Readers already familiar with ancient Iranic languages and Iranic linguistics are likely to have become frustrated by now that I have not yet addressed a glaring fact: Middle Persian is not the only Iranic language that was shaped by drastic reduction of grammatical inflection. Two other known but less well-attested ancient Middle Iranic languages – Parthian and Bactrian – clearly have much in common with Middle Persian in their grammatical structures.Footnote ¹ This chapter addresses this phenomenon. I claim that both Parthian and Bactrian were shaped by a similar process – grammatical reduction through mass adult acquisition – as well as by convergence due to individuals who spoke more than one of these languages at once. I also posit that the Achaemenian period is the time when these regional varieties of Old Iranic were grammatically reduced along the same lines, or at least the period in which this process was well underway. In other words, Parthian and Bactrian took on most of their characteristic grammatical shapes in tandem with the transformation of Old Persian to Middle Persian, and they continued to be used in contact with each other thereafter. I argue that the clear discrepancy in morphological complexity between these three ancient attested Iranic languages, on the one hand, and other ancient and modern Iranic languages used outside of or on the periphery of the Achaemenian Empire, on the other, is due to this social aspect of the ancient history of their speakers.Footnote ²

It was necessary to base the study so far on Persian alone because there are no contemporary Old Parthian or Old Bactrian texts illustrating the regional varieties of Old Iranic from which they descended or a securely dated set of changes in their inflectional systems. It may be that such regional varieties in the Achaemenian period, including Old Persian, were mutually comprehensible dialects of Old Iranic. This study had to begin, rather, with the observation of attested, chronologically anchored changes to the Persian language and the puzzle posed by the rapidity and severity of those changes evident within the Achaemenian Persian inscriptions over a century and a half. Combined with the synthesis of various findings of recent research in linguistics (Chapter 2) and a new look at long-known sources of Achaemenian history (Chapter 3), the Achaemenian Persian inscriptions and their variation over time are what has made it possible, and indeed necessary, to argue that mass adult acquisition of Old Persian did occur and that the result was a semicreolized version of Persian, the directly attested descendent of which we call Middle Persian. Integrating Parthian and Bactrian into this analysis is more difficult, because the oldest extant texts in these two Iranic languages are later by centuries than the latest Old Persian inscriptions and they are both more scantily attested than Middle Persian. When they do appear in the record, however, they stand out as conspicuously simplified in their inflections as compared with all other attested ancient Iranic languages except Middle Persian. The linguistic model presented in Chapter 2, by itself, could support the hypothesis that mass adult learning led to the grammatical reduction of Parthian and Bactrian, but with small, later corpora of texts and few relevant historical sources, it would remain merely a highly likely hypothesis. As it is, this chapter will necessarily conclude with somewhat less certainty than there is for the case of Persian, but it does support the argument of the previous chapter that the social circumstances of population mixture that characterized the lives of the Achaemenian elite was the factor that induced drastic morphological reduction in Persian.

Parthian and Bactrian

Parthian was a distinct Iranic language attested in scattered texts discovered across much of what is today Iran. When it first appears in our extant records, it is written in a regional variety of the Achaemenian Aramaic script distinct from the one used for Middle Persian. The two are different enough that they can be distinguished on sight. Parthian is assumed to have been adopted and used by the administration of the Arsacid kings, a dynasty that first emerged on the Central Asian steppes in the third century bce and that lasted until the Persian Sasanids usurped their role as overlords in the early third century ce. The Arsacids gradually dominated the lands of modern Iran and Iraq by conquest from the Greek-speaking Seleucids, the dynasty of kings founded by Seleucus (d. 281 bce), one of Alexander’s prominent generals. Eventually, the Arsacids ruled the territory of Iran, Iraq, and adjacent regions, particularly from the time of Mithridates I (regn. 171–132 bce). The oldest evidence of the Parthian language is in scattered texts from the late second century bce onward. These early Parthian texts include a small number of very short inscriptions, coin legends, a single document on parchment from the northern Zagros Mountains, and a few thousand inscribed ostraca. The last of these, tersely recording transactions such as the transfer of food commodities and wine, were found at Nisa, the site of an ancient city in present-day Turkmenistan near the border of Iran. The oldest substantial surviving Parthian texts were written much later. Several of the Middle Persian inscriptions of early Sasanid kings, in the third century ce, are accompanied by Parthian versions.Footnote ³ The bulk of the extant Parthian corpus, however, consists of many hundreds of fragmentary Manichaean manuscripts discovered in Central Asia. Parthian had been used for the Manichaean mission from the late third century ce onward.Footnote ⁴ Mani, the founder of the religion Manichaeism, is supposed to have dispatched Ammo, his evangelist to northeastern Iran and Central Asia from Ḥulwān (today’s Sarpul-i Ḏahāb), at the western edge of modern Iran. Some extant Parthian compositions are attributed to Ammo.Footnote ⁵ The apparent presence of Parthian-speakers from Ḥulwān to Nisa suggests that Parthian was used very widely from east to west in the course of its history. A language with such extensive geography should have exhibited dialectal variation, but the limited texts available do not give us an adequate basis for discerning that.

