I recently proposed that the fission of Proto-Uralic *ä and *e into more open and more close vowels in Samic, depending on the following second-syllable vowels (“stem type”), should be dated already to the dialectal West Uralic era, given that similar developments appear also in their closest relatives: the Finnic and Mordvinic languages. This diverges in a couple of ways from the views in the main handbooks on the historical development of Samic, i.e. Korhonen (1981): Johdatus lapin kielen historiaan and Sammallahti (1998): The Saami Languages: An Introduction.
One basic disagreement is over absolute chronology. While both Korhonen and Sammallahti (henceforth: K & S) agree that at least the merger of the stem types *e-ə and *i-ə [1] should indeed be dated to the earliest phase of the pre-Proto-Samic era, their treatises begin from the now obsolete “Proto-Finno-Samic”, dated as some half a millennia later, reconstructed with cross-reference mainly to Finnic, and usually also located some 1000 km more westerly (in the Gulf of Finland area) than my reference point in common West Uralic (around the upper reaches of the Volga). [2]
Another however concerns the overall relative chronology. K & S present the historical phonology of Samic in a highly tiered fashion that makes for some very attractive charts and graphics, with roughly four distinct periods of development:
- an early phase (K’s “kantalapin I vaihe“, S’s “Pre-Saamic” and “Proto-Saamic 2“) with the loss of several inherited vowel contrasts, and the splitting of this smaller pre-Samic vowel system into several allophones, depending on stem type;
- a complete revamp of the vowel length system (K’s “kantalapin II vaihe“, S’s “Proto-Saamic 3” in parts), depending on earlier vowel qualities;
- a restructuring of the system of unstressed vowels (K’s “kantalapin III vaihe“, S’s “Proto-Saamic 3” in parts as well as “Proto-Saamic 4“)
- late phonetic shifts in the sound values of several stressed vowels (K’s “kantalapin IV vaihe“, S’s “Proto-Saamic 5“).
As I have mentioned in an exchange in the comments section, I am however skeptical of the historical reality of this model. It strikes me as unnaturalistically neat altogether. Only a few of the changes can be explicitly shown to have been in the presented order in relative chronology, and probably most of the distinct “phases” here should be meshed together. Others might even be post-Proto-Samic entirely (though that will be another topic).
In particular I do not think that all Proto-Samic umlaut developments should be considered equally early. The Samic languages are some of the most “umlaut-rich” languages within Uralic, and the individual languages have continued to innovate new changes of this type pretty much as soon as new features arise among the unstressed vowel system. In this context it seems entirely implausible to me that at one point the pre-Proto-Samic speakers would have collectively decided “ok, that’s enough for now, let’s call a 500-year moratorium on umlauts”.
More specifically, while I think that developments *ä-ä > *ȧ-ȧ (> PS *ā-ē) versus *ä-ə > *e-ə (> PS *ē-ë) might be even earlier than has been previously suspected, by contrast I think that the a-umlaut of inherited *e and *o (e.g. PU *pesä >> PS *peasē ‘nest’; PU *kota >> PS *koatē ‘tent, teepee’) must be instead dated to a somewhat later Proto-Samic phase. This is due to some exception cases that appear to be explainable by them having been subject to both umlauts.
Umlaut stacking
It’s been observed already since the earliest reconstruction work on Uralic vocalism that PS *ea fairly often turns up in the Samic languages as a reflex of earlier *ä. Explanations for these cases have varied quite a bit, from considering this the regular reflex of the stem type *ä-ä (this was the opinion of Wolfgang Steinitz), to dismissing all instances as irregular or “sporadic” (thus K & S). Neither extreme is satisfying though, and it would be desirable to identify some conditions for the development. Dating the umlauts of *ä and *e into two different chronological stages seems to offer a lead on this.
