ii. The Phonology of Avestan

Attested forms and stages of development. Avestan is attested in two forms, known respectively as Old Avestan (OAv.) or Gathic Avestan and Young Avestan (YAv.). They differ from each other not only chronologically but also dialectally. Avestan, which is associated with northeastern Iran, and Old Persian, which belongs to the southwest, together constitute what is called Old Iranian. It is possible to some extent to reconstruct Proto-Iranian by comparing Avestan with Old Persian. This Proto-Iranian is closely related to the Vedic language of ancient India. Both Proto-Iranian and Vedic go back to Proto-Indo-Iranian or Proto-Aryan, which in turn descends from Proto-Indo-European.

By comparison with Vedic, whose phonemes are consistently recorded, Avestan in the form in which it has been handed down in manuscripts from 1288 A.D. onwards is attested in a very irregular notation. Apart from errors introduced in the post-Sasanian period, the essential features of the manuscript tradition of the Avesta must have been present already in the Sasanian archetype. When the Avestan texts were first recorded, perhaps as early as the fourth century A.D., each sound of the current Avestan pronunciation was designated by a special letter. The fact that a phonetic notation was used rather than a phonemic one means that it is possible to assess the linguistic significance of the individual spellings with regard to both the synchronic description of the language and its historical development.

Every Avestan text, whether composed originally in Old Avestan or in Young Avestan, went through several stages of transmission before it was recorded in the extant manuscripts. During the course of transmission many changes took place.

For Old Avestan the following stages may be assumed: 1. The original language of the Zarathustrian Gāthās, the Yasna Haptaŋhāiti, and the four sacred prayers; 2. Changes involved by the practice of slow chanting; 3. Changes due to transmission by YAv. priests, who introduced many YAv. sound forms into the OAv. texts; 4. Deliberate alteration of the text in the course of an orthoepic revision (“School text”); 5. Continued transmission of the OAv. texts along with the YAv. texts.

Young Avestan went through the following stages: 1. The original language of the composers of grammatically correct YAv. texts; perhaps in Marv or Herat; 2. Dialect influences as a result of the transfer of the Av. texts to Southeast Iran (Arachosia?); 3. Transfer of the Avesta to Persis in Southwest Iran, possibly earlier than 500 B.C.; 4. Transmission of the Avesta in a Southwest Iranian theological school, probably in Eṣṭaḵr: Old Pers. and Mid. Pers. influences, the insistence on fantastic pronunciations by semi-learned schoolmasters (Av. aēθrapaiti-), the composition of ungrammatical late Av. texts, the adaptation of portions of texts taken from other regions where they were recited; 5. The end of the oral transmission: phonetic notation of the Avestan texts in the Sasanian archetype, probably in the fourth century A.D.; 6. Post-Sasanian deterioration of the written transmission due to incorrect pronunciation (Vulgate); 7. In the ninth and tenth centuries A.D. the manuscript copies of individual texts were made on which the extant manuscripts are based; 8. Earlier manuscripts were copied in manuscripts dating from A.D. 1288 till the nineteenth century by scribes who introduced errors and corruptions. These are the manuscripts extant today.

Many phonetic features can not be ascribed with certainty to a particular stage since there may be more than one possibility. Every phonetic form that can be ascribed to the Sasanian archetype on the basis of critical assessment of the manuscript evidence must have gone through the stages mentioned above so that “Old Avestan” and “Young Avestan” really mean no more than “Old Avestan and Young Avestan of the Sasanian period.”

The vowels. The Proto-Indo-Ir. vowels a, ā, i, ī, u, ū and the diphthongs ai̯, āi̯, au̯, āu̯ ( = Vedic e, ai, o, au) remained unchanged in Proto-Iranian. Proto-Indo-Ir. i that arose from Proto-IE. ə (the vocalization of a consonantal laryngal H) is attested by such forms as Av. pitár– “father;” OAv. sīšā “teach,” cf. Vedic śiṣat, from Proto-IE. *k̂əse-; OAv. –maidī, cf. Vedic –mahi, 1 plur. verb ending, from Proto-IE. *-med^hə. But Proto-IE H was maintained under certain accentual conditions in Proto-Ir. and was lost in Av. Hence we find such contrasting forms as Av. draonah- “possession” beside Vedic dráviṇas-; OAv. dugədar- “daughter” beside Vedic duhitár-; OAv. vərəṇtē “he wishes” beside Vedic vṛṇīte; OAv. fəδrōi “to the father” beside OAv. piθrē. Proto-Indo-Ir. sonant ṛ ( = Vedic ṛ) became in Proto-Ir. and Av. ər but before š the tradition introduced the YAv. spelling ar(š) also into OAv. texts. The corresponding long vowel ṝ from Proto-IE. ṛH developed in Proto-Ir. and Av. to ar whereas Vedic had either ir/ur or īr/ūr.