Bactrian was an Iranic language widespread in what is today northern Afghanistan, a populous and prosperous region in antiquity. The corpus of Bactrian texts now known has mostly become available and comprehensible only in the last thirty years, the texts having emerged during the relentless wars in Afghanistan.Footnote ⁶ Bactrian was written in the Greek script, in handwriting a distinct semi-cursive variety that had evolved locally from a ligatured Seleucid Greek hand.Footnote ⁷ This writing system had been brought there originally by large numbers of Greek-speaking colonists in the wake of Alexander’s conquests.Footnote ⁸ The Greek language did not persist there, but the script remained to be used for the Iranic language of regional currency. Bactrian is attested in a small number of inscriptions, graffiti, and words on coin legends and seals from the second century ce to the ninth.Footnote ⁹ More substantial are more than 150 handwritten instrumental and communicative texts (contracts, correspondence, and the like) mostly on leather, from the fourth to the eighth century ce, in a tradition of literacy that continued through many different local dynasties of different origins: Kushans, Persians, Huns, Hephthalites, Turks, and Arabs. A new batch of Bactrian texts on birch bark, from southern Afghanistan, has now come to light.Footnote ¹⁰ There is also one small fragment of a Bactrian text in the Manichaean script found in Central Asia with more extensive Manichaean texts in other languages, such as Middle Persian and Parthian. All these texts, which were recently deciphered, edited, and translated largely through the philology and erudition of Nicholas Sims-Williams, not to mention others, have resulted in the exposition of this hitherto almost unknown Middle Iranic language, its grammar and lexicon, and have triggered recent efforts to build a new, revised family tree of the Iranic languages (discussed further in this chapter under the heading “The Hypothesis of Central Iranic and Areal Features”).

A very recent development must be mentioned here. Scholars have made a breakthrough in the decipherment of hitherto unreadable ancient inscriptions in the region of Bactria, evidently older than the Bactrian texts in Greek letters, written in an abugida script derived from the Aramaic, for now known as the “(Issyk-)Kushan script.” The language has now been determined to be an old form of Bactrian, although the decipherment is ongoing.Footnote ¹¹ We can now anticipate that it will be possible to augment our understanding of the earlier history of Bactrian grammar in the future. For now, however, and until the inflectional morphology of this older Iranic language of the region becomes clear, we must base our analysis on what is clearly known and understood.

Bactrian, Parthian, and Middle Persian are three different languages, not dialects of one language.Footnote ¹² Although they share innumerable cognate words, they became strongly differentiated and were surely mutually incomprehensible before the second and third centuries ce, a period from which enough known texts survive to compare them meaningfully. Corresponding perhaps with their geographic distribution, they exhibit a continuum of features, with Bactrian and Middle Persian as the most mutually dissimilar of the three and Parthian sharing features with the other two but tending more toward Bactrian in resemblance, at least superficially. A major characteristic of all three together, though, is the striking reduction of their grammatical inflection. Not only is it extensive, but the grammatical reduction they share is of the same kind in each, as compared with the features we know for certain to have characterized three other sets of Iranic languages: (1) their reconstructed shared ancestor, late Common Iranic, the grammar of which is best represented in its oldest attested offshoots, Old Persian and Avestan; (2) Middle Iranic languages contemporary with them used in Central Asia: Sogdian, Chorasmian, Khotanese, and Tumshuqese; and (3) most other modern languages of the Iranic family, from Turkey and Iraq to Pakistan. Although the ancestors of most of the modern Iranic languages shared in the loss of much of the same inflectional complexity – so that, for example, they have just two inherited noun cases and two verb stems of the same ancient derivation like Middle Persian – they also frequently preserve grammatical features lost in Middle Persian, Parthian, and Bactrian, like distinct masculine and feminine grammatical genders. This demonstrates that most modern Iranic languages besides New Persian must have descended from extensively reduced varieties of Old Iranic, but not from varieties that had been reduced to the extent that Parthian or Bactrian were.Footnote ¹³

One of Paul’s studies of the taxonomy of Western Iranic languages remarked on this discrepancy between modern Iranic languages of western Iran and the relative simplicity of Parthian morphology, saying “a still unanswered question of West Iranian dialectology” is “why Parthian, spoken nearly 2000 years ago, should be in its noun morphology more modern than any of the closely related N(orth)W(estern) dialects spoken today.”Footnote ¹⁴ Paul’s question, expressed casually, was not new. Otto Jespersen (1860–1943) wrote in 1894 on the distinction between “modern” and “ancient” languages, the difference being that the latter had many more numerous forms of a single word.Footnote ¹⁵ We saw that Meillet, in 1912, considered the Middle Iranic languages to seem “modern.” In 1927, Hans Reichelt (1877–1939) also wrote that “decay” caused the ancient Middle Iranic languages – especially the western ones, Middle Persian and Parthian – to enter “the state of almost modern languages” already before the Christian era.Footnote ¹⁶ Likening these languages to modern languages implies that modern languages constitute a morphologically analytical and inflectionally bare type. The concept of a distinct modern type of language is ill-founded, however. It is not modernity per se but widespread adult acquisition that induces language change in the direction of inflectional simplicity. This is surely the reason for the character of Parthian grammar, just as it was for Middle Persian. If we restate Paul’s unanswered question, though, it is effectively the same one that I am attempting to answer here: why was ancient Parthian more morphologically simplified and less characterized by contextual inflection than its closest living cousins more than a thousand years later? Now that its ancient neighbor and sibling, Bactrian, is much better known, we can pose the same question of Bactrian, too.

