the Iranian language of ancient Bactria (northern Afghanistan), attested by coins, seals, and inscriptions of the Kushan period (first to third centuries A.D.) and the following centuries and by a few later manuscript fragments. Bactrian is the only Middle Iranian language whose writing system is based on the Greek alphabet, a fact ultimately attributable to Alexander’s conquest of Bactria and to the maintenance of Greek rule for some 200 years after his death (323 B.C.).


BACTRIAN LANGUAGE, the Iranian language of ancient Bactria (northern Afghanistan), attested by coins, seals, and inscriptions of the Kushan period (first to third centuries A.D.) and the following centuries and by a few manuscript fragments from a much later period, perhaps the eighth or ninth century. Instead of “Bactrian” some scholars have preferred terms such as “Greco-Bactrian” (emphasizing the use of a modified Greek script to write the language), “Kushan,” or “Kushano-Bactrian.” The name “Eteo-Tocharian,” despite its eloquent defense by A. Maricq (JA 248, 1960, pp. 162ff.), can hardly be justified (see W. B. Henning, BSOAS 23, 1960, pp. 47f.); in any case it is to be avoided in view of the risk of confusion with the non-Iranian language already generally known as “Tocharian.” A similar trap, into which some unwary bibliographers have fallen, results from the (long obsolete) use of the term “Old Bactrian” to refer to the Avestan Language.

Historical background. It is noteworthy that Bactrian is the only Middle Iranian language whose writing system is based on the Greek alphabet, a fact ultimately attributable to Alexander’s conquest of Bactria and to the maintenance of Greek rule for some 200 years after his death (323 B.C.). Soon after the middle of the second century B.C. Bactria was overrun by nomads from the north, notably by the Yüeh-chih or Tokharoi, who settled in northern Afghanistan and subsequently gave their name to the area (medieval Ṭoḵārestān). Early in the Christian era a tribe or family named Kushan obtained supremacy over the rest of the Tokharoi. The Kushan empire founded by Kujula Kadphises soon expanded into northern India.

Nothing is known for certain of the language of the Tokharoi/Yüeh-chih; in view of mounting evidence in favor of the much disputed connection of the Tokharoi with the inhabitants of Agni and Kucha in Chinese Turkestan, it is not unlikely that it was in fact related to the language which modern scholars have named “Tocharian.” (For some recent contributions to the long debate see Henning in G. L. Ulmen, ed., Society and History: Essays in Honor of K. A. Wittfogel, The Hague, 1978, pp. 215-30; H. W. Bailey, JRAS, 1970, pp. 121f.) It seems that the Kushans, upon becoming masters of Bactria, at first continued the traditional use of Greek as a medium of written communication. As a spoken language they had adopted Bactrian, the native idiom of the country, which they afterwards elevated to the status of a written language and employed for official purposes, perhaps as a result of increasing national or dynastic pride. The earliest known inscriptions in Bactrian (the “unfinished inscription” from Surkh Kotal, cf. A. D. H. Bivar, BSOAS 26, 1963, pp. 498-502, and the Dašt-e Nāvūr trilingual) belong to the reign of Vima Kadphises. A few decades later, early in the reign of Kanishka I, Bactrian replaced Greek on the Kushan coins. After this period Greek ceased to be used as an official language in Bactria (although later instances of its use by Greek settlers are known, cf. P. M. Fraser, Afghan Studies 3-4, 1982, pp. 77f.).

