On the epistemology of sound change, part 1

Continuing from the last post, and toning the meta-ness of the discussion down just a little…
What does it, at the level of everyday research, mean for me to request “justification on the basis of more elementary phenomena” for the concepts of historical linguistics? Say, from the viewpoint of sound change?

The foundations of the concept

The concept of sound change is already implicit in the concept of cognate words. If we assert that a word such as Hungarian ősz ‘autumn’ is cognate to Finnish syksy ‘id.’ (I will not go into unpacking what it means for a “language” to have a “word” that is expressible by a string of letters, although these are good questions to ask as well) — then this means that at one time, a common proto-form of the words existed. The self-contained apparatus of historical linguistics can also produce a graphical representation of this; according to my chosen system of Proto-Uralic reconstruction, it will be *sükśə or perhaps *sükəś. We may also propose a phonetical value for this. Indeed, most linguistic transcription systems already do this implicitly. Despite occasional use of cover symbols for difficult-to-reconstruct segments such as Proto-Uralic *d₂, or in words whose precise phonemic content cannot be resolved from the available evidence, I cannot say I have ever seen a purely abstract presentation of a proto-language.

Usually, the suggested pronunciation of the word will end up different from the real attested words on whose basis it was posited, in at least some respects. This thus requires that changes in pronunciation, i.e. sound changes, have occurred at some point in the evolution of Hungarian, Finnish, and any other Uralic languages. In this particular case, we have the loss of palatalization, *ś > *s, in both Finnish and Hungarian; the loss of *k, plain *s, and the second-syllable vowel entirely in Hungarian; the lowering of *ü to *ö in Hungarian; and its acquisition of length. It would be possible to shuffle some of these changes around (e.g. perhaps it is not Hungarian that has lost a /k/, but Finnish that has gained one? perhaps originally a third type of sound yet occurred here?), but the fact that ősz and syksy are not identical in their pronunciation will remain.

In clearer words yet: that sound change somehow, for some reason happens is clear already from the idea that etymologies exist; that non-identical words can have a common origin. Interestingly, note also that relationships existing between entire languages is not a required assumption at this point.

By the way, note that I do not claim this to be the actual history of how the concept of sound change was developed (that story is much more complex yet). This is only an observation on the internal logical structure of the modern-day theory of historical linguistics.

There is thus “downwards inference” involved here. Instead of tinkering with empirical research on articulation & such and discovering that a certain series of events can add up to the large-scale phenomenon of sound change, we have looked at higher-level data yet and found patterns that can be effectively explained by assuming the existence of sound change — despite not yet knowing the first thing about how it works. As a scientific theory, this is adequate insofar as it can still provide predictions, but naturally it leaves us asking: what exactly are these sound changes? Can we actually see one happening somewhere? Could a detailed understanding of them enhance our understanding of etymology as well?

Motivating the instances

There do exist disciplines like phonetics and sociolinguistics that are directly tackling the questions of the evolution of language on the scale of years, weeks, milliseconds rather than generations. However, the theory of sound change can be further sharpened already by more careful investigation of etymology.

There is of course also the tiny snag that detailed sociolinguistic data or phonetic records do not exist for the pre-modern histories of languages (let alone the entirety of prehistory). So we are mostly unable to directly observe sound changes in their full historical context, and indirect inference from etymological data remains our almost sole option for finding out about them. This means that care is required to not drift off to pure speculation.

What, then, is sufficient evidence for assuming a specific sound change to have occurred?

The “naïve method“, I could call it, is to simply indiscriminately collect sound correspondences that “seem to” exist between some given language varieties, claim as cognate any pairs of words that can be linked by some application of these, and then present some sound changes that can account for the correspondences. This has been, and continues to be, used as the usual first step in investigating the sound correspondences within a group of “obviously related” language varieties, such as dialects of a single Language. (We still do not need to take a stance on whether “non-obvious relationships” can exist between languages.) Or, if we’re investigating loanword etymology, we might look at any pair of languages we believe to have been once somehow involved with each other.

Back to the previous Finnish vs. Hungarian example, e.g. two original s-ish consonants can be assumed right off the bat, that we could preliminarily call *s₁: defined as becoming /s/ in Finnish, vs. zero in Hungarian; and *s₂: defined as becoming /s/ in Finnish and Hungarian both.

I will skip for now the wider problem of how to determine what segment exactly corresponds to what. For Uralic languages this is known to be, 99% of the time, a simple task: an initial stressed syllable corresponds to an initial stressed syllable, consonants correspond to consonants, vowels correspond to vowels.

The naïve method is much too powerful though, and let to run on its own, will inevitably lead to an an unfalsifiable system where anything can be related to anything else. This is because under it, word-level and and sound-level relatedness are translative. If we claim that sz in Hungarian ősz corresponds to the 2nd s in Finnish syksy, then it follows that the words are related in general, and that also Hungarian ő corresponds to Finnish y. If this is taken as an excuse to now relate any word that has y in Finnish to any word that has ő in Hungarian, and so on forth — this allows eventually racking up a correspondence library that allows relating everything to everything else.

There are, in principle, two ways of avoiding the problem. The first is a purely statistical approach: if two words don’t share at least some proportion X of known sound correspondences, we do not accept the comparison and do not accept any new sound correspondences that it would imply. This algorithm requires a “seed” of correspondences though — if you sic it on languages of which you know nothing, it will detect no related words, what with no correspondences being accepted yet. A “seed” must be instead generated by some other method. Likely ideas for this might be:

  • Having to build up a set of word comparisons that is closed with respect to sound correspondences, and where every correspondence occurs at least n times. An n = 2 example might be Finnish kala, pala, kesä, pesä ~ Northern Sami guolli, buolli, geassi, beassi (‘fish’, ‘bit’, ‘summer’, ‘nest’).
  • A set of correspondences that occur highly often and/or are between identical segments.

These seed methods, I believe, probably won’t manage to uncover everything that can be uncovered all by themselves, but let’s leave a closer analysis of their pros and cons for some other time.

A more interesting point is that these methods, phrased solely in terms of sound correspondences, are mainly focused on binary comparison. Correspondence-counting of any kind however runs into some rather nasty mathematical problems when a larger number of language varieties is involved. Consider for example: should a correspondence set t ~ t ~ t ~ t ~ θ ~ t be counted as a completely different entity from a correspondence set t ~ t ~ t ~ t ~ t ~ t?

  • If yes, we hit what is called the curse of dimensionality. Say we have two languages with 20 consonants each: there are then 20² = 400 possible correspondences between these, and we can well expect a decent bundle of etymological data to not only highlight which of these correspondences are highly recurring, but also which are noticably rare or absent. But if we rather have as few as six languages, the count of possible correspondence sets becomes 20⁶ = 64,000,000. The lexical stock of even the best-documented languages only reaches a fraction of this, and a typical etymological data set is unlikely to exceed a couple of thousand words. Given a space of millions of possible correspondences, most data points will perhaps cluster at some stable points, or in their vicinity. Any correspondence sets that turns up outside of these islands (say, an apparent correspondence h ~ t ~ t ~ d ~ s ~ z) will be difficult to assess. Also, by far most possible correspondence sets will be entirely absent, and we’ll have no chances of telling if the absense of one particular correspondence carries any statistical significance.
  • And if no — if we treat comparison sets as built from pairwise correspondences — then the transitivity problem pops up again. If we find a correspondence t ~ t ~ d ~ d ~ d ~ d, and we already know the existence of a correspondence t ~ t between the first two languages, can we really count it as evidence for the unity of this larger correspondence set in general?
  • And what of incomplete correspondence sets? Suppose we have p ~ b between languages 1 and 2; b ~ v between languages 2 and 3; p ~ v between languages 1 and 3. Can we really take this as sufficient evidence to unite them to a single correspondence set p ~ b ~ v? What if a correspondence p ~ p between 1 and 3 exists as well?

Instead of getting stuck on fine-tuning these problems, it’s however possible to change gears. There is a second, fundamentally different method possible as well: the chronological approach, whose nature I will be elaborating in the next post of this series.

Tagged with: , , ,
Posted in Methodology

Indo-Iranisms galore?

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.

Tagged with: , , , ,
Posted in Commentary, Etymology

The rooting of historical linguistics

Most of the harder problems in the methodology of historical linguistics seem to come from it being a fairly “high-order” discipline, and a relatively isolated one at that.

To an extent, this true of all humanities. With the levels of computational power currently available to us, it’s not possible to start with a couple of known physical laws and derive exact predictions about human behavior from them. The best we can do along these lines is to establish boundary conditions. And of course, most of these are sufficiently obvious from our daily experience as humans that they sound more banal than profound when spelled out: e.g. language families usually have a distribution limited to the surface of the planet, and they fail to extend up to the stratosphere, or down to the oceanic crust. :ɪ

But the historical angle complicates things. Most of the historical sciences rely heavily on evidence preserved from the past: history itself is based on written sources, and the “auxiliary historical sciences” such as archaeology on other objects preserved from the past.

And yes, historical linguistics also builds on preserved evidence from the past, mainly via philology and epigraphy. But this has only been a small initial inspiration. Most of our historical insights are instead derived from from the observation and analysis of attested modern languages, and the application of a general theory of linguistic evolution. This model is, I think, quite alien to all other humanities. (Even plenty of non-historical lines of humanities research seem to remain stuck in a pre-scientific “there are no theories, only paradigms of discourse” mire.) In this sense historical linguistics has much more in common with evolutionary biology, although I suspect that also that discipline would not be doing as well as it is without the more direct evidence from an extensive fossil record. [1]

The inevitable implication is that nothing in historical linguistics can be understood without a good grasp of the underlying theory. And yet, it seems to me that many of its premises have not often been even stated aloud. No dout this is due to how the theoretical foundation seems to have been developed on a need-to-know basis by its users, as the discipline has expanded, not by any separate class of theoreticians. Yes, starting from the Neogrammarians, many of the surface phenomena have been described, from old’uns like “regularity of sound laws” to innumerable newer achievements like “typology of semantic change in body part terminology”… but the nuts and bolts of it, that really “root” historical linguistics to its sociolinguistic foundations, not so much. There has been so much work in cataloguing the “whats” that we have had not much success yet in uncovering the “whys”.

I am not sure if my term “root” is readily understandable, or if there might be a better term available. It seems likely that this could be confused with a discipline’s internal history, at least. Which is not what I mean: I refer here to by how various sciences can be ordered in how far removed they are from the basic laws of the universe. The typical example being how all biological processes can be broken down to individual biochemical processes; all biochemical processes to chemical ones; all chemical processes to particle physical processes. The reason that biology looks very different from chemistry, or from particle physics, is that studying the behavior of macroscopic masses of particles requires very different methods from studying 10 or even 1000 of them. A phenomenon such as embryonic development could in principle be modelled in terms of individual protons and electrons, but this would require enormous amounts of efforts wasted on reiterating problems like “how does a water molecule hold together” or “what happens to a protein when it encounters a water molecule”, that have already been solved to sufficient precision for us to instead model an embryo as being built from cells that are built from cellular organelles that are built from macromolecules. A biologist — or a geologist, or a cosmologist — is not interested in the whereabouts of individual particles, but rather in their patterns of distribution at a specific scale in space and time.

The same exact principle holds for the humanities. Say, all psychology is at a certain fundamental level about neurons; but in analysing the overall behavior of the brain, built from a hundred billion neurons, the beliefs, feelings, etc. that they encode can (and must) be treated as entities in their own sake. And similarly, while the speech of one human can be studied by phonetics, neurolinguistics, and similar disciplines, it again takes different tools to study the speech of a hundred million humans sprinkled across five thousand years. We need concepts such as “isoglosses” and “etymologies” that exist only as generalizations about the idiolects of individual speakers.

Our tools, however, do not seem to decompose easily into insights about smaller and smaller groups. How exactly does sociolinguistic variation in speech end up producing clean and neat sound laws, or patterns of loanword dispersal, or language areas sharing grammatical features? I do not think we have much more than loose guesses about the workings of these processes, so far.

This type of disconnect is, of course, quite common at the biology/humanities interface, and can be sometimes found elsewhere as well (e.g. in the absense of a working theory of quantum gravity). But to see it within a single discipline — linguistics — seems to me like a situation that ought to be resolvable.

This also means that historical linguistics knowledge rests, to an extent, on questionable ground. If we do not name our implicit starting assumptions, and end up making little effort to justify them on the basis of the more elementary phenomena they emerge from, is there not a risk that our edifice of knowledge stands askew, and ends up being an excercise in the construction of an essentially abstract theory, rather than a real description of the past?

Some philosophers would at this point certainly retort that all historical inquiry, being both unverifiable and unfalsifiable in the absense of a time machine, does not exist for the purpose of creating a real description of the past, but to create compelling stories about it. OK, I say, but some of us happen to consider truth an essential component of what makes a story “compelling”. Moreover… any model of the past will also make predictions about some parts of the present that we have not examined yet, which grants all historic theories a limited degree of falsifiability.

I do not claim to have a dossier of answers to issues of this sort prepared. Perhaps one or two sketches of solutions. But, of course, questions have to be asked before they can be even begun to be answered.

[1] Arguably though one could claim that the majority of our planet’s biodiversity exists at the microscopic level, and that most of biologial history must be thus similarly approached via comparative reconstruction. But in my understanding this is a relatively new approach in evolutionary biology; while historical linguistics dove headfirst into reconstruction already back in the 19th century.

Tagged with: ,
Posted in Methodology

Some sunny words

A recent blog post from Christopher Culver brings to my attention an apparent family of Turkic word roots showing irregular variation in form: *künäš ~ *qujaš ‘sun, day, heat’. Aside from the alternation *n ~ *j (for which *ń seems to be a standard explanation), these seem to make up a neat pair of front/back variants.

I am wondering however if this relationship might be illusory, and if there might be an old Uralic loanword in Turkic involved here instead. There are a few Uralic word roots (themselves probably in some sort of an obscured correlative relationship) that seem quite relevant here:

  • *kaja ‘sun, to shine’ (> Finnic *kajasta- ‘to dawn, to shine’, Lule Sami guojijdit ‘to rise (of sun or moon)’, Samoyedic *kåjå ‘sun’, etc.)
  • *kojə ‘dawn’ (> Finnic *koi ‘dawn’, Hungarian hajnal ‘dawn’, Mansi *kuj ‘dawn’, etc.)

Of particular interest is the Hungarian word, which seems to show the exact same “suffixal” elements as Turkic. This even has a formal equivalent in Khanty: *kuuńəɬ´ ‘dawn’ (apparently showing a change *jn > *ń, in neat parallel to the change *jt > *ć that was proposed by Aikio recently [1]), coming closer yet to the Proto-Turkic form.

It’s hard to say though what the dangling element -nal is here. It’s neither an independent word root on its own, nor a regular derivational affix. If I had to speculate, a compound *kojə-n‿alŋV > *kojnal- ‘beginning of sun’ could be assembled… but this seems a bit contrived semantically. Also I am not convinced if Khanty *aaLəŋ ~ Mansi *aaɣəl ‘beginning, end, point’ is an inherited root at all. [2]

And while phonetically the Khanty form in particular seems like a prime loan original, the semantics are a bit off. Is the meaning ‘dawn’ in Hungarian and Khanty perhaps secondary, from earlier ‘sun’ or the like? Or was there instead a shift ‘dawn’ > ‘sun’ in some transmission language along the way?

Some Turkologists, I’m sure, could also see it as an obstacle that this etymology seemingly requires adhering to sigmatism (reconstructing a Proto-Turkic lateral” *l₂ that later shifts to *š in Common Turkic) over lambdaism (reconstructing PT *š that shifts later to *l in Oghur Turkic). Now, yes, from what evidence I’ve seen, I lean on the view that sigmatism is the better solution [3]… But it is, however, not an entirely inescapable assuption here. Say we instead assumed that early Oghur maintaines *l₂ for some time apart from *l (perhaps indeed as a lateral fricative [ɬ]? [4]) Then we could posit an etymological sound substitution to have occurred during propagation to the other Turkic languages: Khanty *kuuńəɬ´ → Oghur #qujal₂ → Common Turkic *qujaš.

Independent loaning to different Turkic varieties might also be chronologically preferrable to assuming loaning already to unitary Proto-Turkic. Christopher notes that *qujaš seems to have a kind of northerly-leaning distribution across the Turkic languages… not bad news for an attempted Uralic loan etymology, I’m sure.