Systemic Similarities between Middle Persian, Parthian, and Bactrian

Middle Persian, Parthian, and Bactrian appear, superficially, to be cut from the same material. Of course, most of their similarities are due to their derivation from a Common Iranic ancestor, but that is not what I mean. It is not just a matter of lexical and formal cognates of the sort found in all Iranic languages. It is morphological reduction on a similar pattern. The noun systems of all three were reduced to one declension of two numbers and two cases with the same functions and the same case-marking suffixes (further eroded in Middle Persian and Parthian; Bactrian noun inflectional endings became as simple as those of the other two in the third century ce).Footnote ¹⁷ Members of the same small class of nouns for family relations (father, mother, sister, etc.) constitute a special type in each, as expected from the regular patterns of syllable loss. Grammatical gender disappeared from Middle Persian and Parthian; it appears evanescent in a few second-century Bactrian inscriptions, absent thereafter. The verbal systems of all three languages lost the ancient perfect and aorist verb stems (like all Iranic languages classified as Middle and New Iranic) and then lost the imperfect verb (the normal one used for past narration in the Old Persian corpus), with the exception of imperfect forms of “to be” and trace examples of imperfect verbs in two specific Middle Persian texts of the third century ce.Footnote ¹⁸ This left only a completely uninflected perfective participle combined with present-stem copulas as the basis for past-time reference.Footnote ¹⁹ In Middle Persian and Parthian, the plural forms of the clitic pronouns are formed agglutinatively with the addition of the same suffix, -ān, which was imported to the pronouns from the nouns; Bactrian shows a closely parallel development.Footnote ²⁰ The present-stem verbs of all three were conjugated with the same generalized suffix, -áya-, a single common morphological reduction out of several possibilities from the inherited verbal inflection classes.Footnote ²¹ Examples of verbs conjugated in the third-person singular present without this generalized suffix do appear in Bactrian and possibly in Middle Persian (as well as other Iranic languages).Footnote ²² All three retained a subjunctive mood, exhibit an uncommonly used optative (more productive in Bactrian) restructured along mutually parallel lines,Footnote ²³ and imperative verbs, all used and formed nearly the same way, based on an unchanging present verb stem, with a few exceptions.

These languages each reflect sweeping changes leading to an outcome of extraordinary mutual similarity in type, in each case apparently formally correlating with the position of the stress accent. (I discussed the relative value of structural explanations of this kind in Chapter 3.) Sims-Williams refers to the “severe simplification of the inflectional system” in Bactrian,Footnote ²⁴ the least reduced of the three, and we have already seen similar assessments for the other two languages. The features mentioned in the preceding paragraph, moreover, concern inflectional morphology alone. More changes common to the three could be listed, as well as many features in which they diverge significantly from one another, differentiating the three as distinct languages. It is unlikely, however, that all three languages, spoken in adjacent regions, evolved just so independently of each other and happened to end up so very alike fortuitously, while other Iranic languages did not. There are only two plausible explanations to account for their uncannily common developmental outcomes. Either they took their shape in the same period under the same or closely similar circumstances to those described in the last chapter, as contemporary shared phenomena due to common causes in their historical social context, or they are the result of linguistic convergence over time due to extensive social contact between their speakers. Each of these two explanations requires that real, specific social phenomena occurred: mass adult acquisition in the one case and large-scale intergenerational bilingualism in the other. Either explanation has a strong bearing on ancient social history. Importantly, they are not mutually exclusive explanations. Both may be true, but they should be considered separately for the sake of clarity of analysis.

Comparisons with Languages from the Outer Achaemenian Domain and Beyond

Comparison of the inflectional morphology of Middle Persian, Parthian, and Bactrian with their other known contemporary Middle Iranic languages to the east offers a strong indication that it was social mixture and consequent mass nonnative acquisition fostered by Achaemenian society that generated grammatically reduced regional Iranic languages.Footnote ³⁵ We see, roughly, that the farther they lay from the centers of Achaemenian power, the more inherited inflectional complexity they retained. Such observations do not provide for an exact science, but they nevertheless show a striking correspondence between the grammars of the specific languages and the social-linguistic principles introduced in Chapter 2.

Sogdian was an Iranic language spoken to the north of Bactria in what is today eastern Uzbekistan, in a prosperous country of developed towns along the Zerafshan and Kashka rivers. The region of Sogdia was thus on the northeastern outskirts of the Achaemenian domain. Sogdian came to be written in a derivation of the Aramaic script, like Middle Persian and Parthian. There are a few short Sogdian inscriptions surviving from as early as the second century ce,Footnote ³⁶ but the oldest extant substantial texts are remarkably well-preserved letters from the early fourth century ce, found in the ruin of a Chinese frontier fortification. Still in the time of these letters, contemporary with Middle Persian, Parthian, and Bactrian texts and their reduced grammar, Sogdian retained two to six of the Common Iranic noun cases, depending on the noun, as well as the three common Indo-European grammatical genders (masculine, feminine, and neuter), and a specialized function of the dual inflection for counted nouns.Footnote ³⁷ Like Old Persian and Avestan, Sogdian retained the imperfect verb and elaborate distinctions of verbal mood. Sogdian also underwent various other modifications conditioned by the position of the ancient lexical stress accent: when it fell on inflectional endings, those endings tended to be preserved, but when it was on noun or verb stems, the endings were reduced or lost, as in Middle Persian. This meant the development of two separate inflectional classes of verbs and nouns depending on the position of the accent, one more like that of Middle Persian and the other more conservative. Along with this, Sogdian exhibits further modest reduction as well as distinct innovation.Footnote ³⁸ The conservative features of early Sogdian as well as its complexifying innovations imply that it was a language learned by relatively few outsiders before the fourth century ce. When Alexander’s armies arrived in Sogdia, they reportedly required a special interpreter for the speech of the people of Samarkand, suggesting that the Iranic speech of this outlying country remained distinct at the end of the Achaemenian period.Footnote ³⁹