Sources. 1. Coins. The coins of Kujula and his immediate successors give the king’s name and titulature in Greek, often with a Kharoṣṭhī version on the reverse. The earliest issues of Kanishka likewise bear on the obverse his name and title in Greek (basileus basileōn kanēškou “[coin] of Kanishka, king of kings”), while the reverses portray divinities named in Greek as Hēphaistos, Hēlios, Nanaia, and Selēnē. Later issues follow the same pattern, but the legends are henceforward in Bactrian rather than Greek. Kanishka’s titulature appears (in its fullest form) as šaonano šao kanēški košano “of Kanishka, king of kings, the Kushan,” that of his successor Huvishka as šaonano šao ooēški košano. The forms kanēški and ooēški (also ouoēški, ooēške) are in the oblique case, cf. the Greek gen. kanēškou; a few coins of Huvishka have the nom. form (ooēško), as do those of the succeeding rulers, Vasudeva (bazodēo), Kanishka II (kanēško), Vasishka (bazēško, cf. R. Göbl, Dokumente zur Geschichte der iranischen Hunnen in Baktrien und Indien, Wiesbaden, 1967, III, pl. 8), etc. Of the numerous divinities depicted on the coins of Kanishka I and Huvishka, most are given Iranian names, e.g., ardoxšo (Av. ašiš vaŋuhi), aθšo “fire,” farro (Av. xᵛarənah-), lrooaspo (masc.; cf. Av. drvāspā-), mao “moon,” miiro (in many spellings; Av. miθra-), nana (Sogd. nny), oado “wind,” oaxšo “Oxus,” oēšo (Av. vayuš, conflated with the Indian Śiva, see H. Humbach in Monumentum H. S. Nyberg I, Acta Iranica 4, Tehran and Liège, 1975, pp. 402-08), ōoromozdo (Av. ahurō mazdǡ), orlagno (Av. vərəθraγna-), and teiro (Mid. Pers. tīr). See further F. Grenet, “Notes sur le panthéon iranien des Kouchans,” Studia Iranica 13, 1984, pp. 253-62. The coinage of Kanishka includes issues portraying boddo “Buddha,” sakamano boudo “Śākyamuni,” and mētrago boudo “Maitreya” (see J. Cribb, “A Re-examination of the Buddha Images on the Coins of King Kaniska. . .” in Studies in Buddhist Art of South Asia, Delhi, 1985, pp. 59-87), while that of Huvishka attests foreign gods and demigods such as ērakilo “Heracles,” sarapo “Sarapis,” maasēno “Mahāsena,” and skando komaro “Skanda Kumāra.” After Huvishka the repertoire of reverse types contracts sharply, the only deities named on the coins of the last Kushans being ardoxšo and oēšo.

With the eclipse of the Kushan dynasty their lands west of the Indus fell into the hands of the Sasanians, under whom the administration of these provinces was entrusted to a governor styled Kušānšāh. In addition to coinage inscribed in Pahlavi and (occasionally) in Brāhmī, the Kušānšāhs issued coins with legends in cursive Greco-Bactrian script. The language of these latter is usually said to be Bactrian. However, most of the “Bactrian” legends are virtually identical in vocabulary and phraseology to their Pahlavi equivalents. A legend such as bago pirōzo oazarko košano šauo “Lord Pērōz, great Kušānšāh” may be compared with the Pahl. mazdēsn bay Pērōz wazarg Kušān šāh, while the usual reverse borzaoando iazado (variously spelt) corresponds precisely to Pahl. burzāwand yazad. Thus it is possible to regard these legends not as Bactrian but as Middle Persian, superficially adapted to local orthographic norms. It should therefore not be too readily assumed that the Bactrian language possessed words such as oazarko “great” or iazado “god,” which are absent from the Bactrian of the Kushan period.

From the middle of the fourth century Bactria and northwestern India were overrun by Hunnish tribes, of whom the Hephthalites proved the most durable, maintaining their rule in parts of Afghanistan up to the Arab conquest in the seventh century. The “Bactrian” coin-legends and countermarks of the Hunnish period display a rich linguistic diversity, with titles deriving from Indian (sri “Śrī”), Turkish (kagano “Khaghan,” tarxano “Tarkhan,” todono “Tudun”), and even Latin (fromo kēsaro “Caesar of Rome”), as well as from Bactrian and Middle Persian (bago, šauo, xoadēo).

