Now we return to Old Persian and its evolution into Middle Persian. As shown in the first chapter, between about 500 bce and 350 bce the grammar of Old Persian changed radically. The frequent irregularities and omissions in the written inflections of the late Achaemenian Persian inscriptions show that its complex inflectional endings had mostly been lost by the end of that period in the living language. The scribes responsible for these inscriptions succeeded in obscuring the extent of this loss from our view when they copied older inscriptional formulae or spelling exactly. They increasingly often failed in this, however, for they also wrote inflectional endings that were not etymological, misapplied inflectional endings by misconstrued analogy, made mistakes by hypercorrection, and, most tellingly, omitted inflectional endings reflecting their living language. As discussed in Chapter 1, there are also signs of general phonological changes affecting Persian words regularly that show through the inscriptions. Mancini has even suggested that the late Achaemenian Persian writing comprised a somewhat regular system of conventional spellings indirectly representing a phonological state of the language much changed from that of the earliest inscriptions.Footnote ¹ These sorts of changes in morphology and phonology are represented directly in the next attested stage of the language, Middle Persian, apparent fully only in texts surviving from centuries later, written in a different script. The late Achaemenian Persian inscriptions reveal, however, that those changes were well underway, if not substantially complete, by the time of the disastrous invasion of Alexander III of Macedon. The change itself is widely acknowledged among specialists in ancient Iranic languages. In short, we may regard the language of the late Achaemenian Persian inscriptions as Proto-Middle Persian or by a similar term.Footnote ² That is, this form of Persian expresses the earliest specific characteristics of the language we conventionally call Middle Persian.Footnote ³

As seen in Chapter 1, the conspicuous omission in Middle Persian of the complex inflections of Old Persian has drawn many remarks and observations but few explanations. Middle Persian appears to have the simplest inflectional morphology of any known ancient Indo-European language (except for the two most closely related known ancient relatives, Parthian and Bactrian, to be discussed in the next chapter). As I remarked, and as specialists in Iranic languages have observed many times, its inflectional morphology is simpler even than that of the great majority of modern Indo-European languages. How did this happen? This chapter answers that question using the explanatory model set out in Chapter 2.

A Nation Morally Corrupted?

There are wrong answers to the question. Let us deal with an old one now. Different strains of modern nationalism have construed a connection between concepts of linguistic vitality and national strength. Often, this is rooted in the implicit notion that conservatism, or resistance to change, signifies the vitality of a people (whatever that would mean), and that real social or cultural power ought to include the ability to keep culture stable and unchanging. From this point of view, if a language changed, or was simplified, it was losing its strength and becoming corrupt. This was a sign of social decay and perhaps even the gradual weakening of a nation. This way of thinking was common in nineteenth-century linguistics, but it is false. These values are not inherent to language of any kind but arise rather from the individual’s aversion to social change or from romantic dreams of a history of continuity of a nation’s imagined essence as represented by its language. These are the same kinds of attitudes that have made the study of creoles, discussed in the last chapter, a politically charged arena. In an early stage of the study of creole languages, it was normal to dismiss radically restructured languages as uncouth bastard tongues. This judgment was reinforced at times by racist disdain for the populations that lived in colonial sites, whether colonized peoples or imported slave labor populations or mixtures of hitherto unmixed peoples. This old attitude about language change was not peculiar to the study of creoles. Early pioneers in modern linguistics often regarded change from a pristine, ideally unmixed and grammatically complex state as “corruption.”Footnote ⁴ Such terms, blending moral judgment with linguistic description, have long intruded into the study of the late Achaemenian Persian inscriptions. In 1852, Oppert regarded later Achaemenian inscriptions as representing a language in decline (“décadence”) practically as soon as they were deciphered.Footnote ⁵ Meillet regarded the early inscriptions of Darius I and Xerxes as written in a language “coherent and manifestly correct,” whereas the later ones were “very incorrect” and finally “simply barbarous.”Footnote ⁶ For Kent, the late Achaemenian inscriptions represented a “corrupted form of the language.”Footnote ⁷ Olmstead, even less cogently, held that the late Achaemenian inscriptions “betray a degeneration in literary style.”Footnote ⁸ Even for Schmitt, one of the most insightful interpreters of Old Persian, when writing only a few decades ago, late Old Persian had “degenerated” and was “barbaric.”Footnote ⁹ These modern philologists were writing descriptively about language forms and not necessarily about moral change, but the two are easily connected by the use of such descriptions. This was typical of nineteenth-century and even twentieth-century linguistics.

This is a sensitive issue for a few different reasons that are mutually entangled. One pertains to the history of the study of the Achaemenian Empire. Some ancient Greek authors – a few of whom appear later in this chapter – created a narrative of moral decay for their contemporary Persian kings and their subjects that European historians, dependent on Greek sources, have echoed until the most recent generation.Footnote ¹⁰ Ancient Greek authors imagined the Persians, who were an abiding political and military threat to them, as a nation increasingly spoiled by their power and wealth and corrupted thereby in their morals and conduct. These Greeks defended their stubborn independence as the alternative to what they regarded as slavery under eastern despots. While Greek thinkers argued, disagreed, and warred among themselves about the best government for their small-scale states and cities, they used the vastly more powerful Achaemenid Persian dynasty as an example in their own arguments about what worked and what did not, and they defined their people and culture by contrast with those who resided in the immense Persian domain. Modern historians of antiquity have struggled to shake off ancient Greek characterizations of the ancient Persians, which they imbibed from their Greek sources. It was all too easy to connect these old attitudes about the gradual moral corruption of the ancient Persians, which is entrenched in modern and ancient literature, with the “corruption” of their language, as if moral corruption and the reduction of grammar were two aspects of the same phenomenon.Footnote ¹¹

Meanwhile, two more modern tendencies have allied in defending the reputation of the Achaemenian Persians against these attitudes, endowing recent historiography on ancient Persia with different, reactive, recent political values. The last generation of specialists in Achaemenian history made great gains in historical interpretation, largely by using a much wider variety of non-Greek sources and perspectives. In doing so, they thoroughly defeated the Hellenocentric and implicitly anti-Persian historical perspective that prevailed earlier. On occasion, this has been carried out to the point of overkill, leading to caricatures of some earlier scholarship on the ancient Persians and even, occasionally, to their own neglect of relevant Greek sources.Footnote ¹² The aspiration to write Persian history from the tacitly essentialized “Persian point of view,” rather than a more obviously essentialized “Western” one, seems to assume that sources such as royal inscriptions really represent a true, general, national, “Persian point of view” or even “Iranian” point of view rather than a specific Achaemenian royal point of view.Footnote ¹³ Alongside this development, Iranian studies today are thoroughly imbued with strains of Iranian nationalistic sentiment that no participant can overlook. The Pahlavi shahs (1925–1979) and their supporters, the leaders of the present Islamic Republic of Iran, and many individual Iranians today in Iran and abroad have fostered their own selective narratives of “ancient Iran,” in which “ancient Iran” is a glorious property revealing something wonderful about modern Iranian identity, facilitated by silently passing over historical episodes inconvenient to the master narrative of national continuity.Footnote ¹⁴ In rejecting the point of view of Greek sources, historians of the Achaemenian Persian Empire are able not only to appear as opponents of Western bigotry but also to play the part of cultural champions for modern Iranians by defending their claimed ancient heritage from the Eurocentric bias of classical scholars and the judgmental specter of colonial-period Orientalists.

This present study eschews these nonanalytical tendencies in linguistic history. Whatever the causes of the changes that turned Old Persian into Middle Persian, they did not “corrupt” the language, for no language is corrupt or degenerate. Reduction of Old Persian morphology neither simplified the minds of Persian-speakers nor exalted them. It does, however, reveal something important that will affect our understanding of the Achaemenian Persian Empire. Moreover, the antipathy of ancient Greek sources toward the Persians will not obscure the problem but shed light on it.

What Happened to Old Persian? The Linguistic-Historical Model Alone

If we had no other information about the Persians in the period of the Achaemenian Empire apart from the inscriptions, then the linguistic-historical model presented in the previous chapter should, by itself, lead us to conclude confidently that there had been an extended episode of mass adult acquisition of Old Persian. The inflectional morphology of the language was profoundly simplified within five generations, and the only known, amply documented cause of such a change is mass adult acquisition. Under no other known historical circumstances has this taken place. The speed of the change itself is characteristic specifically of mass adult acquisition, rather than merely a situation of prolonged societal bilingualism or multilingualism, in which the drastic reduction of grammatical features is not expected, but rather two-way borrowings would proliferate leading to grammatical convergence. Grammatically reduced, nonnative speaker varieties of languages are ubiquitous in human societies among adults, but what is also quite remarkable here is that a grammatically simplified, nonnative variety of Persian became the common one for subsequent native speakers of Persian, just as Afrikaans became the normal variety of speech for the descendants of the Dutch colonists in South Africa. Middle Persian, however, is much more reduced from its Old Persian source than Afrikaans is from Dutch. That is partly because Old Persian was much more complex in its inflections than Dutch at the outset of contact. Old Persian had more inflectional complexity to lose, making the transformation much more dramatic. The degree of change is nevertheless remarkable.

Middle Persian, as we know it, does not seem to have evolved from a stabilized pidgin variety. Unlike all modern creoles, it retained a little bit of the contextual inflection from the immediately ancestral form of the language, however minimal. Middle Persian verbs have two stems each that are, synchronically, not transparently derived from the other. Three persons and two numbers are distinguished in the present stem of the verb. The preterit is entirely uninflected by itself but requires an inflected copula to correlate a patient outside of the third-person singular. A simple set of subjunctive verb endings, slightly different from the indicative, is used to indicate the expectation of occurrence (futurity or eventuality) or undefined occurrence (as in indefinite relative clauses). Simple optative forms occur in verbs in the third century ce. An oblique case existed alongside the direct case (former nominative). However minimal this inflection was, based on the linguistic-historical model alone we would assume that Middle Persian, as compared with Old Persian, underwent semicreolization, a term that refers not to its status over the centuries of its existence, but to the social process that shaped it. Middle Persian was drastically reduced to the point that it looks close to a creole but without clear evidence of having emerged from a pidgin. We can say with complete confidence that exoterogeny played a decisive role in the formation of Middle Persian.

The case is not so neatly closed, however. The existence of decreolized “acrolects” of creoles, discussed in the previous chapter, which have emerged through the transfer of grammatical features of the lexifier into the creole derived from it, does prompt the question of whether Middle Persian could have undergone a similar history. For example, the singular and plural of the direct (nominative) case are identical in most, but not all, nouns, as illustrated in Table 1.2. Only plural verb inflection or semantic context marks the plurality of a Middle Persian grammatical subject. In these instances, the noun lacks contextual inflection that cross-references the plurality of the noun phrase with that of the verb. Creole languages do not inherit contextual inflection from their lexifiers. In this specific respect, Middle Persian resembles a true creole. Yet a small number of nouns, like the word for “father,” do exhibit a distinct singular nominative (direct case): nom. sing. pid, nom. pl. pidar. Inherited contextual noun inflection therefore does exist in Middle Persian, at least marginally. For another example, of the two verb stems of Middle Persian, only the present stem takes personal endings, as just mentioned. These present-stem endings are strange, from the point of view of early Old Persian, in deriving from the endings of one verb class of several (the one with the formant -áya-), even on the most frequently used verbs (like “to be”), creating an entirely regular, agglutinatively inflected verbal system. Purely hypothetically, this could represent the reapplication of endings generalized from one class of verbs, in a process of decreolization of verbs that had become entirely uninflected in a pidgin form of Old Persian. Only if this sort of speculation were to be verified, then Middle Persian would cross the line from a semicreole into the territory of a true creole in acrolect or decreolized form. It seems impossible to verify or refute this hypothesis, however, so it is not worth pursuing here further. Although it is practically certain that a Persian pidgin of some kind existed at least at royal sites (to be discussed), any popular pidgin based on an Old Persian lexifier is inherently very unlikely ever to have been recorded. With the limited sources of information at our disposal, the more analytically secure solution is to understand Middle Persian as a semicreole, being drastically reduced in its inflectional morphology without having emerged from a prior widespread pidgin, because it does retain some contextual inflection (inflection governed by syntax). If we consider Holm’s demographic data on semicreole formation charted in the last chapter, this may give us a vague sense of the proportion of native and nonnative speakers of Persian in Persian households and settlements in the fifth century bce: something like half and half. This chapter will show that that estimate is quite plausible.