There is a wide variety in the representation of the vowels in the manuscripts. Most of these features were already present in the Sasanian archetype.

Numerous anaptyctic vowels, represented mostly by ə but also by a, ō, and other vowel signs, were used to simplify consonant clusters especially after r : arəθa-, karapan-, vīžibiiō, θβarōždūm, etc. These anaptyctic vowels were introduced during the course of transmission in order to account for the pronunciation used in the slow chanting of the texts.

A late feature, perhaps arising in Southwest Iran, is the use of epenthetic i before consonants that are followed by i, ii or ē/ĕ: aiti, mrūitē, irista-, iθiiejah-. This epenthesis is not found before ń, ŋ́, st, št, m, hm, but it does occur before rm: zairimiia- “house,” cf. Vedic harmⁱyá-; airime “quiet” beside armaē^o. Anaptyxis and epenthesis may occur together: YAv. kərəiti-; OAv. daibitā.

Epenthetic u occurs only before ru, rṷ: uruθβarə, pouru. It is a genuine YAv. development in the case of –uri– arising from older *-urṷi-: YAv. paoiriia- “first” from *paṷrii̯a- from older *parṷii̯a-, cf. OAv. paouruuiia-; YAv. tūiriia- “father’s brother” from *təurii̯a- from older *tərṷii̯a- and ultimately from Proto-Indo-Ir. *pHtṛṷii̯a-, cf. Vedic pitṛvⁱya-. (On the phonological status of epenthetic i and u, see Morgenstierne, “Orthography and Sound-system,” pp. 55-58 par, ix.)

There is a consistent pattern in the representation of the quantities of the vowels a, ā, ə, ə̄, i, ī, u, ū, e, ē, o, ō in final position: in OAv. they are always long, that is, both original a and ā are written ā, etc., while in YAv. they are always short, except for -ō, -ə̄ (the YAv. final -ą is always long) and in monosyllabic words. The short vowels were probably closed, the long open, as in Attic Greek. Hence in the Sasanian archetype short and long vowels were often used to indicate degrees of openness of the vowels rather than their quantities. Thus we find vīspa- “all” with ī indicating a (short) open vowel: cf. Vedic víśva-. Similarly explainable are the spellings of ahura- “lord,” with (long ?) closed u, beside the derivative ahūiri- with (short ?) open u (through dissimilation with the closed i ?); note also ao from *aṷ beside aē from *ai̯.

Qualitative changes are seldom found in the case of ī/ĭ and ū/ŭ but note OAv. ə̄nəiti- from *əniiti- and drəguuaṇt- from *drugṷant-. As in East Iranian dialects, Av. *-īṷ- became -uṷ- in juua- “living,” cf. Vedic jīvá-, and cuuaṇt- “how much, how big” from *čīṷant-.

Proto-Ir. a suffered very many changes: to ā, ə, ə̄, e, o, ō. These came about partly due to phonological development caused by the surrounding sounds, partly due to the liturgical chanting, and partly due to dialect influence. Before final -n and -m, a always became ə̄/ə̆. This was originally the case also in word-interior position but ə was often replaced by a in this position in YAv., from where it was introduced also into OAv. Thus both OAv. and YAv. have nəmah- beside manah-. Before -ŋh- (-ṇgh-) where the nasal is etymological (-ŋh- from *-ns-), OAv. has only ə̄/ə̆ in sə̄ṇgha- “pronouncement,” cf. Vedic śáṁsa-. But before -ŋh- where the nasal is secondary (-ŋh- from *-s-), OAv. has only a, e.g. in manaŋhā, cf. Vedic mánasā. In both cases YAv. has a: saŋha-, manaŋha. In final position *-ans became -ə̄ṇg in OAv. and -ə̄ in YAv.

In YAv. ə developed further to i after i̯, č, ǰ: YAv. yim beside OAv. yə̄m; YAv. drujim, beside OAv. drujə̄/ə̆m. Postconsonantal *-i̯ə- became first *-i̯i-, then YAv. -ī/ĭ-, which was introduced from YAv. into OAv.: YAv. ainim, OAv. aniiə̄m, ainīm, cf. Vedic anyám “other.” Similarly *-ṷə- became -ṷu and then ū/ŭ: YAv. tūm, OAv. tuuə̄m, cf. Vedic tvám “you.” Note that -aii̯ṷə- became –ōiiu- by umlaut: YAv. ōiium from aēuua- “one;” YAv. vīdōiium from vīdaēuua- “abjuring the devils;” YAv. Harōiium beside OPers. Haraiva-. An exception is daēum (not *dōiium!) from daēuua- “devil.” Proto-Av. *ai̯ə, *āi̯ə, *aṷə, and *āṷə before n, m were reduced in YAv. to the disyllabic diphthongs aē, āi, ao, āu respectively: YAv. aem beside OAv. aiiə̄m “this;” YAv. daēnā- “religion” (from *dai̯ənā-) was introduced into OAv.