Chorasmian, a close cousin of Sogdian, was spoken in the towns and farms along the river delta that formerly existed to the south of the Aral Sea, in northwestern Uzbekistan. This was the northernmost province of the Achaemenids, virtually an oasis of settlements surrounded by desert and seashore, connected to the rest of the empire by the Amu Darya river flowing from Bactria to its south, north of the oasis of Marw. The extant ancient Chorasmian texts are very few and limited, telling us little about the grammar of the language in antiquity. Whatever Chorasmian texts were written are now mostly lost. Much later, a few thirteenth- and fourteenth-century Chorasmian texts preserved in the Arabic script are the most extensive and latest known texts in language. These attest to the state of Chorasmian after centuries of Arab, Turk, and Mongol invasions and many changes of government. Even in this much later period, Chorasmian still retained two genders, five noun cases (though with only two or three inflectional endings between them for any declension), the imperfect verb along with the present stem as well as the expected preterit stem, and numerous innovations.Footnote ⁴⁰

Because extensive, sudden grammatical reduction is induced by mass adult acquisition, any hypothesis about the nonreduction of Sogdian and Chorasmian should be contextualized with evidence that these two languages were more isolated from numerous outsiders than Persian, Parthian, and Bactrian. In these cases, it is obvious: Sogdia and Chorasmia were remote northeastern regions of the Achaemenian Empire and were probably not places receiving frequent visitors from the main part of it. La Vaissière characterizes Achaemenian Sogdia as “a distant frontier province” governed from Bactria “in constant contact with the nomads of the steppe.”Footnote ⁴¹ Its provincial isolation would limit the causes of linguistic reduction that prevailed in imperial centers to the west and south. Chorasmia may even have been more in the nature of a client country peripheral to the Achaemenian domain rather than a tightly integrated province.Footnote ⁴² A few scholars have supposed that it became independent sometime during or after the reign of Artaxerxes II, which would naturally cut it off from some of the causes of grammatical reduction proposed here.Footnote ⁴³ In the end, we know very little about the social history of the speakers of Sogdian and Chorasmian, apart from what the conservatism of their inflectional morphology shows us. Here, it is the linguistic theory applied to these two languages that offers insight into the history of the populations of their speakers. Yet there are many possibilities. Perhaps the movement of nomadic speakers of relatively unreduced Iranic languages into these regions could have displaced older, more reduced varieties of regional Iranic speech that are not otherwise attested. Such speculation aside, what the extant data do show is a couple of languages infrequently learned by adults from other countries.

Two other Middle Iranic languages are attested from a region never controlled by the Achaemenids. Khotanese and Tumshuqese were Iranic languages used outside of the Achaemenian domain, spoken by ancient immigrants who founded settlements at the desert oases in what are today the western edges of the Xinjiang Uygur Autonomous Region of the People’s Republic of China. These languages remained in use there for centuries. Tumshuqese is too poorly known to form a useful comparand here, but Khotanese illustrates the present point clearly. The Iranic language of Khotan was used by a local monarchy and for an extensive Buddhist literature in the first millennium of the common era, contemporary with texts extant in Middle Persian, Parthian, Bactrian, and Sogdian. Substantial Khotanese texts survive. Early Khotanese texts distinguish two grammatical genders, masculine and feminine, with traces of the old neuter gender, and at least six different ancient noun declensions (a-, ā-, i-, nd-, n-, and r-stems) as well as innovated new declensions. Nouns differ by as many as six inherited cases. Although the ancient imperfect verb has been lost, verbs generally are inflected for all the moods attested in Avestan, with separate active and middle endings for several conjugations. The level of inflectional complexity is very high compared with Bactrian, Parthian, and Middle Persian.Footnote ⁴⁴

Modern Iranic languages that evolved in places remote from the centers of ancient Achaemenian power show similar distinctions in inflectional complexity. Yaghnobi has already been mentioned, but we can add the testimony of Pashto, the Iranic language of a numerous population who had been, until the eighteenth and nineteenth centuries, “people at the margins” of empires, as Thomas Barfield put it, living in tribes based on egalitarian lineages.Footnote ⁴⁵ D. N. MacKenzie sketched the “remarkably complex nominal and verbal morphology” of Pashto, spoken today by millions in southern and eastern Afghanistan and western Pakistan.Footnote ⁴⁶ Pashto retains the ancient Iranic masculine and feminine genders, the ancient class of ā-stem nouns as a separate declension, and, remarkably, traces of the ancient pattern of accentuation on some words. Georg Morgenstierne wrote, “Nowhere in Iran, and hardly in any modern I(ndo-)E(uropean) language, do we find such a bewildering variety of nominal plural endings” as in Pashto.Footnote ⁴⁷ Much of this variety is innovation, not due to inheritance from Common Iranic, but MacKenzie remarked casually that “If we compare the archaic structure of Pashto with the much simplified morphology of Persian, the leading modern Iranian language, we see that it stands to its ‘second cousin’ and neighbour in something like the same relationship as Icelandic does to English.”Footnote ⁴⁸ Such remarks by specialists in Iranic languages reveal unstated suspicions about the social factors that play a role in conditioning the morphology. MacKenzie’s remark is cited by McWhorter in his study of the contrast between Pashto and New Persian, which singles New Persian out as having been reduced by nonnative acquisition (see Chapter 2).Footnote ⁴⁹ Pashto looks like a language removed from the morphology-reducing effects of the sudden arrival of populations of foreigners at Iranophone sites, induced by the reign of the Achaemenids.

