A recent volume (1/2013) of Acta Linguistica Hungarica sports an interesting article from longtime Ugricist László Honti, “Comments on Uralic historical phonology”. I’m tempted to add some thoughts of my own. (Who knows, maybe I’ll make a habit out of this. Too few journals have even irregular comment sections…)
Stop voicing
The article covers three distinct topics. The first section seems to be some kind of an excercise in arguing that an irregular sound development (in this case, initial voicing *p, *t > b, d in a small number of words in Hungarian) is actually perfectly OK. I am firmly unconvinced. Honti mainly appeals to homonym avoidance as a mechanism. While this might indeed explain the lack of a change… this seems rather unlikely to motivate an otherwise unknown development! Not a single context exists where initial voicing would be regular in Hungarian. As for the two cases where a Uralic etymology suggesting irregular voicing competes with a Turkic loan etymology suggesting *b-, *d- to start with, let it be noted that Finnic *tuŋke-, Mordvinic *toŋgə- ‘to press, to cram’, which have been proposed as the cognates of Hungarian dug- ‘to push’, have recently been explained by J. Koivulehto as being rather loaned from pre-Germanic *θweŋgan-. [1]
Moreover my impression is that these voicing irregularities appear to be concentrated in the less secure parts of the Uralic comparative material, which seems likely to contain loanwords from various sources…
I also don’t see much logic in the argument that the lack of cases with *k > **g is to be expected on account of *k already splitting to k, h. For one, Honti explicitly notes that this split would have occurred allophonically at the Ugric stage, and claims that thus the actual expected parallel would require a voiced AND backed reflex as well. This conclusion seems unnecessary at first glance — the alleged voicing split could AFAICT actually be sufficiently late that the entire chain *k > *[q] > *[χ] > h had already occurred, and thus plain *k was left as the only initial dorsal stop to potentially become voiced. (Similar irregular voicing developments occurring in the Permic languages, for which similar explanations have been advanced, definitely suggest that this at least would have come about during Hungarian’s initial steps westward, i.e. after the Ugric period; but I do not know an exact dating for the origin of /h/.) Moreover I really cannot see how the absense of not one but TWO predicted developments would help the case that the voicing split had “partially regular” motivation of some sort? This section raises more questions than it seems to answer.
Laterals or spirants?
Section 2 gets more interesting. I’ve seen Honti bring up his reconstruction of *ɬ, *ɬʲ in place of traditional *ð, *ðʲ for Proto-Uralic in a number of places, but not the reasoning behind this. Indeed, as he says “[the original] papers were published in less readily accessible places”. (At least I now have the exact references!)
The arguments, as presented here (and seemingly a direct English translation of the original in-Hungarian paper), turn out to be slightly disappointing though… even if they are definitely better than some other proposed pipe dream solutions he reviews. He does not really establish his reconstruction as anything more than a viable alternative. I get the impression that a basic point is to take Khanty *ɬ rather than Samic *ð as an “anchor” reflex, but this is all still left as implicit rather than explicit, and with no attempt to review which of these models would fare better typologically or in terms of the number of developments assumed.
(I have an opinion of my own on this matter as well, though. This is not the place to go into it, but very loosely, I think that too much focus has been put on these consonants merging in West Uralic, and that it is not absolutely necessary to reconstruct these as a palatalized-unpalatalized pair of phonemes.)
The suggestion that a PU *θ might be reconstructed to explain Finnic *sulka ‘feather’ vs. PU *tulka, and Finnic *soo ‘swamp’ vs. PU *toxə ‘lake’ strikes me as much too hypothetical to actually support anything about the reconstruction of *ð.
*m-lenition
The largest part of the article is in the third section, which examines the Hungarian reflexes of PU *m, either by itself or in clusters preceded by a resonant (*l, *r, *ð, *ðʲ, *j). A long-standing issue of Hungarian historical phonology is unexplained reflexation of *m as /v/ in a number of words, which Honti attempts to shed some light on. Another is a great degree of variability in if the first member of such a cluster is retained or not, and this issue actually spills beyond Hungarian.
After 15 pages of recapping data, Honti’s first observation draws attention to suffixation: that in words showing the development to v, no derivational suffixes appear. I.e, before a derivational suffix, *m would have been preserved. This seems like a fairly promising condition:
- PU *emä ‘mother’ → H em-e
- PU *kama ‘crust’ → H hám-lik, hán-t ‘to peel’ (*kama-lə-, *kama-kta-?)
