arts. Everyone speaks, but not everyone tells tales: active bearers of any narrative tradition are vastly outnumbered by passive recipients, and the inventory of tales in any speech community is held by a small number of outliers.Footnote ³³ Consequently, individual storytellers can have an enormous impact on cultural traditions, and stories can leap from one speech community to another through individuals at home in both. Even where multilingualism is not widespread, one multilingual can make a narrative available in a new language, after which monolinguals can inherit and transmit it by exactly the same mechanism as their ‘native’ tradition. A rich body of scholarship demonstrates that the norm in this domain is continuous and multilateral transfer between geographically proximate communities, regardless of language.Footnote ³⁴

Thus stories preserved in unrelated languages can share an ancestor in a story that travelled between speech communities. In the narrative domain, ancestry should be examined in a regional rather than linguistic framework. The IE language family diverged into its main surviving branches in Central and Western Eurasia. The section following argues that at every stage in this process PIE and its descendants were spoken by at least some multilingual speakers of other languages of this region.

5. THE SPREAD OF PROTO-INDO-EUROPEAN

Every linguist knows William Jones’s momentous inference that Sanskrit, Greek and Latin were ‘sprung from some common source, which, perhaps, no longer exists’.Footnote ³⁵ Less famous, but no less influential, is his inference about the speakers of these languages: ‘we may fairly conclude that they all proceeded from some central country’ (3.46).Footnote ³⁶ At the dawn of comparative linguistics, migrations from a common homeland were a speculative solution to the puzzle of IE distribution: an IE language was spoken in some region because PIE-speakers once migrated there, and their speech subsequently diverged. Most linguists recognized that migration and divergence could not be the whole story, but accepted it as the normal case: in the mid 1980s Indo-Europeanists could still state that ‘a group of related languages is formed when an original linguistic system disintegrates due to disruption of contacts between speakers of individual dialects; the languages are spread to their historically attested territories by migrations of the speakers.’Footnote ³⁷

Since the mid 1980s this model has come in for critique.Footnote ³⁸ Linguists reassessed IE distribution in light of empirical research on how speakers adopt new linguistic behaviours.Footnote ³⁹ In the same period, archaeologists were settling on a broad consensus that IE spread from the Western Eurasian Steppe, during and after the fourth millennium b.c.e., with the rise of mobile pastoralism.Footnote ⁴⁰ We now have a better understanding not only of the social and economic circumstances under which IE came to occupy its historical range but also of the mechanisms of language spread plausible in such circumstances.

The Eurasian Steppe was opened to human exploitation by a suite of technologies developed in the fourth millennium b.c.e. Horseback-riding allowed humans to traverse greater distances and control larger herds than they could on foot.Footnote ⁴¹ The development of milking made herds a source of renewable protein and, with the advent of the woolly sheep, of textiles.Footnote ⁴² Wheeled wagons, pulled by animal traction, could transport shelter, food and water into previously uninhabitable regions, and enabled trade in bulk and at long distance.Footnote ⁴³ A new form of subsistence, transhumant pastoralism, was now possible and indeed lucrative: booming textile industries linked urban centres to networks of mobile herders.Footnote ⁴⁴ Horse- and cattle-breeding grew to meet commercial, agricultural and military demand.Footnote ⁴⁵ Taste for metals, precious stones and exotica among emerging elites generated mining outposts and entrepôts—and between them roamed the nomadic stock-breeders and herders now exploiting the vast pasturelands of the Eurasian Steppe, interacting with settled communities along the river valleys and over the mountain passes.Footnote ⁴⁶