Before *-ṷi-, a became ə: səuuišta- “strongest,” cf. Vedic śáviṣṭha-. In certain environments a became e: between i̯ and j, cf. iθiiejah- “abandonment;” between i̯ and a syllable containing ī/ĭ, ii, or ē/ĕ, cf. yesne, loc. sing. from yasna– “veneration.” In some environments a became o: between p, m, ṷ and a syllable containing u (but not ṷ): pouru “much;” mošū “soon;” vohu “good,” but there are exceptions: vaŋhuš “good;” pasu- “cattle;” maδu- “wine.”

Proto-Ir. *ā/ăn became ą before spirants x, θ, f, s, z, š, hi: mąθra- “sacred utterance,” cf. Vedic mántra-; ąsa– “party,” cf. Vedic áṁśa-; mąsta “he thought,” cf. Vedic maṁsta; ązō “narrowness,” cf. Vedic áṁhas-; dąhišta– “most versed,” cf. Vedic dáṁsiṣṭha-; vąs “he prevailed” (from *ṷānst); sąstā (2 plur. imv.) “appear,” cf. Vedic á-chāntta; frąš “forward” (from *prāŋ^kš).

In OAv. final -ah (cf. Vedic -aḥ from -as) became -ə̄ but it has in most cases been replaced by YAv. -ō. That even YAv. originally had -ə̄ (cf. Khot. -ä[ə] from *-ah) is indicated by such forms as YAv. vacə̄bīš (instr. plur.) based on nom. sing. OAv. vacə̄ (= Vedic vácaḥ).

We often find long, that is, open (back) ā instead of closed (front) a in initial position: ārmaiti- “right-mindedness,” cf. Vedic arámati-; kāuuaiiō “princes,” cf. Vedic kaváyaḥ “seers;” srāuuahiieitī “he desires fame,” cf. Vedic śravasyáti; hātąm (gen. plur.) “of the existing (ones),” cf. Vedic satāˊm. Note also ā for a after i and u: vii-ādarəsəm “I have seen,” cf. Vedic adarśam; vərəziiātąm, nīdiiātąm, višˊiiātā, hə̄miiāsaite, paitii-āmraot̰, aiβii-āma-, drəguuātā, drəguuāitē (but drəguuatō), etc.

Proro-Ir. *ai̯ usually becomes aē in open syllables (vaēdā/ă “he knows”) but ōi in closed syllables (vōistā/ă “you know”). In final position it appears as -ōi in OAv. but as -e in YAv. The spelling -ē in OAv. is due to YAv. YAv. has -ōi only in yōi and mai’iiōi. The dat. sing. forms OAv. axtōiiōi and YAv. anumatə̄e point to an original *-ə̄/ə̆i̯ə̄/ə̆i̯ from Proto-Indo-Ir. *-ai̯ai̯, cf. Vedic. -aye.

Proto-Ir. aṷ became ao, but before final -š it usually became ə̄u in OAv. and YAv.: gə̄uš, mańiiə̄uš, mərəθiiaoš. In final position -aṷ became sometimes -ṷō, sometimes -ō (cf. Vedic -o): OAv. huuō “yonder” (from *haṷ, cf. OPers. hauv); ərəzuuō (voc. sing.) “O straight one;” huxratuuō (voc. sing.) “O skilful one,” cf. Vedic sukrato; but mainiiō (voc. sing.) “O spirit,” cf. Vedic manyo; aŋhuuō (loc. sing.) “in the life;” gātuuō “in the place;” daŋ́huuō “in the land” beside daŋ́hō; haētō “on the bridge;” šˊātō “in peace;” vaštō “in the wish;” həṇtō “in gain,” cf. Vedic sánitau.

Many changes are found in the case of Proto-Ir. ā, e.g. ą: uruuąnō “souls” beside uruuānō (in final position always ąm, ąn); ǡ: mazǡṇtəm (acc. sing.) “great;” mǡŋhəm (acc. sing) “moon;” e: aiienī “I shall go,” cf. Vedic ayāni; zbaiiemi “I call,” cf. Vedic hvayāmi. Final -āh became -ǡ (cf. Vedic. -āḥ from -ās); sāsnǡ “commandments” (by analogy also sāsnǡs-ca “and commandments”).