The Hypothesis of Central Iranic and Areal Features

Alternatively, one could imagine that both Parthian and Bactrian became grammatically reduced only after the downfall of the Achaemenian Empire and before both languages are attested. For this to work in the case of Parthian, we would have to posit that it shed its complex inflections between Alexander and the earliest attested Parthian texts from the first century bce, under Seleucid rule. This seems much less plausible, unless we assume that few foreigners learned the ancestor of Parthian during Achaemenian times, and the reduction occurred only when many Greek-speaking newcomers in the wake of Alexander learned Parthian at once. In that case, we should expect to find features from Greek transferred into Parthian, and Parthian words adopted in Greek. But we do not find them. The unattested Old Parthian is therefore likely to have been restructured into the Parthian we know in tandem with Old Persian’s transformation into Middle Persian, in the Achaemenian period. The morphology of Bactrian may have diminished from its prior Common Iranic state during the comings and goings of invading peoples and new dynasties from Alexander to the Kushans. The linguistic model suggests that people at social crossroads should be expected to speak languages with reduced inflection, regardless of other factors. But it is less plausible that Middle Persian, Parthian, and Bactrian lost the same features in the same way, and generalized the same features in morphological reduction, for entirely separate reasons that left no trace in each instance. This suggests common circumstances in the formation of each of these languages.

All offshoots of a prior single language initially share in its earlier innovations. What they retain of these innovations is a sign for historical linguists of the former unity of their speech community. That is, shared morphological innovations in two related languages strongly indicate common ancestry in a society of people who could speak to each other and understand each other because they spoke more or less the same language, a language that changed eventually for all its speakers systematically, as languages do. Therefore, shared morphological innovations serve as a strong criterion for determining the affiliation and the order of separation of the descendant languages from their common stock. When shared innovations cannot be accounted for by the comparative method, however, because they seem to have crossed language boundaries or to have arisen separately in two related but adjacent languages, historical linguists invoke areal features as described. Areal feature is a term describing an outcome, not the process by which they originated. As already discussed, to call on areal features as an explanation is to assert a specific social scenario of multigenerational bilingualism, because only bilinguals can transfer features like morphological innovations from one language into another. It is now generally agreed that even fundamental features of morphology are susceptible to transfer,Footnote ⁵⁰ but one expects this only when bilingualism with the two languages involved is widespread in a society.

In 1976, Robert Hetzron rightly insisted that shared morphological innovation is an especially strong criterion for determining the affiliation of the Semitic languages.Footnote ⁵¹ His article expressing this had the effect of rewriting the family tree of the Semitic languages, relocating Arabic away from the branch of the family called South Semitic, where it had been regarded as closely connected with the Ethiopic languages, into a new Central Semitic group, together with Hebrew and Aramaic. In the decades since then, his revised model has become the consensus on the reconstructed Semitic family tree and Central Semitic is a generally acknowledged subgrouping of Semitic. Not dissimilarly, in an article of 2016, Agnes Korn used the new data from Bactrian and the principle of shared morphological innovations to propose a Central branch of the Iranic languages.Footnote ⁵² Her partially revised family tree, which focuses on the Middle Iranic languages, is based on phonological and morphological innovations shared between Parthian, Bactrian, and Sogdian. These three appear to share not just grammatical reduction but also certain common innovatory features. Applying the principle that affinity is demonstrated by shared innovations, she shows that Sogdian probably diverged earlier from Bactrian and Parthian. (Previously, Bactrian had been assumed to belong to an Eastern Iranic branch of languages, a subgrouping that no longer has any sure empirical basis.)Footnote ⁵³ Korn’s proposal of a Central Iranic subgroup is promising, and it appears that more insights are yet to come from this direction.

One noteworthy aspect of her proposed configuration of the branching trees leading to Middle Iranic is, however, the inference of areal features. In Korn’s analysis, Middle Persian clearly belongs to a separate branch of Middle Iranic. This coheres with the long-standing recognition of decisive phonological isoglosses that mark Persian (and its phonologically closest cousins, called Perside languages) apart from its neighbors to the north, discussed in Chapter 3. For example, Middle Persian has dil for “heart” whereas Parthian has zird, both from an earlier *źṛd (cf. Avestan zǝrǝd-, Balochi zird,Footnote ⁵⁴ Pashto zṛǝ).Footnote ^55, Footnote ⁵⁶ The initial consonant of the Persian word sets the history of other Iranic languages that systematically share this feature into a group apart from their relatives but together with Persian. At the same time, however, Korn recognizes affinity between Persian and Parthian, and to a lesser extent Bactrian, crossing between the separate branches and making the lines of descent unclear. Without many specifics, she turns to areal features in her explanation for these phenomena: “there is sufficient motivation at hand to justify the attempt of a different family tree, and of an approach that combines it with areal features.”Footnote ⁵⁷ Not only does she hypothesize areal features shared between the three languages under discussion here – Middle Persian, Parthian, and Bactrian – but also another set of areal features shared separately among Bactrian and “some Pamir languages” of the Iranic family. In her second contribution directed at this problem, Korn proposes that a wave model, based on the geographical distribution of isoglosses among the Iranic languages, may describe their interrelationships better than the traditional branching tree model. She suggests that apparent waves of features that spread from Persian are “possibly attributable to the periods of Persian as official language in Achaemenid and Sasanian times.”Footnote ⁵⁸ Although in neither time was Persian the “official” language, strictly speaking, of the domain ruled by the Persians – there was no constitution or decree stipulating an official language – the basic idea has been corroborated by my findings in the present study. All one needs to do is change the word from “official” language to “ruling-elite” language.

Korn has opened a new and exciting phase in the study of Iranic historical linguistics that will require more research for its ramifications to be traced. It is likely to change our comprehension of many specific and general features in the historical relationships of the Iranic languages. Readers can join me in holding high hopes for more results along these lines. The present concern, however, is with reconciling social history and linguistic history. Again, areal features automatically imply long-term bilingualism or multilingualism in some past period. Korn’s hypothesis of widespread areal features shared in the formation of Middle Persian, Parthian, and Bactrian, which she arrives at by the methods of comparative historical linguistics, matches the present thesis based on correlation between the linguistic models and the history of the inflection of these languages. Areal features point to social factors and language contact, specifically bilingualism, and the restructuring and diminishment of inflectional morphology in these Iranic languages was due to social factors. The period in which these factors played a role must have been one in which the populations that spoke the immediate ancestors of Middle Persian, Parthian, and Bactrian were united in one political domain, before they came to be attested directly. That can be only the Achaemenian period.

Like the concept of “Central Iranic,” or indeed that of any proposed linguistic group, the argument I have advanced here is ultimately inferential,Footnote ⁵⁹ but it does find support from these distinct lines of inquiry. It helps us to reconcile known ancient history with specific characteristics of the languages involved. In the comparative method of historical linguistics, shared innovations typically indicate a common social historical stage. The common loss of final syllables after or in tandem with the common repositioning of the accent in Persian, Parthian, and Bactrian therefore suggests a common social history. A major theoretical question requiring further research is whether co-occurring reductions or losses of the same sets of features in different systems can be diagnostic in the strong way that shared phonological and morphological innovations are.

As discussed here previously, languages do not inherently and automatically undergo a cycle of decay and regeneration of inflectional morphology. These variations over time correspond only with the social history of the speakers of the languages. This may make sense of the evolution of the generality of the Iranic languages. For a time, many Iranic language-speakers, with the Persians leading them, presided over and participated in the most extensive empire in the history of the world until that time. An age of population movement and mixing set in, not among all the subject peoples but among the beneficiaries of the empire and their polyethnic servitors, leading to grammatical reduction and widespread areal features among the Iranic languages of the masters. In subsequent ages, speakers of these languages were only rarely united in one political domain. Meanwhile, the Iranic languages spoken in places remote from population movement remained more morphologically conservative. Some, like (Proto-)Balochi, took on a much-reduced form – in this case extremely close to ancient Parthian – but subsequently evolved into new levels of inflectional complexity, “mutating normally,” in McWhorter’s catchy phrase,Footnote ⁶⁰ among relatively insular pastoralists during and after their migration to the region now called Balochistan. On the hypothesis that the Achaemenian Empire created the circumstances for the reduction of the Iranic languages, as postulated first by Meillet, these modern Iranic languages should serve as important examples in future studies of the gradual growth of inflectional complexity from reduced states.

An Achaemenian Common Tongue?

The ancient Greek scholar Eratosthenes (circa 275–circa 195 bce), in a report preserved by Strabo, stated that the geographical name Ariana (ἡ Ἀριανή), which referred to a region of what is today western Afghanistan, could be extended to include the Medes, Persians, Bactrians, and Sogdians, “for they speak just about the same language, with small differences.”Footnote ⁶¹ The remark is clearly based on an awareness that the term Arya referred also to a larger group of people, whom we designate as ancient Iranic language-speakers. Eratosthenes surely derived this report from an earlier source, so it probably reflects the linguistic situation around the end of the Achaemenids’ dynasty. As Panaino rightly recognizes, this ambiguous statement may be interpreted variously.Footnote ⁶² The possibility that jumps to mind first is that the four peoples named here, for each of whom a distinct Iranic language is attested in later centuries, were indeed recognized as speaking closely cognate languages with a common name, Arya. Inscriptions in two of these languages, Old Persian (Bisitun, 521 bce) and Bactrian (Rabatak, second century ce), both refer to their own language as “Arya,”Footnote ⁶³ suggesting that this was indeed the name of the common ancestor of the Iranic languages (spoken by the people known in Avestan as airiia-), and that at least two of these languages retained that common name as they diverged into separate species over time. Eratosthenes says that this name may apply to all of them. Panaino doubts this interpretation because he holds that Sogdian, Bactrian, Median, and Persian could not have been mutually comprehensible at the end of the Achaemenian period. Amid other speculation, Panaino offers a second plausible hypothesis to explain the statement of Eratosthenes, and this is the interpretation that he prefers. If the Iranic languages of the four peoples named must have been mutually incomprehensible, as he supposes, then the claim that they all shared more or less the same language would imply the existence of a “prestigious passpartout” variety of the Persian language, a common Iranic tongue.