Such an economy spreads mobile pastoralist speech traditions not only by conquest but also as settled populations shift language in response to new incentives. To the nomads of the Chalcolithic, farmers could be useful allies. Pastoralism is high-risk—plague or severe weather can devastate herds—and good relations with farming groups insure against disaster.Footnote ⁴⁷ Pastoralists also trade with farmers to secure agricultural products and outlets for their own goods, and such trade is advantageous to mobile groups, who can access multiple markets and control exchange between them. Since the speech of horseback nomads is learned as a second language by settled populations across a wide territory, it often functions as a lingua franca in the region: farmers have good reasons to teach their children nomad.Footnote ⁴⁸ Population transfer in either direction favours pastoralist speech traditions: dispossessed pastoralists settle among agriculturalists in large kin-groups, where their language lives on in the home, while herders absorb agriculturalists as individuals: farmer women may marry steppe men, and farmer men seeking fortune may turn nomad, but both must acculturate to their new speech community.Footnote ⁴⁹ Even hostile relations spread pastoralist speech: when settlements are raided and abandoned, displaced farmers must integrate into a mobile economy under the patronage of wealthy herders.Footnote ⁵⁰ These dynamics are so powerful that the rapid spread of a language family across the Steppe by a horse-riding, nomadic, pastoralist culture has recurred many times in Eurasian history. After PIE, the Iranian, Oghuric Turkic, Common Turkic and Mongolic language families spread one after another westward over the Steppe, in each case as settled populations shifted to the language of an expanding nomadic pastoralist group.Footnote ⁵¹

The spread of PIE has therefore been linked to a Late Eneolithic and Early Bronze-Age expansion of nomadic horse-riders and stock-breeders, visible in the archaeology of the Steppe. The case is made in extensive detail in Anthony’s The Horse, the Wheel, and Language (263–457), whose title I have pluralized for this article. Anthony associates PIE with a cultural complex that developed between the Don, Volga and Ural Rivers in the early and middle fourth millennium. Between about 3300 and 2500, this cultural complex spread a package of customs and technologies across the Steppe and beyond: pottery-making and bronze-working techniques, kurgan cemeteries without accompanying settlements, brief and episodic use of sites, the use of horses and wagon transport and transhumant grazing of herds. This assemblage, known as the Yamnaya horizon, not only tracks the expected sequence and direction of PIE spread but also explains it in terms of the processes outlined above: during this period of expansion, PIE probably functioned as a language of interregional communication. Travellers could expect to find PIE-speakers in every entrepôt and urban centre, and interregional utility encouraged language shift.Footnote ⁵² At the frontiers where Yamnaya pastoralists interacted with settled agriculturalists, local variants developed among bilingual populations. The model explains many linguistic features of the IE family: the distribution of innovations in the central clades suggests a long period of areal development within a dialect continuum; the unusually large number of primary branches is a typical result of gradual differentiation within such a continuum (as opposed to sequential splitting of populations); and the ‘substrate’ features of many IE clades are common outcomes of language shift involving communal bilingualism.Footnote ⁵³

The model is therefore a powerful explanation of IE distribution, and a significant advance in our understanding of Eurasian prehistory. But its implications for cultural comparison have not been fully appreciated. PIE spread as a medium of interregional exchange, and early IE speech communities were composed partly of second-language learners who did not necessarily abandon their own cultural traditions or even their other language(s) in shifting to PIE. Such communities are perfect incubators for multilingual transmission. There is no reason to assume that narratives current in this cultural continuum were bound to one speech tradition, nor to look for their descendants in IE languages exclusively.

It remains only to see how this bears on our focus, the relation between IE narrative traditions and those of the Ancient Near East. The puzzle of the Cypria concerns specific speech traditions: Atra-ḫasīs is written in Akkadian, a Semitic language, and emerged from an intellectual culture where diglossia with Sumerian was typical. In the model above, PIE would have interacted most intensely not with Sumerian or Semitic, but with unknown languages it replaced in the course of its spread. Still, there is linguistic and archaeological evidence of interaction—less intense but still significant—between PIE-speakers and speech communities in Mesopotamia.

