Prothesis in Permic (and related matters)

A certain type of soundlaw that frequently seems to occur across the Uralic languages is the prothesis of a labial semivowel before “medium-strong” labial vowels. Two clear examples are Samic and Finnic, where this has occurred before the long vowel ⁽*⁾/oː/, but not short ⁽*⁾/o/, nor ⁽*⁾/u/ of any length. In Samic this comes up particularly frequently, due to the unconditional raising of previous *a to *ō.[1] Thus e.g. PU *ëppə “father-in-law” → West Uralic *appə → pre-S *ōppë → proto-S *vōppë → Southern Sami voehpe, Northern Sami vuohppa, Skolt Sami vuõpp. (Contrast Finnish appi, Hungarian ipa, etc.) Or for an example from Finnic: PU *uwa “flow” → eLPF *oo → LPF *voo → Estonian voo, Finnish vuo.[2] The effect, I suspect, is in these cases related to the diphthongization of *oo to *uo, which has independently occurred in both Samic and most of Northern Finnic.

This phenomenon also turns up in Permic, though with some interesting quirks. Three broad features set this case apart as remarkable among the Uralic languages:

  1. The appended consonant is *w, remaining distinct from PU *w → Proto-Permic *v. In Komi and several Udmurt dialects, the distinction has been lost with this secondary *w also spirantizing to /v/; but a genuine distinction between /w/ and /v/ turns up in some other dialects of Udmurt, alone of all Uralic languages (and, indeed, most other languages of western, central and northern Eurasia).[3]
  2. The change occurs not only word-initially, but also following the consonant *k. This change seems to represent a major departure from the nearly pan-Uralic syllable structure template, which allows only a single onset consonant.[4]
  3. Not all instances of the change occur in the vicinity of labial vowels, leaving the phonetical motivation of the change slightly unclear.

#1 and #2 are related features, I suspect. Tackling the 2nd of these first, the reconstruction of a lone initial consonant cluster *kw- in Proto-Permic seems in fact sufficiently odd to be untenable… this should probably be instead analyzed as a labiovelar monophoneme *kʷ! Its development back to /k/ in Komi might make better sense from this starting point too.[5] Moreover an analysis as /kʷ/ should possibly be applied to Udmurt as well. Although my raw data sources on the language seem to incidate simply kw, one argument in favor of analyzing /kʷ/ is that w remains here even in the dialects where bare word-initial *w has been merged into /v/. The existence of a Permic *kʷ would also have a nice parallel with the existence of *kʷ in Mansi to the east.

Once the introduction of a labiovelar stop is understood as an areal feature, the same turns out to be applicable to the labiovelar glide. Permic originally seems to have been a “frontier” language with respect to the spirantization of original *w to *v: the subfamily is flanked by Mari and Mansi, both of which have to this day yet to shift their /w/ to /v/.[6] Hence the appearence of secondary *w rather than *v can be explained as areal interference. It could even be speculated if the rendering of original *w as *v indicates Permic having originally occupied a more western location among the early Uralic dialect chain (not necessary perhaps, since the once-southern neighbor Hungarian also shows /v/)?[7]

This explanation of the first two oddities, though, seems to only deepen the mystery of the third. A particularly clear example of *w before illabial vowels is Udm. *war ~ Komi /ver/ “slave”. Other similar instances are findable as well; this is however complicated by the fact that Udmurt seems to retain its labiovelars only before the open vowel /a/. But there is evidence that previously these occurred also before other vowels. Consider the following three words:

  • Udm. /ɨm/ ~ standard Komi /vom/ ~ Komi-Permyak /ɤm/ “mouth”
  • Udm. /ɨn/ ~ std. Komi /von/ ~ KP /ɤn/ “drape”
  • Udm. /ɤsk-/ ~ std. Komi /vos-/ ~ KP /ɤs-/ “to vomit”

I would suggest that *wëëm, *wëën and *wësk- should be reconstructed here (modulo interpretations of the Proto-Permic vowel system), with *w lost in Udmurt and Komi-Permyak, but retained as /v/ in standard Komi. Moreover, these cases also suggest a soundlaw Proto-Komi *ë →  Standard Komi /o/ after labiovelars (but not PP *v: cf. e.g. /vɤj-/ “to sink”, /vɤl/ “horse”, /vɤr/ “woods”, /vɤź/ “wigeon”), which allows bringing in several less direct cases before /k/:

  • Udm. /kɨz/ ~ std. Komi /koz/ ~ KP /kɤz/ “spruce”; ← PP *kʷëëz?
  • Udm. /kɤl-/ ~ std. Komi /kol-/ ~ KP /kɤv-/ “to spend a night”; ← PP *kʷël-?
  • Udm. /kɤs/ ~ std. Komi /kos/ ~ KP /kɤs/ “dry”; ← PP *kʷës?
  • Udm. /kɤt/ “belly” ~ std. Komi /kot/ “peritoneum” ~ KP /kɤt-a/ “pregnant (of cow)”; ← PP *kʷët?
  • std. Komi /kod/ ~ KP /kɤd/ “who”; ← proto-Komi *kʷëd?

The Udmurt doublet /kɤj/ “fat” ~ /kʷajɨ-/ “to become fat” also adds some support for a loss of labialization before central vowels.

The mystery still isn’t particularly deep, though. Rewinding all these words backwards toward Proto-Uralic turns up mainly *o: *orja “slave”, *oŋə “mouth”, *owd₁əm(ə) “drape”, *oksɜ- “to vomit”, *kowsə “spruce”, *koksɜ “dry”, *koktɜ “uterus, belly”… The main exception is the interrogative pronoun, which seems to go back to *ku-. “To spend a night” has cognates only in Ugric and appears to be compatible with both *kola- and *kala-.[8] It is also *o that can be reconstructed in the best-attested cases where Udmurt /kʷa-/ occurs:

  • /kʷa/ “house” ← *kota
  • /kʷal-d-/ “to split” ← *kolɜ “cleft”
  • /kʷaś-/ “to dry” ← *kośkɜ-
  • /kʷaśi/ “wild duck” ← *kośkɜ? (~ Khanty *kaas, Selkup *kʷëćə)
  • /kʷar/ “leaf” ← *korwa? (~ Samic-Finnic *korva “ear”)

Hence it would appear that what originally happened here was very similar process as in Samic and Finnic: *o was broken to *wo (or labialized a preceding *k), and only later did the vowel diverge to various other reflexes.

Oddly though, cases of *wa- seem to instead mainly go back to original *a-ə:

  • *wa “year” ← *ëd₁ə
  • *waĺ- “to spread out, to cover” ← *ad₂ə “bed”
  • *wadź “early” ← *anśə? (~ Mari *ažnə; also *onśə might be reconstructible)
  • *wadźer “tusk” ← *anśər(ɜ)

Should an early labialization to *o be reconstructed here? It’s worth noting that PU *a, *o and *ë have very similar reflexes in Permic — all three seem to have PP *u as their main unconditional reflex, for starters.

Other questions remain as well.

  • Can any Permic-Mansi loanwords be identified where *w or *kʷ would occur? If this is to be an areal feature, such cases could be suspected to exist.
  • How shall cases where *k does not labialize before original *o be explained? Examples include at least *kočka → *kuč “eagle”, *kod₂ka → *kuĺ “spirit”, *kopa → *ku “skin”, *kośɜ “long” → *kuź, *kojə “male” → *koom “man”. Postulating any original distinction such as *k versus *q doesn’t seem to be supportable.
  • Udmurt /a/ after labiovelars corresponds to Komi /o/. After a non-labializable onset, in words of this type (← PU *o-a and *a-ə) we however find PP *u = /u/ in both: this includes both the immediately previous examples, and many others, e.g. *tolwa → *tul “wedge”, *worka- → *wurɨ- “to sew”, *ńalə- → *ńul- “to lick”, *śarwə → *śur “horn”. Presumably *o was retained after labiovelars, and was furthermore lowered to /a/ in Udmurt. But the correspondence U. /a/ ~ K. /o/ (let’s denote it as PP *å for the moment) also sporadically turns up in some other words: e.g. *låjɨ- “to knead”, *måď “riddle” (← *mojə- “to remember”?), *pårś “pig” (← *porćas). What’s up with these?
  • Could labiovelars have once occurred also before other Permic vowels yet, only to have been later lost? This is not obvious in any way, if they can be mostly traced back to original *o. But this would provide an interesting new mechanism for perhaps explaining some aberrant vowel correspondences.
    In particular, Udmurt /ɨ/ frequently corresponds “irregularly” to Komi /u/ — and this looks rather analogous to the /ɤ/ ~ /o/ and /ɨ/ ~ /o/ correspondences brought up above. Several cases with original *ku- seem to turn up here: e.g. PU *kupsa- “to extinguish” → U. /kɨsɨ-/ ~ K. /kus-/ (= PP *kʷɨsɨ-??); PU *kuńa- “to blink” → U. /kɨń-/ ~ K. /kuń-/ (= PP *kʷɨń-??). But there are also cases of the normally expected *ku- → /kɨ-/ in Komi, e.g. *kud₁a- “to plait” → /kɨ(j)-/; and cases of this exception correspondence to be found after other consonants, e.g. PU *salka- “to stand” → U. /sɨl-/ ~ K. /sulal-/.

