Phonology squib: *ë in Kamassian

Another word of previously notably unknown etymology recently has a new lead for it: Finnic *sana ‘word’, suggested by one Otso A. Bjartalíð (in a draft that was briefly posted on Academia.edu but seems to be currently down) to have cognates in Kamassic: Kamassian tenü, Koibal tano ‘word’. Looks possible enough, though I have some lingering douts about a comparison with just one branch of Samoyedic.

Otso holds that the Kamassian form /tʰenü/ (thênü (C), tʰenʉ (D)) would point to PSmy *tänü, slighly off from the back-vocalic forms in Koibal and Finnic, which he takes to suggest Proto-Uralic *sana. Actually, this does not seem to be necessarily a problem: the correspondence Kamassian e ~ Koibal a can also continue PSmy *ë, especially before nasals. Compare:

PSmy *cën (? *tën) > Km. then (C), tʰeǹ (D) ~ Kb. танъ (Sp.) ‘sinew’
PSmy *jëpsə > Km. ťepsü (C), ťeps^ʉ (D) ~ Kb. джяпсы (Sp.) ‘cradle’
PSmy *këm > Km. khem (C), kʰə̑m̀, kʰɛm̀ (D) ~ Kb. камъ (Sp.) ‘blood’
PSmy *pën- > Km. phelľim (C), pʰel̆ľem (D) ~ Kb. паллямъ (Sp.) ‘I put’ (the presumable stem pʰen- does not seem to be attested)
PSmy *wën > Km. men (C), mɛǹ (D) ~ Kb. мянъ (Sp.) ‘dog’

(Usual primary source abbreviations: C = Castrén, D = Donner, Sp. = Spasski.)

As PSmy *ë regularly corresponds with Finnic *a, both coming from PU *ë, a better first-pass PU reconstruction for ‘word’ would therefore seem to be *sëna, ^[1] especially given that PU *a most typically gives PSmy *å > Kamassic o. (A number of cases of PSmy *å > a are still known though; even a number of > u.)

The stem vocalism and overall morphology could call for more thinking too, since PU *sëna would be actually expected to give PSmy **tïnå (regardless of if one thinks *ë > *ï is conditioned by syllable closure or by an *a-stem). Maybe the root was instead a verb *sën(ə)- ‘to say’? Despite the semantic identity, both reflexes could then turn out to be parallel deverbal nouns: in Finnic *sën-ma > *sëna (with *nm > *n as also in the 1PS oblique possessive suffix -ni), in Kamassic *sënə- > *tën- → *tën-u (or *tën-o).

We have already in the data above still some variation within Kamassic though. For one, Spasski’s ‹я› could indicate fronting in ‘cradle’, perhaps *ďapsə > *[dʑæpsə]. The same appears in PSmy *jalä > Kb. джяла ‘light, day, sun’ and PSmy *japsə- > Kb. джяпсьы- ‘to roast’, which retain back /a/ in Km. ťala (C), dʑåɫ̀å (D) and ťapse- (C). Donner’s variant with ə̑ could be just unclear articulation, as he has many variant forms with this.

PSmy *ë also does not always result in these reflexes. Firstly there are also cases of retained /ë/ (= FUT e̮), though this is rarer and is only found in a few of Donner’s records. Castrén’s materials as published do not even recognize this phone, ^[2] and for Kamassian we seem to find in them ä for the short version, ö̂ for the long version.

PSmy *ëlə- > Km. è̮lɛgɛn (D) ‘wet snow’
PSmy *ëptə > Km. äbde (C), e̮ʔʙᴅᴉ, eʔʙtə, ɛʔʙte (D) ~ Kb. абде (Sp.) ‘hair’
PSmy *kë ‘year’ > Km. khä (C), kʰɛ̮, kʰe̮ (D) ~ Kb. ка (Sp.), kôa (Klaproth) ‘winter’
PSmy *ëməl- ‘to forget’ > Km. nö̂mel- (C), nē̮məl- (D) ~ Kb. нумил- (Sp.) ‘to forget’ (with unclear initial /n-/ and Koibal /u/).
perhaps also PSmy *këpə > Km. khö̂b (C), –kʰ^ʉʔʙ (D) ~ Kb. копъ (Sp.) ‘wasp’.

This representation could suggest that Kamassian short /ë/ was phonetically mostly an open-mid [ʌ] or [ɜ]. Spasski’s ‹a› here and above could also actually stand for such a sound; but this is unknowable in the absense of phonetically exact records of Koibal.

On the other hand, I do not suppose Kamassian e to also stand for a mistranscribed back /ë/, as frontness is corroborated by the 2nd-syllable vocalism in thênü, ťepsü and phelľim. Kamassian seems to have had stronger vowel harmony than Proto-Samoyedic or even Proto-Kamassic had, reasserting harmony in words like *jalä > ťala ‘light, day, sun’ as mentioned above, or also e.g. *kålä > Km. kola (C), kʰōɫă (D) ~ Kb. кола (Sp.) ‘fish’; *səjmä > Km. sima (C D) ~ Kb. сима (Sp.) ‘eye’. We see this reharmonization also in some words that show PSmy *ï > Km. /i/, e.g. (*pujå >) *pïjå > Km. phîjä (C), pʰʉj̀e (D) ‘nose’; *sïrå > Km. sirä (C), s^erɛ (D) ‘snow’ (back vocalism retained in Koibal?: сыра (Sp.)). This development is likely to be due to Turkic influence.

Annoyingly however, I cannot assign any particular conditions to this apparent *ë > /e/. As mentioned, to some extent it looks like the typical development before nasals (tʰenü, tʰen, kʰem, men; also lem (D) ‘bird cherry’ << PU *ďëmə, though l- is unexpected), but this is contradicted by ‘to forget’. Other examples still occur too. In three cases we find a front vowel also in Koibal:

PSmy *ërö > Km. ere (C), ɛre (D) ~ Kb. ирэ (Sp.) ‘autumn’
PSmy *lë > Km. le (C D) ~ Kb. ле (Sp.) ‘bone’
PSmy *ńëj > Km. neä, njä (C), ńȧ` (D) ~ Kb. не ‘arrow’

while some show Castrén’s ä corresponding to Donner’s e:

PSmy *ëtå- > Km. ädeʔb- (C), edəʔ- (D) ‘to wait’
PSmy *këcɜ- > Km. khä̂demgä (C), kʰēdəmgə (D) ~ Kb. кадума (Sp.) ‘ant’
PSmy *këptu ‘currant’ > Km. khäʔbde (C). kʰɛʔʙᴅɛ (D) ‘berry’

Given furthermore vacillation between e̮, e, ɛ in Donner’s records of ‘hair’ (and with a 2nd syllable front vowel!), perhaps we should rather assume *ë > /e/ in general in Kamassian — also Castrén’s ä might be generally accurate in this case — with some secondary idiolectal backness variation. Turkic loanwords also show occasional e̮ for front ö in the original, e.g. ȫɫʉm, ē̮ɫ^ə̑m ‘death’, qē̮b^ərʉʔ ‘bridge’ (from ölüm, köbrük). ^[3] The two examples of Koibal e might represent independent fronting (in CV stems? though not in ‘winter’).

Some sporadic examples also have still different reflexes of PSmy *ë in Kamassic: /a/ in e.g. *ńërkå > Km. narga (C), nå̆rga, nə̑rʁa (D) ‘willow’; /o/ in e.g. *më- > Km. mo- (D) ‘to take’ (potentially regular after /m/); maybe /i/ in *sër > Km. siri (C), sɪrɛ (D) ~ Kb. сыры (Sp.) ‘white’ (probably influenced by, if not outright rather derived from *sïrå ‘ice’)…

More strange however are cases where we find correspondences similar to the above, but resulting from a few other PSmy vowels. For starters, the following also show Donner’s e̮ ~ Castrén’s ö̂:

PSmy *ńamɜ- > Km. nö̂mür (C), nē̮mə̑r, nȫmə̑r (D) ~ Kb. нёморъ (Sp.) ‘soft’
PSmy *kät- > Km. šö̂- (C), še̮ʔ-, šə̑ʔ- (D) ~ Kb. сод- (Sp.) ‘to sew’ (clearly with an original front vowel given *k- > š-)

The correspondence Kamassian e ~ Koibal a likewise sometimes seems to result from other PSmy illabial vowels. There is (at least) one case of *ï (usually > Km. i, Kb. и, ы), two of probably *ə (usually > Km. ə̑, a ~ Kb. а, о) and four of *ä (usually > Km. e, Kb. е, э):