6. PIE–MESOPOTAMIAN INTERFACES IN THE LATE ENEOLITHIC AND EARLY BRONZE AGES

PIE and its earliest-attested descendants evince a number of isoglosses with the Semitic languages of West Asia, indicating borrowing in one direction or another.Footnote ⁵⁴ Some show very early contact: PIE *k̂(e)r-n(o)- ‘horn’ (Vedic śṛ́ṅga-, Latin cornū, Proto-Germanic *hurną, etc.) matches Proto-Semitic *ḳarn- ‘horn’ (Akkadian qarnu, Ugaritic qrn, etc.).Footnote ⁵⁵ Others indicate continuing interaction during the process of divergence: Proto-Semitic *ṯawr- ‘bull’ (Akkadian šūru, Aramaic tōr, Ugaritic ṯr, etc.) is the source of a widely-attested word for ‘bull’ in IE languages (Greek ταῦρος, Umbrian turuf, Albanian ter, Old Irish tarb, Old Church Slavonic тоуръ, Latvian taurs, etc.).Footnote ⁵⁶ But while these words point to PIE *táuros or *téh ₂ wros, apparent cognates in Germanic (Gothicstiur,Old English stēor, etc.) lead us to expect Proto-Germanic *steuraz < PIE *stéuros; andthe Indo-Iranian evidence is also unusual (Avestan staora- shows initial *s but not *-eu-).Footnote ⁵⁷ This variety of forms indicates multiple borrowings: several Indo-European speech communities adopted Semitic words for bull, and spread them by diffusion within a dialect continuum.Footnote ⁵⁸ A similar pattern of wandering is evident in a word for wine, attested in the central IE clades (Hittite wiyan(a)-, Mycenaean wo-no, Latin vīnum, etc.), in Proto-Kartvelian (Georgian ɣvino, Laz ğvini, etc.) and in Semitic (Ugaritic yn, Hebrew yayin < *wayin, etc.).Footnote ⁵⁹ These words show that PIE existed and began to diverge in a region characterized by long-range cultural and technological exchange between speech communities. They are, however, no proof of multilingualism, only labels borrowed for novel items.

We are, therefore, fortunate to know the PIE word for wheel, *k ^w ek ^w lo-.Footnote ⁶⁰ This noun is formed by reduplication of the verbal root *k ^w el- ‘turn in a circle’ (an English equivalent would be ‘turny-turn’). Words for wheel across West Asia match *k ^w ek ^w lo- in sound and construction: Hebrew gilgal and galgal ‘wheel’ as well as Aramaic galgal ‘wheel’ are formed by reduplication of a Semitic root meaning ‘turn’; Georgian gorgali ‘wheel’ is a reduplicated deverbal from Proto-Kartvelian *gor- ‘roll’; these forms are matched phonetically (and perhaps derivationally) by Sumerian gigir ‘chariot’.Footnote ⁶¹ Clearly, the word travelled with a new technology, the wheeled vehicle, whose spread can be dated to the latter half of the fourth millennium.Footnote ⁶² But this is not just a borrowed label: in the Semitic, Kartvelian and Indo-European forms, a native verbal root has been reduplicated to form a noun which characteristically does the action of the verb. That is, the construction of the word in one language has been repeated (and phonosemantically matched) in others.Footnote ⁶³ This requires multilingual agents of transfer, capable of morphosemantic analysis of both source and target language.Footnote ⁶⁴ Some speakers of PIE were also speakers of Semitic languages of West Asia.

The likeliest zone of interaction was the western range of the Caucasus mountains. This natural barrier between the Steppe and the river valleys was broken when Mesopotamian urban centres to the south sought mineral resources to the north. In the Middle and Late Uruk periods, there developed in the Caucasus a cultural assemblage whose sites are characterized by the blending of Mesopotamian artistic traditions with Yamnaya burial practices: the Maikop culture (c.3700–3100). Northern miners and herders traded with southern farmers through the Maikop passes; it was by this route that the wheeled wagon was introduced to the Steppe (Anthony, 282–99). Thus, while there was no direct interface between large groups of steppe herders and Mesopotamian settled populations, there was an active trade route along which goods, technologies and people travelled between them, and along it, an interstitial culture which blended elements of steppe and floodplain. Exchange lasted for half a millennium, beginning in the formative period of the late-PIE speech community and continuing through the early phases of its expansion, simultaneously with the divergence of Proto-Semitic into its primary East and West branches.Footnote ⁶⁵ We should not expect widespread multilingualism, but we have every reason to think that throughout this period some individuals were at home in both PIE and Proto-Semitic.

Anthropological, archaeological and linguistic evidence, therefore, converges on the hypothesis that for a relatively long period a PIE dialect continuum linked the Steppe from the north Caucasus into the southern Balkans and up the Danube valley, with extensions into Northwest Mesopotamia and around the coast of the Black Sea.Footnote ⁶⁶ Throughout the period in which PIE diverged into its earliest descendants, many speakers of IE languages were also speakers of non-IE languages, and PIE itself was a medium for interregional transmission of culture between speech communities—including the speech communities of West Asia. These circumstances are ideal for the kind of narrative diffusion described in section 4.