Original ā is often shortened, as in dātaras-ca beside dātārō “creators;” -anąm, gen. plur. ending with disyllabic -ąm, cf. Vedic and OPers. -ānām; aētaŋhąm, cf. Vedic etāˊsām: -at̰.haca (instead of *-āt̰.haca), but -āat̰cā/ă in ašāat̰cā/ă.

Before a vowel, āi̯ and āṷ are often shortened to ai̯ and aṷ, a feature shared by Avestan with East Iranian dialects such as Sogdian: vaiiu– “wind,” cf. Vedic vāyú-; zaiiata “he was born,” cf. Vedic jāyata; -aiiǡ, gen. sing. fem. ending, cf. Vedic -āyāḥ; -aiiāi, dat. sing. fem. ending, cf. Vedic -āyai; nauuāza- “boatman,” cf. Vedic nāvājá-; yauuaṇt- “as great, as much,” cf. Vedic yāvant-; aṧauuā “righteous,” cf. Vedic ṛtāvā.

The consonants: (a) Semivowels. In the Sasanian archetype the semivowels i̯and ṷ were always written ẏ and v in word-initial position. These sounds probably represent an intermediate stage in the development of initial i̯ and ṷ to ǰ and b as seen in NPers. In the Indian manuscripts ẏ is replaced by y, whose original value was probably palatal žˊ. In medial position the manuscripts have ii, uu and not y, v as earlier transcriptions seemed to indicate, (e.g., vayu– for vaiiu-). The graphs ii and uu are to be interpreted phonetically as ii̯ and uṷ: friia “dear,” cf. Vedic priyá-; druua– “firm,” cf. Vedic dhruvá-. The fact that jiia “bowstring” and kuua “where” were disyllabic in YAv., cf. Vedic jⁱyāˊ and k^uvà respectively, is proved by their being written with a short final vowel, since the final vowel of monosyllables was regularly written long in YAv. ii̯ and uṷ may have developed in West Iran under the influence of Old Persian, where every postconsonantal ḭ and ṷ became iy and uv respectively: Av. ańiia- (from *anḭa-, cf. Vedic anyá-) like OPers. aniya-; Av. hauruua– (from *harṷa-, cf. Vedic sárva-) like OPers. haruva-. Even intervocalic i̯ and ṷ are sometimes written iy and uv in OPers. Thus Av. dāraiia- and bauuaiti correspond to OPers. adāraiya and bauvatiy. Note too that intervocalic ii and uu may even be etymologically justified: OAv. āiiāt from *āii̯āt, cf. Vedic iyāt, Av. sraiiah– “more excellent,” cf. Vedic śréyas- (from Proto-Indo-Ir. *śrai̯Hi̯as-); gauuāstriia- “belonging to the cattle pasture” from *gaṷ-ṷāstriia-. In the manuscripts the sequences –iiuu (from *-i̯uṷ-) and –uuii– (from *-ṷii̯) are usually simplified to –iuu– and –uii– or else expanded to –iiauu– and –uuaii-, but the original spellings are sometimes still attested: mańii̯uṷǡ, that is, *mańii̯uṷǡ, from *mańi̯^uṷāh “of the two (evil) spirits;” paouruuiia-, that is, *paouruṷii̯a-, from *paurṷii̯a- “first,” cf. OPers. paruviya-.

Internal i̯ was lost in YAv. before e: YAv. vaheḥī- (fem.) “the better,” from *ṷahi̯ehī-, cf. Vedic vásyasī-; -ahe, gen. sing. masc. ending, cf. Vedic –asya; kaine “girl,” cf. Vedic kanⁱyˋā; bāzuβe “with both arms,” from *bāzuβi̯a, cf. Vedic bāhúbhyām; YAv. -ə̄e, dat. sing. ending, cf. Vedic -aye.

A late but consistent change is that of -uṷe (from earlier *-uṷai̯ and *-ṷai̯) to –uiie: OAv. mruiiē, YAv. mruiie “I say,” cf. Vedic bruve; OAv. vīduiiē “to know” from *ṷidṷai̯.

In some cases Proto-Ir. i̯ and ṷ combine with a preceding consonant. Proto-Indo-Ir. *či̯ became *šˊi̯ in original OAv. and then šˊii in the Sasanian archetype. In original YAv. it became šˊ but is mostly written š or ṧ in the manuscripts. Thus we have: OAv. šˊiiāta- beside YAv. šˊāto (mostly written šāto or ṧātō).

