Bactrian phonology

The phonology of Bactrian is not known with certainty, due to the limitations of the native scripts.

Consonants
A major difficulty in determining Bactrian phonology is that affricates and voiced stops were not consistently distinguished from the corresponding fricatives in the Greek script.
 * Proto-Iranian *b, *d, *g have generally become spirants, as in most other Eastern Iranian languages. A distinctive feature of Bactrian, shared within the Iranian languages with Munji, Yidgha and Pashto, is the development of Proto-Iranian *d > *ð further to, which may have been areal in nature. Original *d remains only in a few consonant clusters, e.g. *bandaka > βανδαγο 'servant', *dugdā > λογδο 'daughter'. The clusters and  appear in earlier Bactrian, but revert to ,  later, e.g. *drauga > λρωγο (4th to 5th century) >  δδρωρο (7th to 8th century) 'lie, falsehood'.
 * Proto-Iranian *p, *t, *č, *k have become voiced between vowels, and after a nasal consonant or *r.
 * Inside a word, the digraphs ββ, δδ for original voiceless *p, *t can be found, which probably represent [b], [d]. The former is attested only in a single word, αββο 'water'. Manichaean Bactrian appears to only have had /v/ in native vocabulary. According to Gholami, instances of single δ may indicate a fricative pronunciation,.
 * γ appears to stand for both the stop and the fricative, but it is unclear if a contrast existed, and which instances are which. Evidence from the Manichaean script suggests that γ from *k may have been  and γ from *g may have been . According to Greek orthographic practices, γγ represents.
 * σ may continue both Proto-Iranian *c > *s and *č, and the Manichaean script confirms that it represents two phonemes, likely /s/ and /ts/.
 * ζ may continue similarly on one hand Proto-Iranian *dz > *z, and on the other *ǰ and *č, and it represents at least /z/ and /dz/. This distinction is again confirmed by the Manichaean script. Also a third counterpart of ζ is found in Manichaean Bactrian, possibly representing.

The status of θ is unclear; it only appears in the word ιθαο 'thus, also', which may be a loanword from another Iranian language. In most positions Proto-Iranian *θ becomes (written υ), or is lost, e.g. *puθra- > πουρο 'son'. The cluster *θw, however, appears to become, e.g. *wikāθwan > οιγαλφο 'witness'.

ϸ continues, in addition to Proto-Iranian *š, also Proto-Iranian *s in the clusters *sr, *str, *rst. In several cases Proto-Iranian *š however becomes or is lost; the distribution is unclear. E.g. *snušā > ασνωυο 'daughter-in-law', *aštā > αταο 'eight', *xšāθriya > χαρο 'ruler', *pašman- > παμανο 'wool'.

Vowels
The Greek script does not consistently represent vowel length. Fewer vowel contrasts yet are found in the Manichaean script, but short /a/ and long /aː/ are distinguished in it.

It is not clear if ο might represent short [o] in addition to [u], and if any contrast existed. Short [o] may have occurred at least as a reflex of *a followed by a lost *u in the next syllable, e.g. *madu > μολο 'wine', *pasu > ποσο 'sheep'. Short [e] is also rare. By contrast, long /eː/, /oː/ are well established as reflexes of Proto-Iranian diphthongs and certain vowel-semivowel sequences: η < *ai, *aya, *iya; ω < *au, *awa.

An epenthetic vowel [ə] (written α) is inserted before word-initial consonant clusters.

Original word-final vowels and word-initial vowels in open syllables were generally lost. A word-final ο is normally written, but this was probably silent, and it is appended even after retained word-final vowels: e.g. *aštā > αταο 'eight', likely pronounced /ataː/.

The Proto-Iranian syllabic rhotic *r̥ is lost in Bactrian, and is reflected as ορ adjacent to labial consonants, ιρ elsewhere; this agrees with the development in the western Iranian languages Parthian and Middle Persian.