7. TRANSMISSION OF NARRATIVES ON THE EURASIAN STEPPE

I close by returning to the puzzle of the Cypria. On the Westian model of distinct inheritance and diffusion strata in the prehistory of Greek literature, our candidate sources were incompatible. Although the Cypria shares much with Atra-ḫasīs, parallels with the Mahābhārata are more numerous and exact. If this indicates inheritance from an Indo-European tradition older than Atra-ḫasīs, we must give up the Near Eastern parallels as coincidence. But the alternative model I have outlined shows that here the distinction between ‘Indo-European’ and ‘Near Eastern’ is unwarranted. Cultural exchange with the Near East did not begin after the spread of Proto-Greek in the Aegean, or even after the differentiation of PIE. It was continuous from the initial expansion of PIE itself: the Cypria, Mahābhārata and Atra-ḫasīs may all share ancestry through multilingual transmission during the mid fourth to mid third millennia.

If all three poems emerged from a common Steppe tradition, they are unlikely to be its sole surviving reflexes. Setting them against a broader background of Greek, Indic and Mesopotamian cataclysmic narratives may reveal a richer set of relations among all three traditions.

The strongest point in favour of the Cypria’s IE origin was that in Atra-ḫasīs the chief god culls by flood; in the Mahābhārata and the Cypria, by war. But in a study of Sumerian and Akkadian sources, Chen has shown that, although the motif of a lost golden age is very old in Mesopotamian literature, its termination in a flood is securely attested only from the second millennium b.c.e.Footnote ⁶⁷ Several texts predating or contemporary with Atra-ḫasīs show the chief god bringing a golden age to an end with a war. The Cursing of Akkad (late third millennium) begins in a period of prosperity and wisdom, but when king Naram-Sin demolishes the temple of Enlil at Nippur, Enlil takes vengeance by sending an army from the highlands to raze his city.Footnote ⁶⁸ The fragmentary Lament for Uruk (twentieth century b.c.e.) begins when a divine council is called to address the multiplying human population. Here Enlil decrees the destruction of Uruk, which first falls into chaos and violence as its people pillage their own city, then suffers a military assault by foreign invaders.Footnote ⁶⁹ Thus the flood is only one possibility attested in Mesopotamian narratives of the epoch-marking catastrophe: elsewhere, motifs familiar from the Cypria and the Mahābhārata (the council of the gods, the burgeoning human population, the chief god’s cull) are joined with war. ^{Footnote 70}

What is more, some Indian sources divide cosmic history by a divinely ordained flood.Footnote ⁷¹ The Mahābhārata itself tells the story of Manu (3[37a]185), a pious man who lived in the blessed first age of human existence. Manu takes a small fish into his care. As the fish grows larger and larger, Manu moves it into ever greater vessels, finally releasing it into the ocean. To reward his kindness, the fish tells Manu that a great flood is coming, and instructs him to prepare a ship and outfit it with the seeds of all creatures. As the world recedes beneath the waters, the fish guides Manu’s ship to the peak of the highest mountain of the Himalayas, and finally reveals itself as Brahmā.Footnote ⁷² Here we have many elements familiar from Atra-ḫasīs: a wise mortal is warned by a patron god of the coming deluge, builds a ship to preserve all species, lands on a mountain peak, and offers a sacrifice. Other elements, however, are missing: the Manu tale gives no reason for the flood and does not connect it to the motif of overpopulation.Footnote ⁷³

Finally, Mesopotamian flood tales influenced several literatures of Iron Age Eurasia, most famously north-west Semitic traditions (Genesis 6–9), but also Greek ones.Footnote ⁷⁴ The flood myth involving Deucalion and Pyrrha, first attested in Pindar (Ol. 9), bears a close resemblance to Atra-ḫasīs: one mortal man is warned of the flood by his divine patron, the trickster god Prometheus.Footnote ⁷⁵ He saves himself and his wife from destruction, and they survive on the peak of Mt Parnassus, whence they descend to repopulate the earth. By the time the Cypria was composed, Greek poets were certainly familiar with the mytheme of a flood as a divine punishment.Footnote ⁷⁶

Thus the apparent strength of the Greco-Indic parallel turns out to be an artefact of the texts selected for comparison. Although the Cypria and the Mahābhārata are about war and Atra-ḫasīs about a great flood, each poem emerges from a tradition which generated both war and flood tales. In all three traditions we see several configurations of elements selected from a common stock, used to tell the same story: the end of an early phase of cosmic history in a disaster sent by the gods.