Perhaps some investigation into the Proto-Permic vocabulary not found elsewhere in Uralic could clarify these matters. I’ve yet to read Sándor Csúcs’ 2005 monograph Die Rekonstruktion der permischen Grundsprache; it should be interesting to see if he arrives at any conclusions alike to mine here.

[1] I seem to have somehow acquired a habit of denoting Proto-Samic long vowels by macrons (*ō), yet by geminated vowels (*oo) in other contexts. Originally this was motivated by the fact that “long” vowels in Proto-Samic seem to have been neutral rather than marked (e.g. in loanwords Finnic and Scandinavian *a is substituted by *ā). However, for full consistency I probably should extend this notation, at minimum, also to Khanty, where a similar situation applies; and indeed, its “long” vs. “short” vowels are actually full vs. overshort. Yet I continue to write *uu *üü *ää *a and not *ū *ǖ *ǟ *ă etc. for the language… this time largely out of disdain for stacked diacritics. I hope this isn’t making too much of a mess out of the followability of my writing.
[2] Curiously, Livonian retains this root as bisyllabic: õvā. It follows that the stage *oo → *voo needs to be considered Late Proto-Finnic, later than the splitting-off of Livonian and South Estonian during the Middle Proto-Finnic phase.
[3] A similar but weaker situation occurs in Selkup however, where following the fortition of inherited *w- to *kʷ-, a new epenthetic *w- has developed in a number of cases. For an actual distinction between /v-/ and /w-/, we need to look quite far though; examples of this occur in some Germanic languages (English, Middle High German), some Sinitic varieties such as Sichuanese, as well as some varieties of Chukotkan.
[4] The other main exception is Mordvinic, where retraction of stress from the 1st syllable has in certain cases (but predominantly in newer loanwords) led to its syncope.
[5] A singular counterexample that may actually be in support for this analysis is the numeral “6”, /kvať/. I suppose /v/ in here does not derive from older *kw/*kʷ, but instead from some kind of a syllable contraction, from something like earlier *kuať. An original trisyllabic shape such as #kuwətə is also suggested by the long vowel in Finnic *kuuci.
[6] Some descriptions of Mari and Mansi appear to posit /β/, but I’d guess this is not intended to stand for IPA [β], and is rather simply a case of the confusing UPA notation where fricatives and approximants are not distinguished, and “β” is moreover used as a broad transcription for [w].
[7] Incidentally, the fact that original *w is not lost in Permic in any position (*wülä → *vɨl “upper”, *wud₂ə → *vɨĺ “new”, *wosa → *vuz “ware”) independently suggests that *w → *v occurred very early on in the subfamily.
[8] Though the fact that this yields Hungarian hál-, Khanty *kaal- versus hal, *kuul from *kala “fish” may point to original *kola- being preferrable.

Advertisements
Tagged with: , , ,
Posted in Reconstruction
11 comments on “Prothesis in Permic (and related matters)
  1. I think there are two very good observations here – the monophonemic interpretation of Proto-Permic *kw- must be entirely right, and it’s an excellent explanation to extend *(k)w- to explain the correspondence Udm ï- ~ Komi vo- ~ KP vë-.

    But as you say, the Uralic background of PPerm *kw- and *w- remain very puzzling. The double representation of PU *ko–a seems especially hard to account for in light of minimal pairs like *kwu ‘house’ (< PU *kota) vs. *kŭ 'skin' (< PU *kopa). One could perhaps assume that in the former case the word internal *-p- might have blocked the labialization of *k-, but this wouldn't account for e.g. *kŭčk 'eagle' ( *kwu- I could add one plausible-looking case: Komi kos- ‘turn around, turn back (intr.)’, Komi kose̮d- ‘turn around, turn back (tr.)’, Udm kwasa-. These would seem to be related to PFi *kosta- ‘turn back (tr.)’ (apparently a derivative: *kos-ta-). However, it is not clear to me why the sibilant remained unvoiced in Permic (*kwus- instead of *kwuz-).