- PU *kämä ‘hard’ → H kem-ény
- PU *kuma ‘overturned’ → H hom-orú
- PU *kumɜ ‘cloud’ → H hom-ály ‘darkness’
- PU *ćolmə ‘knot’ → H csom-ó
- East Uralic #jamɜ- ‘to go’ → H in-dul ‘to set off’
- Mari *temə ‘full’ ~ H tem-et ‘to fill’ (PU *tämɜ-ktä-? though I wonder if relating the Mari word to PU *täwðə ‘full’ would be an option)
- Mordvinic *śuməŕďə- ‘to be upset’ ~ H szom-orú ‘sad’ (PU *śomɜrɜ?)
- Permic *śumal- ‘to be hungry’ ~ H szom-jú ‘thirsty’ (PU *śomɜ-?)
One exception still occurs: *ńoma-la ‘hare’, reflected as Hungarian *ńowəl > nyúl.
I’ve attempted some analysis of this same topic before, and an idea of mine might help here. I’ve considered the development *m > v to occur, in one part, when in original Proto-Uralic suffixes, or otherwise following an unstressed syllable. Obviously not many words would end up twice suffixed, so these cases will be predominantly word-final. Words that can be analyzed in this fashion are the following:
- PU *ńälə- ‘to swallow’ → PU *ńälə-mä ‘tongue’ > Ugric #ńälw̃ä > H nyelv
- PU *ńoxə- ‘to hunt, to fetch’ → PU *ńoxə-ma ‘hare’ > Ugric #ńow̃a-l(ə) > H nyúl
- PU *śVðäm(ə) ‘heart’ > Ugric #siw̃ > H szív
- PU *täjə ‘tick’ → *täjə-ktɜmä > Ugric #täktəw̃(ə) > preH *tätɜwɜ > H tetű
- PU *wVðəm(ə) ‘marrow’ → Ugric #wäləw̃(ə) > preH *wälɜwɜ > H velő
Two Ugric words also fit in here, though there is no indication on if their original structure was CVCVm(V) or CVCmV:
- Ugric #äləw̃(ə) ‘other side’ > H elv
- Ugric #eĺəw̃(ə) ‘glue’ > preH *eńw̃(ɜ) > H enyv
I am admittedly required to postulate a ‘weak nasal’ *w̃ at the common Ugric stage here, since in Ob-Ugric, only /m/ appears: Khanty *seem ‘eye’, Mansi *tääkəm ‘louse’, etc. [2] However, in my model, this split could not be placed any later, as common Ugric innovations such as the syllable contraction in the word ‘heart’ eliminate the original conditioning!
Moreover, two cases seem to show m regardless, so there might still be a better formulation to be found…
- PU *aðə- ‘to sleep’ → *aðə-ma ‘sleep’ > Ug #ålma > H álom
- PU *kojə ‘man’ → *kojə-ma ‘male’ > Ug #kojma > H hím
(Still it may also be possible that the latter was formed later, independently of the ethnonym Komi, given the difference in meaning.)
Much like Honti, I also throw my hands up in the air on the topic of simple CVmV > Hungarian CVm / CVv examples, where no pattern whatsoever is evident. :| Until no progress is made here, explanations for other environments will be preliminary at best.
Of palatals and clusters: In Ugric
— Onto cluster issues. An observation I find important is a connection between cluster simplification in Ugric, and word-initial palatal consonants. I recall seeing this position before; in a paper by Helimski IIRC, but I’ve (alas) been unable to relocate it. Again, I’ve looked into the question myself, and it actually works better yet than Honti suggests! He notes three cases:
- PU *ćolmə ‘knot’ > Ug #ćom > H csomó (not in OU)
- PU *śilmä ‘eye’ > Ug #sɪm(ä) > H szëm, Ms *šäm, Kh *sem
- PU *śVðäm(ə) ‘heart’ > Ug #sim (*siw̃?) > H szív, Ms *šim, Kh *sem
We agree that clearly, for this to make any phonetic sense, the development must be split in two stages: palatality assimilation *Ć-l > *Ć-ĺ, followed by a cluster simplification *ĺm > m. Where this gets interesting is that I’ve noticed that the first step can be applied to a number of other words as well, explaining some seemingly unrelated palatalization irregularities (the Mari and Permic cognates for these words point to original unpalatalized consonants, *ð *l):
- PU *śo/ëð(ka) ‘duck’ [3] > Ug #sëĺ > Ms *sëëĺ, Kh *saaj
- PU *śülkə- ‘to spit’ > Ug #süĺɣ- > Ms *süĺɣ-, Kh *söjəɣ-
- possibly PU *śüðə ‘coal’ > Ug #süĺ > Ms *süĺ, Kh *söj (but cf. note 4)
The simplification of the cluster *ĺm is then just the same change Honti, too, notes as occurring in the Khanty-Samic isogloss *śeðʲmä ‘kidney’ > Southern Kh /siimaa/, Southern Sami tjiermie. (Original *-ð- could also be assumed here, in which case we’d have a 7th example of palatality assimilation.) He seems to endorse an alternate explanation presented in the UEW, namely *ðm > *m, given the parallel from ‘heart’ — but this is an unnecessary assumption since the same mechanism that is in any case needed for ‘eye’ and ‘knot’ explains these words, too, perfectly well. The only real complication is having to assume that this assimilation was later than the development of *ð to *l in Ugric — for which Samoyedic *säjəm ‘heart’ presents a problem. (Something like **serəm would be predicted.)