The plates accompanying R. Göbl, “Die Münzprägung der Kušān von Vima Kadphises bis Bahram IV,” in F. Altheim and R. Stiehl, Finanzgeschichte der Spätantike, Frankfurt am Main, 1957, pp. 173-256, offer a wide range of Kushan coin-legends; see also Göbl’s comprehensive System und Chronologie der Münzprägung des Kušānreiches, Vienna, 1984. The Kushano-Sasanian coinage may be studied from Bivar, ed., Corp. Inscr. Iran., part 3, vol. 6, portfolio I, London, 1968, pl. 4-10 (cf. Humbach, ZDMG 121, 1971, pp. 392f.). For the Hunnish coins see Göbl, Dokumente, esp. III, pl. 9-83; cf. also Humbach, MSS 22, 1967, pp. 39-56.

2. Seals. About forty seals inscribed with names and titles in Greco-Bactrian script have been published. The titulature attested on the seals is largely Western Iranian, e.g., asbarobido “chief of cavalry,” oazarko fromalaro “great commander,” šaurabo “satrap,” and uazaroxto “chiliarch.” (See Henning, ZDMG 115, 1965, pp. 80f.) Particularly important is the bulla of mauo kanēško “Moon-Kanishka” (ibid., pp. 85ff.; J. Brough in W. B. Henning Memorial Volume, ed. M. Boyce and I. Gershevitch, London, 1970, pp. 87f.). Almost all of the seals are illustrated by Bivar, op. cit., 1-4, 27 (cf. Humbach, MSS 25, 1969, pp. 65-74).

3. Inscriptions. By far the most important surviving monument of the Bactrian language is the monolith found at Surkh Kotal. The undamaged 25-line inscription records construction work carried out early in the reign of Huvishka by the margrave (karalraggo, cf. Henning, op. cit., pp. 77ff.) Nokonzoko. It was first published by Maricq (JA 246, 1958, pp. 345-440) under the title “La grande inscription de Kaniṣka” (due to an unfortunate misunderstanding of the passage in which Kanishka’s name is mentioned). Two slightly shorter copies of the same inscription, with interesting orthographic variants, were discovered later (see E. Benveniste, JA 249, 1961, pp. 113-52). The key to the interpretation of this text, and hence to the understanding of the Bactrian language, was provided by Henning’s identification of many of the most common words (BSOAS 23, 1960, pp. 47-55). Further refinements are due to I. Gershevitch (BSOAS 26, 1963, pp. 193-96; Asia Major, N.S. 12, 1966, pp. 90-109; Afghan Studies 2, 1979, pp. 55-73), J. Harmatta (Acta Antiqua Academiae Scientiarum Hungaricae 12, 1964, pp. 373-471), N. Sims-Williams (IF 78, 1973, pp. 95-99; BSOAS 48, 1985, pp. 111-16), and G. Lazard, F. Grenet, and C. de Lamberterie (Studia Iranica 13, 1984, pp. 199-232). The aberrant interpretation put forward by Humbach in Die Kaniska-Inschrift von Surkh-Kotal, Wiesbaden, 1960, and later works has now been abandoned by its author.

Of the minor inscriptions of Surkh Kotal only the “Palamedes inscription” (cf. Henning, BSOAS 18, 1956, pp. 366f.) and the end of the “Inscription pariétale” (cf. Benveniste, op. cit., pp. 146-50) add anything to our knowledge of the Bactrian language.

The trilingual inscription of Dašt-e Nāvūr (in Bactrian, Kharoṣṭhī, and an undeciphered script) is of potential rather than actual linguistic significance. Little can be certainly read beyond the name of Vima (ooēmo) and the date (including the Macedonian month-name gorpiaiou). See G. Fussman, Bulletin de l’Ēcole Française d’Extrême-Orient 61, 1974, pp. 1-75, and cf. Bivar, BSOAS 39, 1976, pp. 333-40.

