Elamite

Elamite is our name for the ancient language written in Susiana and Elam. Susiana, also known as Xūzistān (in Middle and Modern Persian) and Ahwāz (in Arabic, the plural of Hūz, a more ancient variant of Xūz), is a distinct region constituted of the marshes, hot plains, and hills east of the lower Tigris and west of the southern Zagros mountains, watered by the Kārūn, Karxe, Jarāḥī, and Mārūn rivers.Footnote ¹ The name Elam is usually used to refer to the highlands to the east and southeast of Susiana, as far as central Fārs, though sometimes ancient and modern authors have subsumed Susiana under Elam. The Elamite language was used in antiquity in both the lowland and highland regions just described.Footnote ² Elamite is known directly from cuneiform inscriptions and tablets dated from approximately the twenty-third century bce to the fourth century bce in these two regions. During those two millennia, as states ruled by Elamite speakers formed, changed, and collapsed,Footnote ³ the language, too, naturally underwent distinct changes, exhibiting different stages of grammatical development. Despite the length of its attestation, Elamite remains relatively poorly understood as a language.Footnote ⁴ Reasons for this include the unevenness of the documentation provided by batches of Elamite texts from different sites dated sometimes centuries apart, the unsuitability of the adapted Akkadian cuneiform script to represent the language, and the absence of known related languages that could contribute to the elucidation of its grammar.Footnote ⁵ Nevertheless, Elamite was the speech and written medium of several kingdoms in distant antiquity in the regions just mentioned. Elamite was last written for the administration and royal inscriptions of the Achaemenian Empire (circa 550–330 bce). The latest extant text written in Elamite consists of stereotyped expressions in a trilingual (Old Persian-Babylonian-Elamite) royal inscription from the reign of Artaxerxes III (regn. 358–338 bce). Thereafter, direct record of the language vanishes.

Of course, the end of direct documentation does not indicate the demise of the language, nor does the formulaic character of the latest texts in which it is known indicate attenuation of the number of native speakers or restriction of function in the lives of native speakers. Like other dead languages, Elamite “died” only when no children learned that language, learning instead only the other language of a bilingual parent, so that with the passage of the last, bilingual, generation of its speakers, nobody remained who could use it.Footnote ⁶ Such events are seldom documented by their very nature. The only hope for ascertaining approximately the time of its demise is to discover a terminus post quem in the latest report of a language's existence.

Xūzī and Its Speakers

Arabic reports of a Xūzī, i.e. “Susian”,Footnote ⁷ language in the eighth, ninth, and tenth centuries ce, to be reviewed presently, indicate with near certainty that a form of Elamite survived in the foothills of Xūzistān until that time. The ten centuries between the latest known Elamite inscription and the first Arabic reference to a Xūzī language do make one hesitate about the identification of the two as different stages of the same language, but the absence of record does not prove its extinction, whereas positive attestations of a “Susian” language are highly meaningful. One should recall that even in distant antiquity, Elamite written in cuneiform is unattested for periods of centuries before it recurs in known, later texts.

More than anything else, it is the Arabic remarks of the geographer al-Iṣṭaxrī, circa 935–943,Footnote ⁸ on the language of the people of the region of Xūzistān, that lead to the confident assumption that the Xūzī language was the latest known stage of a species of Elamite, rather than another Iranian or Aramaic dialect named for the region in which it was spoken. He wrote about the people of Xūzistān, “As for their language, the common people speak Persian and Arabic, although they have another, Xūzī language which is not Hebrew, Aramaic, or Persian”.Footnote ⁹ In Arabic nomenclature of this period, the term Persian is usually generic, and can refer to any Iranian language (though Sogdian was usually treated as distinct). This statement of al-Iṣṭaxrī—whose name indicates his family origin in Iṣṭaxr, Fārs, and his personal familiarity with the region—should be understood to indicate that the Xūzī language was perceived to be what we today would call neither an Iranian nor a Semitic language. Moreover, it was the language of the anciently autochthonous people, or at least those for whom the region was named. Scholars such as von Spiegel, Huart, Spuler, Lazard, Potts, Orsatti, and Tavernier have already suggested or assumed that the language mentioned here is a very late form of Elamite.Footnote ¹⁰ As Potts expresses it well, “there are not many choices apart from a late form of Elamite”.Footnote ¹¹ I believe we may feel quite confident, with them, in identifying the medieval language of the Xūz, “Susian”, bearing the same proper name as those represented by Old Persian u-v-j (^hūja or rather ^hūža), Manichaean Middle Persian Hūžīg, Armenian Xužik, and Aramaic Huzāyā, as the latest reported survival of a variety of the language more commonly known today as Elamite.

