Indo-Iranian is a part of the Satem areal also including the Balto-Slavic languages, Albanian and Armenian. In these languages PIE "palatovelar" consonants (probably originally the plain velars) were characteristically fronted to postalveolar affricates *c, *jʰ, *j, while the PIE labiovelars and "plain velars" (probably originally uvulars) were then merged as new velars *k, *gʰ, *g.
The Law of Palatals introduced to Indo-Iranian a new set of palatal consonants: *k > *ĉ, *g⁽ʰ) > *ĵ⁽ʰ⁾, when preceding the original front vowels *e, *i.
The most characteristic feature of Indo-Iranian may be its simplification of the PIE vowel system: *e, *o, *a, *n̥, *m̥ were all merged as simple *a. This phonemicized the previous set of palatal consonants. — On the other hand, late PIE *ə (the vocalized allophone of laryngeals) was instead raised to *i.
The Iranian group was originally set up to contain all members of Indo-Iranian not derivable from Sanskrit. Only a small number of isoglosses defining Iranian as a whole have been proposed, and most of these isoglosses appear to "leak" in some way or the other (appearing reversed or absent in a number of descendants, or being found also outside the normally accepted boundaries of Iranian). Of these the Nuristani languages are widely recognized as a separate group since the mid-1900s.
An agnostic account could recognize at least six branches of Indo-Iranian:
- Sakan, a small northeastern group of three languages: †Khotanese, †Tumshuqese, and Wakhi.
- Persid ("Southwestern Iranian"), comprising Old Persian and its descendants.
- Narrow Iranian: most languages traditionally considered Iranian. Drawing an exact perimeter remains difficult; further subdividable:
- †Avestan, attested from very early on as a liturgical language. Unlike Sanskrit, it however shows several unique innovations not known from any other Indo-Iranian language, such as *h > ŋ / a_a (rhinoglottophilia).
- Balochi, spoken in southeastern Pakistan, features as an archaism the absense of spirantization of stops.
- Bactric comprises †Bactrian and Munji-Yidgha.
- Pashto (whose Wanetsi dialect is sometimes counted as an independent language) shows similarities to Bactric on one hand (a development *ð > *l), Sakan on the other.
- Yazghulami, perhaps affiliated with the previous.
- Scytho-Sarmatian, today represented only by Ossetic but formerly including several important languages of the Eurasian steppe, such as Alan and the titular Scythian.
- Sogdic comprises †Sogdian and Yaghnobi.
- Nuristani, spoken in the Hindu Kush mountains.
- Dardic, spoken also in the Hindu Kush mountains. Usually considered closely affiliated to Narrow Indic. Its unity is not demonstrated, and some of the languages may be simply archaic Northwestern Zone Indic languages.
- Narrow Indic consists of most languages traditionally considered Indic (Indo-Aryan).
Major phonetical isoglosses connecting these groups include:
- The two PII voiced series (traditionally reconstructed as plain voiced vs. voiced aspirate) merge everywhere except in Narrow Indic. This change is shared also with Balto-Slavic and Albanian.
- Phonation residues of the contrast have been reported from Dardic.
- A shift *s > *h before a vowel is shared by all "standard Iranian" (exlucing Nuristani) branches.
- The PII clusters *ćw, *jw undergo different characteristic development in each of the traditional "Iranian" branches:
- Sakan: *ćw, *jw > ?*św, *źw > *š, *ž.
- Persid: *ćw, *jw > *s, *z. Perhaps best interpreted as an early deaffrication to *św
- Narrow Iranian, Nuristani, Dardic: *ćw, *jw > *sp, *zb
- In other positions the PII postalveolars *ć, *j are fronted to dental affricates *ts, *dz, in all varieties other than Dardic and Narrow Indic. The change must be dated as posterior to the previous changes, and is therefore inapplicable as evidence for a common Iranian or Iranian-Nuristani group (at least if one wishes to include Sakan). The affricates remain in Nuristani, but develop further elsewhere:
- In Persid these are fronted to *θ and *ð > *d. (A similar development is known from Albanian, though this is probably unrelated.)
- In Narrow Iranian + Sakan these are deaffricated to *s, *z. This extends to Slavic and Latvian as well.
- Dardic and Narrow Indic merge the PII postalveolars with the PII palatals *ĉ *ĵ⁽ʰ⁾, with the exception of *ć; this yields Dardic *š, Narrow Indic *ś.
- In Persid + most of Narrow Iranian (but absent from Balochi and Sakan), PII *p *t *k are spirantized to *f *θ *x preceding a consonant.
- Sakan, Dardic and Narrow Indic develop a series of voiceless aspirates from PIE (and PII) clusters of voiceless stop + *h₂. Originally this was posited as a common Indo-Iranian innovation, though one would have to assume later lenition to voiceless fricatives in parts of Iranian (better handled as a part of the previous soundlaw), and loss of aspiration altogether in Nuristani; a more limited soundlaw allows a more economical analysis.