The inscription of Aĭrtam is dated early in the reign of Huvishka and resembles the almost contemporary Surkh Kotal monolith in vocabulary (e.g., bagalaggo “sanctuary,” ma liza “the acropolis”) and phraseology, ending with the clear statement: oti eimo miirozada nibixto pido ia šod[.]a fromana “And M. wrote this at the command of (the?) Š.” See B. A. Turgunov, V. A. Livshits, and È. V. Rtveladze, Obshchestvennye nauki v Uzbekistane, 1981, no. 3, pp. 38-48; J. Harmatta in Studia Grammatica. Festschrift für Helmut Humbach, ed. R. Schmitt and P. O. Skjærvø, Munich, 1986, pp. 131-46.

The inscriptions of Delbarjīn (I. T. Kruglikova, ed., Drevnyaya Baktriya, 2 vols., Moscow, 1976, 1979) include substantial fragments of a monumental inscription of the early Kushan period (published by Livshits and Kruglikova, ibid., 2, pp. 98-112).

Several graffiti in cursive Bactrian script of the Kushano-Sasanian period have been found in the Buddhist cave-monastery of Kara-Tepe. Although they chiefly record names (borzomiro, bōzano, ōromozdo, etc.), words such as odo “and,” kaldo “when,” and malo “here” make it clear that their language is indeed Bactrian. They have been published by various scholars in B. Staviskiĭ, ed., Materialy sovmestnoĭ arkheologicheskoĭ èkspeditsii na Kara-Tepe 2-4, Moscow, 1969, 1972, 1975. For a similar inscription from Afrasiab see Livshits, ibid., 4, p. 50, n. 16.

Of the two inscriptions from Jaghatu, one, containing the triratna formula namōo boda, namōo dauarma, namōo sagga, may be regarded as Indian in Bactrian script. The other is largely incomprehensible, a remark which unfortunately applies to most late Bactrian inscriptions in cursive script, such as those from Uruzgan and the Tochi valley. Two very short rock-inscriptions from Shatial Bridge have been published by Humbach in Allgemeine und vergleichende Archäologie, Beiträge [des Deutschen Archäologischen Instituts] 2, 1980, pp. 210, 220.

4. Manuscripts. Of eight known manuscript fragments in Greco-Bactrian script, one (possibly of the fourth century) was found at Lou-lan and seven (of much later date) at Toyoq. Their content is largely obscure, though at least one, to judge from the words rakšazano šao “king of the rākṣasas” and rakšazanzo rākṣasī,” may contain a Buddhist text. All eight fragments are illustrated by Humbach, Baktrische Sprachdenkmäler II, Wiesbaden, 1967, pl. 28-32. Their reading and interpretation has been appreciably furthered by Gershevitch, IF 72, 1967, pp. 37-56.

More considerable is a unique fragment in Manichean script from Qočo (see idem, “The Bactrian Fragment in Manichean Script,” AAH 28, 1980 [1984], pp. 273-80; cf. also N. Sims-Williams, BSOAS 49, 1986, p. 588). Although the less legible verso side presents severe problems, most of the recto can be interpreted. The fragment contains many new words and grammatical forms (e.g., lh- “to give,” xwyn- “to call,” ʾzʾγdym “I have come”), while the Manichean orthography of words already known in Greek script (ʾs = aso “from,” ṯʾʾd = tado “then,” etc) provides valuable information on Bactrian phonology.

5. Indirect witnesses. The meager remains of Bactrian so far described are usefully supplemented by some identifiable Bactrian loanwords in other, better-known languages. As well as words already attested in Bactrian (e.g., Pers, xidēv < Bactr. xoadēo “lord,” Kroraina Prakrit personal name Vaǵamareǵa < bago “god” plus marēgo “servant”) these include some new vocabulary. Amongst the most certain items are Kroraina Prakrit lastana, Khot. lāstana- “dispute” < *lastano, Toch. A lāstaṅk, B lastāṅk “execution-block” < *lastaggo, Toch. A. ākāl, B akālk “wish” < *agalgo (see M. Schwartz in P. Gignoux and A. Tafazzoli, eds., Mémorial J. de Menasce, Louvain, 1974, pp. 399-411, and cf. L. Isebaert, De Indo-Iraanse bestanddelen in de Tocharische woordenschat, thesis, Leuven, 1980).