Both hypotheses have explanatory merits, despite Panaino’s doubts about the former. It is entirely possible that the observer originally behind this report reached the conclusion that four ancient regional Iranic languages, despite their evident differences, were basically similar enough to consider them a unit, approximately, and that would help to explain why these four were all known as Arya. The Greek expression “with small differences,” παρὰ μικρόν, may simply have been an exaggeration by somebody who noticed shared vocabulary and cognates either without any profound knowledge of any of these languages or, just as plausibly, with knowledge of all of them. In my life, I have heard scholars claim casually that the Romance languages are “basically all the same,” even though we observe their mutual incomprehensibility, or quite limited mutual comprehensibility. Such remarks were also possible for polyglots in distant antiquity. But it is also possible that Panaino is right, and that a common lingua franca variety of Persian was employed across all these domains. A further complication is that, although Eratosthenes or his informant must have intended only one of these possibilities, both interpretations still may have been true.

Underlying Panaino’s judgment against the first possibility is the widespread but unsupported assumption that language changes are uniformly gradual over time. This is a problem discussed already in Chapter 2. It is impossible to be certain about the pace of change in languages from an unattested, prehistoric state (although sweeping simplification implies rapid change in one episode). The social-linguistic considerations suggest that all these languages may have been mutually comprehensible at an early time in the Achaemenian period, and they may have become quite different by the end of the fourth century bce. Yet it is precisely the role of Old Persian as a common medium between communities that must account for the degree and rapidity of its restructuring during the same time. Pavel Lurje and Ilya Yakubovich refer to a “Late Old Persian koine,” apparently positing a similar concept,Footnote ⁶⁴ and recently Yakubovich more explicitly suggests the existence of a “lingua franca” of “the Achaemenid provincial elites, which can be called for convenience Peripheral Old Persian, and which anticipated the evolution of Late Old Persian,” without elaborating its character.Footnote ⁶⁵ This is almost exactly the hypothesis I have been discussing at length. If Old Persian had not been widely adopted by many adult non-Persian-speakers, then there would not be such an enormous discrepancy between the Old Persian of the time of Darius I and that of Artaxerxes III, and we would be at a loss to explain the widely noted early simplification of the inflections of ancient Iranic languages among all the other Indo-European languages. Panaino supposes further, as most plausible among the possibilities, that a form of Old Persian was known widely wherever Persians ruled, and that “necessarily, [this widely used form of Old Persian] should have been known for practical reasons throughout the whole empire, in particular in the higher strata of the Irano-Aryan ethne and in the local administrations of the different satrapies.”Footnote ⁶⁶

This is supported by anecdotes like one in Xenophon’s Anabasis, in which Xenophon reports that he relied on a Persian-speaking interpreter to communicate with an Armenian village leader and a group of Armenian women at a well in 400 bce.Footnote ⁶⁷ Here, a form of Persian was the linking language between a Greek mercenary leader and Armenian locals. Neither of the two parties included native speakers of Persian, but Persian was their common medium. Circumstances like these must have been quite common across the Achaemenian Empire. Indeed, native speakers must have been a diminishing proportion of the total number of Persian-speakers for as long as the importation of fighters, workers, slaves, eunuchs, and concubines by the hundreds and thousands was ongoing. Such persons as Xenophon’s interpreter and his Armenian interlocutors would not require the highly inflected noun and verb endings natural to the language of Darius I. They simply needed to communicate, and to communicate simply. In conclusion, I do not know exactly what Eratosthenes’ remark means, when he says that these peoples could all be called “Arya” because they speak the same language, with small differences, but either of the two most likely interpretations supports the arguments I have been making.

The Putative Role of Aramaic as an Achaemenian Lingua Franca

Meanwhile, some historians of the Achaemenids have held that the language of written communication among the Persian rulers and their agents was not Persian but Aramaic, even claiming that Aramaic was the lingua franca of the Achaemenian Empire. The latter cannot be true in the conventional sense of the term lingua franca.Footnote ⁶⁸ It applies only with respect to writing for bureaucratic purposes, and, even then, certainly not in every domain. The languages of the provinces continued to be written, as extant texts and subsequent linguistic history show. There was no Achaemenian constitution declaring Aramaic to be the “official language,” though many scholars have stated that it was the official language of the empire (thus ignoring the languages of the royal inscriptions and many texts in different languages used for government purposes under Achaemenian rule).Footnote ⁶⁹ One should specify, rather, that it was a “highly standardized supra-regional chancellery language,” as Holger Gzella puts it.Footnote ⁷⁰ Jan Tavernier meticulously shows that Aramaic seems to have been used particularly for writing at the satrapal level of administration, meaning that of provincial governors. Documents serving lower levels of administration were written in regional languages.Footnote ⁷¹ Aramaic letters and documents reflecting the Achaemenian government and its activities have been discovered at opposite ends of the empire, in southern Egypt and in Bactria, where accidents of climate preserved them. These exhibit a remarkably uniform language of business and correspondence. All the Iranic languages that came to be written in antiquity, except perhaps for Bactrian,Footnote ⁷² were rendered in regional varieties of the Aramaic secretarial script of Achaemenian times, which evolved into distinct letter-shapes in the post-Achaemenian age of regional powers. This indicates the widespread use of Aramaic for the ends of the government scribes at a certain high administrative level, but that is all. One might guess that the large-scale importation of eunuchs from Aramaean Assyria and Babylonia may have provided personnel for this kind of standardized Aramaic chancellery with uniform writing, orthography, and style, for some time, but we are not well informed about the subsequent history of this kind of literacy.