PSmy *ïtå- > Km. ä̂de- (C), ɛdə- (D) ~ Kb. ад- ‘to hang up’
PSmy *ət³kä > Km. eši (C), ɛɕĭ, eɕĭ (D) ~ Kb. асе (Sp.) ‘child’
PSmy #kəå > Km. kee (Adelung), ket (Pallas) ~ Kb. ка (Sp.) ‘moose’
PSmy *kärɜ- > Km. šêr- (C), šērə- (D) ~ Kb. сар- (Sp.) ‘to be ashamed’ (contrast Km. šêr-, šēr- ~ Kb. сер- ‘to dress up’)
PSmy *mäktə > Km. bäkte (C), mɛkté (D) ~ Kb. бакты (Sp.) ‘tussock’
PSmy *wäto > Km. bêdü (C), bedʉ (D) ~ Kb. бадё (Sp.) ‘intestine’
PSmy *tättə > Km. thêʔde (C), tʰēʔd̀ə (D) ~ Kb. таде (Sp.), tätde (Pallas) ‘4’

This correspondence also appears in negators: Km. ɛm (D) ‘I don’t’, ei (D) ‘not, doesn’t’ ~ Kb. абы (Sp.) ‘not’, though this is probably morphological rather than phonological: the rest of Samoyedic also reflects a variety of vowels in different negators, e.g. Selkup aša ‘not’, 2PS imp. ïkə ‘don’t!’ (indeed also elsewhere in Uralic, e.g. Finnish en ‘I don’t’ but älä ‘don’t!’ ^[4])

The relationship of these correspondences and *ë > e ~ a could be understood in at least three ways:

there was a common Kamassic change such as *ë > *e, merging with primary *e from at least *ä, and later backing to ‹а› in Koibal;
there was a change such as (*ä >) *e > *ë with later fronting to /e/ in Kamassian;
backing and/or lowering of vowels in Koibal and fronting of *ë in Kamassian are independent processes that coincidentally result in similar-looking correspondences.

My considerations above on vowel harmony would maybe suggest option 1; cf. also асе, бадё = ? /ase/, /badö/ where disharmony might be secondary from *eše, *bedö; however, the two cases of secondary /ë/ suggest that option 2 can have happened too. Option 3 would have the most flexibility of these, allowing for *ë > e ~ а, *ä > e ~ а etc. to each occur under different conditions. Indeed, while the former looks like it might be conditioned by a preceding nasal, the latter shows no such evidence, and counterexamples can be found: e.g. *ken > Kb. сэнъ ‘knife sheath’, *tänä- > Kb. тен- ‘to think’.

So far I have no really firm conclusions on this wider topic, but at least it is quite evident just how much about the historical vocalism of Kamassic is still up in the air. Substantial progress will probably require various lines of attack, such as revisiting the primary data (esp. Castrén’s original records, cf. footnote 2); seeing what processes could be corroborated by loanwords from Turkic; or double-checking with evidence from Selkup and Mator if any of the “Proto-Samoyedic” reconstructions I’ve given (most from Janhunen) might be really only valid for Northern Samoyedic.

[1] Intriguingly we still find a non-open vowel also in Estonian & Votic sõna, Livonian sõnā, and South Estonian syna (with regular pre-nasal mid vowel raising). It is not inconceivable that this is some kind of a retention, though clearly this is not really regular either.
[2] However, this is probably due to the posthumous editing by Schiefner. The preface to Castrén’s only recently fully edited Selkup materials (Manuscripta Castreniana Ostiak-Samoiedica) already has Castrén using ê and î for the nonfront unrounded vowels that we today transcribe e̮, i̮ or /ë, ï/.
[3] Incidentally I only now realized when writing this post, cross-checking Donner’s dictionary with Joki’s treatment in Die Lehnwörter des Sajansamojedischen, that Donner distinguishes ė (‹e› with dot = close-mid /e/) from e̊ (‹e› with ring = labial /ø/), likewise their reduced equivalents ə̣ and ə̥. Atrocious practice in transcription, if you ask me — and this has been copied also in Janhunen’s SW, whose printing quality seems to be too bad for me to even distinguish the two most of the time.
[4] Negation in early stages of Uralic is still in need of a good reconstruction though. The negative verb maybe most likely had the stem *e- in the indicative and *ä- in the imperative (at minimum Finnic, Mari and Khanty all suggest something like this), but there are also forms pointing to a stem *a-; and e.g. Permic moreover shows a split between *o- (< *e-) as the present stem and *e- (probably < *ä-) as the preterite stem.

Source: protouralic.wordpress.com