As for lack of assimilation in ‘tongue’: H nyelv, Ms *ńeeləm, Kh *ńääləm, I believe this to be understandable if the assimilation in question is limited to syllable-final *l. Since this word is a derivative of *ńälə- ‘to swallow’ (see above), we could reconstruct a trisyllabic form *ńäləmä rather than a disyllable *ńälmä at the stage palatality assimilation operated — and maybe even consider the ‘epenthetic’ vowel in Ob-Ugric actually a retention. This will however again be somewhat difficult to reconcile with ‘heart’, where a reconstruction with a cluster **-ðm- is clearly not an option.
An alternate approach could be to suppose that only the sibilants *ś, *ć triggered palatality assimilation. This would be puzzling phonetically — but I’ve noticed some compareable phenomena elsewhere in Uralic anyway; again, full details TBP.
Two Ob-Ugric words can be noted that contain the sequence *-ĺəm, but these could be explained either via the trisyllabicity argument, or by having only been formed after the simplification of primary *ĺm. The first is Ms *kooĺəm, Kh *kaajəm ‘ash’; the only suggested cognate outside Ugric is Mordvinic *kuləŋ, with what appears to be a different suffix element, and a difficult-to-explain correspondence Mo *l ~ OU *ĺ. The second is Ms *eeĺəm, Kh *ääjəm ‘glue’, without any cognates outside of Ugric. Also, the Hungarian cognates for these (hamu, enyv) display different reflexation each, which makes it questionable if these can be considered even common Ugric heritage. (Perhaps either one, but probably not both.)
Of palatals and clusters: …and beyond
So far, so good. Attempting to include forms from the other Uralic languages under this same flavor of development does not lead to anything good however. There is some strange logic presented according to which assimilation towards *j would be the best explanation for all cases in all Uralic languages where *Cm is reflected as simply *m (as if no [+consonantal] segments could ever be elided).
First off, Honti seems to be unaware of a point concerning the development of Mari: while *-ðʲ- merges with *-t- as in Finnic and Mordvinic, the regular medial reflex of plain *-ð- is zero. Loss in *šüm ‘heart’ has thus nothing to do with palatality assimilation in any direction — by contrast, it serves as evidence that *ð, and not *ðʲ, is original here. Similarly, no “sporadic” change of any kind needs to be assumed for *aðə-ma > *omə ‘sleep’, *wiðəmə > /wim/ ~ /wem/ ‘marrow’. The other examples available are *kuða- > *koe- ‘to weave’, *śo/ëð(ka) > *śoe ‘duck’, perhaps also *śüðə > *šü ‘coal’ [4] and *ðäpðä ‘spleen’ > *lepə (as Honti notes at another point in the paper, verifying the cluster *-pð- in this word is quite difficult though). The lateralization in *käwðə ‘rope’ > *kəl ‘handle’ is best handled as a loan from Permic, which also accounts for the irregular vowel development. There is, as far as I am aware, no evidence in Mari for a development *-ð- > ˣ/-ð-/ or *-ð- > ˣ/-t/. In all cases such as *koð⁽ʲ⁾wa ‘time’ (cf. Samic *koaðvē, Finnish kotva ‘awhile’) > *kot, original *ðʲ can be assumed.
Shoehorning Finnic here fares badly as well. Honti advances a position I find very strange for ‘dark’:
Finnic reflexes of (40) U *piľme ‘dunkel; dunkel werden’ (Finn pimeä, Est pime) also confirm the change ľ > j (though subsequently this j was lost without a trace in this group of languages).
