While the likes of Robert Austerlitz and Roman Jakobson did some initial work reconstructing the language history of Nivkh, much remains, especially in terms of the vowel system of Nivkh.
Nivkh, a critically endangered language isolate spoken in the Russian Far East, can roughly be divided into two dialect groups: Amur-Northwest Sakhalin (abbreviated NWN) varieties and Southeast Sakhalin (abbreviated SEN) varieties. Speakers reportedly claim further differences to be small, and no thorough research has been done on the language up to this point (Shiraishi 2006: 10–12).
One interesting correspondence has to do with /ɨ/. It seems that at least two vowel phonemes underwent a merger to /ɨ/ in the Amur-Northwest Sakhalin varieties, but not in the Southeast Sakhalin varieties, which have either /ɨ/ or /a/. Some examples (all data taken from Savel’eva and Taksami 1970):
(1) NWN [mrɨ-] ‘to swim’
SEN [mra-] ‘to swim’
(2) NWN [ŋɨŋg] ‘hair’
SEN [ŋamx] ‘hair’
(3) NWN [vɨlki] ‘chain’
SEN [valki] ‘chain’
(3) NWN [indɨ-] (with no direct object) ~ [ŋr̥ɨ-] (with no DO) ‘to see’
SEN [idɨ-] (w/ no DO) ~ [ŋr̥ɨ-] (/w DO) ‘to see’
PN ?*(i-)ŋr̥ɨ- ‘to see’
(4) NWN [kɨls] ‘length’
SEN [kɨlr̥] ‘length’
(2) is a bit suspicious, due to the number of changes needed between NWN [ŋɨŋg] and SEN [ŋamx]—the deletion of the second vowel, the assimilation of PN *m to NWN [ŋ] due to the following velar, the spirantization of PN *k to SEN [x], and the voicing assimilation of PN *mk to NWN [ŋg], but I think these forms are ultimately cognate.
Another interesting thing is the large number of consonant clusters found in Nivkh. This is unlike the situation found its neighbors, and, both internal variation as well as loanwords into Ainu and Ul’ta (Orok) make it clear that this vowel loss is recent. A good example of this is the word for ‘reindeer’, where we have good internal evidence, as well as good external evidence—the Nivkh word was loaned into Ainu and into Japanese.
(5) NWN [cʰolŋi]
Sakhalin Ainu [tunakaj]
It is likely that Nivkh /ŋ/ has a secondary origin, from earlier *nk. Also interesting, but perhaps not at all surprising, is the correspondence of Nivkh [l] to Sakhalin Ainu [n]—unsurprising because Ainu does not have [l]. The internal evidence points to just final *-i, while the external evidence points to a final diphthong, *-ai. Here I privileged the former kind of evidence over the latter.
Finally, I still have no good explanation for the distribution of [cʰ] and [t], which I reconstruct with the placeholder *T. Compare the above with the word for ‘tree’, which is frustratingly reversed in terms of which group has which consonant:
(6) NWN [tiɣr] ‘tree’
SEN [cʰxar̥] ‘tree’
PN ?*Tikar̥ (n.b., initial and medial fricatives are secondary in Nivkh)
Savel’eva, Valentina Nikolaevna and Chuner Mixaylovich Taksami. 1970. Nivxsko-Russkiy Slovar’ [Nivkh-Russian Dictionary]. Moscow: Sovetskaya Entsiklopediya.
Shiraishi, Hidetoshi. 2006. Topics in Nivkh Phonology. PhD dissertation: Rijksuniversiteit Groningen.