Proto-Ir. hi̯- from Proto-Indo-Ir. -si̯- remained unchanged in original OAv. but became -hii- in the Sasanian archetype. After the change of h to ŋh, Proto-Ir. -hḭ developed in original YAv. to -ŋ́h- from *-ŋhi̯-. Thus we have: OAv. vahiiō beside YAv. vaŋ́hō (wrongly written vaŋhō), cf. Vedic vásyaḥ “better.” In the same way Proto-Ir. -hṷ- (from Proto-Indo-Ir. -sṷ-) developed into OAv. -huu- (from -hṷ-) and original YAv. –ŋᵛh– (often written -ŋuh- or -ŋh- in the MSS): OAv. gūšahuuā; YAv. pərəsaŋha (often written pərəsaŋuha, pərəsaŋha). Initially *hṷ– became in Av. xᵛ-: xᵛafna- “sleep,” cf. Vedic svápna-. On xᵛ see also under (f) below.

After certain consonants Proto-Ir. ṷ underwent further changes. Proto-Ir. śṷ became sp in Avestan and Median: aspa- “horse,” cf. Vedic áśva-. Proto-Ir. źṷ became zb in Av. and Median: zbaiia– “to call,” cf. Vedic hvaya-, from Proto-Indo-Ir. *j^hṷai̯a-. Proto-Ir. θṷ became θβ in Av.: caθβārō “four,” cf. Vedic catvāˊraḥ, Sogd. and Parth. ctfʾr. Proto-Ir. δṷ became in YAv. δβ (ərəδβa- “upright” beside ərəduua- from *ərdṷa-) but initial *dṷi- became OAv. dbi- (daibišiiant- but duuaēšah-, cf. Vedic dvéṣas- “hatred”) and YAv. t̰bi- (t̰bišiiant- and by analogy t̰baēšah-). From initial *dṷi- YAv. has also bi- perhaps by dissimilation; bitiia– “second” beside OAv. daibitiia-, cf. Vedic dvitīˊya-.

Initial *ṷr- was metathesized to *rṷ- and written uruu- in Av.: uruuata– “commandment,” cf. Vedic vratá-.

(b) Liquids (only r).

Consonantal r and original syllabic *ṛ fell together in Avestan, syllabic *ṛ becoming ər. After t the ə was usually dropped: ātrə̄m (acc. sing.) “fire” from *ātərəm; strə̄š (acc. plur.) “stars” from *stərə̄š; striia– “to sin” from *stəri̯a-, where the ə must have been lost before i-epenthesis could take place. Immediately following the Proto-Indo-Ir. accent rk became hrk and rp became hrp: mahrka– “destruction,” cf. Vedic márka-; vəhrka- “wolf,” cf. Vedic vṛ′ka-; kəhrpəm “body” from *kṛ′pam. Instead of the expected *hrt from *rt we find ṧ: maṧiia- “man,” cf. Vedic mártⁱya-; aməṧa- “immortal,” cf. Vedic amṛ′ta-. From the third century A.D. Mid. Pers. loanwords from Av. are attested which have hr/hl for Av. ṧ: Mid. Pers. ʾhlw [ahlaw] from Av. aṧauua. ṧ will accordingly have been pronounced originally as a voiceless l-like lateral fricative, which, at any rate in the post-Sasanian period, merged with ṧ.

(c) Nasals.

On the whole the nasals n and m remained unchanged in Av., but they are regularly written ṇ before t, d, k, g, c, j, b. The letter ṇ probably represents a uvular nasal that was articulated just by lowering the soft palate. It is indicated in this article by N in reconstructions. The same sound no doubt occurred in OPers. but it was not written: Av. aṇtarə “inside” but OPers. a-ta-ra [aⁿrar]. The dorsal nasal was, however, retained in YAv. paŋtaŋhum “a fifth” from *paŋ^ktahṷəm. An unusual metathesis is attested by YAv. mərəγənte “he destroys” for *mərəŋte from *mṛŋ^ktaḭ. Proto-Indo-Ir. ns before ā/ă resulted in Av. ŋ(g)h: YAv. saŋha-, OAv. sə̄ṇgha-, cf. Vedic śáṃsa-.

For discussion of Av. -aŋha-, -aŋ́ha-, aŋᵛha, and -aŋ́hi- see above under (a) and below under (f).

Before ḭ, n was palatalized to ṅ but in the manuscripts ṅ is usually replaced by n: ańiia-, cf. Vedic anyá-. The manuscripts often have ṃ instead of hm, which makes it probable that ṃ was a voiceless m. Final –m is found for –n when the syllable in question had a labial initial: OAv. dāmąm, nāmąm: YAv. uruθβąm, θrizafəm, aṧāum (from *aṧāṷən).