How can a zero reflex of a consonant “confirm” an intermediate change before its disappearence? Particularly when loss of *j in clusters is not a known Finnic feature? Replacing one irregular change with two is not progress. [5]
The situation in Permic is less clear and there indeed seems to be some kind of a dual development here as well. However *śul ‘duck’, *śɔlal- ‘to spit’, *śinm ‘eye’, U /śulem/ ~ K /śɤlɤm/ ‘heart’ demonstrate that in general, palatality assimilation of the Ugric sort cannot be considered having occurred in this subfamily. (Dialectal Udmurt /śińm-/ for the third seems like a late innovation, and presumably rather related to the vowel /i/ anyway.) I also believe U /vijɨm/, K /vem/ ‘marrow’ could be considered a loan from Mari. So whatever the hell is going on with U /peńmɨt, peĺmɨt/ ~ K /pemɨd/ ‘dark’, U /peń(m-)/ ~ K /pɤjim/ ‘ash’, and U /kwiń(m-)/ ~ K /kujim/ ‘3’ (which seem to require something like *-ĺm- in Proto-Permic), it is probably an entirely separate issue from the Ugric developments discuss’d above.
— Finally, speaking of the numeral three… While Honti takes no shots here at the well-known liquid discrepancy between Hungarian három, Mansi *kuurəm, versus *kolmɜ elsewhere in Uralic, on this topic I find a suggestion by Janhunen interesting: perhaps *r is original here and *l elsewhere is an innovation? To this I would add that *ńeljä ‘4’ presents a source for cross-numeral analogy (a powerful force in introducing irregularities in numerals), so this scheme does not even require assuming any “Finno-Khantic” subgroup as Janhunen does. [6] There is also a potential reason for the absense of such analogy from H+Ms: in these languages -lj- in ‘4’ has coalesced into a single consonant *lʲ, possibly early on, since the later Hungarian fortition development applies: PU *ńeljä > preH *ńeĺ > H negy.
Similar coalescence has incidentally also occurred in Permic (U /ńɨĺ/, K /ńoĺ/). If here the palatalization did not stop analogy from kicking in, we should indeed expect something like Proto-Permic *kuĺm, rather than simple **kulm > **kunm. This particular flavor of analogy, though, could easily also occur starting from older *-lm-. Should two layers of analogy be assumed, even? First *-rm- to *-lm- in Finno-Permic, followed by *-lm- to *-ĺm- in Permic? But… this is only really of any extra benefit if Finno-Permic is a valid subgroup of Uralic to begin with, something that remains to be seen.
Notes
[1] Koivulehto, Jorma (2009). Etymologisesti hämäriä -(is)tA-johdosverbejä, lainoja ja omapohjaisia. In: Journal de la Société Finno-Ougrienne vol. 92.
[2] An alternate path for the development *m > v might be an early denasalization to *b, as Hungarian -v- can also reflect earlier *-p-, in addition to earlier *-w-. Cf. e.g. ‘orphan’: PU *orpa > preH *arbɜ > H árva, or ‘thin’: PU *čupa > preH *šʊbɜ > H sov-ány. Yet the assimilation of *ĺ to ny in ‘glue’ seems to show that nasality must have persisted at least during the early pre-Hungarian period.
[3] Incidentally: only West Uralic *śoðka explicitly points to a CVCCV structure for this word. While *k is not expected to be retained in Permic in this context, perhaps not even Mari if a suitable dating of sound changes is assumed (*ð > ∅ before *-k- > ∅), explaining away Ob-Ugric *sëëĺ is harder. Thus, perhaps the original root is *śoðə or *śëðə, extended to *śoð-ka in WU?
[4] Though reconstructing *śüðʲə might work as well, given a precedent for loss in (nominal?) roots of the shape CVCə: *wuðʲə > *u ‘new’. Samoyedic *sijə will be easier to handle from this starting point, too. No Permic reflexes of ‘coal’ are known for verification.
[5] I believe a reconstruction *piðʲmə- may be preferrable. We do not have a precedent for the reflexation of *ðʲ before a nasal in Finnic, but that precisely allows the possibility of loss regularly having occurred here. The predictable reflex would be *-tm- which is not attested in the Finnic languages. — I recall seeing comparisions with *pilwə ‘cloud’ in various sources, but I don’t think this is sufficient to estabilish *-lm- or *-ĺm- by internal reconstruction; the semantic gap alone is considerable, and while there is some scattered evidence suggesting a *-wə suffix in Proto-Uralic, the same does not hold for *-mə.
[6] Janhunen, Juha (2009). Proto-Uralic—what, where, and when? In: The Quasquicentennial of the Finno-Ugrian Society. Mémoires de la Société Finno-Ougrienne vol. 258.
Freelance reconstruction 
Intriguing.
I’d be interested to learn what you think of the Moscow School proposal that PU *ð, *ðʲ were [rʲ], [lʲ] (I forgot which is supposed to be which)? Unfortunately I don’t have a reference, all I’ve seen is secondary or tertiary literature that takes this for granted.
I attempted to reply to this in brief, but it quickly became clear that this idea will have to be covered together with a deeper look at the issue of Honti’s *ɬ *ɬʲ versus traditional *ð *ðʲ. (FWIW, it is the former that has been suggested to have been *rʲ.)
Even briefer though: I find the proposal to be poorly argued and motivated, to the point of being just about ad hoc, but bits of it are perhaps accidentally on to something regardless.
Looking forward to it, then!
The first part of my detailed reply is now up over here.
Intriguing post dealing with multiple questions. Some further thoughts on the issues of Ugric cluster simplification:
The parallelism you noted between the developments of the type *śilmä, *ćolmi on the one hand and *śodka, *śülki- on the other is intriguing. The two phenomena really seem to be related. However, one problematic word not mentioned here is PU *śilkaw ‘pole, stick’ > PKh *saaɣǝl, PMs *sïïɣlaa; wouldn’t one expect palatalization here, too?
The word for charcoal must, however, have had palatalized *d’ already in PU because of PSam *sijǝ ‘charcoal’ (cf. PU *nüdi ‘handle’ > PSam *nir).
The Ugric word for ‘ashes’ is certainly unrelated to Mordvin *kulǝŋ – neither the vowels nor the consonants match. I have another etymology for it which should be published sometime in the future. The words go regularly back to Ugric *kad'(a)ma ‘ashes’. This looks like a regular derivative of PU *kad’a- ‘leave over’. The semantic connection is rather easy to see: “that which is left over (after the fire has burnt)” = “ashes”.
One might, if this were from PU. But I suspect, for unrelated reasons, that this word may have been loaned from Indo-European only (shortly) after Ugric had split off from its western relativs. (Long story, I’ll be covering this eventually.)
Also an alternate explanation could be that the mostly regular metathesis *lɣ → *ɣl (and the loss of *ɣ from Ob-Ugric *sëëĺ, if it ever was there) is older than the assimilation *ś-l → *ś-ĺ. “Spit” seems to not have been affected by this metathesis in the first place. Or: maybe -ɣ is a suffix here? This would require a surprizingly long development from Proto-Uralic to Proto-(Ob-)Ugric, though: *śülkə → *śülɣə → *śüɣlə → *śüɣlə-kä → *śülɣä → *śüĺɣä → *süĺɣ.
…hang on, it occurs to me that assuming suffixation might also solve the previous problem I had with /ü/ turning up in Southern Mansi in this word. Suppose we reconstruct Proto-Mansi *säwĺəɣ, with later loss of the *w everywhere? The Ob-Ugric languages abound with word roots of the shape *CVɣəC, *CVwəC, *CVkəC (I have counted no less than about 80 among the about 800 common Ob-Ugric roots reconstructed by Honti), as well as *CVCəɣ or CVCək (about 100, not counting cases where a velar is suffixed only in either language), but there are barely any cases along the lines of *CVɣCəɣ to be found. (There is Ms *mäɣtək ~ East Kh *mööɣtək ~ West Kh *meeɣtəŋ “a kind of fish”.)
EDIT: A velar C₂ being (eventually) lost when a velar C₄ was suffix’d could also explain the sporadic cases where one language has a root of shape *CVɣəC, the other *CVCəɣ.
But, this idea requires assuming that metathesis is later than palatality assimilation. Hmmm.
I don’t see this as a decisive point, since the same assimilation seems to have affected Samoyedic as well! Most of the affected words are not attested in Smy, but there is *säjəm “heart”.
EDIT 2: of course, also *sə̈jmä “eye”. While *l → *j occurs in several positions in Samoyedic, *l seems to remain in *kalə-ma → *kålmɜ “corpse”.
Sounds entirely plausible.