Script. The seventh-century Chinese traveler Hsüantsang reported that the language of Ṭoḵārestān employed an alphabet of 25 letters, evidently the 24 of the standard Greek alphabet with the addition of ž, sometimes referred to as san, to represent š. (Despite W. W. Tarn, The Greeks in Bactria and India, 2nd ed., Cambridge, 1951, pp. 508-10, the letter is probably not in fact san but a variant of rho, cf. Humbach, Baktrische Sprachdenkmäler I, 1966, p. 41). Of these letters, ksi and psi were not actually used for writing Bactrian, in which the sequences ks and ps do not occur. As in Greek, numerals are indicated by the letters of the alphabet, generally with the addition of some diacritic mark. For this purpose the extra letters digamma (= 6), koppa (= 90) (both possibly attested), and sampi (= 900) will no doubt have been required, as well as ksi (= 60) and psi (= 700).

For writing an Iranian language the Greek alphabet is not ideal. The Bactrians overcame some of its disadvantages by the introduction of the letter š and by the use of u (upsilon) for h. Ambiguities remain: Thus i (iota) represents both i, ī, and y; o (omicron) both u, ū, and w; a (alpha) both a and ā. Since there are no special signs for affricates, it is possible that s (sigma) and z (zeta) stand for c (ts) and j (dz) as well as for s, z/`. As in Greek, gg represents ŋg. Of the vowel digraphs, ei seems to stand for ī and ou for ū (also uh, etc.); the value of final -ēio is not yet clear. Concerning the function of the vowel letters in final position see below.

The decipherment of the Bactrian cursive script (for which see G. D. Davary, Baktrisch: ein Wörterbuch, Heidelberg, 1982, pp. 45-52) is probably not yet complete. In this style, chiefly but not exclusively attested in the post-Kushan era, the forms of certain letters (e.g., a/d/o) are wholly or partially confounded. The reading of the “rectangular” script found on most of the coins and inscriptions of the Kushan period is comparatively straightforward, though confusions due to the influence of the cursive or simply to the engraver’s ignorance are not infrequent.

The Manichean script of the manuscript from Qočo contains two letters with unusual diacritics, š and j. The function of the suprascript points is not clear.

Language. In the following paragraphs the symbols G and M will be employed to identify words cited from texts in Greco-Bactrian and in Manichean script respectively. Proposed phonetic interpretations will be indicated, where necessary, by square brackets.

1. Dialectal position. Bactrian belongs to the northeastern group of the Iranian languages. This is clear from certain features of its phonological development, e.g., ē and i from palatalized ā and a respectively; γd from *xt; c or s from *č; β and γ from *b and *g (even in initial position). The development of *d via *δ to l is as in Pashto and Yidgha-Munji; it can also be shown to have occurred dialectally in Sogdian (cf. Sims-Williams, JA 269, 1981, p. 353). In word-formation and syntax Bactrian shows some particular affinities to Sogdian, for instance in possessing a fem. suffix G -anzo (in rakšazanzo) = Sogd. -ʾnc. But it also has more westerly connexions, as indicated by forms such as M lh- “to give,” lʾhwʾn “gift” (Parth. dh-, dʾhwʾn) or by the employment of G i as eżāfa (cf. Persian and Choresmian). Bactrian thus “occupies an intermediary position between Pashto and Yidgha-Munji on the one hand, Sogdian, Choresmian, and Parthian on the other: it is thus in its natural and rightful place in Bactria” (Henning, BSOAS 23, 1960, p. 47).