While Middle Persian exhibits extraordinary grammatical simplification in inflectional morphology, its lexicon offers hardly any words at all borrowed from the languages of the subject nations of the Achaemenids.Footnote ¹⁵ Critically, the near-absence of words adopted from other languages in Old as well as Middle Persian reminds us of the second part of the linguistic model, concerning the agency of speakers in the transfer of features such as words from one language to another. It indicates that few ancient Persians (native speakers of Persian) were highly proficient in foreign languages during the Achaemenian period or perhaps in any ancient period. Like nonimmigrant inhabitants of the USA today who know only English and lack a motive and a social opportunity to learn any other language at a young age, and have little incentive to do so because of the global status of English as a lingua franca, the ancient Persians enjoyed the restricting, dubious benefits of unilateral multilingualism: many others had need of their language, but they had little need of others’ languages, and so they spoke just Persian, expecting to be understood sufficiently when they spoke. There is, however, one clear exception to the pattern. The presence of some not strictly Persian Iranic words attested as adopted in Old Persian and Middle Persian demonstrates that some ancient Persian-speakers knew at least one closely related Iranic language. Philologists call such words Median when discovered in Old PersianFootnote ¹⁶ and Parthian when discovered in Middle Persian.Footnote ¹⁷ What such words really have in common is that they are Iranic but not Perside. “Perside” is a term of art used to refer to the languages bearing features (usually phonological) reflecting innovations found in Persian and related dialects but not in other Iranic languages. One example is the outcome of Proto-Iranic *ź /dz/ in Persian. In most Iranic languages, this evolved into /z/, but the regular Persian outcome was /d/. As illustrated with a few examples in Table 3.1, however, some Persian words exhibit /z/ where a /d/ is expected by the regular development. Such words in Old Persian have been assumed to be adopted from “Median” because the Achaemenian Persians were political successors of the Arya Medes, making Median (a language scarcely attested directly) the only Iranic language of political significance before Persian. Naming words of this kind adopted in Persian as either Median or Parthian is a convenience, because we do not really know in all cases that they were specifically from one of those two languages, but just that they come from another Iranic language besides Persian.Footnote ¹⁸

In this chart one can see that the Old Persian words in three rows exhibit the so-called Median or simply non-Perside /z/ instead of the expected /d/. We see also that in the Middle Persian, other “Median” or “Parthian” forms with /z/ rather than the inscriptional Old Persian reflex /d/ occur. One could multiply this with more examples in which other unexpected phonemes exhibit traces of non-Persian Iranic borrowings. We need not dwell on this longer here, however, as these phenomena have been discussed adequately in reference works just mentioned in Footnote note 18. The main point for now is that they demonstrate that some Persians spoke other Iranic languages; if they were not familiar with other Iranic languages, they would not have adopted words or pronunciations from them. The near total absence of words in Old and Middle Persian adopted from other languages, including the languages of their non-Persian subjects, is therefore even more striking.Footnote ¹⁹ Few Persians knew the languages of most of their subjects; otherwise, they would likely have adopted at least some of their words, beyond proper names. Yet some Persians did know other Iranic languages closely related to Persian.

The combination of the linguistic-historical model of Chapter 2 with the linguistic facts about Achaemenian Persian reviewed in Chapter 1 diagnoses a situation in which many speakers of other languages all at once had to cope with people who spoke Persian, and to learn some of their language immediately as adults, and that the Persians did not have any need of learning the languages of their subjects: a situation of unilateral bilingualism. Nevertheless, some Persians spoke other contemporary Iranic languages, accounting for the “Medianisms.” This is what we should assume about the early history of Persian from the linguistic model alone. As it happens, many historical testimonies support this interpretation. The remainder of this chapter will survey those testimonies and discuss their ramifications. The subsequent chapter will discuss the relationship of the socially induced restructuring of Old Persian to other, closely related Iranic languages that show similar characteristics. Before that, however, it is necessary to address a major problem of method and perspective.

Purely Structural Linguistic Alternatives?

Historical linguistics provides descriptions of language change over time without taking social factors much into account. Whether that is enough of an explanation of language change or not depends on the kind of history one wants. I posited in the Introduction that linguistic history should be distinguished from solely structural analysis by its primary and complementary attention to the history of a language’s speakers. Taking social factors into account requires the use of contextual primary sources as well as linguistic data. It is possible, however, to construe language change as motivated entirely from within the structure of a language itself. That is what historical linguists normally do. From a strictly descriptive point of view, this is valid. Although many words and word-forms extant in Old Persian are still debated by specialists, the tested methods of historical linguistic analysis of Old Persian work most of the time and have elucidated Old Persian remarkably well. These methods, unfortunately, offer historians almost nothing. The gap between philology and history is not only institutional but a matter of method, too. Before going on, then, let us consider some ways in which this stage in the evolution of Persian has been explained before as the outcome of purely linguistic phenomena, meaning that no social factors play a role in these explanations.

Gernot Windfuhr, a leading specialist in the typology of Iranic languages, has posited a regular pattern of increasing and decreasing morphological differentiation as “typical” of the history of languages. Over centuries, noun case systems would bloom, collapse, and bloom again in a cycle, apparently by the nature of language itself.Footnote ²⁰ At first glance, an example of this would be Balochi, an Iranic language spoken today mostly in southeastern Iran and southwestern Pakistan. Balochi, which bears a very close family resemblance to ancient Parthian, exhibits the same direct and oblique noun cases found in most of the other Iranic languages to the west, but has developed new noun cases along with other inflectional novelties.Footnote ²¹ As mentioned, however, research has shown that the number of noun cases in a language tends to hold an inverse correlation with the proportion of nonnative speakers, so it is clear that this is not merely a cyclical phenomenon.Footnote ²² It is rather a matter of population movement and patterns of language acquisition over the long term. Balochi represents an instance in which an ancient episode of grammatical reduction brought on by population mixture, like the kind under discussion here for Persian, was followed by a subsequent period in which speakers of the language migrated to the south and east and pursued a pastoralist economy in relative social isolation.Footnote ²³ During a long subsequent period of relative social insularity for speakers of Balochi, the inflections of their language became more complex in ways unrelated to the common features of Iranic languages. The inflectional patterns of Balochi reflect not naturally determined cyclical developments but changes in the history of the society of its speakers. I will take up other examples like this in the following chapter.

Geoffrey Haig suggests that the Iranic languages were each somehow preconfigured or programmed to undergo the same sorts of changes because, deriving from a common ancestor language, they reach “remarkably similar” syntactic structures, “although over the course of time the speakers became separated by thousands of miles, and came to speak mutually unintelligible languages.”Footnote ²⁴ The idea that the Iranic languages, or any group of related languages, are genetically predisposed to undergo certain kinds of change by virtue of their cognate relationship is controversial and unproved.Footnote ²⁵ Of course, the common ancestor of the Iranic languages had specific characteristics, and changes affecting daughter branches of Common Iranic were all changes relative to that shared prior state. A language’s grammatical history is, in effect, the descriptive register of the retention, transformation, and omission of the characteristics of the language at an earlier time, as each new generation acquired it, replicated it, or employed it differently. Although grammaticalization does appear to be mostly unidirectional in all languages,Footnote ²⁶ there is no reason to think that the Iranic languages were all bound to experience the same outcome by inherent tendencies, and, in fact, they did not.Footnote ²⁷ Thomas Jügel provides a relevant example in charting the different development of the verbal systems of New Persian and Yaghnobi, and the comparison he illustrates between them is revealing, contradicting the notion of inherent grammatical destiny. Yaghnobi is a modern descendant of an ancient Iranic language closely similar to the Iranic language Sogdian (discussed in the next chapter) that survived in a remote mountain valley of Tajikistan. While its nominal inflection is quite reduced, its speakers retain the ancient Common Iranic imperfect verb, even with the late Proto-Indo-European temporal augment prefix a-, used now in Yaghnobi for the simple past (aorist) function and, with new suffix, for continuous past time, in addition to the forms of the verb retained in Middle Persian: the present verb stem and a preterit formed from the ancient perfective verbal adjective in *-ta-. Persian, by contrast, like most other modern Iranic languages, lost the ancient imperfect verb already in antiquity, leaving New Persian to develop new past-time uses of the verbal stem based on the ancient verbal adjective in *-ta-.Footnote ²⁸ See Table 3.2.

The different social histories of the speakers of Yaghnobi and those of New Persian correspond loosely with the kinds of verbal systems they have inherited. New Persian has been a lingua franca, and it is derived from an ancient language that had undergone semicreolization, Middle Persian. Its verbal system is built on the basis of a drastically simplified ancestor. Yaghnobi’s ancestor was spoken in inaccessible mountain valleys of the Zarafshan range, where relatively few visitors ever came, over many centuries as the sole surviving descendant of a once substantially widespread Iranic language.Footnote ³¹ Its verbal system, structurally, resembles that of Old Persian more closely than that of Middle Persian. The language of the isolated population preserved and developed the ancient verbal tense called the imperfect, whereas the common, formerly cosmopolitan dialects had lost that verbal tense already in ancient times. It is therefore not genetic predispositions in the structure of a language but social relationships with newcomers, or lack thereof, that condition this kind of linguistic change. A pattern of preprogrammed attrition and growth may seem to be typical, as Windfuhr supposed, because the Middle Persian verbal system represents the basic type from which most modern Iranic languages have developed their respective verbal systems, characterized by a great reduction of forms long ago followed by subsequent development in varying conditions. An exception like the verbal system of Yaghnobi suggests, however, that something else is at work: social changes conditioned by demography affecting the evolution of languages. We do not have an exact science to say why Yaghnobi’s nominal system is greatly reduced whereas its verbal system conserves archaic features preserved nowhere else today – just as we do not have an explanation for why its phonology turned out just as it did – but the point is that the idea of a preprogrammed grammatical fate for the Iranic languages, just because they are Iranic languages, cannot be maintained.Footnote ³²

Philologists focus on the purely structural features of language as they change over time to the point that they tend to treat languages as living entities separable from their speakers. When linguists discuss “languages in contact,” rather than the use of multiple languages by one person or in one social setting, they too tend to treat languages as separable from users. This is the very gap that necessitates linguistic history. Languages have no agency and are not entities that can come into contact independent of speakers. Languages do not have “self-preserving power” that can become “diluted and weakened” resulting “in their accelerated development or rather decay,” as Szemerényi put it in his discussion of the origins of Middle Persian.Footnote ³³ It is speakers who reproduce languages through learning and who bring their languages into contact, particularly when the brains of individual speakers harbor more than one of them at once. The method of the historical linguist has been to explain changes in a language by terms descriptive of processes that operate on patterns, such as analogy, without regard for the agency of speakers who make choices and errors for mental and social reasons. This has often been necessary, because human choices and agency, normal considerations for historians, are beyond the reach of most of the established methods of comparative historical linguistics. Nevertheless, while languages’ inflectional systems may change by analogy (for example), it is the speaker that enacts grammatical analogy in a language, consciously or not. If others repeat an individual’s linguistic novelty, an idiosyncrasy may be propagated and adopted by other speakers to become a general feature of the language. It is typically only then that the features appear in written texts available to us, and by the time those texts were written, the inception of the changes was in the unrecorded past.