Phonetically Av. ą was probably nasalized ə̄. Not only did it develop from Proto-Indo-Ir. ā/ăn (ā/ăN) before s and š as seen above on the vowels but it occurs also in OAv. ərąš from *-ərəNš from older *-rNš: nərąš, mātərąš, mərąždiiāi. Note also mərąšˊiiāṱ from *mərəNšˊḭāt from older *mṛṇčḭāt. As in the case of the OAv. and YAv. acc. plur. endings –īš and -ūš from *-iNš and *-uNš, the nasalization is not attested in the acc. plur. of consonant stems in YAv.: nərə̄š, strə̄š, and pairiiaētrə̄š-ca. In the manuscripts these forms are often miswritten, e.g., nərə̄uš for nərə̄š.

(d) Occlusives.

The Proto-Indo-Ir. occlusives p, t, k, became f, θ, x in Proto-Ir. before a consonant. Proto-Indo-Ir. p^h, t^h, k^h also became f, θ, x before a vowel. However, Av. shows certain peculiarities. After s and š it has only p, t, k. Moreover, Av. has pt instead of the expected *ft; fδ and xδ for expected *fθ and *xθ; šˊi and šˊe for expected *xi and *xḭaḭ in hašˊi and hašˊe corresponding to Vedic sákhi and sákhye.

It is characteristic of OAv. that it has preserved b, d, g from Proto-Indo-Ir. b, d, g and b^h, d^h, g^h. In YAv., b, d, g are retained only in initial position while in medial position they were replaced by the voiced fricatives β, δ, γ except after a nasal or a sibilant. Thus, OAv. dugədar- “daughter” contrasts with YAv. duγδar-. There are, however, a number of exceptions. Note OAv. -βž- and γž-, YAv. γž-, γəm-, and γən-. Proto-Ir. -dn- became -n-: OAv. and YAv. bū/ŭna- “bottom,” cf. Vedic budhná-. Proto-Ir. dm- was retained in OAv. but became nm- in YAv.: OAv. dəmāna- “house” beside YAv. nmāna-. In YAv., γ was lost before u and ṷ: Mourum, cf. OPers. Margum; raom, cf. Vedic raghúm; druuaṇt- from *druγṷant-, cf. OAv. drəguuaṇt-. In YAv. driγūm “pauper,” the γ was restored by analogy with other forms of the paradigm such as gen. sing. driγaoš.

The YAv. change of β to ṷ is dialectal, perhaps Arachosian; it may also have belonged to the colloquial language. Examples are: gəuruuaiia- “to seize” from *gəṛβāḭa-, cf. Vedic gṛbhāyá-; the prep. auui “to,” which is also written aoui, aoi, from *aβi contrasting with aiβi in nominal compounds, cf. OAv. aibī and Vedic abhí; the adj. uuaiia, uuaēm “on both sides,” cf. Vedic ubháya-; uua “both,” cf. Vedic ubhāˊ (mase. dual); uiie from *uṷe, cf. Vedic ubhé (neuter dual); nəruiiō “to the men” beside nərəbiiō, cf. Vedic nṛ′bhyaḥ; aṧauuaoiiō “to the righteous” from *aṧaṷaβḭō. In some cases the spellings seem to be arbitrary: YAv. māuuōiia “to me” from *maṷḭa from older *maβḭa, cf. OAv. maibiiā; huuāuuōiia “to (your)self” from *huṷaṷḭa from older *huṷaβḭa, cf. original *-ṷḭ- in hāuuōiia (inst. sing.) “with the left (hand)” from *haṷḭā, cf. Vedic savyāˊ.

The occasional replacement of δ by θ appears also to be dialectal, perhaps West Iranian. In the athematic daδāiti “he puts; he gives,” cf. Vedic dad(h)āti, δ is retained but in the thematic new formation daθaiti earlier δ has been replaced by θ. The gen. sing. of daδuuǡ “creator” is daθušō, which is confirmed by dathousa (in Greek script) in the Cappadocian calendar. Note also h from θ in Parth. dh-, NPers. dah- “to give.” θ is attested also in East Iranian in Khot. parāth- “to sell” from *parā-daθa-, cf. Av. para.daθa-.

Proto-Indo-Ir. t was lost before s: Av. masiia– “fish,” cf. Vedic mátsya-. Similarly, Av. has st from *-t^st- from t/d + t as in vista– “found” from *ṷid^stá-, cf. Vedic vittá and zd from *-d^zd^h– from d^h + t as in vərəzda- “grown” from *ṷṛd^zd^há-, cf. Vedic vṛddhá-.