2. Foreign elements. A non-Iranian element, presumably a remnant of the language which the Tokharoi had brought with them to Bactria, is clearly apparent in Kushan onomastics (Kujula Kadphises, Nokonzoko, etc.). The suffix -šk occurs frequently, especially in royal names: Kanēško, Ooēški, Bazēško, Kozgaški. Its similarity to the Kuchean (Tocharian B) suffix -śke (fem. -śka), the most characteristic name-formative in that language (cf. W. Krause, Ural-Altaische Jahrbücher 25, 1953, pp. 11ff.), can scarcely be coincidental (see V. V. Ivanov in Narody Azii i Afriki, 1967/3, pp. 106-18, English summary on pp. 234-35).

Perhaps surprisingly, Bactrian seems to have adopted little from Greek other than its alphabet. The month-name Gorpiaiou and the Greek divine names mentioned above come into the category of foreign vocabulary rather than that of loanwords. A more significant borrowing, if its derivation from Greek khrónos is correct (A. Thierfelder, apud Humbach, Baktrische Sprachdenkmäler I, p. 24, cf. also W. S. Allen, Vox Graeca, Cambridge, 1968, p. 41 n. 1), would be G xšono “year” (Khot. kṣuṇa-, Tumshuqese xšana-, Toch. B kṣuṃ, Kroraina Prakrit kṣunamy). The title G xoadēo “lord” (Sogd. xwtʾw, Mid. Pers. xwadāy, xwadāwan, etc., all from *xwa- “self” plus forms of *taw- “to be powerful”) is a calque on Greek auto-krátōr (A. Meillet, MSL 17, 1911, pp. 109-12).

The presence of Indian divinities on the coins of Kanishka I and Huvishka has already been noted. The much later manuscript fragments, like other texts from Central Asia, contain many Indian (Buddhist) loanwords, e.g., M mwwl “root,” nrḫ “hell,” pwwn “merit,” G *rakšazorākṣasa” (in plur. rakšazano) and perhaps *marano “death” (in adj. maraniggo, cf. Khot. maraṇīnaa-, Parth. mrnyn?).

3. Phonology. The development of *b, d, g to β, l, γ may be illustrated by M βrg “fruit,” G lizo “acropolis,” and M γʾʾw “cow.” Non-initial *t and *k become d and g (M ʾwd and ʾwṯ [ud] “and,” ywg “one”), while *-p- has further developed via *b to β (M yʾβ “water”). The affricate *č is de-palatalized, giving G s (e.g., sido “which,” aso “from”), which, to judge from M ʾcyd and ʾs, may sometimes represent [c] and sometimes [s]. Voiceless stops become voiced after nasals (G oanindo “victor,” bagolaggo [-ŋg] “temple”), while *ṇč presumably gives [ndz], written M nj (yʾwydʾnjyg “everlasting”), G nz or ns (cf. uastiloganzeigo beside astiloganseigi, meaning uncertain). Intervocalic *θ survives in G iθa “so;” the h of M lh- “to give,” etc., may be interpreted as a later development. However, the change of *θr to hr is early, cf., G miuro [mihr] “Mithra,” uirso [hirs] (< *hris) “30.” Bactrian h, of whatever origin, is unstable, so that one finds G arougo, mao, miiro beside uarougo [harūg] “whole,” mauo [māh] “moon, month” and miuro.

Certain consonant groups deserve special mention. While *xt becomes γd (e.g. G pidorigdo “abandoned”), *xšt becomes xt (G nobixto, M ṇβyxt- “written,” cf. Munji nəwuxt-). Simple *št sometimes remains (M rštyg “just,” etc.), but elsewhere it gives t (G xoto “dried up,” etc.), as does *t + t in secondary contact (G oto, oti “and,” etc.); see Sims-Williams, “A Note on Bactrian Phonology,” BSOAS 48, 1985, pp. 111-16. The place-name G lrafo (< *drafša-) indicates that * gives f. Initial * becomes š, cf. G šao “king,” etc. On (h)r < *θr see above, but note also rl < *rθr in G orlagno “Vərəθraγna” and xl < *xθr in G baxlo “Bactra” (via *x’(r), cf. Av. bāxδī-?). Metathesis, as in G alošxalo beside alaxšalo “merciful,” albargo “water-tank” (ālβārg < *āβ-lārg), or uirso “30” (cf. above), is not uncommon.

Palatalization affects both vowels and consonants. On the possible interpretation of G -ge, -gi, -ki as [gyy), ky] see below. The development of s and z after palatal vowels to M š and j/j, as in wyšp “all” and prdyjg “orchard,” may be late or dialectal, contrast G oispo etc. Palatalization of ā to ē and of a to i may be exemplified by the present stem -lēr- “to hold” (in G ablēr-, meaning unknown, M lynlyryg = Mid. Pers. dēndār) and the preposition G pido, M pyd; cf. Khot. pader- and väte respectively. A remarkable form is G nokonziki (i.e. [-iky] < *-əki?), obl. of the personal name nokonzoko. If the title G loixobosaro “helper of the country (?)” does indeed contain *dahyu-, the first syllable shows both i- and u-umlaut.

Long *ā generally remains unchanged in non-final syllables; *ai, aya, (i)ya give G ei [ī] and ē (e.g., ei(i)o, eido “this,” frei- “dear,” but zēnobido “armorer,” marēgo “servant”); *awa- gives ō- (G ōsogdo “pure”), while *-wa- sometimes remains uncontracted (G froxoaš- beside froxōš- “to withdraw,” xouzo, xozo [xūž] “well,” uarougo, uarogo (?) “whole”). G iōgo, M ywg “one” < *aiwaka-, cf. Munji yūγa, is a special case.

Vocalic * gives ir or ur according to phonetic context: G kirdo “made” vs. borzomiuro (personal name) < *bṛza-. The remaining short vowels *a, i, u are generally well preserved in stressed syllables (but note G boo- [βow-] “to become”). In unstressed syllables they tend to be weakened or lost (G -maggo [-māŋg] “-minded” < *-mānaka-, proclitic article mo < *ima-, zooasti, azooasto, ozooasto “led out” < *uzwā/ăsta-; before nd unstressed *a often becomes i as in G oanindo “victor,” M wyšpzʾnyndyg “all-knowing”). In the Manichean fragment most polysyllabic words (other than those ending in -y(y)(ḫ) = G -ēio < *-ai(C) have lost their former final vowels. It is not clear whether this development had already taken place at the time of the inscriptions. In these, every word ends in a vowel-letter, but the original final vowels are largely confused, -o tending to usurp the place of all others (hence iθo, sido, (plur.) bago beside iθa, sidi, bage, etc.). It has been proposed to interpret this -o, and similarly the -o- which occurs internally in place of an earlier unstressed *a or * i (e.g., G nošalmo beside nišalmo “seat,” bagolaggo beside bagalaggo, preverb poro- < *pari-), as a reduced vowel [ə] (thus: nəšalmə, βaγəlaŋgə, pərə[?]). But the fact that -o is added even to words already ending in a vowel (as in G -ēio < *-ai(C), namōo for Skt. namo) suggests that, at least in some instances, G o has no phonetic value, functioning merely as a word- or morpheme-divider.

See also G. Morgenstierne, “Notes on Bactrian Phonology,” BSOAS 33, 1970, pp. 125-31, and I. M. Steblin-Kamenskiĭ in Osnovy iranskogo yazykoznaniya II: Sredneiranskie yazyki, Moscow, 1981, pp. 335-41. For a different view on the interpretation of Bactrian vocalism, see G. Lazard, Studia Iranica 13, 1984, pp. 219-22.

4. Morphology. In its simplified morphology Bactrian resembles Western Middle Iranian. The only inflected form of nouns attested in the manuscript fragments is the plural in G -ano, M -ʾ(ʾ)n. Coins and inscriptions up to the time of Huvishka show vestiges of a two-case system: sing. -o, obl. -i or -e (e.g. karalraggi/°ge “margrave,” ooēški/°ke), plur. -e (bage “gods,” asagge “stones”), obl. plur. -ano (e.g., oadobargano “living beings;” note also šaonano, obl. plur. of the n-stem šao “king”). Since the obl. and plur. endings -i and -e occur only after k and g (γ), it is possible that they indicate palatalization of the preceding consonant rather than distinct vowels. Fem. ā-stems are sometimes spelt with final -a but more often with -o: fromano/°na “command,” lizo/liza “acropolis,” etc. Similarly, the articles i and mo have fem. forms ia and ma, but these are not used consistently. The neuter survives only in G sido, sidi, M ʾcyd “which” beside G kido, kidi, M kyd “who.” Other notable pronominal forms are G pideino = NPers. baḏ-īn and the third person enclitics, sing. G -ēio, M -y(y) (ḫ), plur. G -ano, which display generalization of *-hai, *-hā/ănām (as in Choresmian and Khotanese) rather than of *-šai, etc. (see Sims-Williams in R. E. Emmerick and P. O. Skjærvø, eds., Studies in the Vocabulary of Khotanese II, Vienna, 1987, pp. 74-75).

The verbal inflexion is very imperfectly known. Best attested is the 3 sing. pret., which consists of the simple past stem: G kirdo, kirdi, M qyrd “made,” etc. The 1 sing. (M ʾγdym “came”) and 3 plur. (G froxortindo/°di “withdrew,” etc.) are formed with the enclitic auxiliary verb “to be.” Note that the pret. of a trans. verb agrees with its logical object (the “ergative” construction). Of verbal forms based on the pres. stem, the following are securely attested: pres. indic. 3 sing. M -y(y)d (e.g., ʾβyryyd, “produces”), 3 plur. M -y(y)nd (e.g., ʾβyryynd, cf. perhaps G na-tirindo “do not go(?)”); opt. 3 sing G -ēio (e.g., booēio “may be”), 3 plur. G -ondēio, -indēio (froxoašondēio, froxōšindēio “may withdraw”).

5. Syntax. Three characteristic features, all of which appear also in Sogdian, may be mentioned here. (i) Two demonstrative elements may be used together, e.g., G eiio mo (whence also eimo?), eido ma. (ii) A preposition governing a series of appositional phrases may be repeated before each phrase, as in: G amo borzomiuro amo kozgaški pouro “with B. son of K.” (see Sims-Williams, “A Note on Bactrian Syntax,” IF 78, 1973, pp. 95-99). (iii) Throughout the great Surkh Kotal inscription the enclitic particle *uti “so” (> -do, -di, etc.) is attached to the first word or phrase of every clause, giving rise to a series of compounds such as kaldo, kaldi “when” (= Sogd. k’wty), to which alone enclitic pronouns can be added (e.g., kaldano). See Sims-Williams, “A Note on Bactrian Phonology,” BSOAS 48, 1985, pp. 111-16.



Most important publications are cited in the text. Fuller bibliography is given in G. D. Davary, Baktrisch: ein Wörterbuch, Heidelberg, 1982 (a work not without merits, but to be used with caution).

I. M. Steblin-Kamenskiĭ, “Baktriĭskiĭ yazyk,” in Osnovy iranskogo yazykoznaniya II: Sredneiranskie yazyki, Moscow, 1981, pp. 314-46, gives a balanced analytical survey of studies up to about 1977.

Cf. also I. Gershevitch, “Bactrian Literature,” in Camb. Hist. Iran III, pp. 1250-58, and N. Sims-Williams, “Bactrian,” in Compendium Linguarum Iranicarum, ed. R. Schmitt, Wiesbaden (forthcoming).

Figure 6. Sites at which Bactrian inscriptions have been found

(N. Sims-Williams)

Originally Published: December 15, 1988

Last Updated: August 19, 2011

This article is available in print.
Vol. III, Fasc. 4, pp. 344-349