If we assume that the pre-Samic dialect of late West Uralic — I will call it “pre-Samic” or “preS” for short — had already raised *ä-ə to *e-ə, as in words like the following:
- PU *jäŋə [jɛŋə] > preS *jeŋə > PS *jēŋë ‘ice’
- PU *kälə [kɛlə] > preS *kelə > PS *kēlë ‘tongue’
- PU *mälkə [mɛlkə] > preS *melkə > PS *mēlkë ‘breast’
— then at this point, a derivational process turning one of these *ə-stem words into an *ä-stem word would allow it to be later subject to a-umlaut just as inherited *e is, yielding PS *ea. There appear to be some clear examples that involve the syncope of *-ə- upon the addition of a derivational suffix: PU *CäCə > preS *CeCə → *CeCə-Cä > *CeCCä > PS *CeaCCē. Some other examples involve a derivational process that leads to a pre-Samic *o-stem (which similarly trigger a-umlaut): preS *CeCə → *CeC-o > PS *CeaCō. [3]
This mechanism appears to explain a reasonable number of the cases of PU *ä yielding PS *ea. Thus far, I have identified seven possible front-vocalic cases (including one somewhat speculative new etymological proposal):
- *keaćō ‘medium-sized whitefish’ (only in Lule Sami: getjuk) < preS *keć-o
← *kećə < PU *käśə(ŋ)
Cf. Mansi *kääsəŋ, Hungarian keszeg ‘bream’, which both indicate earlier *ä. (Finnish keso ‘white bream’ has also been considered cognate, but is better derived from kesä ‘summer’.) - *leapō- (Lule Sami lehpagis ‘nice’, Old Swedish Sami leppotet) < preS *lep-o-
← *le(p)pə < PU *lä(p)pə
Cf. Moksha /ľäpä/ ‘weak’, Mari *lewə ‘warm, mild (of weather)’, Khanty *leepət ‘weak’, which indicate *ä. Finnic *leppedä ‘balmy’ again looks like the odd member out in the cognate set. The similar *leepedä ‘mild’ could be instead compared here just about as easily. [4] - *meanō- ‘to become evasive’ < preS *men-o-
← *menə- < PU *mänə-
Cf. Mordvinic *mäńə- ‘to dodge, to get free’, Komi /mɨn-/ ‘to get free’, Hungarian mentes ‘free’, which indicate *ä. The verb *mänə- ‘to get free’ is probably ultimately somehow related to *menə- ‘to go’, but the cognates suggest the two having been distinct already at the PU level. (I additionally wonder if contamination from the former could perhaps explain the irregular vowel in Savonian/Karelian mäne- ‘to go’.) - *peajō- ‘to shine’ < preS *pej-o-
- *peajvē ‘day’ < preS *pejwä < *pejə-wä
both ← *pejə < PU *päjə ‘bright, shining, etc.’
The bare root does not appear to unambiguously survive anywhere (perhaps in Komi /bi/ ‘fire’?), but numerous other derivatives generally indicate *ä, e.g. Finnic *päivä ‘day, sun’, Hungarian fehér ‘white’. - *pealkē ‘thumb’ < preS *pelkä < *peləkkä < PU *pälə-kkä
Cf. Mordvinic *päĺka, where the unvoiced cluster *ĺk must be secondary (PU *lk would have yielded **ĺg). Komi /pel ~ pev/ also suggests *ä. The underived root could be identified with *pälə ‘side’, as has been proposed by Janhunen. The messy Finnic words for ‘thumb’, often included here, mostly point to PF *peikala or *peikoi; and they probably need to be kept separate (at best some kind of secondary contamination of the original Uralic word with some other source could be involved). - *veakkē ‘help’ < preS *wekkä < *wekə-(k)kä
Formally, this might be a derivative of PU *wäkə ‘power’ > preS *wekə (> PS *vēkë ‘people’). A semantic intermediate ‘activity with several people, work bee’ could be involved.
It is however necessary to also assume similar but even earlier syncope in some other old derivatives, which do show regular a-umlaut in Samic.
- *ńālmē ‘tongue’ < *ńälmä (~ Hungarian nyelv ‘tongue, language’ etc.) ← PU *ńälə- ‘to swallow’
- *ńālkē ‘tasty’ < *ńälkä ← id.