As pharmacists offer drugs to solve medical problems and surgeons offer the scalpel, historical linguists are predisposed to find phonological solutions to linguistic problems, because the phonological isogloss is rightly their fundamental criterion. The discovery of the regularity of sound change was the foundation of diachronic linguistic research, one of the most important modern insights into the nature of language itself. For all their importance, however, sound laws, being descriptive, do not tell us why a language has changed, but only how the forms and structures differed regularly from period to period.Footnote ³⁴ Along these lines, the standard explanation for the loss of inflectional endings in Iranic languages like Persian is that it was simply a matter of the position of a strong stress accent.Footnote ³⁵ Iranic language-speakers are thought to have stressed certain syllables so heavily that vowels in specific post-tonic phonological conditions were syncopated and the final syllables were dropped. Note that the position of the stress accent is inferred by the loss of syllables, not the other way around. In practice, the argument has been to explain the evolution of Persian by drawing parallels with the evolution of Latin and its Romance descendants.Footnote ³⁶ These parallels are real, but pointing to parallel phenomena is explanation by comparison, not by positing a cause. The loss of final syllables in Persian eradicated much of the inflectional morphology, blurring or eliminating distinctions between noun classes and grammatical gender, the distinction between the present and imperfect verbs, built on the same stem but with different verb endings, and in all the other ways discussed in Chapter 1. Understood as a natural and practically inevitable phonological change, the loss of final syllables can even be construed as triggering the rise of the “split-ergative” verbal system common to Middle Iranic languages by leaving only one distinct medio-passive past-time construction standing along with two noun cases, direct and oblique. In short, this purely descriptive explanation based on phonological factors alone indicates that the inflection of Iranic languages came to be extremely pared down, and the verbal system reorganized, just because of a consistently emphatic manner of intonation. One regular sound change would eventually transform the entire grammar.Footnote ³⁷

Such an explanation looks clean and effective, with a single phonological solution to a very complex problem. It is also descriptively correct, as far as it goes, and I endorse it as such. That said, it is insufficient to the point of being misleading. It is not that the stress accent caused the loss of subsequent syllables. It was, however, a major condition of their loss that followed a regular pattern.Footnote ³⁸ Not all languages drop or syncopate post-tonic syllables. Some do so, but only after hundreds of years of intergenerational replication and apparent stability. Why do some speakers of languages reduce post-tonic syllables while others do not? Why did they lose those syllables at the specific time in which they did, and not sooner, and not later? The actuation of language change has been extensively discussed for decades; it was one of the main questions that motivated the genesis of sociolinguistics.Footnote ³⁹ As Rudi Keller remarked in his treatise on explanations for language change, “Historians of language change have traditionally focused on the aspect of change, perhaps tacitly assuming that ‘Where nothing changes, there is nothing to be explained.’”Footnote ⁴⁰ That is a limit of the structural argument about language change. Stasis remains unexplained without taking social factors like those I discussed in Chapter 2 into account. The apparent arbitrariness of the actual occurrence of a generally common tendency to drop post-tonic syllables and its timing must be understood not just as a matter of accentuation but also as the product of speakers’ choices and habits and their accommodation to circumstances, including differences in language acquisition. For all we know, the assignment of the stress accent to the received position was simultaneous with the loss of final syllables, not its cause. Ancient records do not otherwise indicate the position of the stress on Old Persian words, but we know that the accentual outcome was not inherited from Proto-Indo-Iranic. The two changes – reassignment of stress and loss of syllables – could have been concomitant, enacted by the very people who dropped final syllables. Researchers like William Labov, who has painstakingly scrutinized phonological change in living progress, have demonstrated that social factors play a primary role in sound changes that originate on idiosyncratic, personal scales before adoption by larger groups, even as the sound changes that are successfully propagated ultimately do follow regular patterns on a large scale that can be described as rules.Footnote ⁴¹ The social subtleties arising in the communication of individuals with different social roles are usually inaccessible to us who study ancient languages, but, as shown in Chapter 2, large-scale social factors can and should be part of our large-scale analysis. Mass adult acquisition of a language has predictable effects on that language, indeed more predictable than the consequences of a stress accent on its phonological environment. Social factors allow reliable inferences pinned to knowable historical events. Mass nonnative acquisition is the only known condition under which contextual inflection is drastically reduced in a language. The correlation is so strong that the reverse is true, too: those predictable effects on language allow reliable inferences about the history of a language’s acquisition. In other words, social and phonological explanations for language change need not be mutually exclusive. They cannot be mutually exclusive. Both rely on the application of expected tendencies to explain empirical phenomena that are attested on a scale that represents a society. A more comprehensive explanation of systemic language change will consider both our reasonable expectations about what commonly happens to the sounds of languages over time (such as palatalization of consonants before front vowels, or the lenition of postvocalic occlusives)Footnote ⁴² and the social circumstances in which language change of different kinds is facilitated by living speakers responding to their personal circumstances. One may easily regard the omission of all final syllables in Old Persian, most of which were markers of inflection, as a necessary ad hoc choice to simplify a language made by ordinary speakers, people who had no concept of linguistics. Separating the fixation of the stress accent and the utterance of words without final syllables may be impossible, because, as just suggested, they may have occurred simultaneously, the stress shift being enacted when the syllables were dropped, perhaps as its audible cue.Footnote ⁴³

It is clear, however, from the detailed research of scholars such as Sims-Williams and Korn that the loss of final syllables in Old Persian did not occur overnight. This is to be expected when many native speakers of Old Persian were still around. Rather, some ancient Persian final syllables underwent stages of phonological reduction prior to complete omission. This is discernible through later traces. Sims-Williams demonstrated that holdovers of Old Persian final vowels were sometimes preserved in Middle Persian words before bound enclitics (words that are attached to preceding words in a single accentual unit).Footnote ⁴⁴ In a few instances, the combination of inflected form and enclitic may have been in effect lexicalized. For example, the attested Middle Persian form gawā-m <gwʾm> “tell me,” evolved from Old Persian *gawbā-mai. The vowel -ā- here, originally an inflectional ending for the imperative singular, was not lost, despite being the final syllable of a word. It was preserved at least occasionally in this expression dating back to Achaemenian times – whereas the final diphthong of the enclitic -mai was lost, as usual, as a word ending. This is not surprising, as the command “give me!” must have been frequent enough to be treated as a unit. Compare the frequent English vernacular “gimme!” with the English enclitic pronoun “-me” (where the outcome is not the preservation an archaic vowel, but elision of the /v/). Other examples, likewise preserved in Middle Persian before enclitics, show a different phenomenon of change before loss: the endings -im and -am changed to -u before the vowel disappeared entirely.Footnote ⁴⁵ The Middle Persian enclitics thereafter sometimes could occur in a form including the preceding vowel -u, now a linking vowel, etymologically derived from the inflectional ending of the preceding word, but later treated as a vowel accompanying the enclitic. Thus Manichaean Middle Persian <qyrdwš> kerd-u-š, “he made, it was made by him,” would have evolved from Old Persian *kǝrtam-šai, where Sims-Williams regards the -u- linking vowel as a trace of the ancient neuter singular nominative ending -am. Although these vowels survive only because they are preserved at the ends of words before enclitics – to my knowledge, never in Persian without enclitics – it is not unreasonable to assume that the change was taking place even in the absence of an enclitic, before or during the loss of all final syllables. That is, kǝrtam was becoming *kǝrtu before the final syllable was lost. These are examples of what Sims-Williams dubbed “isolated archaism” in Middle Persian, and they are valuable for demonstrating that final syllable loss was not a uniform and instantaneous phenomenon simultaneous for all users – as variationist linguistic studies of sound change in living progress also have led us to expect.Footnote ⁴⁶

A remarkable challenge to the present argument emerges from the fact that final syllables were omitted from all words, even uninflected words such as conjunctions and discourse markers. If the issue was that mass acquisition by nonnative speakers will generalize the omission of inflectional morphology, why should uninflected words lose final syllables? This is provisionally best understood as a process of analogy, one of the mainstays of explanation in historical linguistics.Footnote ⁴⁷ When a process of change should affect one instance or several instances but speakers extended it to all other parallel instances, that is dubbed analogy. If, according to the descriptive phonological approach, a strong stress on penultimate or antepenultimate syllables rendered final syllables omissible – at a time that could not be predicted by phonology alone, prompted by unspecified factors other than “it just happened”Footnote ⁴⁸ – then for which speakers were they most omissible? There is no motive for native speakers to have enacted this sweeping omission. We know that native speaker children among native speaker adults generally learn the inflections used by adults quite well, because languages with highly complex morphology continue to exist. Bentz and Winter also verified an inverse correlation between the number of nominal cases in a language and the proportion of L2-users.Footnote ⁴⁹ In light of this and everything discussed in Chapter 2, by far the most likely culprits for omission of final syllables are nonnative learners of the language, who benefited pragmatically from such omission, or for those attempting to communicate with them in foreigner talk. If speakers were already propagating a pattern of pronunciation whereby the final syllables on most words – nouns, pronouns, and verbs, including some adverbial inflections – were reduced and then omitted, it is reasonable to suppose that speakers carried out this pattern of clipped pronunciation indiscriminately by analogy on small classes of uninflected words regardless of their grammatical categories, being unconscious of such grammatical categories. It is already by analogy that specialists explain the generalization of the plural oblique noun ending -ān (from genitive plural -ānām after it lost its final syllable) to nouns that belonged historically to noun classes where that ending is not etymological.Footnote ⁵⁰ I must admit, however, that we are reaching an epistemological limit of diachronic linguistics here.Footnote ⁵¹ While variationist sociolinguistics has shown how speaker choice and conformity in relationships drive the widespread acceptance of a phonological change, the interaction of speaker agency and the omission of a language’s features is an overlooked topic.Footnote ⁵²

It is not new to propose that multiple explanations coexist validly for one phenomenon. During the reign of Artaxerxes III, the philosopher Aristotle distinguished different kinds of causes, or reasons for things becoming as they are. Two of his four causes are known as “formal” and “efficient.” A stool at a bar requires at least three legs and a stable, flat surface for sitting. That shape, which is one inevitable reason that it is a stool, is its formal cause. In this sense, its form is the cause of its being a stool. But there was also a carpenter who fashioned it, who caused it to be in another sense. The carpenter is the “efficient cause” of the stool. In the evolution of Middle Persian, the position of the strong stress accent is a part of the formal cause of the reduced inflectional system. The strong stress accent can be linked, by regular rules that describe the pattern of sound change as it occurred, with the loss of subsequent syllables. There is a formal correspondence expressing a determined condition. But there were living speakers who reassigned accents in Iranic languages, giving simple uniformity to a complicated, arbitrary, and various inherited accentual and inflectional system, as they reduced the syllables following those accents. They did so in an effort to communicate. These people were something like an efficient cause of the reduction of Persian inflection. In this way, descriptive (formal) and social (efficient) causal explanations for language change – even if they are not exactly as Aristotle proposed in his theory of causation – must coexist. Both are valid and entail different kinds of causation. Understanding both simultaneously bestows greater comprehension of the problem of language change in history.

Historians do not hesitate to accept that major events, such as wars or famines, do not have just one cause. Philosophers have debated the nature of scientific explanations and how they interact. It is reasonable to conclude, with philosopher of science Chrysostomos Mantzavinos, that explanatory pluralism is not only possible but should be embraced, for different explanatory games each have their own rules without contradicting each other.Footnote ⁵³ In linguistics, Haspelmath has argued convincingly against partisanship between different linguistic methods that “different explanations in grammar are mutually compatible.” Structural explanations are not necessarily contradicted by functional explanations or biocognitive explanations.Footnote ⁵⁴ In this study of the evolution of ancient Persian, we have all three of these kinds of explanations operating at once.

• Structural explanation: the placement of a heavy stress accent correlated with the reduction and eventual omission of the final syllable of every word. This change follows a regular pattern and has parallels in the evolution of Latin and its descendants. This is the explanation usually offered, partly because it has been the easiest to apply, requiring only intuition and a minimum of linguistic comparison.

• Functional explanation: nonnative speakers and their Persian interlocutors omitted most inflectional syllables because it helped them to communicate more easily by drastically reducing the number of forms needed. The reduction of inflectional morphology induced by mass nonnative acquisition is amply demonstrated in languages around the world.

• Biocognitive explanation: adult humans generally do not learn contextual inflectional morphology as well as children. This is attested universally.

None of these explanations need stand alone.Footnote ⁵⁵ The linguistic history of Persian is more robustly supported by all three at once and is insufficiently explained by just one. The functional and biocognitive explanations are, moreover, necessary for linguistic history, which follows the history of languages as used by speakers.

Already in 1906, Antoine Meillet foresaw that this multiplex kind of explanation would be necessary as linguistics developed into a fully fledged field of inquiry. He anticipated the need for social and biocognitive explanations together with structural linguistic ones.

Imperial Persian, Heterogeneous Persians

The rapid transformation of Old Persian into a radically simpler form of speech was due to the very success of Persian-speaking rulers in dominating so much of the world. Meillet, author of the passage just cited – the same scholar who excluded the late Achaemenian Persian inscriptions from his grammar because they were “incorrect,” who deemed the latest inscriptions “simply barbarous,” and who insisted that the only possible explanation for the conspicuous palatial inscription of Artaxerxes III (A³Pa) was that a foreigner had written it – was also the first scholar who understood and expressed the cause and the social context of the changes in ancient Persian correctly. In an essay of 1912, published in a semipopular French scientific and literary journal, Meillet reflected on the recent discovery of Middle Iranic and Tocharian texts in Central Asia, which was shedding new light of the history of the Iranic languages and Indo-European generally. About the ancient Iranic languages, he pointed out that, “In no part of the area of Indo-European was the evolution [of the daughter languages] so rapid as in the Iranic. … The speech of the Sasanian period is of an entirely modern type.”Footnote ⁵⁷ His explanation was a surmise:Footnote ⁵⁸

He added that whatever forces had preserved linguistic standards were soon destroyed with the advent of Alexander. This last point is somewhat unclear, but it seems that he is referring to the orthography of Old Persian inscriptions. The veneer of the formulaic inscriptions was lost and the real, transformed character of Persian would be henceforth evident, whenever it would be written thereafter.

This brilliant intuition of Meillet was rooted in his comprehensive familiarity with the linguistics of his time, the engine of which was la grammaire comparée. Despite his correct hunch, however, Meillet was not able to explain the processes by which language “imposed” as a common tongue might lead to its very rapid evolution into a “modern” type of language.Footnote ⁵⁹ That is because several of the subfields of linguistics, findings of which were synthesized in the preceding chapter, had not even come into existence in Meillet’s time.Footnote ⁶⁰ The industry of research devoted in the last several decades to second-language acquisition, bilingualism, language contact, grammatical transfer, creoles, and related problems had not yet been conceived when he wrote this. Nevertheless, Meillet’s solution, almost forgotten today and only once partially endorsed,Footnote ⁶¹ was basically correct: Old Persian – and much of Old Iranic, a point to which I will turn in the next chapter – was transformed when it became a necessary medium of communication for many nonnative speakers gathered at Persian-dominated sites from across the vast territories ruled by the Achaemenid Persians. It metamorphosed abruptly in the direction of inflectional simplicity through mass adult acquisition over a few generations. Perhaps this is something close to what Meillet had in mind when he said that the latest inscriptions were barbarous: they reflected nonnative, foreign acquisition. Whatever he really had in mind, Meillet seems to have abandoned or omitted the idea when he published his grammar of Old Persian a few years later. In any case, now, with the benefit of great advances in the understanding of the interaction of society and language, we can be much more precise about the motivating social context for the phenomenon that Meillet intuitively understood.

Meillet’s intuitive solution was mistaken in one subtle but important way. Old Persian was not imposed on the subjects of the Achaemenids as a policy decision. As has been well noted, the local languages of the subjects of the Achaemenids continued in use and in writing.Footnote ⁶² The Achaemenids did, however, create a need for many others to learn to speak the Persian language with them. When the Persians and their close Arya allies ruled many other nations, they integrated them, over seven generations of kings, in a vast, durable, ideally unitary network of dominion and interregional contact and exchange, presided over by “one king of many, one commander of many,” as Old Persian inscriptions of Darius, Xerxes, and Artaxerxes III declare.Footnote ⁶³ We call this an empire. At least two subject regions, Egypt and the Ionian coast, were perpetually restive and prone to rebel, but most of the remainder, from Anatolia and Judaea to Afghanistan, stayed within the realm of the Persian Achaemenid monarchs, even regions that were sometimes almost independent in practice. One result of the unprecedented power of the Persians was a degree of Persification, meaning acculturation to the Persians, and Persianization, meaning the acquisition of some manner of Persian speech. These two allied processes, which should be distinguished from each other, did not much affect the vast majority of the estimated 17 to 35 million regional subjects of the Achaemenids, but rather it mattered most to the elite ruling-class culture of the Achaemenids’ delegates, friends, and family that the subjects supported and toward which many of them gradually assimilated.Footnote ⁶⁴ Speaking Persian mattered near the top of the social pyramid. Fostering an interregional ruling-class culture with common values and expectations is itself a source of social power besides typical imperial strategies such as military conquest, direct rule, and rule through clients.Footnote ⁶⁵ Some subjects found opportunities to strive upward socially toward the status of the Persians. Being Persian meant being identified with the ruling regime, but for this reason the ruling regime was not unchanging. It was gradually transformed as subjects actively assimilated themselves to Persian ways. There are always those who aspire to reach new levels of prestige, wealth, and power, and pursue them wherever they are located, opportunity allowing. Many people imitated the manners of the Persian elite – ways which thereafter became synonymous with the behavior of rich rulers, for whom the Persians became the generic model in every country they ruled. Likewise, the bureaucrats and servants of the Persians had to understand their commands. Quite a few of the same Old Persian expressions are found calqued in writing in Aramaic, Elamite, and Babylonian in the Achaemenian period, reflecting the common reception of Persian statements by literate members of these different language communities.Footnote ⁶⁶ This unidirectional borrowing diagnoses a situation of unilateral bilingualism: Persians did not need the languages of others, but others needed to understand the Persians.

What did it mean in practice to be Persian or to do things as the Persians did? From the early fourth century bce we have the explicit testimony of Xenophon, who briefly served a Persian satrap, closely observed Persian aristocrats, and traveled as a mercenary fighter within the Persian Empire. He describes the imitation of Persian rulers as a custom he believed to have been ordained by Cyrus, the renowned founder of the imperial dynasty.

Whether his account of the origin of this custom with an edict of Cyrus is correct or not, it describes what was believed, and the practice Xenophon directly observed, in his own day. Each man who was participating in Persian rule seemed to be expected to imitate (μιμεῖσθαι) the Persian he served.Footnote ⁶⁸ The locus of imitation was at courts, arranged in a hierarchy of power. Each satrap’s court was to mirror the great king’s court, and young men of landholding families were expected to be socialized in that environment of shared elite pursuits. Of course, speaking Persian was a basic criterion and requirement for participation. Without a common language, social interaction in any court setting would be extremely restricted. It is difficult to maintain a meaningful relationship solely through a translator. The emphasis on raising boys in this environment suggests that language learning was an implicit part of the education entailed; adults do not learn to speak languages as well as youngsters do. Xenophon did not need to state explicitly that imitating Persian satraps meant at least attempting to speak Persian. There are, nevertheless, clear testimonies in Greek and other sources to bilingualism with Persian and the learning of Persian by adult individuals, most famously by Themistocles of Athens (circa 529–455 bce) in political exile. When he sought refuge with King Artaxerxes I, he asked for a year to learn “whatever he could” of the Persian language and was then admitted to the court. Of what use could he be if he could not communicate with those around him in the Persian setting? He lived out the rest of his life under Persian rule.Footnote ⁶⁹ We can only imagine how many other individuals like Themistocles there were from lands ruled by the Persians or from their frontiers, who came to the court of the king of kings to continue a life in association with power. Assimilating to life among Persian masters must also have meant adopting Persian attire and traditions and participating in the customs that arose in the new political reality of Persian mastery. What such participation required, beyond imitation, learning to speak at least some Persian, ostentatious and obsequious service to Persian kings and lords, and receiving magnificent rewards in return, is not clear, because sources scarcely inform us about Persian culture before the Achaemenids. Certainly, as Xenophon wrote, for the men who wanted to socialize with Persian elites, it entailed avid horsemanship and interest in the gear necessary for itFootnote ⁷⁰ and martial prowess with spear and bow – skills practiced in frequent hunting. It also probably meant the conspicuous profession of certain moral ideals such as honesty and self-restraint,Footnote ⁷¹ although these were also values held in common with other ancient peoples. At least from the time of Darius I and Xerxes, it could perhaps entail support for the cult of Ahuramazdā and patronage of the magi, the priests presiding over that god’s cult, something held in common with some other Arya peoples,Footnote ⁷² but it is also clear that the early Persian kings and their immediate followers supported the cults of other gods.Footnote ⁷³

More particularly, whatever the Persian rulers and elites adopted from their subjects became Persian in effect because it now belonged to them. It is not that there was ever a Persian essence, before or after the Achaemenids. New peoples and nations are always taking shape; old ones are always evolving. This applies to the Persians at the rise of the Achaemenids, too. Some historians of this period hold that the Persian people from whom Darius I emerged were themselves the product of a hybrid that took shape in the immediately preceding centuries of Arya settlement among Elamite-speakers. The degree and manner of their intermixture and cohabitation before Cyrus and afterwards is a subject of debate.Footnote ⁷⁴ “Persian” came to be, in effect, the term that designated this specific “hybrid” as it developed. Recent scholarship strongly emphasizes the preexistence of Elamite language and culture in pre-Achaemenian Persis. Elamite was the first language of the Bisitun inscription of Darius I and a major language of the administration of Persepolis during his and his successor Xerxes’ reign. The name Cyrus was an Elamite name, Kuraš,Footnote ⁷⁵ and Cyrus was designated as king of Anshan, an Elamite highland region, in documents contemporary with him, an office he inherited.Footnote ⁷⁶ The earliest Achaemenids are supposed to have worn Elamite formal attire, with a fringe of tassels, but the royal fashion changed to Median robes “after 538, but before work started on the Persepolis reliefs.”Footnote ⁷⁷ While some scholars, such as Henkelman, have envisioned a fairly seamless blend of Elamophones and Iranophones giving birth to a vigorous new ethnic group that created an empire, those who regard the Elamite and Persian cultures as distinct components in close contact have given reason to identify Darius I as a transition-figure from predominantly Elamophone rule to Arya Persian rule, a transition obscured by later history as the Elamite background was forgotten and Persian identity was projected back to prehistory. Cyrus, then, would not be a Persian in the sense connoted at a later time, but rather he would come from an Elamite-speaking background, rising to power in service of the Medes.Footnote ⁷⁸

Less controversially, several textual signs point toward linguistic effects when Elamite-speakers started to learn Old Persian even before the time of Cyrus. Tavernier’s detailed study of Iranic names in pre-Achaemenian Elamite administrative texts dating circa 590–555 bce shows that some of the phonological changes later characteristic of Middle Iranic were taking place, at least in one subregion, already a few decades before the Bisitun inscription.Footnote ⁷⁹ These include monophthongization of diphthongs (e.g., ai > ē),Footnote ⁸⁰ contraction of vowels separated by evanescent consonants (e.g., iya > ī, ahya > ē), and loss of final vowels or short unstressed vowels between two syllables. This may be due to “more developed” “Elamite-Iranian acculturation” in the locale from which many of these texts come.Footnote ⁸¹ The omission of final vowels on a small number of Persian proper names in Elamite texts from the decades before Cyrus suggests that this phenomenon is due to nonnative use. We should expect to see more changes in Old Persian where more individuals learned it as adults as a second language.Footnote ⁸²

The goal here, however, is not to resolve a debate about the ethnic makeup of the Persians at the beginning of the Achaemenid dynasty. The main point is that this was always changing and that there is no such thing as a pure and unchanging ethnicity. Whatever the reality, and however Elamites and Persians had intermixed, in the time of Darius I the outcome was still local and regional. Pārsa was a specific region. The drastic change came with the conquests of Cyrus and Cambyses and the consolidation of Darius and Xerxes. Then Persian people mastered millions of other people. They became a ruling minority. As his tomb inscription shows, Darius I proudly wanted those after him to know that, although the Persian man was formerly a merely local figure, now he ruled every other nation. Darius declares that when you look at the rock relief of his tomb, where one figure for each of his many distant subject nations is depicted physically supporting his lone royal figure, “then you will recognize, then it will be known to you: the spear of a Persian man went forth far; it will be known to you: a Persian man fought the foe far from Pārsa,” adā xšnāsāhi ada-tai azdā bawāti Pārsahyā martiyahyā dūrai ǝrštiš parāgmatā ada-tai azdā bawāti Pārsa martiya dūrai hacā Pārsā parataram patiyajatā.Footnote ⁸³ This seems to indicate how hitherto unexpected it was for a Persian, as somebody from a formerly minor nation, to have achieved such extraordinary power. (It also makes Darius appear to be the first truly Persian king.)Footnote ⁸⁴ Representative figures of every other nation are depicted in relief on the rock face of his tomb, elevating the Persian king and contributing to a new royal foundation. But as representatives of the people of one region had become practically universal rulers, they were exposed now to innumerable close contacts with people of other cultures. This had great consequences for the evolution of the Persian language far more than for the languages of the subject peoples.

Several kinds of testimonies indicate the subsequent mixture of the Persians with their subjects, creating circumstances in which many adults had to learn to speak Persian. I turn now to review them. They provide strong primary-source contextual support, beyond the preceding section, for the case that Old Persian became a target of mass acquisition by adult learners, and that the Achaemenian period witnessed the most sweeping changes in the language, giving rise to Middle Persian.

Fighters and Laborers for the Persians

The inscriptions of Darius I and Xerxes I attest to their rule over lands “of people of many kinds,” paruzana-, or “of all kinds,” wispazana-.Footnote ⁸⁵ The Akkadian equivalent of this expression, where Akkadian versions survive, says lands “of people of all tongues.”Footnote ⁸⁶ At the palatial complex of Persepolis, Xerxes ordered the construction of a portal named in the accompanying trilingual inscription as “the Gate of All Lands,” Old Persian duwarθi- wisadahyu-. In the early fourth century bce, Xenophon wrote that Cyrus had “made so many peoples dependent on himself that it was quite a task to traverse them.”Footnote ⁸⁷ For the first time, all of these different peoples with different languages now served a common master in circumstances of compulsory cooperation.Footnote ⁸⁸ Members of all these peoples now found themselves face to face at royal sites in large labor crews and marshalled together in massive polyethnic fighting forces.

It is almost needless to say that the military adventures, army camps and marches, and pageants of the Achaemenids and their Persian delegates must have created circumstances in which many individuals from different countries, who spoke different languages, had to learn some Persian as adults. Even if this involved merely individual officers representing the military contingent from one subject country each, the mustering of forces on behalf of the Persian king must have entailed language learning by scores of individuals, if not many more, from different speech communities who now had one thing in common above all: their shared service to the Persians. Such individuals would usually not have been native speakers of Persian, as they came from different countries, and they would themselves then have been a major conduit for information about Persian speech to their followers and companions. Herodotus’ famous description goes to great lengths to convey the dazzlingly polyethnic character alleged of the army of Xerxes.Footnote ⁸⁹ Though the numerical scale is not to be believed, the heterogeneous character of the Persian kings’ forces cannot be doubted. The contingents of Ethiopians, Thracians, Libyans, Phrygians, Egyptians, Caspians, Indians, Assyrians, Bactrians, Greeks, Arabs, Lydians, and many other groups would each have required at least one interpreter capable of comprehending enough Persian speech to cooperate with the rest. It would have been pointless, let alone impossible, for any of them to learn all the languages of the others. The Persian rulers did need translators, as Cyrus the Younger needed one to communicate with his Greek mercenaries, whereas a Greek officer in his employ, the commander Clearchus, appears to have understood Persian.Footnote ⁹⁰ For anybody who wanted access to the aristocratic and higher military domains of power, the target of acquisition must have been the Persian language – or rather, an adequate, functional, basic, minimally ornate form of Persian that an adult might learn without modern textbooks and teachers trained in today’s communicative methods.

Military organization and the congregation of elite boys at Persian courts thus probably provided one conduit for Persianization, but craftsmen, traders, and many lowly laborers surely participated in their own way, too.Footnote ⁹¹ The clearest direct Old Persian testimony about the sudden intermixture of people of different tongues occurs in one of the major inscriptions of Darius I at Susa.Footnote ⁹² This inscription has long served modern historians as an illustration of the cultural workings of early Persian imperialism. It was issued in Babylonian, Elamite, and Old Persian, and survives in different copies at the site of Darius’ palace at Susa, city of Elam, the region known in the first millennium ce as Khuzistan. Designated DSf today, the inscription commemorates the construction of the palace there, which must have been gorgeous but is now a bare ruin. One portion describes the ethnic varieties of personnel involved in this undertaking:Footnote ⁹³

Darius boasts about his palace that it was constructed using rare and costly commodities derived from all parts of his domain: cedar from Lebanon, timber from Gandara (eastern Afghanistan), gold from Lydia (western Turkey) and Bactria (northern Afghanistan), silver and ebony from Egypt, ivory from Kush (northeastern Africa) and India, stone from local quarries, and more.Footnote ⁹⁴ More important for the present argument are Darius’ specific statements about the labor force employed in the construction. Assyrians, Babylonians, Carians, Greeks, Medes, Egyptians, Lydians, Hūžes (known today as Elamites) – representatives from many of the nations ruled by Darius – participated in the building project and its adornment. Most of the workers drafted for this purpose will not have used the languages of the others in their collaboration. The one language that would provide the most advantages to all these workers from abroad, who resided together at least for a time at Susa, was Persian, the language of their masters and employers.

The same goes for construction and work at Persepolis. The thousands of Elamite administrative tablets discovered at Persepolis, surviving portions of archives with dates ranging from 509 to 457 bce, have provided material for a large batch of scholarly articles and books specifically on non-Persians at Persepolis. Judging by their names, at Persepolis there were Babylonians, Cappadocians, Carians, Carmanians, Gandharans, Indians, Lydians, and others.Footnote ⁹⁵ Besides texts in Elamite, the site has provided plentiful short examples of Aramaic writing and individual examples of Greek and Phrygian texts. Other royal bases of operation must have been like Susa and Persepolis, sites of highly heterogeneous populations of immigrant workers serving Persian overlords, generating a new mixed culture. We should expect that the diverse assortment of laborers from different far-off countries would produce an intercommunal language modeled on that of the masters, just as a common Melanesian Pidgin English, discussed in the previous chapter, was created by laborers from distant islands and villages in the South Pacific, from New Guinea to Samoa, and just as South and Southeast Asian laborers, of various origins, working today in countries of the Arabian Peninsula have created a new pidgin, called Gulf Pidgin Arabic, based on the Arabic of their employers, a pidgin in formation even now.Footnote ⁹⁶

Darius’ inscription about his palatial construction at Susa is intended to convey the vast extent of Darius’ power, but it also indirectly documents an important event in the history of the Persian language.Footnote ⁹⁷ Under the circumstances described, many non-Persian adults will have learned the Old Persian language with limited morphology and native speakers of Persian will have needed to simplify their speech to communicate with many nonnative speakers. At Persepolis, moreover, newly arrived laborers from abroad will have depended on Persian-speakers in their immediate environment for their immediate needs, for they were in Persia proper and so surrounded by local Persian native speakers, but they will not have mastered Persian as native speakers do. Some of them must have stayed in Persia proper and exposed young people with whom they interacted to a nonnative variety of Persian. Perhaps some of them returned to their home countries, where they would be able to serve as representatives of their locality to Persian authorities, because of their experience with Persians. Nonnative learners of Persian who returned home from Persia proper would bring with them their adult-learner, morphologically reduced variety of the language, likely a somewhat restructured nonnative variety acquired during a finite period among the Persians of Pārsa. In this they would resemble the local constables of British Papua who brought Pidgin Motu with them on their return to their villages upon their retirement from government service, where they shared it with young people to groom and prepare them for interaction with the new interregional power and for access to goods and opportunities available outside their own locale.

Alternatives to the hypothesis that the newcomers to Persian royal sites would have learned Persian badly, or even a pidgin variety of Persian, are unlikely. It is not as if the laborers imported to Persepolis and Susa from many different countries were given formal night-classes in Old Persian with instructors trained in modern methods. It is not as if they could all resort to pantomime and would not bother to attempt basic verbal communication, either. It must have been vital for many of the workers to make themselves understood to their Persian-speaking masters. The facts of human language acquisition, seldom considered by ancient historians, practically require the view proposed here. The inscription of Darius at Susa and the Persepolis administrative archives testify to only two major examples of situations in which non-Persians would have needed to acquire a working, basic ability with Old Persian. There must have been innumerable other such sites like this that have remained undocumented by written sources.Footnote ⁹⁸ Everywhere that Persian rulers and their immediate followers dealt with subjects of different nations, some of those subjects would have required Persian language as the medium of intercommunication. Most of them in the first generation or two will have done so as adults. Thereafter they will often have learned it from fellow countrymen who themselves had learned it as adults.

The large-scale migration of soldiers and workers in imperial labor markets has been a major factor in the genesis of pidgins, creoles, and semicreoles of modern languages. The Persian Empire, a state formed by conquest on a scale unmatched at that time, could not have been an exception to these ordinary and regular social processes. On the contrary, the Persian Empire seems exemplary of this process in ancient history, as illustrated by the evolution of Persian in the inscriptions of the Achaemenids. Achaemenian Persian inscriptions may provide the earliest clear example of morphological reduction through mass nonnative acquisition. The imperial context makes sense of this.

Domestic Personnel of the Persians

Important testimonies to the population mixture of Persians with others survive in the writings of their neighbors on their western frontier, the ancient Greeks. Not all Greek authors were overtly hostile in their accounts of the Persians. Many collaborated with them even in times of war, inspiring Greek accusations of “Medism” against collaborators. Yet even the hostility expressed by well-known ancient Greek authors shed light on what was special about Persian society as compared with that of Greek observers, as I will show.

Around 425 bce, Herodotus wrote explicitly about the cultural mixture fostered by the Persian rulers.Footnote ⁹⁹

This is to say that the Persians selected what they regarded as best from every subject population, including many non-Persian women and boys. They adopted what was best in their view as their own and they made it Persian. Racial purity was evidently not an ancient Persian concept, although legitimacy of patrilineage and family affinity were all-important. Every male Persian aristocrat is evidently supposed to have begotten many children of mixed ancestry raised, often, by non-Persian mothers. Many female mates were brought to them from different home countries where different languages were spoken. Herodotus adds an observation critical for considerations of language acquisition: a Persian child “does not come into his father’s sight before he turns five years old but lives with the women. They do this in order not to strike the father with grief if it should die young.”Footnote ¹⁰⁰ Strabo, centuries later, reports the same details, apparently on the basis of an earlier source of the Achaemenian period shared with Herodotus, and adds that Persian men were rewarded for having many babies.Footnote ¹⁰¹ “They marry many women, and they keep more concubines at the same time, in order to have many children. The kings set annual prizes for those who beget many, but the children are not brought into their parents’ sight until they are four years old.”Footnote ¹⁰² Other ancient Greek authors on the Persians, such as DinonFootnote ¹⁰³ and HeracleidesFootnote ¹⁰⁴ (both mid fourth century bce), remark on the hundreds of concubines of the Persian kings.Herodotus elsewhere adds the detail that “the women sleep with the Persian men in rotation.”Footnote ¹⁰⁵

Many, perhaps most, of these numerous concubines – women enslaved for sexual companionship, a practice that was normal in many societies including the one under discussion – were imported from non-Persian-speaking societies into wealthy Persian households as young women who had entered sexual maturity and likewise had passed the early age of sensitivity to fluent language acquisition, as adolescents or adults. The wives, who may have been mostly Persian by birth, were outnumbered,Footnote ¹⁰⁶ and like the European wives of Dutch colonists at Capetown in the seventeenth and eighteenth centuries discussed in the previous chapter, Persian women would have been at the center of domestic language contact, surrounded by speakers of other languages, adult learners of Persian, and widespread linguistic variation.Footnote ¹⁰⁷ One thinks also of the portrayal of concubines in the fictional Hebrew Book of Esther, set in the empire of the Achaemenids, in which numerous selected good-looking girls from various countries, like the book’s Judaean heroine Esther, were carefully prepared at the city of Susa for their roles as royal mates.M. Brosius, the author of the only monograph on women in the Achaemenian Empire, concludes on this matter, “To have concubines was not the privilege of the king alone. By all accounts they were to be found in the palaces of the satraps and Persian nobles. The presence of concubines therefore must be seen as a characteristic of Persian society.”Footnote ¹⁰⁸ Foreign concubines and maidservants necessarily came late to the acquisition of Persian speech, as nonnative speakers. If the reports of Greek authors are correct, some Persian kings were routinely accompanied by dozens of pretty foreign girls of this kind. These attractive young women not only contributed to the ostentatious display of social power by a small number of men who increasingly owed their high rank to mere patrilineage rather than to force of arms, but they also bore the babies of the Persian men who acquired them. The children of Persian men will therefore have had many nonnative speakers as their first models for Persian language acquisition. If the report shared by Herodotus and Strabo is true, then concubines, wet-nurses, nannies, and eunuchs, all of various origins and language backgrounds, were caring for the children of the Persian men in women’s quarters along with what was probably a minority of Persians who spoke Persian natively.Footnote ¹⁰⁹ Nonnative speakers of Persian must have predominated numerically in many Persian homes, and they were also more responsible for early childcare.Footnote ¹¹⁰

As mentioned, eunuchs must have played a role in this, too.Footnote ¹¹¹ They were gatekeepers and attendants of night quartersFootnote ¹¹² and companions and servitors of kings and lords.Footnote ¹¹³ Enslaved eunuchs were given as the payment of tribute to the Persian kings. Herodotus mentions an annual tribute of five hundred boys from Babylon as well as boys from Ethiopia and Colchis; he also mentions the punishment of Ionian rebels through the castration and enslavement of the best-looking boys (τοὺς εὐειδεστάτους).Footnote ¹¹⁴ If the Babylonian tribute just mentioned had lasted just as described even for twenty years, that would populate Persian sites with 10,000 castrated young speakers of Aramaic, possibly as well as Akkadian, who had to learn some Persian speech to adapt. The Platonic First Alcibiades informs us that expensive eunuchs of good character, and not female “nurses of little value,” raised the king’s own eldest heir and groomed him, but eunuchs, too, were probably mostly non-Persian by origin, and often must have been nonnative speakers of Persian.Footnote ¹¹⁵ One expects that more ordinary Persian household lords would have had female nurses and the mothers of the babies themselves to care for their children. In any case, Persian boys from powerful families reportedly would not have had much interaction with their Persian-speaking fathers during the initial period of language acquisition, their first four or five years according to the Greek sources just mentioned. We must not believe that this alleged five-year rule was strictly observed at all times just as described. Rather it was a noticeable cultural difference from the Greek point of view, remarked upon because Persian fathers tended to have so many children that they could not know them all well. The point is again that nonnative speakers played an outsized role in providing linguistic models of a variety of Persian that the children of Persian fathers would learn.Footnote ¹¹⁶ The large place of non-Persian domestics in the household inevitably must have had major effects on the language of each new generation of Persian elite youth. Such Persian boys would be fluent from childhood in nonnative varieties of Persian, with reduced morphology, learned from the women and eunuchs who nurtured them, from servants who supplied their needs and cleaned up after them, and from laborers who built and repaired their domiciles.

Plato clarifies these matters. Hardly any more explicit testimony could be wanted for the mixture of population and culture taking place in the homes of Persian lords, and the role of foreign women and eunuchs in that mixture, than that provided by Plato in the third scroll of his dialogic treatise Laws. He wrote it circa 357–347,Footnote ¹¹⁷ during the reign of Artaxerxes III. This is the very Persian king for whom the latest extant Old Persian inscription was carved, exhibiting Old Persian grammar at its most divergent from that of Darius I and at the attested point of greatest development toward Middle Persian. At just that time, Plato characterizes the nations (γένη) ruled by the Persians, which “are settled, now spread around and jumbled together,” as ethnically mixed (μεμειγμένα) and thus “ill-begotten” (κακῶς ἐσπαρμένα).Footnote ¹¹⁸ The crass expression “ill-begotten” here is literally “ill-seeded,” referring to the father’s seed in the combination of parents of different ethnic origins. Plato is proud that Athenian and Spartan defenses against the Persians had kept the Greeks pure of such mixing in previous generations, so that the Greeks remained a distinct people. This, at any rate, is Plato’s ideal, also expressed in the Platonic speech called Menexenus.Footnote ¹¹⁹ In his view, the Persians in his day presided over a people now blended into something new, indicating that their former national distinctions were to an extent lost. For Plato, the blending of peoples had consequences for the Persian rulers themselves: it corresponded with a dilution of Persian manliness. The Persian kings lost their original Persian virtue, he went on to say, when they were raised in comfort by women and eunuchs. Plato’s main purpose in this passage was not to discuss the Persians, but to discuss the proper education of the young. He condemned royal luxury as a source of corruption in the Persian princes’ upbringing. This is an example of the ancient Greek antipathy toward the Persians that influenced modern historiography, a tendency that predominated as long as classically trained historians relied almost entirely on Greek perspectives. But Plato’s words, suited for modern nationalism, too, were not merely expressing anti-Persian sentiment without an argument. Plato could have chosen traits to criticize besides their jumbled ethnic mixture. For example, he could have followed the diatribe at the end of his acquaintance Xenophon’s Cyropaedeia, a work he knew,Footnote ¹²⁰ against the latter-day Persians, and found fault with them for such unmanly luxuries as wearing gloves in winter, using carpets on the floor, or sitting too long at meals, or for their perceived cowardice in combat or the mismanagement of charioteers, or some other particular.Footnote ¹²¹ Plato rather deemed virtue to have been lost by the Persian dynasty through the effect of women and eunuchs who raised the royal boys in childhood. As this passage follows immediately his observation that the Persian subjects are an ethnically mixed nation, it is evident that here, too, Plato was thinking of mixing. For he must have known the obvious fact that Persian concubines, nurses, and slaves were drawn from different subject nations. Plato judged the outcome to be degeneration and corruption of character, whereas the Greeks, he held, avoided such corruption by holding off foreign rule and consequent intermixture of Greek with foreigner. We will dismiss the negativity of Plato’s evaluation as merely a matter of his personal judgment, but we should not regard his observation about the mixture of the subject nations of the Persians as creative invention, especially as he presented it in the manner of a fact that his contemporary reader is expected to know already. For Plato’s critical evaluation to be meaningful, it had to be founded on a generally accepted observation about Persian society. It is this mixing that necessitated large-scale adult acquisition of the Persian language. More than army camps and teams of ethnically mixed laborers, foreign women and eunuchs in Persian homes must have played a large role in generating a drastically reduced form of the language, ultimately the beginnings of Middle Persian. If an adult learners’ form of Persian had remained restricted to the barracks of warriors and builders, Persian boys would not likely have acquired it as their own, except when involved in war. It is the domestics who must have played the biggest role.

Historians Amélie Kuhrt and Josef Wiesehöfer, knowing sources like these, both emphasize that local elites intermarried with Persian royals and that Persian royals took many provincial concubines.Footnote ¹²² Herodotus records a case in which a Greek man married a Persian woman, and their children were regarded as Persian.Footnote ¹²³ Modern historians wonder about the “identity” of such people, but events like these necessarily have more salient linguistic effects. The Persian kingdom promoted the integration of local elites in a common ruling-class culture. Being allowed to join the Persians and to sit with them was regarded as an honor. Although in the time of Darius I being Persian seems to have become a special and exclusive privilege, it was possible to join the ranks of the Persians, and especially to have one’s sons raised as Persian, through imitation of and socializing with Persian aristocrats. That was one function of the “king’s gate,” a term usually rendered as “the court” today.Footnote ¹²⁴ The most promising young men from local elites, many of whom will not have spoken Persian at home, were sent to attend at the “king’s gate,” where they vied for royal favor through assiduous service to the king. Here, too, was another site in which a common variety of Persian speech would have been shared by men of different origin interacting with servile non-Persians. The royal family and its privileged Persian and Iranic-language speaking counterparts such as Medes, in and around their own homes, must have been outnumbered, perhaps drastically, by servants, workers, concubines, and others who came from other language communities. Persians purchased, imported, attracted, and accommodated outsiders and the result was a hybrid that soon became the elite standard.

The rapid grammatical transformation of the Old Persian language indicates that the successful practice of populating elite Persian households with foreign domestics and sexual mates who raised Persian children contributed to a transformation of the Persians themselves. Persians actively drew into their midst visitors, followers, and household members from every land, many of whom became permanent attendants, courtiers, and even close relatives. It was simply expected that Persian leaders would promiscuously enjoy the company of the most pleasing of the conquered peoples. They chose slaves, brides, and concubines for themselves (not to mention boys) in numbers quite surprising by today’s most widespread standards, and they had very large numbers of children by them.Footnote ¹²⁵ A special title applied to the male and female scions of so many royal children: “son/daughter of the (royal) house,” a term widely used and subsequently retained in local languages from Armenian to Sogdian and Bactrian.Footnote ¹²⁶ But Persian elite men were not changing diapers or chasing defiant toddlers. They brought non-Persian-speakers into their palaces, audience halls, and homes, as companions, flunkies, and mates, so their own children were frequently raised, at least in their early years, by women and eunuchs who were foreigners, who therefore must have learned Persian mostly only as adults, and who thus certainly spoke a nonnative sort of Persian. As discussed in Chapter 2, it is nonnative learners, not native speakers, who reduce and even drop contextual inflection in a language, the very effect observed in Middle Persian.

Modern scholars have sometimes seemed embarrassed about the hundreds of beautiful women and castrated males reported to abide in the fancy halls of Persian lords.Footnote ¹²⁷ Either they have been interpreted as a source of corruption, as Plato saw it, or they need to be explained away or minimized, as a misunderstanding, a distasteful exaggeration, or a mirage arising from bigoted perceptions and orientalist bias. But all the sources together have not lied. Such human beings have existed in many societies, for many different reasons. Close to the epicenter of Persian power, concubines and eunuchs, who evidently greatly outnumbered their masters in their private quarters, surely played a large role, not only as foils for lordly Persian manhood, but also in the nurture of Persian boys, the creation of elite Persian culture, and the evolution of the Persian language. Plato did not like it, but that does not mean that we should dislike it – or deny the sources that testify to it.

The Testimony of Material Culture

Besides the ample testimonies from sources pertaining to the world of men employed at war and in construction work, on the one hand, and the domestically employed women and eunuchs, on the other, a third testimony to the cultural mixture of rulers and ruled comes from the material remains of durable precious objects and architecture from the Achaemenian Empire. These pertain less to language, so I will linger here more briefly, but they buttress the case that the Persians were not just willing but interested in fostering a new cultural blend of the things that pleased them. Material remains have testified to art historians and archaeologists that the Achaemenids produced a mixture of hitherto regional cultural forms. Mixture has long been a keyword in the history of Achaemenid royal material culture. Formerly this mixing ran against the tastes of art historians seeking neat, distinct national types of art. Herzfeld, the pioneering archaeologist of Iran, wrote that Persian art was “a hybrid art, if art it can be called, worthy to be studied only out of scientific and historical, not of aesthetic, interest.”Footnote ¹²⁸ Roes complained that “if Achaemenid art was not regarded [by modern scholars] as the work of foreigners, it has generally been called a hodgepodge of foreign elements that hardly deserves the name of art and seems to have been regarded by many scholars as an affront to their artistic sensibilities.”Footnote ¹²⁹ Ghirshman admitted but defended the apparently “dependent” character of Persian art, which relied on “borrowing from other peoples”: “There is a tendency to criticize Achaemenian civilization for depending too much on the achievements of others, particularly in the sphere of art. But even if a study of the surviving monuments lends support to this view, it only illustrates the old truth that peoples living in close relationship inevitably exert a reciprocal influence.”Footnote ¹³⁰ A half century later, Boardman showed that art historians finally had to acknowledge that “Persian” meant “mixture,” when it came to material culture. He wrote, “It must soon have become a matter of indifference whether any given Mesopotamian, Egyptian or Anatolian was carving a figure composed of elements of Anatolian, Egyptian or Mesopotamian style. It was by then all Persian.”Footnote ¹³¹ For Henkelman, “Persepolitan art is willfully synthetic,” and art at Pasargadae synthesizes a “coherent vision” appropriated from elements of “eclectic origins.”Footnote ¹³² From points of view beyond imperial centers, Khatchadourian, using sophisticated language, emphasizes the agency of local elites in adopting forms of material goods in architecture and ornament that emphasize their participation in a larger imperial culture,Footnote ¹³³ and Colburn probes the possibilities of describing such Achaemenian material production with the term globalization.Footnote ¹³⁴ We see in such statements from over decades an evolution from distaste for the mixture of regional forms into understanding based on appreciation, but the basic idea of synthesis remains.

In my view, some historians of art go too far in supposing that every aspect of Persian art and architecture was part of a master plan of “imperial ideology,” as if the kings and their ministers were informed architects and decorators who carefully planned the “ideological” import of every statue’s minute characteristics. Instead, the mixture of forms evident to art historians were more probably the result of a collaboration of intelligent artisans of different regional origins who developed something new and distinctive to please their kings, deliberately suited to the purpose of their employment. Ultimately, for the present argument, one should note that although art historians have disagreed over the decades in their personal evaluation of artistic quality in the mixture of regional styles and forms in objects from imperial Persian sites, they agree in the fact of that mixture.Footnote ¹³⁵ Considering these consistent observations by archaeologists and historians of art, the material environment of royal Persian sites likewise shows that Persian soon meant hybrid. This was not regarded as something bad, but as prestigious. It was a demonstration of the universality of Persian power, but it was human individuals who had to bring the regional expertise required to fashion such objects synthesizing previously separate forms. Population contact of this kind also naturally and necessarily had striking linguistic effects.

The Testimony of the Language of the Late Achaemenian Inscriptions

To summarize the foregoing, three sorts of testimonies to the mixture of peoples brought about by Persian rule have long been available. First, the early Achaemenian royal inscriptions and Elamite administrative texts tell how the people of one region became rulers of many far-flung peoples who spoke different languages. Persian kings commanded workers and craftsmen from distant countries and united them all in sites of royal display as ostentatious demonstrations that they had indeed taken the best of all and used them together in an awe-inspiring mixture. Elamite administrative texts on clay tablets likewise reveal how workers from many different subject populations, who did not originally share a common language, collaborated in managing the Persian state and found employment in labor forces at sites ruled by Persians. We can add to this the obvious role that shared military pursuits must have played in fostering a common speech among martial servants of the kings. Fighting men from different countries serving Persian masters would need to learn just enough Persian to do their jobs, and more if they wanted to communicate directly with their masters. Second, with regard to the Persian domestic domain, Greek authors comment explicitly on the willingness of the Persians to accept the customs of others and to mix socially with them, especially their taking numerous foreign females and eunuchs into their homes. Herodotus states it as an obvious fact. Plato and other Greek authors disdained such mixing as impure, fearful of the mixing that was the outcome of the worldly success and power that inspired their jealousy and especially their feelings of political insecurity. But it was obvious to them that the mixture was real. Third, modern art historians and archaeologists testify to a blend of regional cultural forms into something new and synthetic as the outcome of the material production that they fostered, as far as durable material media preserve them for us to see today. These varieties of available evidence support the argument that by the fourth century bce, the Persians, as people with imperial privilege, had willfully taken the best of everything they ruled and incorporated choice objects and people into their own. Thus, they had evolved and changed over several generations.

The ruling elite called Persian in the fourth century bce, whose domain stretched from the Mediterranean to Afghanistan, were different from the Persians local to Pārsa of the early sixth century bce. Their empire had promoted and homogenized local elites into a new synthesis with less conspicuous regional variation among themselves and, as argued here, a new form of Persian language. One might say, speaking loosely, that they fostered a hybridized pan-Near Eastern elite culture, while retaining the name and meaning of “Persian,” and they bequeathed its amalgamated forms to later elites as cultural norms for the rulers of successor states. We can call this tendency Persism,Footnote ¹³⁶ a term on the model of “Hellenism,” to focus the investigation of the impact, influence, and reception of ancient Persian people and their culture in the world around them. As a descriptive neologism serving as an umbrella for several different phenomena, Persism can be used to characterize the elite cultural blend that arose as an effect of Persian power and practices of rule during the period of the Achaemenids and lingering long after them. The phenomenon of population change here should not come as a surprise. No nation today is the same as it was one hundred and fifty years ago. Immigration and population movements have created great demographic changes, even when the names and sometimes even governments of these nations remain apparently consistent and stable. The Achaemenian Persians clearly experienced such changes on a scale never seen before them in history but seen many times since then. Those changes did not affect all their provincial subjects nearly so much as it affected their own Persian people.

The clinching testimony comes from the Persian language itself. This is the focus of this study, and it also offers the decisive argument that shows that the other testimonies just reviewed reflect not just the perspectives of onlookers but substantial social changes evident to contemporaries. Corresponding to the changes in population, Old Persian grammar exhibits drastic changes in the direction of inflectional simplicity within a few generations. By the late fourth century bce, the old inflectional system of the Persian language had been reduced to a minimal state (Chapter 1). Middle Persian had emerged in some sense. Knowing now what we know about the social factors in language change (Chapter 2), we should already expect this to be due to the sudden mixture of populations and the concomitant acquisition of Old Persian by large numbers of adult learners. The combination of clear linguistic effects and explicit ancient sources is conclusive. The changes in the Persian language become, in effect, yet another witness to the changing constitution of Persianness in the Achaemenian period, from numerous regional sources. The testimony of the changes to the grammar of Old Persian is even more important than the contemporary witnesses, in that it is based not on third-party description but on modern grammatical analysis combined with a robust theory of language change based on facts about human language acquisition.

Dominant Ethno-Class or Domestic Melting Pot?

Pierre Briant, a leading historian of the Achaemenian Persians, posed a model to understand the culture of Achaemenian Persian rule in which the Persians constituted a “dominant ethno-class” that insisted on Persian superiority to and difference with those ruled. Being ethnically Persian would be a prerequisite to true participation in the ruling elite.Footnote ¹³⁷ In this view, Persian ethnicity was an essential attribute for members of the elite ruling class. It is hard to argue against this. Briant goes further, however, in holding that Persianness was exclusive to Persians and the Persians remained quite distinct. He holds that the Persians did not preside over “an imperial melting pot” (his expression).Footnote ¹³⁸ Even the Persian language is supposed to have been exclusive to this dominant ethno-class. “With only a few exceptions, only the Persians spoke Persian,” Briant writes.Footnote ¹³⁹ Perhaps on the basis of a similar estimation, Mancini recently interpreted the history of Old Persian with the idea that the Persian language persisted in “a rigidly close-knit network” of speakers.Footnote ¹⁴⁰ Part of Briant’s argument is based on the analysis of the names of the satraps, generals, and high officials from the reigns of Darius I and Xerxes; these names are overwhelmingly Old Iranic names (Persian or Median). This is indeed striking.Footnote ¹⁴¹ The present discussion, however, shows that there is a limit to which we should accept this analysis.Footnote ¹⁴² Although “the Persians” were always identified as the rulers of the Achaemenian Empire, the six royal generations comprising the kings from Darius I to Artaxerxes III entailed great demographic changes that transformed what it meant to be Persian. That is because the Achaemenian state, as the kings’ inscriptions explicitly state, was not only a state of Persians or Aryas (Ariyā), but it included many peoples of different kinds (paruzana-). The early testimonies from the times of Darius I and his son Xerxes do clearly express Persian difference and dominance. They did so in the decades immediately after the formation of the empire by conquest and by crushing subsequent resistance raised by local rulers, when differences between rulers and ruled were at their most conspicuous. Their inscriptions and rock reliefs, especially those of Darius I, depict subject ethnic groups in carved relief in stereotyped appearances and attire. This striking overt display of ethnic domination supports the idea that this “ethno-class” was a political regime characteristic of Persian rule for its duration. Others have adopted Briant’s idea. It is now a standard view about the history of the Persians of the Achaemenian period. Just before Briant began to publish his views on the dominant ethno-class, however, an earlier view was that Achaemenid rule “induced racial and cultural fusion,”Footnote ¹⁴³ and that “the Achaemenid period is characterized by intensive processes of ethnic mixing and syncretism of the cultures and religious concepts of various peoples,”Footnote ¹⁴⁴ just as Plato said that the Persians ruled people who had been “spread around and brought together,” διαπεφορημένα καὶ συμπεφορημένα. This chapter reminds us that the rock relief images of Darius’ tomb, clearly depicting Persian dominance over other many distinct ethnic groups, were established near the beginning of Achaemenid rule, and cannot be used to explain the meaning of Persian ethnicity for its entire duration. After one hundred years, and still more after two hundred years, members of Persian aristocratic families and households surely included many men and women of mixed descent and heterogeneous family origins. If, in a later period, a Greek name used by a follower of the Seleucids does not necessarily convince us that its bearer was a native speaker of Greek – or even Greek in ancestry at all – or an Arabic name borne by a servant of the caliphs that he was of pure Arabian ancestry, why should the Persian name of an Achaemenian satrap necessitate that its bearer belonged to an ancestral exclusive Persian ethnic class that always excluded non-Persians? What made someone Persian was success in assimilation to the ruling class, which itself changed over time. As Xenophon said, those who served the satraps were supposed to imitate their Persian rulers. Persian nomenclature from the Achaemenian period therefore reveals little about the family background of individuals. It tells us only that people adopted Persian names or gave such names to their children in hope of their future social success.

Of course, this is not to say that the Persians shared their supreme power with non-Persians. In a sense, it was lastingly true that there was a dominant class of Persians, for the rulers were always known as Persian.Footnote ¹⁴⁵ That was their identity. Briant and others are surely right that they had no specific, deliberate imperial policy of programmatic Persism or Persianization, nor any intention of making the generality of their subjects Persian. Nevertheless, they created the circumstances through which Persian demography and culture were bound to change drastically. The melting pot happened more covertly, from the point of view of the sources surviving today: in the night quarters and the halls of Persian manors and palaces, in the army camps between ethnically different units, and in the work yards where men from different nations collaborated in creating monuments to the Persian kings and the royal family. The preceding discussion demonstrates that what it meant really to be Persian was not static. By the time of Artaxerxes III, Persian men were probably mostly of an ancestry that would be considered mixed by comparison with their forebears of the time of Darius I, just as Plato described them with his contemporary negative evaluation. This is not a matter of biological race, a false concept.Footnote ¹⁴⁶ The point is that the Persians readily accepted cultural goods and customs from other peoples. They took on new customs and acquired new luxuries, just as happily as we today adopt new technologies that transform our lives with hitherto unimagined comforts and conveniences, without our considering much how our culture is thereby transformed. Most importantly for the history of the Persian language, Persians accepted thousands of non-Persians into their households and made thousands of children with them, and these numerous children, raised in large part by non-Persian servants, were often accepted as Persian. The Persians were thus, in fact, assimilated to those they ruled, in their own palaces, just as local elites gradually Persified themselves to maintain and promote their status. It could hardly be otherwise in the multilingual, polyethnic social conditions of the Achaemenian centers of power. The transformation of the Old Persian language into Middle Persian is a clearer testimony to this internal metamorphosis than even the direct statement of contemporary observers. The concept of a Persian “ethno-class,” if one wishes to retain it, cannot refer to something static, and so becomes much less useful for historians today. Individuals could join the Persians, intermarry with them, and their progeny could be raised as Persian, speaking a new, current, simpler kind of vernacular Persian language. There was indeed a melting-pot of heterogeneous ingredients right in the homes of Persian kings and lords. While most local peoples under Persian rule retained their own languages and regional identities, what being Persian looked like and sounded like in 350 bce must have been quite different from what it was in 500 bce. This is inherently plausible or even obvious when we consider the rearrangement of ethnicities in the one hundred and fifty years of imperial processes and population mixtures preceding our own time. It seems, then, that historians should allow for more social and cultural change among the Persians of the Achaemenid period. Historians have followed the two largest textual corpora from royal centers within the Persian Empire – the royal inscriptions and the Elamite administrative tablets from Persepolis – but these have skewed our attention toward the earliest decades of the empire and away from the ensuing changes that imperial organization brought with it. One cannot expect the snapshot of a few major early inscriptional sources, despite their inestimable importance, to account for change across two centuries of Persian rule.

In the end, what makes the hypothesis of population mixing accompanied by unilateral nonnative bilingualism not just plausible, but certain, is the combination of the evidence of the history of the grammar of the Achaemenian Persian inscriptions with a host of testimonies about the social facts of the Achaemenian period.

Modern Terms for Imperial Population Mixture

Recent historians of ancient empires have adopted comparative approaches to discover common patterns and special characteristics in empires. This has required a definition of empire to establish a set of comparable cases. They have usually defined empire as a state formed by one people or ethnic group through conquest of others, typically on a large scale. Empires, understood in this way, are distinguished by conspicuous ethnic difference between the rulers and the ruled. One weakness of this definition is that these differences erode over time as the populations remain in contact, cooperate, and share goods and practices, so that by this defining feature, the older the empire, the less “imperial” it will be.Footnote ¹⁴⁷ Accordingly, the idea that difference between ruler and ruled is the core feature of empire has turned the attention of these historians toward the terms in which that difference is characterized and how imperial states maintained their integrity along with the ethnic differences. If empire is defined by ethnic differences along a scale of power, then should it not be that a long-lived empire maintained ethnic difference? That would be a false assumption arising from this concept of empire created for comparative purposes.

One attempt to solve this problem of historians’ own making, by insisting on this definition of empire, relies on a concept of “cosmopolitanism.” A group of ancient historians have recently taken “cosmopolitan” to designate “persons and polities that freely cross cultural boundaries.” They posit an imperial cosmopolitanism that works in two ways: by assimilation, whereby the differences between local elites and universal rulers are blurred, and by subordination, whereby the preeminence of the culture of the rulers over others is continually reasserted.Footnote ¹⁴⁸ Either way, cosmopolitanism entails the creation of a new, common ruling-class culture with elements that pass over and erode former boundaries. Ancient empires fostered cosmopolitan ruling classes by necessity because their control of many ethnically different populations required individuals who could mediate between the rulers and the subjects. Rulers rely on such people. In this sense, the regime of the Achaemenids clearly produced a cosmopolitan culture – one that created a model for ruling elites that lasted for centuries after them – and they belong in this comparative discussion.Footnote ¹⁴⁹ But because of the widespread acceptance of Briant’s idea of an ancient Persian dominant ethno-class, Achaemenian Persian cosmopolitanism is deemed to have been strictly one of subordination, wherein the Persians always remained sharply distinct masters of the rest, not an empire of cosmopolitan assimilation.Footnote ¹⁵⁰ The preceding discussion of Persism and Persianization undermines the meaning of this by showing that the lines between subject populations and masters were blurrier than has been assumed. The conduits for assimilation were personal and often intergenerational, formed by marriages and concubinage and the birth of children into a Persian identity assigned mostly by patrilineage, carried out by foreign-born eunuchs and servants who surrounded Persian masters, not merely through the occasional promotion of non-Persian individuals through the favor of the king. Without understanding the very large scale of servitude and of sexual and reproductive access to women enjoyed by privileged Persian masters, we cannot understand the scale of the corresponding Persianization.

Linguists, by contrast, do not normally discuss “cosmopolitan” languages. Their category is “contact languages,” with the creole as a distinct phenomenon among them. Creolization is a specific linguistic process, as discussed in the previous chapter, and not a moral evaluation. Along with linguistic creolization, however, scholars specializing in the cultures of speakers of modern creoles sometimes discuss “creolization of culture.” The use of the term as applied to language and culture differs very distinctly, though, because the discussions of creolization of culture do not usually refer to reduction of inflected forms through nonnative acquisition, as discussions of creolization of language do. “Creolization of culture” usually refers to mixing and hybridization, not to simplification or reduction of a culture’s complexity (whatever that might be) on the model of grammatical, morphological reduction.Footnote ¹⁵¹ This makes discussion of linguistic creolization and cultural creolization fundamentally incompatible at an analytical level, even though they can be said to happen simultaneously in one population. One could, however, very nearly equate the concept of cultural creolization in some form with cosmopolitanism. Both are terms of description referring to different kinds of hybridization of cultures; both lack analytical precision.Footnote ¹⁵² At the same time, the two terms carry distinct values. Cosmopolitanism refers to the creation of an elite ruling-class culture bridging formerly separate groups, whereas creolization is typically conceived as a phenomenon among subordinate, conquered groups who have been forcibly mixed. These connotations are due to the modern histories of the two words. Creole-speakers were colonial subjects of European empires; European imperialists generated a cosmopolitan, colonial culture for themselves.Footnote ¹⁵³ Despite the difference in valuation, the names nevertheless refer to similar processes. Both arose through mixing and homogenization of various components: practices, habits, products, words, relationships. Hybridity is one outcome of both. It is arresting to think that by this similarity we might invert the values of these terms and designate the later Achaemenid Persian rulers “creole kings,” thinking of the imperial mixture of cultures that they propagated for their own glorification and private enjoyment. One specific example jumps out of the pages of history. Darius II (regn. 423–405) had a Babylonian mother; with this in mind, his contemporary Greeks entitled him Nóthos, νόθος, “bastard,” “born of a foreign concubine.” Would his royal great-grandfather, Darius I, have thought, if he could have met him, that this descendant of his spoke a “bastard tongue” as well?Footnote ¹⁵⁴ Through this bastard king every subsequent Achaemenid king shared Babylonian ancestry. Of course, mixed ancestry alone does not make a subject “creole” in any of the senses current today. The royal environment of the Persians was, nevertheless, supported by non-Persian subjects who, coming from other backgrounds, brought new cultural goods to enhance the majesty of Persian rulers. Thereby we simultaneously force ourselves yet again to think of ancient Persian culture not as pure but as a hybridFootnote ¹⁵⁵ – like nearly all cultures and peoples, but in relatively sudden and high degree – in which the relationship of Persian cultural contributions to the cosmopolitan culture fostered by the Achaemenian regime is analogous to the relationship of the “lexifier” to its pidgin and creole offspring. It must be emphasized that this is merely an analogy. One must not speak of a corresponding “simplification” of a “cultural grammar.” Thinking about imperial hybrids is not new, but few such ruminations are genuinely useful here for my purposes.Footnote ¹⁵⁶ For historians, designating the culture of the Achaemenids as hybrid, or perhaps better, suddenly heterogeneous, or even exoterogenic (to use Thurston’s neologism discussed briefly at the end of Chapter 2), facilitates two elements in our account of the Achaemenids. For us, it acknowledges the blending of local elite cultures that resulted in what was effectively an evolving new elite superculture with an old name – Persian – and that this blending occurred within the period of a few generations, even while Persian people were regarded as distinct and retained privileges under the Achaemenid kings. The Persian language itself testifies to this. In the end, this experimental line of thought is not especially productive, because it is based on terms that lack analytical power. As with many arresting ideas, that of the Achaemenids as living in a creole culture offers a perspective that does not bear satisfying fruit. It does not change the facts of history but only the terms of our description. The linguist S. Mufwene, active in the analysis of creolization, rightly warns against the overgeneralization of a historically specific term for basically unrelated phenomena.Footnote ¹⁵⁷ Yet, as history creates a narrative, the terms of our narration do matter, and sometimes new perspectives are illuminating. I leave it to the readers to contemplate what “Persian” may have meant by the time of Artaxerxes III, and how that meaning has changed, and continues to change, over time.