Final –t was lost after n, probably already in Proto-Indo-Ir., and also after s. Examples are: YAv. ās “he was” from *āst; OAv. cinas “she assigns” with -s from *-st; vąs “it prevailed” from *vān-s-t; OAv. sąs “it seemed” from *sśānd-s-t. However, both -st and -št are also found: OAv. urūraost “he wailed (?)” from *ruraud^st; YAv. nāist “he cursed” from *nāid-s-t; OAv. vaxšt “he made grow,” cōišt “he assigned,” tāšt “he shaped.” In all other cases -t became -ṱ (probably an implosive): YAv. baraṱ, OAv. cōrəṱ from *čart; OAv. yaogəṱ “he harnessed” from *ḭaugd (?) from older *ḭaukt. The graph -gət may represent an implosive –k / –g in YAv. paragəṱ “apart from,” cf. Vedic párāk; YAv. aṧiš.hāgəṱ “following Aṧi;” OAv. paitiiaogəṱ “responding.”

(e) Affricates.

The palatal affricates of Proto-Indo-Ir. č, ǰ, ǰ^h, which in Vedic became c, j, h, survived in Av. as c, j, j. On the development of Proto-Indo-Ir *čḭ to OAv. šˊii and YAv. šˊ see above on the vowels. The YAv. change of j to the palatal *žˊ, always written ž, is dialectal, perhaps Arachosian: druža- “to deceive” from *drujḭa-, cf. OAv. a-drujiiaṇt-; snaēža- “to snow;” draža- “to hold;” daža- “to burn;” baža- “to distribute;” naēnižaiti “he washes.” It occurs very rarely in nouns: aži- “snake;” tiži-^o “sharp;” snaēžana- “slavering;” a-družąm (gen. plur.) “of the deceitless” (otherwise only druj-).

The primary palatal affricates of Proto-Indo-Ir., namely ć, jˊ and j^h from Proto-IE. kˆ, ĝ, ĝ^h, developed via Proto-Ir. ś, ź, ź to Av. s, z, z corresponding to Vedic ś, j, and h respectively: Av. satəm “hundred,” cf. Vedic śatám; zaoša- “pleasure,” cf. Vedic jóṣa-; zaotar “priest;” cf. Vedic hotár-. Before t, d^h, and b^h, ć and j′ developed already in Proto-Indo-Ir. to š and ž respectively: OAv. vaštī “he wishes” beside vasəmī “I wish,” cf. Vedic váṣṭi beside váśmi; OAv. važdra- “pulling” from vaz- “to pull,” cf. Vedic voḍhár- “draught (i.e., pulling) animal” from vah-; OAv. and YAv. vīžibiiō, abl. plur. from vīs- “tribe,” cf. Vedic viḍbhyáḥ from víś-. In initial position źn- became žn- in YAv. ( = OPers. xšn-): žnātar- “knower,” cf. Vedic jñatár- “knower,” OPers. xšnā- “to know;” žnu- “knee,” cf. Vedic jñu-^o. Internally both śn and źn became šn: YAv. frašna- “question,” cf. Vedic praśná-; YAv. baršna “in height, depth” ( = OPers. baršnā) from *barźnā from older *b^harj^hnā. But sn is found instead of šn in some cases due to the influence of other forms: OAv. vasnā “according to wish” ( = OPers. vašnā) from vas-; OAv. and YAv. yasna– “veneration” (cf. Vedic yajñá-), from yaz-.

The Proto-Indo-Ir. clusters sć and šć from Proto-IE. sk developed via Proto-Ir. sś and šś to Av. s ( = Vedic ch): Av. pərəsa- “to ask,” cf. Vedic pṛchá-. Similarly, ćš and jž^h, from Proto-IE. kˆs, kˆρ and ĝ^hs, ĝ^hρ respectively, developed via Proto-Ir. śš and źž to Av. š and ž: Av. šōiθra- “dwelling-place,” cf. Vedic kṣétra-; uz-uuažaṱ “he drew out,” cf. Vedic vákṣat (subj.) from Proto-IE. ṷeg^h-se- (see next paragraph).

(f) Sibilants.

Proto-Indo-Ir. s and z were maintained in Av. before n and occlusives, and after t and d, which were lost in that position as noted above. Thus we find: YAv. snāuuarə “sinew,” cf. Vedic snāˊvan-; asti “he is,” cf. Vedic asti; masiia– “fish,” cf. Vedic mátsya-; YAv. mazga– “marrow” from Proto-IE. mozg^ho-; YAv. aspas-ca “and the horse,” cf. Vedic áśvaś-ca; OAv. zdī (2 sing. imv.) “be,” cf. Vedic edhi, from Indo-Ir. azd^hí; vərəzda- “grown” from *ṷṛd^zd^há-, cf. Vedic vṛddhá-.