In a treatment of this issue in Greek and Mesopotamian literatures, Haubold has argued that the real parallel is a shared set of generic alternatives.Footnote ⁷⁷ In the flood tablet of the Standard Babylonian Gilgameš, Ea scolds Enlil for choosing the flood when milder alternatives were available: he might have sent a great beast, a famine, or Erra, god of pestilence and war (11.188–95). The scholiast on Il. 1.5 likewise notes that Zeus chose the Trojan War, ‘since he could have destroyed them all by thunderbolts or floods’. For Haubold, these texts confirm that ‘the flood narrative of Mesopotamia can be translated into the war narrative of Greece (and vice versa)’, and testify that Greek and Mesopotamian poets worked within ‘one and the same international genre of literature’: a cycle of narrative poetry treating cosmic history from the birth of the gods down to the cataclysm, conceived as a war or flood (54–8).Footnote ⁷⁸ It is only necessary to expand the scope of Haubold’s analysis. The matrix in which this international genre took shape includes India, and the Indian material reveals aspects of its development not apparent from the Greek and Mesopotamian material alone.

Though Haubold avoids speculating on dates and mechanisms of transmission, his genre of cosmic history in Greece depends on the formation of a Hesiodic and Cyclic canon, not long before the sixth century. There is no question that the Greek flood narratives depend (if only indirectly) on a mature Mesopotamian tradition. If we looked only to Greek and Mesopotamian sources, it would be reasonable to guess that the Greek parallels formed under the influence of the Mesopotamian literary canon in the Iron Age—that is, through late and basically unilateral Near Eastern transmission. But attending to India shows that the epoch-marking cataclysm is very old in Greece. Greek tradition has not adopted a Mesopotamian conception of cosmic history wholesale: precisely because the two traditions shared this broad narrative framework, motifs arising in one were easily attracted into the other. The features of the Trojan Cycle that made a Near Eastern origin so persuasive—the identification of Zeus’s advisor Momus with the divine vizier Mummu from Enūma eliš, and the Iliad’s adaptation of Atra-ḫasīs’ myth of the gods casting lots for their domains—need not be given up as coincidences. These details do appear to reflect an Iron Age Greek encounter with Mesopotamian tradition, in the form of a textualized canon of Akkadian literature. But it was not a first encounter: both traditions were already shaped by mutual assimilation.

The manifold resemblances between the Cypria, Mahābhārata and Atra-ḫasīs have no single explanation. They evolved in a recurring pattern of areal diffusion, independent development and reassimilation.Footnote ⁷⁹ Greek, Indian and Mesopotamian traditions each inherited from Early Bronze-Age Steppe tales the framework of an epoch-marking battle, decreed by the king of the gods to reduce the human population. Parallels multiplied over several phases of contact. When flood variants arose within this macro-region, they spread: in the Aegean, the motif of a divinely ordained flood was readily integrated into an existing tale of an epoch-marking catastrophe, the Trojan Cycle.

A comprehensive study could not stop with Greek, Mesopotamian and Indian sources: it would have to take into account Northwest Semitic flood narratives, the myth of human overpopulation preserved in the Avestan Vendidad and much else besides. Such a study is beyond the scope of this article, but I hope to have shown that it is a desideratum for prehistorians of Greek literature.Footnote ⁸⁰ The processes that brought about this mosaic of correspondences are regional, ongoing through several millennia and are not bound to any speech tradition. They are not neatly divisible into strata of inheritance and transmission, and we cannot simply collate the results of Indo-European and Ancient Near Eastern comparative studies. We must bring their material within the scope of one enquiry.