    As a small note, Selkup *kwëćə cannot have anything to do with Udm kwaśi because Selkup *ć reflects Proto-Samoyed *j.

    Udm war ~ Komi ver ‘slave’ looks like an irregular correspondence. Since ver ‘slave’ is only attested in Old Komi, I assume it has some special explanation. I think the other Komi word mentioned here (verës ‘husband’) must be unrelated; it seems rather evident that we have here a different PPerm word *virës ‘man’, which must be an Indo-European loan (cf. PIE *wiHros ‘man’). Could there have been some sort of contamination between the reflexes *virës ‘man’ and *wur ‘slave’ in Old Komi?

    • Juho says:

      The double representation of PU *ko–a seems especially hard to account for in light of minimal pairs like *kwu ‘house’ (< PU *kota) vs. *kŭ 'skin' (< PU *kopa)

      There’s some further indication that these two do not actually constitute a minimal pair. For one, Estonian koda vs. kõba, where õ in the 2nd does not seem to be explainable by any of the usual conditioning factors. Mokša /kud/ vs. /kuva/ : /kuvə-/ might also be relevant — except this time it’s the 1st that diverges, with a loss of the stem vowel, so perhaps this is a secondary development.

      Perhaps some or all of the words listed under *kopa derive not from PU, but from Turkic *qab (→ Turkish kav, Bashkir /qaw/ etc.) “skin”?? Starting from *kapɜ in pre-Permic would work, and although this word does not seem to be attested from Chuvash, loaning via Oghuric might be capable of explaining the *o in Samoyedic *kopå.

      (*kočka “eagle” also has an interesting parallel in Turkic: *küčigen (→ Kazakh /küšigen/, Siberian Tatar /köčögän/ etc.), though this does not seem to be straightforwardly connectible.)

      I could add one plausible-looking case: Komi kos- ‘turn around, turn back (intr.)’, Komi kose̮d- ‘turn around, turn back (tr.)’, Udm kwasa-. These would seem to be related to PFi *kosta- ‘turn back (tr.)’ (apparently a derivative: *kos-ta-). However, it is not clear to me why the sibilant remained unvoiced in Permic (*kwus- instead of *kwuz-).

      Given askel : as-t-u- (and known parallels for *-Ck-ta- → *-Cta-), this looks like *koskɜ- could be reconstructed.

      • Actually, õ in Estonian kõba would seem to be the result of a regular change *o > õ adjacent to labials. There was apparently also a change *o > õ before i/e in the next syllable. These two conditions seem to account for the vast majority of the attested cases of *o > õ, even though a few exceptions remain.

        As for Finnic *kosta-, derivation from *kos[k]-ta- would seem to work. But then we would need to assume analogical loss of *k in Permic to account for formations like Komi kos-ëd- (the expected form would be kosk- before vowel-initial suffixes).

        • Juho says:

          Actually, õ in Estonian kõba would seem to be the result of a regular change *o > õ adjacent to labials. There was apparently also a change *o > õ before i/e in the next syllable. These two conditions seem to account for the vast majority of the attested cases of *o > õ, even though a few exceptions remain.

          The second change you mention is quite clear indeed, and *o → õ before u, v is also unambiguous — but extending the latter to all labial consonants does not seem to bring in many new cases at all; and there are some clear counterexamples including poeg “son”, oma “own”, and even the inverse *õ → o in hobune “horse”.

          I’d suggest that perhaps the change *o-u/v → õ-u/v is not to be understood as labial dissimilation, but as spreading of [+HIGH], followed by a shift *o̝ → *ɤ̝ → /ɤ/ (much as happened with Livonian ȯ), where labiality is lost to avoid a merger with /u/? This mechanism has the benefit of also explaining the change *o-i → õ-i. (By contrast *o-e → õ-e seems best explained as assimilation via a stage *ë-ë, still retained in Votic and South Estonian õ-õ.)

          There is, on the other hand, a rather frequent change *ko- → kõ- in eastern Estonian (also extending to Votic), which might be relevant for this case.