[1] Aikio, Ante (2014): Studies in Uralic etymology II: Finnic etymologies. Linguistica Uralica 50:1.
[2] There is, yes, a rather similar word root in Finnic: *alka- ‘to begin’ — but this does not quite correspond regularly to the Ob-Ugric words, esp. on account of the discrepancy between *ŋ and *k. The vowel correspondence Kh *aa ~ Ms *aa is not typical of inherited Uralic vocabulary either.
[3] But note that this does not compel me to take a stance on the similar rhotacism/zetacism debate, nor to consider *l₂ of “Altaic” inheritance.
[4] Which even brings to mind the East Uralic shift *š > *ɬ, rather similar to the shift *š > *l posited by the lambdaist side of the Turkic debate.

Tagged with: , , , ,
Posted in Etymology

Similar Place Avoidance in language history

An interesting paper I’ve found a couple days ago: Pozdniakov, Konstantin & Segerer, Guillaume (2007). Similar Place Avoidance: A Statistical Universal. In: Linguistic Typology 11:2.

The main thesis is relatively simple: most languages of the world disfavor word roots where the word-initial and word-medial consonants have the same place of articulation; and, more generally, word roots combining two peripheral (labial, velar) or two “central” (dental, alveolar, postalveolar, retroflex, palatal) [1] consonants.

I have also independently discovered this principle some time ago in my exploration of statistical properties of phonotactics in the Uralic languages. Unlike P&S, though, my first reaction was not to assume status as a defining characteristic of Uralic in general. Certainly its occurrence in well-separated branches of the family seems to require its occurrence in Proto-Uralic as well… but who knows how much further back does it go? I do not recall seeing very many word roots shaped anything like √kag- or √bomp- in almost any Eurasian language at all, really. I have had an impression they’d be slightly more common in some Niger-Congo languages — but apparently not. (Seeing what the results are for Japanese might be also interesting; the language seems to be quite rife with words like tatami, tsunami, kami, fuku, fugu. But I am not sure if Internet Japanese™ constitutes a representative sample.)

Some further observations on the topic:

The maintenance of SPA

A question that I did not see covered in the paper is that the maintenance of SPA in languages requires a degree of diachronic stability of consonant POA classes. Now indeed, as a first approximation, while fluctuations between different types of e.g. coronals (ts > tθ > θ > ð > d > l > …) or velars (k > x > ɣ > g >…) are commonplace sound changes, it’s much rarer to see consonant evolutions such as *p >> *d or *d >> *x.

But the boundaries are still not impermeable. Quite a few relatively general sound changes are known across the world [2] that convert consonants from peripherals to centrals, or vice versa:

  • Labial > palatal: e.g. *w > *j in Hebrew
  • Coronal > labial: e.g. *θ, *ð > /f/, /v/ in Latin and the other Italic languages; similarly *t > *θ > /f/ in Rotuman
  • Coronal > velar (or uvular): e.g. *š > *x in Finnic, Spanish, Pashto…; *t > *k in Oceanic languages such as Samoan, Hawai’ian; *r > ʀ/ʁ in continental Western Europe; *ɫ >ʁ in Armenian
  • Velar > palatal: *k, *g > *c, *ɟ > tɕ, dʑ — a frequent change: Satemic IE, Romance, etc.

This raises the question of how the strength of SPA evolves in languages. Changes of the above sort, applied to a language that follows SPA, will necessarily decrease its SPA-compliance. If *š frequently co-occurs with velars, and rarely co-occurs with coronals, then a change *š > *x will introduce a larger number of velar-velar roots than velar-coronal roots. It follows that there must also exist some mechanisms that increase the SPA-compliance of a language.

A naive assumption that P&S summarily dispatch would be sound changes running in the opposite direction: place dissimilation to re-establish SPA, a la (? *kaša >) *kaxa > *taxa. Yet this is not a commonly attested type of change at all (the only example I can think of of is *t > *k only when a 2nd *t follows; attested IIRC from one of the Oceanic languages [3]), and it clearly cannot be a relevant factor.