The speakers of this language called al-Xūzīya are stereotyped in the Arabic sources as lowly menials. Ibn al-Muqaffaʿ (d. circa 757), another Persian, famous as a translator from Middle Persian into Arabic, is the earliest of the Arabic authors to mention the language. His remark about Xūzī occurs in his frequently cited sketch of the languages of the Sasanian kingdom, surviving as an excerpt from an otherwise lost work. Among five languages in use among the people of the Persian kingdom, he lists Xūzī, and defines its social function with respect to the lives of the kings. “As for the Xūzī language, the kings and nobles used to speak it in private quarters, places of recreation and pleasure, and together with attendants”.Footnote ¹² The remark suggests that some servile personnel among royal attendants in the late Sasanian period were recruited from Xūzistān. This should not be surprising, as Xūzistān was the region between Pārs and Sūristān—Persia proper and Iraq—both regions frequented by the Sasanid monarchs and their aristocratic supporters. If some Sasanian nobles spoke the language, as Ibn al-Muqaffaʿ claims, they may have learned it from nannies, wet-nurses, maids, and the like, as children. It was, in any case, according to Ibn al-Muqaffaʿ's characterisation, a language used with attendants in the privacy of the home and among intimates. Another version of the text, preserved by Yāqūt centuries later, but which may lay claim to originality, specifies the domain of Xūzī rather as that of servants in the lavatory, and it would thus be the language used by Persian aristocrats with attendants responsible for baths and bodily waste.Footnote ¹³

We learn more about the Xūzī language from the prolix al-Ǧāḥiẓ, writing a few generations later than Ibn al-Muqaffaʿ. He was from Baṣra, immediately adjacent to Xūzistān, and, apparently, he heard this language himself. In the 840s, he seems to single out “the language of the Xūz” (luġat al-Xūz) as having an exceptionally rich inventory of sounds. This is in the context of his discussion of sibilant phonetics, notably an area of special uncertainty and difficulty in the linguistic description of Elamite phonology today.Footnote ¹⁴ In another work, Xūzī serves as al-Ǧāḥiẓ's example of a language most difficult to learn, even for those who had plenty of contact with Xūzī people, trading with them and living as their neighbour for a long time.Footnote ¹⁵ Its reported difficulty accords with what linguists today maintain about languages learned by few outsiders: they tend to remain or even gradually become more grammatically complex, that is, difficult to learn non-natively.Footnote ¹⁶

One half century after al-Ǧāḥiẓ, we are assured of the continuing life of the Xūzī language by a fragment of a lost portion of the geographical work of al-Yaʿqūbī, written in 891, mentions that “the people of this region”, Xūzistān, “have a language peculiar to them that resembles gibberish, although Persian predominates among them”.Footnote ¹⁷ A generation later than al-Iṣṭaxrī, whose report has already been mentioned, the traveller and geographer al-Muqaddasī (wr. circa 988) provided a bounty of information about the Xūz, our latest original reports about them. Although he characterises the people of Xūzistān generally as bilingual in Arabic and Persian, and frequently alternating between the two and blending them, he specifies that in his time the people of Rāmhurmuz, a town in eastern Xūzistān at the foothills of the Zagros, spoke a language that was incomprehensible to others.Footnote ¹⁸ Rāmhurmuz was a Sasanian foundation that enjoyed renewed prosperity in the late tenth century, after the Buyid ruler ʿAḍud ad-Dawla (regn. 949–983) founded a market there.Footnote ¹⁹ The town's flourishing should likely be implicated in inducing the death of the Xūzī language at that site, for an influx of wealthier foreigners is just the sort of condition that can cause a young generation to neglect their parochial home language and to prefer instead the language that gives them access to better opportunities outside the family; then the children of the next generation would not need to use the grandparents’ home language.

This is especially the case if being, or appearing to be, a local Xūzī was stigmatised in that period. Indeed, being Xūzī did connote low status to others. As seen, Ibn al-Muqaffaʿ characterised the Xūzī language as used among the menials of the Sasanian royal household and aristocracy. Al-Ǧāḥiẓ relates a report, from al-Aṣmaʿī (d. 828), from Abū Ẓubyān, that the name Xūz was derived from the Persian word for pig, xūk.Footnote ²⁰ A ḥadīṯ preserved by al-Muqaddasī and others alleges that the prophet Muḥammad said, “The most hateful language in God's view is Persian; the devils speak Xūzī; the people of Hellfire speak Buxāran; the people of Paradise speak Arabic”.Footnote ²¹ The date of origin of this ḥadīṯ, at least in this form, is not likely to antedate the invasion of Buxārā, where Sogdian was spoken, because it refers to a language special to that city.Footnote ²² Al-Muqaddasī further reports a ḥadīṯ of the prophet on the authority of Ibn Masʿūd (d. circa 653), “Do not marry the Xūz, for they have roots that invite disloyalty”. ʿAlī ibn Abī Ṭālib (regn. 656–661) is credited with the statement that “There is no people on the face of the earth worse than the Xūz. They have never had a prophet or a man of distinction”. The caliph ʿUmar (regn. 634–644) is supposed to have said, “If I live, I shall sell the Xūz and deposit the proceeds in the state treasury!” It was also said, “Whoever has a Xūzī neighbour and needs the sum of his worth, let him sell him!” Al-Muqaddasī then delivers an anecdote clearly expressing the gradual geographical restriction of the Xūzī population in the late tenth century. He says,

Al-Muqaddasī goes on to say that Iraqis in his time used the address “Hey you Xūzī!” (yā Xūzī) as a form of personal abuse.Footnote ²³ This exchange, reported by an eyewitness, conveys both the stigma attached to being Xūzī and the gradual diminishment of the domain of the Xūz themselves, or of those willing to claim a Xūzī identity. The Acts of Mār Mārī, a Syriac text written during the last century of Sassanid rule, specifies that the major cities of pre-conquest Khūzistān were inhabited by Xūzian merchants (Aramaic taggāre Huzāye, in this text distinct from Persians and others),Footnote ²⁴ but by the tenth century, the major town of Xūzistān was populated by immigrants from neighbouring territories, and the real Xūz, in the view reported here, were located only upland, near the Dez (Dizful) river's confluence with the Kārūn. If inhabitants of Xūzistān who spoke Arabic and Persian were no longer “real Xūz”, then the distinct Xūzī language must have been one of the most salient features of their identity, but the geography of the “real Xūz” was shrinking.

Later Arabic sources do not mention the Xūzī language, except when authors recycle the statements already related. The passages reviewed here give us important clues about the evanescence of this language. Being Xūzī bore a stigma.Footnote ²⁵ The Xūzī language was not learned by outsiders because it was not a vehicle of opportunity for others. It was perceived as difficult to learn, or, at least, only “real Xūzīs” spoke it. So long as Xūzī-speaking communities reproduced themselves with Xūzī-speaking children, in locations not inhabited by many speakers of another language having more general utility for those children, Xūzī would continue to be spoken natively and remain alive. Because Xūzī is now extinct, we know that this condition did not last forever. At some time, the last speakers of Xūzī were bilinguals whose children had no special stake in the use of Xūzī, or refused to use it, and used only their parents’ other language with their own children. The complete shift of Xūzī-speaking families to Persian or Arabic, or both, and the corresponding extinction of the Xūzī language—apparently the latest known stage of what we call the Elamite language—suggest an enduring pattern of this sort of social relations. If merely being Xūzī was a stigma, one should expect that many young Xūzīs on the frontiers of social contact with other groups would acquire a negative attitude toward the Xūzī language and would have been quick to shed markers of their parochial identity, especially when more powerful foreign colonists moved into their vicinity.Footnote ²⁶ The eleventh century probably witnessed the demise of this language.

A Language Leaving Little Trace

F. de Blois devoted an article to the problem of contact between the Elamite language and Western Iranian languages, of which Persian is the longest and best attested branch.Footnote ²⁷ There he rightly notes that it should be surprising if Elamite had not left a mark on Western Iranian languages, and Persian in particular, because speakers of the two languages were manifestly in contact in antiquity for many generations, and there must have been persons bilingual in both languages to transfer features from Elamite to Iranian. De Blois's conclusion, however, is that, apart from some minimal traces, scarcely any clear sign of Elamite features can be found in Persian or other Western Iranian languages. It is remarkable that the contact of these two languages should not leave more traces, especially in the form of loan-words, in both of them and not just one of them. Vocabulary is a feature most liable to transfer from one language to another, and people sometimes adopt loanwords even from languages with which they have little contact,Footnote ²⁸ but there are exceedingly few Elamite loan-words in Persian, or any Iranian language, at any known stage. Excepting some proper nouns, practically the sole example is a word attested as Old Persian dipi-, “inscription”, which comes from Elamite tipi-, “clay tablet” (ultimately, via Akkadian, from Sumerian).Footnote ²⁹ It survived also in later Iranian languages: Middle Persian and Parthian dib, Bactrian λιβο, “document, epistle.” The loan-word in Old Iranian provided also the first part of the later-attested Persian compound words dibīr, “scribe”, and dīwān, “archive, bureau”.Footnote ³⁰ Where nearly no other loan-words are found in such a situation, it must mean something about the relationship of the speakers of the two languages in antiquity.

This is especially so when the reverse influence, of Persian on Elamite, is obvious to all readers of Achaemenian Elamite. As long noted by scholars, Elamite texts of the Achaemenian period contain many Persian loan-words and even Persian syntagms.Footnote ³¹ Recent contributions by W. Henkelman have made this meaningful for the cultural history of ancient Persia. Following a suggestion by I. Yakubovich,Footnote ³² and employing part of a model of linguistic interference through varieties of bilingualism elaborated by F. van Coestem,Footnote ³³ Henkelman has argued that Elamite grammar of the Achaemenian period exhibits a degree of restructuring induced by the imposition of features of Iranian grammar by speakers of Old Iranian. He regards this species of Elamite as the result of “the (imperfect) acquisition of a second language and the resulting restructuring of that language on the model of the speakers’ first language”.Footnote ³⁴ This is initially plausible, by itself, with respect to both the description of Achaemenian Elamite and the linguistic theory. Henkelman proposes that the state of Achaemenian Elamite grammar, divergent from that of earlier periods and exhibiting features transferred from Old Iranian, is evidence of “the widespread usage of Elamite by Iranophones”.Footnote ³⁵ There are, however, problems in this application of the model of language contact. The speakers’ agency should affect both languages, if Elamite-Iranian bilingualism was widespread and generally characteristic of the population in which both languages were used. As Henkelman himself notes, “Language contact is hardly ever a one-directional process”,Footnote ³⁶ but his application of the model accounts only for one of the directions. We should expect to find, rather, if Iranophones employed “widespread usage of Elamite”, significant Elamite borrowings in the form of loan-words in Persian, at least according to van Coetsem's model, which Henkelman invokes. De Blois, whose article on the nearly total absence of Elamite words and features in Persian was not noted by Henkelman, showed that these are not to be found, with a few marginal and doubtful exceptions.

This becomes a problem in one component of Henkelman's greater argument about Persian ethnogenesis. Following de Miroschedji,Footnote ³⁷ he has posited a fusion or blend of Elamites and Iranians at the origins of Persian ethnicity, a view that has been endorsed by others.Footnote ³⁸ This model of Persian ethnogenesis is not in itself refuted in a general way by the absence of significant Elamite loan materials in Persian; I think that the hypothesis of a blend of indigenous and immigrant populations in pre-Achaemenian Pārsa is likely to be correct, in some form or other. The alternative, after all, would be total population replacement, for which there is no evidence, and which is inherently unlikely. What is at stake is rather how the fusion will be correctly characterised and for which period it should be so characterised. Henkelman has rightly dismissed racial interpretations of an Arya takeover of Elamite populations who thereafter served their foreign masters, assumed by scholars of a bygone generation.Footnote ³⁹ In place of such a model, he emphasizes “centuries of cohabitation, acculturation, and integration” between the originally separate populations, giving rise to a “Persian identity” that was “inclusive”.Footnote ⁴⁰ These pacific terms are chosen to contrast with the older view of Persian conquest and domination of sedentary subject peoples who spoke Elamite. Henkelman's argument runs against an older view of a “dichotomy” of the two peoples, which he holds to have become fused by the time of Cyrus, the conqueror from Anšan, into one Persian community.Footnote ⁴¹

The merits of this idea of Persian ethnogenesis notwithstanding, there are obstacles to both attractive hypotheses 1) that an Iranian-Elamite population merger had a bilingual outcome for Iranophones and 2) that speakers of Old Iranian learned Elamite and imposed their Iranian grammatical structures on it. The mere use of a secondary language besides one's dominant language is hardly the sole factor involved in changes to languages in contact. When it comes to grammatical “imposition,” it is specifically the large-scale acquisition by adults—non-native speakers—that has been shown to cause the most serious effects of this kind, often with grammatical attrition.Footnote ⁴² Children, by contrast, learn languages to which they are regularly exposed perfectly well. In van Coetsem's model, children bilingual from a young age will experience what he calls “neutralisation” of linguistic dominance, in which they are comparably proficient in both languages.Footnote ⁴³ Typically, highly proficient speakers of a language do not impose features of other languages on it because they do not need to rely on material from other languages to meet their communicative needs. They already know that language quite well. They may impose features of another language, however, when bilingualism is shared with their interlocutors and code-mixing becomes useful in communication, but we should expect grammatical convergence affecting both languages if that is so.Footnote ⁴⁴ The kind of change for which Henkelman is arguing, to account for the history of Achaemenian Elamite grammar, by contrast, often takes place within a few generations, in the midst of or wake of a demographic upheaval resulting from intense, large-scale population contact. The inconsistency is that while Henkelman posits “(imperfect) acquisition of a second language”Footnote ⁴⁵ as the cause of the imposition of Iranian features in Elamite, he simultaneously maintains that the linguistic effects were due to “centuries” of acculturation. In the latter, hypothetical scenario of pervasive and long-lasting Elamite-Iranian bilingualism, children would, one supposes, be exposed to both languages from childhood, in which case imperfect adult acquisition should not be such a factor—whereas code-mixing and grammatical replication would be—but adult acquisition is, of course, what is meant by “imperfect acquisition of a second language.” In a scenario of pervasive bilingualism in an “inclusive” population of the kind posited by Henkelman, we should expect to find substantial features, especially lexical, transferred from each language into the other. But we do not. Instead, the transfer of features is largely unidirectional, from Persian into Elamite.

Both van Coetsem's elaborated linguistic model and other related models offer explanations that can account better for the unidirectional transfer of features from Iranian into Elamite and the absence of the reverse. The cause can be construed not as the imposition of Iranian linguistic structures on Elamite by Iranophone adults, but as structural borrowing or grammatical replication by Elamophones (those for whom Elamite is the dominant language). That is, Henkelman's examples of grammatical “imposition” in ElamiteFootnote ⁴⁶ may be understood as structural borrowing by Elamite speakers, such as the reassignment of imported grammatical functions to Elamite words and morphemes.Footnote ⁴⁷ The importation of such features is known to occur during a population's gradual shift from one language to another. As bilingual members of one linguistic community become more proficient in the language of another, neighbouring community, beginning in special domains of speech use and progressing to all domains of life outside the ancestral home, the bilinguals can induce structural change on their home community's language in a kind of reverse “imposition” from the social linguistic frontier.Footnote ⁴⁸ This is what Heine and Kuteva call “L2>L1-replication,” a well-documented phenomenon, as for example when Basque speakers, most of whom are multilingual, replicate French or Spanish constructions in Basque, while French and Spanish speakers, who almost universally do not know Basque, do not do the reverse.Footnote ⁴⁹ In other words, the Iranian features of Achaemenian Elamite can be interpreted, with the benefit of hindsight, not as the symptom of a stable- and reciprocal-bilingual society in which Iranophones spoke Elamite, but rather as the virtual harbingers of attrition of Elamite and, ultimately, “language death”—here, locally, the eventual total shift of the Elamite-speaking population of Pārsa to dialects of Persian.

If Persian of any stage exhibited significant features, particularly vocabulary, transferred from Elamite, the sociolinguistic component of Henkelman's argument could be more easily endorsed. As it is, the abundant presence of Iranian features in Elamite and the absence of Elamite features in Persian together suggest a less uniform and mutually acculturated sort of society in Pārsa at the time of Cyrus and Darius. Van Coetsem's model rightly insists that bilingual speakers are the main agents of the transfer of features between languages.Footnote ⁵⁰ It is this feature of van Coetsem's model that Henkelman duly emphasizes (“agentivity of the native speakers”).Footnote ⁵¹ But Henkleman's idea that “the late, morphosyntactically restructured form of Elamite is unlikely to have emerged among native speakers of the language”, because these changes are “too radical” to be caused by native speakers, overlooks plenty of research on grammatical replication of the kind just described, including even van Coetsem's own notion of an “extended mode of borrowing”, in which bilinguals do “borrow” grammatical and phonological (and not just lexical) “material” from another language and incorporate it into their dominant (native) language.Footnote ⁵² When applied to de Blois's negative findings, van Coetsem's model, taken into account in its entirety, suggests rather that “unidirectional bilingualism”Footnote ⁵³ applies here: Persian-Elamite bilingualism was probably mostly limited to persons for whom Elamite was the dominant language, i.e. from Elamite-speaking families, but who also knew or learned Persian for the purpose of interacting with the people outside of their home community. This bilingualism was widespread enough among Elamite speakers that they could both code-switch with (Old) Persian and replicate Persian grammatical patterns in Elamite, deliberately or unconsciously, within the limits of comprehensibility, when communicating with other bilinguals like themselves. This accounts for the large number of Persian words attested in Achaemenian Elamite texts as well as the replication of features of Old Persian grammar with Elamite materials. These are normal kinds of grammatical innovation, here employed in Elamite texts to be read by other Elamophones who were probably assumed by the authors to have some proficiency in Old Persian as well. Simultaneously, the extreme paucity of Elamite loan-words and other features in Persian, noted by de Blois, indicates that few speakers for whom Persian was the dominant language learned Elamite. Perhaps some Elamite-Persian bilinguals transferred some Elamite features into Old Persian, when they used it with speakers of the latter, but the generality of Old Persian speakers, those for whom Persian was the dominant language, were not bilingual with Elamite, and they did not adopt those foreign, Elamite features if they heard them. Elamite features were not, in any case, transmitted in Persian, as de Blois showed. (De Blois's main exception is the possible transfer of a feature of Elamite phonology into Old Persian with regard to nasal consonants occurring before stops, but this is uncertain).Footnote ⁵⁴ The discrepancy in the effects on the two languages on each other—through the medium of individuals who used both languages—thus diagnoses a scenario in which Persian-speakers were distinctly socially aloof from Elamite speakers, typical of settings of unidirectional bilingualism.Footnote ⁵⁵ This also happens to resemble the situation suggested by the Arabic sources reviewed above, pertaining to a period more than a thousand years later. Coincidentally, there are no known Xūzī loan-words in Arabic, either, despite the presence of Xūzīs in the vicinity of Arabic speakers for centuries. The two negative observations together suggest that the social domain of the Elamite language was gradually contracting over many centuries, probably at some times more rapidly than at others. Compare the history of Irish vis-à-vis English. As it happens, van Coetsem holds that the “essential characteristic” of situations in which grammatical patterns—and not just loan-words—are borrowed (rather than imposed—his “extended mode” of borrowing) is that speakers of the recipient language (the borrowers) “feel culturally subordinate” to the speakers of the language from which they are replicating grammatical structures.Footnote ⁵⁶

A new contribution by Yakubovich (which likewise does not refer to de Blois's negative argument) proposes that Persian nominal syntax does show a sign of remodelling on the pattern of Elamite nominal syntax. He relies on linguistic typology—here, the observation of tendencies for groups of linguistic features to occur together more frequently or less frequently in the known languages of the world—to make the argument that Persian is more like Elamite and less like other Iranian languages with respect to a combination of two syntactic features. Specifically, the world's languages infrequently combine right-branching noun-phrase constructions, like the so-called ezāfe (iḍāfa) found germinating already in Old Persian, with the order subject-object-verb (SOV), also found in Persian. Yakubovich notes both the rarity of the coincidence of these two basic syntactic features among the world's languages and especially its rarity among the Iranian languages, which otherwise share abundant features by virtue of their common descent from a prehistoric ancestor language.Footnote ⁵⁷ Yet Elamite and Persian (typically but not exclusively) share this combination of features.Footnote ⁵⁸ This leads him to suggest that right-branching noun syntax in Persian, specifically the ezāfe, arose due to early contact with Elamite. It is a difficult argument because it is hard to control all the factors. For example, it was not possible to address the syntactic ordering in all the languages in contact with other Iranian languages, from Turkey and the Caucasus to South Asia. Nevertheless, Yakubovich's hypothesis warrants serious consideration. I am not able to discuss the entire argument here, but I note that his conclusions, at least, largely correspond with my own. Although he invokes Henkelman's idea that Elamite was restructured due to “imperfect second-language acquisition” by Persian speakers, a view I have just given reason to doubt, Yakubovich concludes that Persian itself was the target language for Elamite-speakers who induced changes in the syntax of Persian through imposition of elements of their own grammar.Footnote ⁵⁹ If Elamite learners were truly the efficient cause of the ezāfe, imposing the nominal syntax of their Elamite on Persian, it is significant for our inferences about the social reality behind this language contact that the argument specifies grammatical and not lexical transfer. The relative absence of Elamite loan-words in Persian strongly suggests that few Persians used Elamite, because loan-words are typically features adopted by speakers into their personally dominant language, whereas the imposition of foreign grammatical patterns, such as the syntax of noun phrases, tends to come from non-native learners who make up for the shortcomings of their fluency in a secondary language using materials of their dominant one.Footnote ⁶⁰ In brief, if it is correct, then Yakubovich's argument from the relative frequency of typological bundles of features generally supports the present argument, that this was largely a situation of unilateral bilingualism leading to Elamite speakers’ shift to Persian. Yakubovich likewise connects the demise of Elamite with language shift to Persian. This de-emphasises the role of Persian speakers’ learning of Elamite in the contact between the two languages.

There are clear ancient parallels to the scenario that I have been describing as well as modern ones. A. Butts has similarly explained the unidirectional transfer of features between Syriac (the Aramaic of late ancient Syria) and Greek as due to the unequal status of the speakers of the two languages. More than 800 Greek words were taken over into Syriac, and basic Aramaic grammatical words came to replicate the function of Greek grammatical words, but very few Aramaic features, even loan words, were taken into Greek.Footnote ⁶¹ The degree of Greek material in Syriac is one of the specific differences separating the Syriac dialect from the rest of Aramaic. Butts argues that those whose dominant language was Greek seldom learned any Aramaic, whereas Aramaic speakers often learned to function in Greek.Footnote ⁶² The case of Coptic (Egyptian) and Greek in Egypt should probably be explained in the very same way. The Coptic lexicon contains about 5,000 Greek loanwords, including many basic syntactic coordinators, whereas the number of Egyptian words borrowed and maintained by speakers of Greek, barring proper names, is tiny by comparison.Footnote ⁶³ This suggests that those in Egypt whose dominant language was Greek seldom learned Egyptian, and perhaps rarely learned it well, but many Egyptians learned enough Greek to induce profound effects on the Egyptian language. The different cross-effects of Greek and Coptic together suggest asymmetrical, unilateral bilingualism arising from the unequal status of the two language communities.

The foregoing observations about the Persian and Elamite languages may affect the historiography on the ancient mutual acculturation of Aryas and Elamites. De Blois's findings, when van Coetsem's model is applied fully, indirectly suggest social dominance on the part of the speakers of Old Persian and not a pervasively or reciprocally bilingual society forming a harmonious blend. This somewhat accords, happily, with one of Henkelman's newest comments on Persian ethnogenesis: “The ethnogenesis model does not, it should be stressed, project a homogenous cultural landscape as the outcome of acculturation”.Footnote ⁶⁴ The linguistic evidence suggests, however, that Elamite speakers adapted to accommodate the Iranian-language speakers who moved into their territory (at an unknown prehistorical time) by learning their language, and that Elamophones gradually shifted to the use of Persian, but Persians usually did not need to learn Elamite much, at least not enough to affect the Persian language, as generally used, even with loanwords that Persian speakers might have adopted. We should suppose, then, that the ancient, pre-Achaemenian Arya immigrants to Pārsa, whose language soon evolved to be distinctly Old Persian, admitted Elamites into their Iranian-speaking social midst, but Elamite speakers did not bring Iranian-language speakers into their Elamite-speaking social midst in the same way. The Persian language became a vehicle to achieve opportunity for Elamite speakers, whereas Elamite never became such a source of opportunity for Persian-speakers, or at least not for more than a very small number. Again, this suggests a discrepancy in social status.

There are also non-linguistic hurdles for the hypothesis of widespread Persian-Elamite bilingualism (bilingualism by speakers for whom Persian was dominant but who also spoke Elamite). The very coexistence of both languages at the time of Darius I suggests that Elamite speakers and Old Persian speakers were not a fused community without social boundaries between them. If the Persians of the early Achaemenian period were largely bilingual in Old Persian and Elamite, and that was a part of their “identity,” then one should require an explanation about why Elamite did not survive in any part of Pārsa as it did in the foothills of Xūzistān. If it existed, this hypothetical inclusive identity must have changed, with respect to its linguistic medium, at some time, to account for the demise of Elamite, but we lack an explanation for this. Another factor arises from the very multilingualism of the Achaemenian royal inscriptions. This has sometimes been cited as evidence of a multilingual population, but one must be careful about the meaning of the expression ‘multilingual population’. While it is true that the Achaemenids ruled lands of many peoples speaking different languages, as they declared in their inscriptions, so that they ruled a multilingual empire, it is a common mistake of historians to assume that the intended audience of a multilingual inscription must themselves have been multilingual in the languages of the inscription. On the contrary, multilingual individuals need a text only in one language common to all of them. A multilingual text is required only when there is no single language common to the generality of the audience. The multilingualism of the Achaemenian inscriptions, at least the early ones—the later ones being mostly traditional formulaic reiterations of the early ones—indicates an audience of individuals who largely spoke different languages, but not all of them at once. Of course, there were some multilinguals among them, who acted as social intermediaries between the communities, but these were exceptions in the view of the designers of these inscriptions. The distinction to be made is a basic one between bi- and multilingual individuals and bi- and multilingual societies, the two expressions referring to different things. Consider modern scholarship for comparison. Historians and philologists in Europe and America are expected to read English, French, German, Italian, and more modern languages besides. If they did not, then a translation of every important article and book into different languages would be necessary. The individual multilingualism of modern scholars means that we can and do publish our findings in only one language at a time. Mass individual functional (passive) multilingualism in this specific audience of shared texts means that these shared texts can be monolingual in any one of the shared languages. A multilingual text, conversely, such as the trilingual Achaemenian royal inscriptions in Elamite, Babylonian, and Old Persian, suggests the expectation of an immediate audience of mostly monolingual literate individuals, who would convey the meaning to illiterate monolingual persons. If potentially literate speakers of Old Persian were already bilingual with Elamite, there would have been no need of an Old Persian version of the Bisitun inscription. The Elamite would have sufficed. The very creation of the Old Persian version of the Bisitun inscription, as an addition to the initial bilingual Elamite-Babylonian inscription, indicates the existence of plenty of Persian speakers who knew very little or no Elamite (or Babylonian) but for whom the author of the text intended his message. This does not rule out the symbolic importance of using multiple languages in imperial monuments, but it is clear that the contents of the texts, such as that of the Bisitun inscription, were meant to be known. The contents of the Bisitun inscription were deliberately circulated, as the text of the inscription declares itself, and as the Aramaic version extant in fragmentary state and Herodotus's retelling at some removes prove.

Did any Iranophones learn Elamite? Of course, some probably did. Maybe some of the authors of Elamite administrative texts were individuals of the kind that Henkelman and Yakubovich have proposed. But all the factors just discussed give reason to doubt the extent of Persian-Elamite bilingualism in the Achaemenian period. There is no evidence that the generality of Persians of the Achaemenian period spoke Elamite, but there is strong evidence of the reverse. This generally unidirectional bilingualism implies different social statuses for the respective communities. In successive empires ruled by speakers of Persian, Greek, Parthian, and Arabic, the Elamite language was not a vehicle for opportunity for those not raised speaking it, except perhaps for a few scribes for a few generations. Elamite probably disappeared from the highlands of Pārsa first and then from the lowlands of Xūzistān later. In the foothills of the latter country, it persisted longest, against different waves of immigration, and it survived until the end of the tenth century ce, at least, as the speech of what was, in the last centuries of its distinct existence, a marginalised community looked down upon by outsiders.