After Proto-Indo-Ir. ī/ĭ (ḭ), ū/ŭ (ṷ), r (ṛ), kˆ/ĝ/ĝ^h, and ć/j/j^h (from Proto-IE. k/g/g^h), Proto-Indo-Ir. s and z became š and ž: Av. vīša- “poison,” cf. Vedic viṣá-; mīžda- “reward,” cf. Vedic mīḍhá-; zušta- “loved,” cf. Vedic juṣṭa-; aršti- “spear,” cf. Vedic ṛṣṭi-; uxšan- “bull,” cf. Vedic ukṣáṇ-; OAv. aoγžā “you say” from *aṷg^h-sa; vašī “you wish,” cf. Vedic vakṣi, from Proto-IE. ṷek-si; tašan- “fashioner,” cf. Vedic tákṣan-, from Proto-IE. tékρon-. In Proto-Ir. this development took place also in clusters with labials. Thus Av. has fš from *ps and *pś: Av. drafsa- “banner,” cf. Vedic drapsá-; fšu- from *pśu- to pasu- “cattle.” Similarly Av. has βž from *bz^h: diβža- “to deceive,” cf. Vedic dipsa-, from Proto-Indo-Ir. d^hibz^ha-; vaβžaka- “wasp” from *ṷabz^ha- from Proto-IE. uob^hso-.

In all other positions Proto-Indo-Ir. s became Proto-Ir. h. This h was kept initially before a vowel: hafta “seven,” cf. Vedic saptá. But *hi became x́ ii- in OAv.: x́ iiāṱ “he should be,” cf. YAv. hiiāṱ, Vedic syāt; and *hṷ became xᵛ in both OAv. and YAv.: xᵛafna- “sleep,” cf. Vedic svápna-. Medial h was unchanged only before i and u: ahī/ĭ “you are;” uhura– “lord.” In OAv. medial h remained unchanged also before ḭ and ṷ: OAv. ahiiā, cf. Vedic ásya: gūšahuuā with the ending -ahuuā corresponding to Vedic -asva. In the sequence ā/ăha, h probably became voiced and resulted in ŋh: aŋhaṱ, cf. Vedic ásat; ǡŋharə, cf. Vedic āˊsur. That this ŋ was phonemically significant is shown by the fact that it was extended from the gen. sing. vaŋhə̄uš from *ṷahaṷš ( = Vedic vásoḥ) to the nom. sing. masc. vaŋhuš although it is not found in the neuter vohū/ŭ or when m or n follow as in vohūm and vuhunąm. In medial position hḭ and hṷ developed in YAv. to ŋ́h and ŋᵛh; see (a) above. (See also Hoffmann, Aufsätze II, pp. 595-96).

In OAv. the gen. sing. ending -ahiiā is always written with x́ before enclitic -cā “and:” -ax́ iiā-cā. This pronunciation may reflect the secondary accentuation *ahḭá-ca. -x́ ii- is also found elsewhere for -hḭ-: OAv. dax́ iiə̄uš “of the land” but YAv. daŋ́ə̄uš; both OAv. and YAv. dax́ iiūm (acc. sing.) and dax́ iiunąm (gen. plur.).

The use of -xᵛ- for internal -hṷ- in YAv. Haraxᵛaitī- “Arachosia” and OAv. nəmaxᵛaitī “respectful” may be dialectal, perhaps Arachosian. The same applies to the use of xᵛ for unaccented syllabic huṷ- in the following: Xᵛāstrā-, name of an Arachosian river, from *hu-ṷāstrā-; OAv. xᵛāθra- “welfare” from *hu-āθra-, cf. duž-āθra “discomfort;” xᵛə̄ṇg (gen. sing.) from *huṷə́ŋh to huuarə “sun,” cf. Vedic suvár; xᵛaēta- “easy to walk along” from *hu-ā-ita-; xᵛīti- “easy walking.” (See also Hoffmann, “Das Avesta in der Persis,” pp. 92-93.)

Proto-Ir. hm is retained internally as in ahmi “I am” but the h is lost in initial position: mahi “we are,” cf. Vedic smasi. Proto-Indo-Ir. sr appears to have become θr in YAv. in initial position: θraotō.stāc- “flowing in rivers,” from *srautas-tāč-, cf. Vedic srótas- but OPers. rautah-. Medially hr became ŋr in YAv.: aŋra- “evil,” cf. Vedic asrá- “painful;” daŋra- “knowing,” cf. Vedic dasrá-. These forms were introduced from YAv. into OAv., where one also finds the spellings aṇgra- and daṇgra-.

For the loss of final -h see above on the vowels.