Currently I am making my way through a fascinating and peculiar book: Hartmut Katz’s posthumously released Studien zu den älteren indoiranischen Lehnwörtern in den uralischen Sprachen (Heidelberg: Universitätsverlag C. Winter, 2003).
Fascinating, in that the book’s ~700 loan etymologies, some of them providing novel and quite believable solutions to various etymological problems, are obviously much food for thought.
Yet also peculiar, in that Katz’s goal seems to have been to ascribe Indo-Iranian origin to as many words as remotely possible; while operating with a bizarrely dated framework of Uralic history, cut off from modern research. His reconstructed Proto-Uralic appears to be based mainly on the work of Wolfgang Steinitz around the 1940s, and views newer than the 60s are just about not even cited.
It would not be too hard to spend a while discussing the shortcomings of his reconstruction scheme in detail, but I suspect Katz’s ghost will not be appearing to defend, amend or recant his views. So for now, to simply note some strange ideas included, he posits e.g.
- Mobile stress for old stages of Uralic, on the leftmost non-reduced vowel in the word.
— To be fair, this is a system that can be indeed found in Uralic languages of the Volga-Kama area, and some hints of it can be seen in Hungarian and Mansi as well. But the complete absense of evidence for it in Samic and Finnic on one hand, and most of Samoyedic on the other, definitely suggests a secondary areal innovation. Katz fails to match this stress system with Indo-Iranian either.
- A relatively extended system of “ablaut” that seems to be essentially projected backwards from Khanty (whose vowel alternations are usually thought to be umlaut anyway), and then employed to explain some exceptions that fail to fit his vocalism framework.
- A reconstruction of Proto-Permic that contrasts eight labial mid vowels: short and long *ɔ, *o, *ɔ̇ [ɞ], *ȯ [ɵ] — yet has no plain *a of any length.
- Four different sibilant series, with the new fourth one generated by splitting standard PU *ś in three. Both splits are established on the basis of evidence from a single language (Mator for *ć versus *ś; Mansi for *ś versus *š´). He still seems content to consider the traditionally similarly problematic retroflexion contrast in Khanty (*n *l versus *ɳ *ɭ) as entirely “affective”, however.
Some of the alleged sound substitutions in loanwords are puzzling as well. One claim that I am definitely not buying offhand is that PIE labiovelars could have been substituted by plain labials — sometimes, even, with different ways in the same word. For example, Katz attempts to derive both Finnic *pöörä ‘wheel’ and *käkrä ‘curved’ from pre-II *kʷekʷra- ‘wheel’. The latter strikes me as at least possible (there is moreover *kekri ‘year, yearly feast’ which has been explained from the same source as well); the former as requiring rather too many phonological assumptions, despite the seemingly straightforward semantics. Especially since the usual PU reconstruction: *peŋər- ‘to turn, to rotate’, establishable by comparison with the Ob-Ugric cognates, seems unproblematic.
At other times incredibly archaic PIE sound values are claimed preserved in Uralic. E.g. *e reflected as a front vowel even when adjacent to *h₂; *ḱ reflected as *k. It is unclear to me on what grounds these could be analyzed as specifically Indo-Iranian loans. If anything, such etymologies (in case they’re not illusory) might rather speak for loanword transmission through Late Proto-Indo-European, before diversification; from the adjacent IE branches, say Baltic or Tocharian; or some lost Indo-European languages, perhaps intermediate between Balto-Slavic and Indo-Iranian.
—Or even languages from different families entirely. I suspect taking e.g. Turkic better into account might help for resolving some oddities. There is, for example, some evidence for a correspondence II *bʰ : Uralic *m, which Katz claims would have been a phonetic accommodation to retain the voicing of the original consonant (rather a priori suspicious, since there was no contrastive voicing in early Uralic). At least one of the cases might allow for a different explanation: this is the Ob-Ugric word he reconstruct as *māŋkɜ, meaning ‘hammer’. He compares here Indo-Aryan *bʰangá- ‘to break’. But there is also Turkic *böŋk- ‘to kick, to buck’, with reflexes such Uyghur /möŋkü/-, Tuva /mög-/, with the assimilation *b-N > *m-N. This seems like at least a plausible intermediate for reaching the Ob-Ugric words (though accounting for the development of the semantics and vocalism would take some extra work).
Another problem should be a warning sign even for modern-day researchers. Katz’s scheme of separating the Uralic etymological material into about four stages of development — Uralic; Finno-Ugric; Finno-Permic vs. Ugric; Permic vs. Ob-Ugric — appears to not produce any particular benefits. Under his framework of historical phonology, there are essentially no differences to be found between the three early stages, only a number of more or less trivial rewriting rules such as “Finno-Ugric” *ə̑ > “Finno-Permic” *ă. The vast majority of the proposed loanwords fail to show support for the dating in their distribution either. “Proto-Uralic” words frequently turn up in Samoyedic and nowhere else (a few even, within Samoyedic, in Selkup only!); “Finno-Ugric” ones similarly in only a single sub-branch such as Mari, Finnic or Hungarian; etc. Chronological paradoxes arise too, when e.g. some words are posited to have been loaned before the Indo-Iranian shift *l > *r, yet during a later era such as “Ugric”, while others are posited to have been loaned after the change, yet during an earlier era such as “Finno-Ugric”. In other words, the traditional taxonomy of Uralic is here treated as a fact that has been given ex cathedra, and not critically engaged at all.
There is still one particular insight that I am happy to see appearing. Quite a few words are present in the data seemingly as etymological doublets (often even triplets, sometimes as much as sextuplets!) across the different Uralic languages. When this happens, Katz does not insist on shoving them under a single Uralic proto-form and deriving all the forms as “variants” or by “sporadic sound changes” (though these mechanisms are still a minor part of his toolbox). Instead he is on board with concluding that the one and the same word can have been loaned several times — as an areal rather than a genetic innovation — and perhaps in different shapes in different proto-dialects. If taken to its full conclusion (comparative reconstruction cannot be based on loanwords), I believe this method seems even likely to be able to resolve various lingering problems of reconstruction. Not that that’s quite happening yet in the book. I might cover some exemplary cases in detail in future posts, though.
At any rate, for now this remains one messy bunch of comparisons. Yet clearly on a valid topic, deserving of critical treatment. Perhaps one day we will see a more fruitful analysis of this corpus.