- *pāŋkē ‘reindeer’s headgear’ < *päŋkä ← PU *päŋə ‘head’
The first of these words has a very wide distribution, and the bisyllabic form *ńälmä could perhaps be assumed already for PU… though this would get in the way of a partial rule for *m-lenition in Hungarian that I have sketched some years ago.
I can also think of a slightly different mechanism to account for one of the remaining high-profile examples of *ä >> *ea. This is *pealē ‘side; half’. The polysemic meaning suggests that this may have come about as a blend of two originally distinct PU words: the above-mentioned *pälə ‘side’ (> Finnic *peeli, Mordvinic *päľ, Mari *pel), and the evidently closely related but distinct *pälä ‘half’ (> Finnic *pooli, Mordvinic *päľə, Mari *pelə).
The two words also seem to merge in Ugric: compare Hungarian fél : fele- ‘side; half’; Mansi *pääl ‘side; half’; Khanty *peeɭək ‘side; half’. But while this development can be simply due to the loss/reduction of 2nd-syllable vowels, the Samic development would require assuming contamination: the stem vowel seems to continue preS *pȧlȧ ‘half’, while the *e-type 1st syllable vowel seems to continue preS *pelə ‘side’. The two would led to the creation of a preS “compromise” form *pelä, from which then regularly > PS *pealē.
Finnic parallels
Worth noting is that in the case of ‘day’, a similar exception development is also found in Finnic. PF *päivä ‘day, sun’ (and not **paivi) has likewise escaped the early lowering/backing of *ä-ä, perhaps for the same reasons too: contraction from *päjə-wä taking place only after the a-umlaut of primary *ä-ä.
This pattern seems to extend further: among the remaining cases with *ä > PS *ea, Finnic cognates usually have *ä-ä as well. At least five cases can be identified that have correspondences in the more eastern Uralic languages:
- ‘lichen’: PS *jeakēlē ~ PF *jäkälä (~ Permic)
- ‘paw’: PS *keapēlē ~ PF *käpälä (probably related to *käppä ‘paw’ > Finnic, Mordvinic)
- ‘bog’: PS *jeaŋkē ~ Fi. jänkä (~ Permic, Mansi, Khanty)
- ‘flap, cover’: PS *leappē ~ PF *läppä (~ Mari, Permic, Hung., Mansi)
- ‘smoke hole’: PS *reappēnē ~ PF *räppänä (~ Permic)
While this same correspondence is also common enough in loanwords (PS *(h)earkē ← PF *härkä ‘bull’; PS *kearnē ← PF *kärnä ‘crust’; both originally from Baltic), and this approach has in the past been applied to ‘bog’ (S → Fi) and ‘paw’, ‘flap’ (F → S) as well, nothing seems to outright require considering these words later than the pre-Samic / pre-Finnic period. If *ä-ä [ɛ-a] had in both groups been lowered to *ȧ-ȧ [a-a] by then, new lexical innovations of the time could reintroduce also a new, secondary *ɛ-ä in pre-Finnic (*jɛkälä ‘lichen’, etc.); while in pre-Samic, only *e-ä would have been available.
Conveniently enough, there is also one word of this type for which early loaning in the West Uralic period is assured: PS *keavrē ~ PF *käkrä ‘bent’, which probably derives from Indo-Iranian *čakra- ‘wheel’ (or from a slightly earlier *ḱɛkra-). [5]
To be sure, I still generally hold that if two competing etymologies are available for a word, then all other things being equal, the more recent explanation should be preferred. But this is only a probabilistic rule-of-thumb. So while several of the words here (and also many of the more numerous similar cases yet that are restricted to Samic & Finnic) probably have indeed been loaned between Finnic and Samic at a later date, I would not rule out the possibility of some of them still going back to different parallel preS and preF sound substitutions in the late West Uralic era.
For now I’m still sketching out the situation with back vowels. In particular it’s not clear to me how the raising PU *ë-ə and *aj(C)ə > preS *a-ə > PS *ō-ë should be dated: this is attested from numerous Germanic loanwords, and thus could be newer than *ä-ə > *e-ə. It may well be the same change as preS *a-a > PS *ō-ē (likewise attested from Germanic loanwords); and thus not triggered by stem vowels at all.
[1] In their view the result of this merger would not have been quite [i], but a near-close vowel they mark as *ḙ. I would suggest the sound value [ɪ] for this (similarly [ʊ] for their *o̭), reflecting the common tendency of close short vowels to reduce and centralize. Initially this probably would not have had any phonological signifigance though, so I will continue to use *i and *u for the early pre-Samic and early pre-Mordvinic era.
[2] “Common” rather than “proto”: while West Uralic at least seems like a defensible subgrouping to me (unlike its traditionally assumed kin like “Proto-Finno-Volgaic”, “Proto-Finno-Permic”, etc.), the common innovations are not many, and it remains effectively only a dialect of Proto-Uralic itself. This being the case, an accurate picture of West Uralic can only be gained by starting from Proto-Uralic and “reconstructing upwards”, not by presuming the existence of the group and attempting to compare Samic/Finnic/Mordvinic in isolation (a method that has traditionally generated rather Finnocentric models, further muddled by conflicting evidence from areal later-diffused vocabulary). It would also be premature to rule out entirely the possibility of WU being an “areal-genetic” group of dialects after all, since non-exact parallels for a few of the characteristic innovations (e.g. *ë- > *a, *åĆ > *aĆ, *-d₂- > *-d₁-) can be found in Mari, Permic and even Hungarian as well.
[3] It does not seem clear to me if these cases should be assumed to involve the suffixation of a consonantal suffix such as *-w and a later development *-əw > *-o, or simply the addition of *-o as a suffixal element right away, but this does not really affect their validity. If the former though, then this has some implications for the history of the PS stem type *ā-ō; they could not descend from *ä-o at the West Uralic level, but would have to go back to preS *ȧ-ȧ < PU *ä-ä, with the labial suffix only as an incidental addition.
[4] The irregular vowel correspondences in the Finnic words could perhaps be accounted for by assuming contamination from the Germanic loanword *leevä ‘slight; temperate’. This is one of the very old Germanic loans in Finnic that shows *ē → *ee. While both sides appear to point to a mid vowel [eː], I believe this is illusory. PIE *e was probably closer to [ɛ], and the eventual lowering and backing of *ē to *ā in Northwest Germanic suggests that even an intermediate [æː] existed at one point; as is also shown by the existence of a couple of loanwords in Finnic that have *ē → *ä (e.g. PGmc *wēgaz ‘lever, scales’ → PF *väkä ‘hook’). Pre-PGmc *klēwas ‘lukewarm’ was thus probably loaned as pre-Finnic *lääwä, later raised to PF *leevä together with inherited words like PU *lämə > preF *läämi > PF *leemi ‘broth’. At this period it could be assumed that pre-F *läppətä ‘mild’ was adjusted to *lääpətä, on the model of *lääwä; with later raising then giving PF *leepedä. — Even slightly earlier *leeppedä could perhaps be assumed, with PF *leepedä and *leppedä representing two ways of naturalizing the overheavy syllable structure.
[5] Another highly similar word family exists as well: PF *käprä ‘rolled up’; PS *kēpr-ë- ~ PF *käpr-i-stä- ‘to roll up’. As has been proposed by Katz, in principle this might represent an earlier parallel loan with PIE *kʷ still retained on the loangiving side, substituted by pre-S / pre-F *p. Dating *l > *r in Indo-Iranian as already this early seems unlikely however, and I suppose a more probable explanation would be that this is Uralic-internal descriptive variation. Note also a number of obviously secondary formations in Finnish such as käkkyrä, käppyrä ‘curved thing’.
Freelance reconstruction 
Your scenario of umlaut stacking looks quite plausible, but I would propose an alternative explanation for some of your examples. Namely, *meanō-, *peajvē, *pealkē and *pealē (and possibly *peajō- and *veakkē as well, if the etymologies are correct) can be accounted for by assuming a regular sound law *PäCä > *PeaCē, where P is any labial consonant. This explanation has two advantages: first, *pealē can be traced back directly to *pälä without any blend whatsoever; second, we do not need to assume late independent syncope in *päjwä in Saami, Finnic and Samoyed. The rule also explaines PS *veaškē < *wäśkä.
Possible counterexamples are few and not very convincing. For example, *pāŋkē ‘reindeer’s headgear’ may be a loan from Finn. panka 'handle' and so does not necessarily go back to PU *päŋə ‘head’.
As a sidenote, I cannot believe that *ńälmä ‘tongue' is derived from *ńälə- ‘to swallow’. The semantic side seems unclear to me and I do not know any parallel cases where the word for 'tongue' is derived from 'to swallow'.
One could even add a few more examples supporting this labial consonant rule:
– SaaN beađŋŋis ‘place for foot’ on ski’ ? < PU *pätŋVs (cf. PSam *päŋtüt 'place for foot on a ski' – metathetic?)
– SaaN veadjit 'manage to, be able to, have the energy to' < *wäjä- (cf. Finnic *voi- 'can, be able to' < *vooje- < *vaje- < *väjä-)
– SaaN meaddit 'miss (the target); make a mistake' *peä > *pää.
As regards *päjwä ‘sun, day’, the suggestion that this would be a derivative (*päj-wä < *päji-wä) appears implausible. First, it is unclear what suffix *-wä is supposed to be. Second, there are hardly any forms really supporting the reconstruction of a shorter root *päji-. In MSFOu 259 (p. 259) Saarikivi lists many kinds of forms to support this, but most of his examples are misanalyzed:
– Veps päi ‘sunny weather, day’ is merely a dialectal variant of Veps päiv (< *päivä).
– Estonian päikene clearly reflects earlier *päivkene < *päiväkkäinen, and moreover it may have been influenced by äikene ‘thunder’ < *äijäkkäinen
– Finnish päilyä does not even belong here, as it must be a loan from Swedish spegla ‘to mirror’.
– Khanty *pääj ‘thunder, lightning’ proves nothing, as it may simply have regularly lost postconsonantal *w. Moreover, the vowel is not regular.
– Hungarian fehér can reflect multiple proto-forms, and anyway only the short part part fe- is comparable.
This leaves us only with SaaN beadjut, which has a very narrow distribution (the only cognate is found in Lule Saami); I would just exclude it from the entire etymology.
It looks like you may have lost some material from between meaddit and *peä here. < and > signs in close succession often seem to get interpreted as delimiting HTML tags by the WordPress software. (One workaround would be to use the HTML entity codes < and > instead.)
The count of examples after *p and *m is starting to look interesting by now, but I’m still not very impressed by the examples after *v. The cases of PS *veajē- ~ PF *voi- and PS *veaškē- ~ PF *vaski show no corroborating evidence for original *ä-ä, and in the second, most other Uralic cognates such as Mordvinic *uśkə, Mari *wåž, Hungarian vas instead point to original *a. I find it more likely that these examples show a shift of *a to *ea before palatal consonants: this appears to be regular at least before *ś, *ć, for which four other examples seem to exist (mostly loanwords, but this may simply reflect the fact that native Uralic vocabulary favors medial sonorants over medial obstruents in roots of the shape CVCV):
— Koivulehto’s original proposal of positing loans of different ages from earlier NWGc *gǣtijan- vs. later *gātijan- would be theoretically possible; but in clear cases of parallel loans of this type, it seems to be the case that Finnic usually has an older loan and Samic a newer one, e.g. PS *riŋkē ~ PF *rëngas ‘ring’; PS *tiljā ~ PF *teljo ‘seat in boat’.
— Though it might also be possible to assume two original verbs that have become homonymous in Finnic: *läśkä- ‘to pour’; *laśkə- ‘to let down, let go’, especially since we do find also PS *lōštē- ‘to let free’ < *laśk-ta-?
Re *päjwä, postconsonantal *w does not appear to be lost in Khanty, but rather preserved as *-ɣ:
Es. päikene and Hu. fehér, even if they might be explainable also from *päivä, also do not show any actual evidence for earlier *-wä- either.
I agree that more evidence for a nominal suffix *-wA would be needed, but we have so little solid knowledge about word derivation in Proto-Uralic that I do not think a lack of parallels constitutes an especially strong counterargument against a reconstruction.
Yep, it’s not the first time my post has become garbled by WordPress. So maybe this comes through:
SaaN meaddit ‘miss (the target); make a mistake’ from PU *män-tä- (a causative of *mäni- ‘evade, escape’; cf. SaaN meannut). — The same causative is found in Moksha Mordvin mäńďǝ- ‘let go, let away, let escape (unintentionally)’ and Proto-Samoyed *mäntä- ‘pass; miss’). I would further suggest that Komi me̮d- ‘go, start moving, set off’, Udm medi̮- ‘intend to, be going to, want to; (dial.) procrastinate’ also reflect PU *män-tä- (to my knowledge this has not been earlier suggested).
The connection of SaaN bággi ‘reindeer headstall’ to the PU word meaning ‘head, end’ seems dubious to me because most of the forms do not actually suggest a reconstruction *päŋi. At least Permic and Mansi would rather point to *peŋä, and also in Moksha Mordvin there is /e/ which does not really agree with PU *ä. So perhaps in Finnic the development was from *peŋä to *pe(v)ä and ultimately to *pää (but note Estonian pea).
Deriving *meantē- ‘to miss’ from ‘to get free’ would surely make sense (and your new cognates look good too). This leaves open though the same mechanism of development I just suggested for ‘thumb’, ‘day’ and ‘help’. Also, any opinions on the comparison in UEW with Vakh/Vasjugan Khanty /mintəɣtə-/ ‘to shoot off target’, Tremjugan /mäntəkint-/ ‘to happen upon something’? It’s noticable that similar umlaut-upon-derivation (from *ä to *ii ~ *ää rather than expected *ee) seems to appear here too, and could be assumed also for *ńääləm ‘tongue’, *śänä-kkä > *sääɳəɣ ‘polypore’.
There indeed seems be more evidence in favor of *e than *ä for ‘head’, but at least the Proto-Finnic form is most likely just *pää. The diphthongization *ää > eä > ea is regular in various Estonian dialects, and ‘head’ seems best explainable as having ended up standardized the other way around from most other words; cf. also seadma ‘to adjust’ < *säätä-, teal ‘here’ < *tägällä. Against reconstructing *peŋä we have also *e rather than *i in Mordvinic, and the case of Fi. kevät etc. ~ Mari *keŋəž (and Permic *gož ‘warm’, *gožem ‘summer’?)
Yes, I now noticed the Khanty form, and also that Mansi has *mänt and *mäntǝl ‘through; (dial. also) past’. This makes the etymology somewhat shaky, as Ms. *ä goes back to PU *e-ä and the form *mäntǝl is strikingly similar to the North Saami adverb/postposition meaddel ‘past’ (quasi-PU *mentä-lä-n). So maybe one should rather reconstruct *men-tä- (causative of *meni- ‘go’) and remove the Mordvin, Permic and Samoyed items from the etymology… Hard to say.
Regarding *w > *ɣ in Khanty, the case does not appear convincing to me. The recontruction *parǝɣ ‘flock’ (not ‘herd’) is not right, the forms clearly go back to PKh *pi̮rā. As regards ‘female’ and ‘winter’, *-ɣ could be a suffix. The Khanty words for ‘winter’ go back to PKh *tilǝɣ (not *tälǝɣ!), so there clearly was also a lost I-umlaut trigger here – which also points to some kind of suffixation (? Pre-PKh *täl-iɣ or something like that).