          (By the way, apologies if any readers are getting confused by the numerous different notations in use here: UPA *e̮ = my historical transcription *ë = Estonian orthographic õ = IPA /ɤ/ all refer here to the same sound.)

          we would need to assume analogical loss of *k in Permic to account for formations like Komi kos-ëd- (the expected form would be kosk- before vowel-initial suffixes).

          Hm, true. There are cases for Permic nominals with a “thematic consonant” drifting from one declension type to another (e.g. -S : -Sk- to -S : -Sj-), but complete loss of the consonant is a rarer phenomenon I think.

          Perhaps a better option might be *kopsV: unlike *k, it looks like *p is regularly lost entirely in Permic in clusters with a sibilant (*jepśɜ, *kupsa-, *küpsɜ → Komi /joś/ “spit”, /kus-/ “to extinguish”, /kɨs/ “skin on reindeer leg”), and the same reduction is again known from Finnic before a suffix (lapsi : las-ta).

  2. Hmm, it seems that there is some sort of bug on WordPress that occasionally garbles postings with asterisks. So the garbled part (without asterisks) was:

    …this wouldn’t account for e.g. kŭčk ‘eagle’ ( kwu- I could add one plausible-looking case…

  3. Oops – that didn’t help either, some text still missing. But I guess you’ll figure out the content.

  4. David Marjanović says:

    Interesting that asterisks get garbled. Were there any < or > in there that were interpreted as HTML? (I had to spell out the HTML entities to prevent that.)

    the diphthongization of *oo to *uo, which has independently occurred in both Samic and most of Northern Finnic.

    Could that be an areal feature (a substrate or adstrate effect)? Compare the same sound change in Romance (Italian and Spanish, with traces in Old French) and Old High German.

    • Juho says:

      I don’t think it can be the asterisks — ASCII symbols that do not have a HTML function usually behave perfectly cleanly. Some particular diacritic is what I would suspect.

      Could that be an areal feature (a substrate or adstrate effect)? Compare the same sound change in Romance (Italian and Spanish, with traces in Old French) and Old High German.

      Yes, it’s quite obviously a sound change that has spread areally. Interestingly, the change’s occurrence also in Western Slavic (later → /uː/ as in High German, but still graphically retained as Czech ů), the Baltic languages, Livonian, and South Estonian even leaves a possibility of connecting the Central and Northern areas of this as part of a single chain. I have not seen studies on if there is actually a connection, but I’d like to investigate this at some point in more detail.

      One complication is that Ingrian, Votic and Standard Estonian seem to form a kind of a barrier zone where /oː/ is retained (similarly /eː/ and /øː/) — but as there are several marginal Estonian dialects that show the change, I’ve for a while now suspected at least Estonian to have actually reversed the change.

      The Samic diphthongization is commonly considered to go back to Proto-Samic, together with diphthongization of *ɛː and *ɔː to *ea and *oa, a dating which would preclude any influence from central Europe. But there’s no necessary reason for these two changes to be of the same age — and actually at least the Southern Sami merger of *ie and *ea, yielding *ea before *á but *ie elsewhere, looks older than either diphthongization. (*oa also merges with /uo/ in various places, but in light of an /ua/ stage attested from Inari Sami, this could be secondary.)

      • David Marjanović says:

        later → /uː/ as in High German

        Only its northern half (Middle German). Almost all Upper German dialects have kept the diphthong ([ʊɐ] in my case, merged with |ur| since non-rhoticity).

        even leaves a possibility of connecting the Central and Northern areas of this as part of a single chain. I have not seen studies on if there is actually a connection, but I’d like to investigate this at some point in more detail.

        Intriguing.

        I’ve for a while now suspected at least Estonian to have actually reversed the change.

        Even more intriguing.

  5. David Marjanović says:

    Oh, a distinction between /w/ and /v/ occurs in westernmost Dutch (West Flemish) and in the Dutch of Suriname, if you want to count that. However, I’ve seen the Dutch /v/ described as a voiced fortis…

  6. David Marjanović says:

    Update: I got lost in the Alemannic Wikipedia for half a day and found there’s an endangered Walser dialect in the Aosta valley that preserves the MHG /v/-/w/ distinction unchanged, with real [w] and apparently real [v]! Blew my mind.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Enter your email address to follow this blog and receive notifications of new posts by email.

%d bloggers like this: