Māryanyā sound changes can be expressed as changes from Proto-Indo-Iranian, as changes from Proto-Indo-Aryan (though it does not share Dravidian influences common to languages of the Indian subcontinent, such as the development of the retroflex series), or as a comparison to Vedic Sanskrit. A combination of these techniques will be used here, as well as evidence of certain changes in the real Mitanni-Aryan script from Akkadian, Hurrian, and Hittite texts. Changes will be presented in a roughly chronological order, with commentary, though exact order cannot always be determined.
As in Sanskrit and Avestan, the Proto-Indo-Aryan/Proto-Indo-Iranian laryngeal *H lengthens preceding vowels, vocalizes to short /i/ after or between consonants, and then is lost. It does not have the u-coloring effect on *r̥ as in Sanskrit; e.g. Skt. पुर púra, Mry. parā “fortress, walled city”, from Proto-Indo-European *(t)pĺ̥H-.
As in Sanskrit, the two palatal series in PII are merged, except *ć, which becomes *ś. Unlike Sanskrit, which preserves a distinct *š from the RUKI sound change as ष ṣ, Māryanyā merges the two as /ɕ/.
PII *ȷ́ʰ ǰʰ merge into a breathy-voiced fricative, /ʑʱ/, as reconstructed for PIA; this can become an affricate in certain clusters or a plan voiced fricative by Bartholomae’s law. An example is found in the Hurrian substrate word ⟨mištannu⟩, from PII *miždʰám, Mry. miždham, Skt. मीढ mīḍhá, all meaning some kind of reward.
Though the Sanskrit palatal series च छ ज झ c ch j jh are theorized to have been palatal stops [c cʰ ɟ ɟʱ] in an earlier stage, the Māryanyā/Mitanni-Aryan equivalents (barring ch which never developed in Māryanyā) were certainly affricates by the 1300s BCE, as evidenced by the Hittite transcription of the Mitanni-Aryan numeral five, ⟨panza⟩; the Z series in Hittite cuneiform represented an affricate.
The “thorn clusters”, clusters of a dental and a palatal or velar stop in Proto-Indo-European, as well as the sequence of a palatal or velar followed by PIE *s (> PII *ćš), which both become the common cluster क्ष kṣ in Sanskrit, instead become palatal affricates in Māryanyā, which are geminated intervocalically; e.g., PII *Háćšas, Skt. अक्ष ákṣa, Mry. accas “axle”, and PII *Hŕ̥ćšas, Skt. ऋक्ष ṛ́kṣa, Mry. arcas “bear”. The sequence *-ts-, mostly in substrate words like PII *mátsyas “fish”, also became c, as in Mry. macyas.
The PIE liquids *l and *r, are merged into *r, as in Iranian, surfacing as /ɾ/ in Māryanyā, whereas Sanskrit occasionally preserves the distinction; e.g., PII *HlawpaHćás, Skt. लोपाश lopāśá, Mry. rupāšas, Modern Persian روباه rubâh.
Similarly, PIE syllabic liquids *r̥ *l̥ are both inherited as *r̥, then the r is desyllabized into /aɾ/. This latter change is seen in Mitanni-Aryan substrate names such as Artatama, Mry. Artadhāma, “truth-abode”. /l/ in Māryanyā is only present in loanwords from Akkadian, Ancient Egyptian, etc.
At some point after *r̥ > ar, the vowel /a/ was inserted epenthetically to resolve certain clusters, including onset clusters of more than two consonants (e.g. Skt. स्त्री strī́, Mry. astrī “woman”) as well as *arn > aran /# in the accusative singular of masculine r-stem nouns.
Māryanyā maintains the three-vowel system of PII, and like Vedic but not Classical Sanskrit, does not smooth the dipthongs *ai and *au to /e/ and /o/. Vedic ए ऐ ai āi are both simple /ai̯/ in Māryanyā. PII *aw, Vedic ओ au is simplified to /u/, then the equivalent of Vedic औ āu (from *āw or *aHu) is shortened to /au̯/.
A very late change, possibly still in the process of occurring during the Amarna era when Mitanni-Aryan is attested, is *wa > u. This only occurs in the middles of words, or occasionally at the end; e.g., in a Hittite text, ⟨na-a-wa-ar-ta-an-na⟩, Mry. nau vartāni. Cuneiform texts in general are inconsistent in the representation of *wa; two different versions of a Mitanni-Hittite treaty invoking the god Varuṇa, for example, call him in one Aruna and in the other Uruwana. My version of Māryanyā treats the sound change *wa > u except word-initially as complete, and nominal declension classes are adjusted as such. Since this is a much later change than the earlier vowel changes, it results in *awa > au and *wai > *ui > vi.
Māryanyā acquires the voiceless aspirated stops /tʰ kʰ/ in the same places where Sanskrit does, from the loss of the PIA laryngeal *H. Voiceless aspirated stops are also used to loan plain voiceless stops from Egyptian and Semitic languages, whereas unaspirated stops are used to loan ejectives; e.g., Mry. pītathi, Egyptian pḏtj, Akkadian piṭatu “archer”. /tʰ/ is also used to loan ⟨θ⟩ from Hurrian and Ugaritic.
Additionally, as mentioned elsewhere, Māryanyā acquired the phonemes /l/ and /x/ only in loanwords. /l/ is used to loan ⟨l⟩ from cuneiform languages and, in some environments, Egyptian ⟨ꜣ⟩ (reconstructing the Middle to Medio-Late value of ⟨ꜣ⟩ as [ɫ] rather than [ʀ]).
/x/ is used to loan anything represented in cuneiform as ⟨ḫ⟩, which was something like /x/ in Akkadian but could most likely represented a variety of Hurrian and Amorite phonemes.
Māryanyā does not conserve Proto-Indo-European pitch accent, instead innovating a stress accent similar to that reconstructed for Akkadian. See the Phonology section for more.