Considering *käd’wä female, I think it would be rather odd if *-d’w- developed to PKh *-jǝɣ- if at the same time *-d’k- was reduced to plain *-j- (cf. PU *tud’ka ‘tip’ > PKh *tuj ~ *tüj). Also Mansi has a single consonant in this root (PMs *kääľ ‘female’), so I’d assume PKh *-ǝɣ in *käjǝɣ ‘female’ must be a suffix (cf. PMs *kïïľ ~ PKh *kaaj ‘goldeneye’ < *śod'ka).
As regards PKh *pääj, this word does not only mean 'thunder' but also ‘heap, mass’, ‘hayrick’, ‘hillock’, and ‘(large, dark) cloud’. The likely path of semantic development probably was from ‘heap’ or ‘hillock’ to ‘(thunder) cloud’ (cf. Latin cumulus ‘heap, pile; (large and puffy) cloud’), and then further to ‘thunder’, so the comparison to PU *päjwä seems also semantically unjustified.
…Hm, so it seems: there’s a Vasjugan form with /-a/, missing from UEW. (Vakh /-ə/ would have to be irregular in either case.)
I do not find it plausible that the only inherited Uralic words showing the correspondence Mansi *CVC ~ Khanty *CVCəɣ would be purely by coincidence exactly those that derive from PU roots of the shape *CVCwV. Now that you suggest that *pääj ‘thunder’ belongs instead together with the various other Khanty words of the same shape (sounds reasonable), there does not seem to be any counterevidence for PU *-Cw- > Kh *-Cəɣ on the table either.
The development *-d₂k- > *-d₂ɣ- > *-j is an interesting contrast, but does not have to present a conflict. I would think palatalization is the likely reason for why this particular development happens in the first place: *d₂ɣ > *d₂j, then cluster simplification > *d₂. This is neatly paralleled by the Hungarian development of *d₂ + velar clusters: > *ľɣ > *ľj > *lj > lgy; as well as by the development *-ix- > *-ij- seen in the Ob-Ugric reflexes of *kixə- ‘to rut, be in heat’. This palatalization of “primary” *ɣ could date to the common Ugric or East Uralic era already.
Typo for *e, actually (you may recall I follow the Steinitz-Honti-Sammallahti notation for PKh vocalism, not the Tálos-Helimski-Zhivlov one).
Regarding ‘raft’: Actually, the V form /parǝ/ is entirely expected, the correspondence V -ǝ ~ Vj -a being regular in nouns. Cf. V ăjǝ ~ Vj ăjā ‘luck’, V ăsmǝ- ~ Vj ăsma- ‘pillow’, V kăčwǝ ~ Vj kăčwa ‘strap for binding cargo’, V lăŋwǝ ~ Vj jăŋwa ‘den’, V kălǝŋʔsǝ ~ Vj kălǝŋʔsa ‘graveyard’, etc. Verbs in *-ā-, however, show a different correspondence.
Regarding the development of *d’k, you might be right that this involved a palatalization of *k (as in Hungarian). But note that Hungarian seems to have changed also *d’w to -lgy (in hölgy (?) from *käd’wä, even though initial h- is irregular); cf. völgy and tőgy (old Hung tőlgy, tölgy) which originally had *d’k.
Interestingly, the Hungarian development seems strikingly similar to Selkup where Proto-Samoyed and *jw and *jk changed to *ľć. First there must have been a change *w > *k (which also happened in initial position in Selkup).
As regards Mansi *CVC ~ Khanty *CVCəɣ, I am not even aware of many examples of such a correspondence. Five seem to be found in Honti 1982:
Ms *ūj ~ Kh *wājǝɣ ‘animal’
Ms *kǟl’ ~ Kh *käjǝɣ ‘female’
Ms *ńǟr ~ Kh *ńǟrǝɣ ‘raw’
Ms *wās ~ Kh *wǟsǝɣ ’duck’
Ms *tǟl ~ Kh *tilǝɣ ’winter’
Three of these have a Uralic etymology. The word for animal is probably cognate with Komi ve̮j, Jazva Komi vu̇·j ‘timid, wild (esp. animal)’ and Hill Mari wojǝr ‘timid (animal)’, but there is nothing explicitly pointing to postconsonantal *w in this case (although the possibility cannot be entirely excluded: PU *wojV / ? *wojwV).