In any case, it is quite clear from several facts that Aramaic was never a general vernacular lingua franca across the domains of the Achaemenids.Footnote ⁷³ First, Aramaic did not become a common language of any non-Aramaean population, Iranic-speaking or otherwise, under the Achaemenids, except the Judaeans and Babylonians, peoples who already had intensive contacts with Aramaic-speakers under the Assyrian and Babylonian Empires.Footnote ⁷⁴ As just mentioned, Xenophon used a Persian-speaking interpreter to communicate with an Armenian village leader, not a Syrian-speaking (Aramaic-speaking) interpreter. No language of the populations ruled by the Persians, such as Armenian, or indeed any Iranic language, is full of words adopted from Aramaic from this period, as we should expect to find if Aramaic was the common spoken tongue.Footnote ⁷⁵ Instead, Armenian is full of words adopted from Persian and Parthian – a sign of widespread unilateral bilingualism, wherein Armenians knew Western Iranic languages but not vice versa.Footnote ⁷⁶ Old Iranic words were adopted and naturalized in Aramaic already in Achaemenian times.Footnote ⁷⁷ By contrast, Middle Persian exhibits practically no Aramaic adopted words except specifically in Manichaean texts, many of which are found in translations from Aramaic and so probably do not represent ordinary Middle Persian usage; rather it is the transfer of Aramaic terms for specifically Manichaean phenomena, including technical terms of the religion, into Manichaean Middle Persian.Footnote ⁷⁸ It seems likely, therefore, that outside of Syria and Iraq, where it was becoming the common vernacular already before Cyrus, Aramaic was, as far as the government was concerned, no more than a language of professional scribes and secretaries. At most, it was a language used to link levels of administration under the Achaemenids, but even this is not certain.Footnote ⁷⁹ The reports that are available about the use of language in the Persian Empire do not characterize Persian elites as using Aramaic as their primary language. Nevertheless, there is no doubt that Aramaic was used for correspondence, record-keeping, and formal documents. Such texts do survive in sufficient numbers to be sure of this.

Szemerényi proposed that the “decay” and “disintegration” of inflections in western Iranic languages was due to the role of Aramaic, which, he supposed, formed the influential member of a Sprachbund (linguistic area) with Middle Persian and Parthian.Footnote ⁸⁰ Although we can assume that Babylonians and other Aramaic-speakers played a role, among many others, in the formation of Middle Persian, as adult learners of Persian, the case is not so decisively clear as Szemerényi held. More recent research has amply demonstrated that Aramaic was affected by the transfer of features from western Iranic languages into Aramaic, but it is not apparent that the reverse was true.Footnote ⁸¹ We can abandon the hypothesis that Old Iranic languages were restructured due specifically to bilingualism with an “official” Aramaic. Mass nonnative acquisition reduced the Old Persian morphology, not areal phenomena due to bilingualism with Aramaic.

Socially Remote and Socially Common Iranic Languages

Viacheslav Chirikba remarks about the languages of the Caucasus that “It has been noted that multilingualism in the Caucasus has a certain vertical dependency: the population of the mountains knew the language(s) of their heterolingual neighbours in the foothills and lowlands, but not vice versa.”Footnote ⁸² This condition of unilateral bilingualism is likely also to have been common throughout the domain of the Iranic languages in antiquity as it is today, for the signs of this occur in their relative grammatical complexity and simplicity and the direction of the movement of word-adoption (“loanwords”). Most, if not all, of even the most socially remote modern Iranic languages – remote by geography or by community boundaries, such as endogamous religious or tribal boundaries – have received vocabulary from New Persian (in its different regional varieties, including the ones called Tajik and Dari today), because speakers whose dominant language was not New Persian have been likely to learn New Persian as a vehicle for socially transcending their locale and gaining access to a wider range of goods and opportunities. New Persian-speakers, by contrast, generally have not needed those parochial languages used only in a remote place or in a cluster of villages. In a way supportive of the present argument, New Persian inflectional grammar is famously simple, while the Iranic languages not learned by outsiders often retain older grammatical inheritances, such as distinct grammatical genders, lost in Persian and Parthian already two thousand years ago. This is due to the distribution of multilingualism in Iran. Learners of Persian as a second language have steadily maintained its inflectional simplicity, while the relative lack of learners of the socially remote modern Iranic “dialects” and non-Iranic languages of the region have not subjected them to imposition by outsiders to such an extent.Footnote ⁸³ This is a description of universal tendencies, which will certainly vary in each case, according to their respective social histories. It is possible, nevertheless, to see a correspondence between the social connectedness and remoteness of the speakers of Iranic languages today and their grammatical complexity.

This chapter has shown that Iranic languages other than Persian were affected by changes very close to those that befell Persian in the Achaemenian period. The causes of the morphological reduction in these cases were the same: mass nonnative acquisition triggered by the demographic upheaval induced by empire. Areal features are clearly a factor, too, but differently. The strongest argument for the role of the Achaemenian Empire in the common linguistic effects is that the Iranic languages spoken on the Achaemenian periphery or beyond its domain were not affected in the same way, retaining a relatively high degree of morphological complexity instead. These observations firmly support the thesis of Chapter 3.