Only for ‘female’ and ‘winter’ posticonsonantal *w can be securely reconstructed in PU. I would not conclude much on the basis of these two examples, for two reasons. First, postconsonantal *w was nevertheless definitely lost in PKh *pilǝŋ ‘cloud’ from PU *pilwi (*-ŋ of cource cannot be a reflex of *w, but it is rather a derivational suffix corresponding to Hung -g in felleg ‘cloud’). Second, there are cases where a word-final *-ǝɣ in Khanty obviously must be a derivational suffix that was added to the PU root, such as PKh *sǟṇǝɣ ‘bracket fungus’ (cf. Fi sieni) and *kälǝɣ ‘strap’ (cf. Fi köysi). (These two show *-ǝɣ also in Mansi, though, so the suffix must be a common Ob-Ugric addition here.)
I find this less certain. Geminate -ll- in Hungarian isn’t derivable from *-lw-, and in light of the parallel form fëlhő, we might need to segment fëlleg as √fël+leg (from proto-Hungarian *filɣäg?). Mordvinic forms suggesting PMo *pejəľ (or even *peŋəľ?) are hard to reconcile with the Samic-Finnic reconstruction *pilwə as well.
Possible, but IMO not quite obvious either. This formant appears to be 1) unproductive, 2) without any semantic function, and 3) typically no ‘underived’ root can be found alongside. Gerald Ganschow (e.g. in the C5IFU proceedings) may not have been entirely wrong in supposing these formants to have been generally a part of the original root structure, as it definitely is in cases where other cognates indicate earlier *k (e.g. *jepəkä > Ms *jipəɣ ~ Kh *jepəɣ ‘owl’; *särkə > Ms *täärəɣ ~ Kh *ɬäärəɣ ‘ruffe’). I would not rule out entirely the possibility of reconstructing PU *śänäx ‘polypore’ (?? perhaps with *-äx > *-ə in pre-Finnic, much as also *-A-k > *-Ek in deverbal derivatives), *käwd₁əx ‘rope’, and furthermore also e.g. *käčVx ‘knife’, *čončəx ‘netstring’, *śülkəx- ‘to spit’.
It is not clear to me that Hungarian -ll- could not reflect PU *-lw-. Also the reflexes of PU *-lk- show unexplained duality between -l- and -ll-, sometimes in the same root (cf. világ and villám from PU *wi̮lki-). And at least PU *d’k and *d’w seem to have identical reflexes (Hung lgy), so couldn’t the same be true of *lk and *lw?
As regards Ob-Ugric root-final *-ɣ, I don’t think it’s very probable that it reflects a part of the original root in words like *sǟṇǝɣ ‘bracket fungus’ etc. Of course, much remains unknown about PU root types and root-final elements, but postulating something like *śänäx etc. on the basis of Ob-Ugric evidence alone strikes me as an “obscurum per obscurius” type of argument. The fact remains that there are many quite obvious cases of added suffixes even though 1) the suffix is unproductive, 2) the suffix seems to lack a semantic function, and 3) no underived root survives in the same branch. Just a couple of examples from the top of my head:
– Finnic *kama-ra ‘crust (of the earth); pig skin’ and *viherä ‘green’ (~ Mari *ŭžar) from PU *kama ‘skin, shell’ and *wiša ‘green’
– Saami *koackē-mē ‘eagle’ from PU *kočka ‘eagle’
– Saami *pean-e̮k ‘dog’ and *kām-e̮k ‘shoe’ form PU *penä ‘dog’ and *kämä ‘shoe, boot’
In the last example the derivatives must have been formed in Proto-Saami after the metaphonic vowel shifts (because PU *pen-ik and *käm-ik would of course have yielded **pe̮ne̮k and **kieme̮k). I’m sure a lot more similar examples could be found.