My hypothesis is that lexical loss is not random. Suppose a language had two synonyms /maba/ and /suba/ for expressing a given concept; then over time, as the language splits into descendants, SPA-violating /maba/ would be more likely to be lost than the SPA-compliant /suba/. A motivation for this could be that SPA-violating roots are generally found to be more “childish” or “non-serious” in sound, and that they’d be more likely to go “out of fashion”. (Pop quiz: which of the sets {boob, dude, google}; {duty, goop, boogie}; {duke, good, butte} do you find the funniest-sounding?)

This is, in principle, a testable proposition. Take for example the interdental > labial shift in Latin. I would predict that PIE roots that display the change *dʰ >> /f/ ~ /b/ are more likely to be lost or marginalized in Latin (both in early Latin and later on in Vulgar Latin) when there is an original labial or velar consonant in the root as well. Or, in the other direction: I would predict to be able to trace the ancestry of words like duke, on average, a longer way back than that of words like dude.

Affricate co-occurrence

P&S further divide the SPA principle into a couple statements of different strength. The “general” version is that peripherals avoid any other peripherals, and centrals any other centrals; while the “strict” version is the rarity of, especially, word roots with two consonants of the same exact POA. They discover, however, one major divergence from even the last: the Bantu languages apparently feature a high number of word roots with two palatal consonants. I’d guess this represents an assimilation development of some sort. Perhaps the palatal series represents the merger of former palatalized alveolar and palatalized velar series? This relatively frequent development would easily leverage the apparently universal abundance of TK and KT roots to produce instead an abundance of CC roots.

— In Uralic we find no evidence for an especially strong co-occurrence of palatals. However, the postalveolar affricate *č has a strong tendency to “repeat”. There is a remarkable number of  old Uralic roots (some of these more, some less secure) such as:

  • *čača- ‘to be born’
  • *čača- ‘to walk’
  • #čEnčä ‘back’ ~ ‘tail’?
  • *čänčä ‘goose, duck’ (from Baltic *džans- < PIE *ǵʰans-)
  • *čëčə ‘duck’ (perhaps also from the above PIE root somehow)
  • #čečə(kä) ‘moment’
  • #či(n)čä ‘little bird’
  • *čoča- ‘to sweep’
  • *čo(n)čə ‘netstring’
  • *čučkə ‘block of wood’

Perhaps a partial explanation would be some sort of consonant assimilation phenomena. At least the 3rd word seems to have involved an assimilation *č-s > *č-č. And a couple of these roots are reflected in Finnic and Samic as if coming from original *ć-č  — yet not all, as shown by Finnic *häntä ‘tail’, *hetki ‘moment’, Samic *cōccë ‘netstring’ (provided the Uralic etymologies for these are valid: they all involve some irregularities). And maybe the “dissimilating” roots should hence be similarly reconstructed as dissimilar to begin with.

We could also wonder if this should be taken as evidence for an origin of some cases of *č via palatalization from earlier velars.

…and other reduplications

P&S also find, though, that at least some languages can have a tendency to favor “reduplicated” roots (their example is Wolof), with the exact same consonant in the root-initial and root-medial positions. Obviously in a language with several consonants per POA, this effect will be overshadowed by the numerous other combinations possible — so /b-b/ could end up relatively frequent, but cases like /b-m/, /b-v/, /b-f/, /b-ɓ/ etc. will still remain rare.

From my initial observations, though, this does not keep up in Uralic, where classes like “labials” are frequently limited to only a single obstruent *p, the nasal *m and the glide *w or *v. The Proto-Sami lexicon, [4] for example, contains less than two dozen PP roots, and most of them are either of the shape *m-v, *p-v; *v-m, *v-p; or *v-v. There is only one root of the shape *p-m; none of *p-p, *m-m or *m-p.

The occurring cases incidentally can be shown to be in large part secondary innovations. E.g. the 2nd class contains *vāpsē ‘blade of mitten'; *vipsë ‘skein'; *vēpsēs ‘wasp'; *vōmë ‘width'; *vōmē ‘woods'; *vōmā- ‘to notice'; *vōmtë *body cavity'; *vōmtē- ‘to sell'; *vōpējē ‘narrow bay'; *vōpērēs ‘three-year-old reindeer bull'; *vōppë ‘father-in-law'; *vōppō- ‘to pluck'; *vōpsë ‘mesh in a fishtrap'; *vōptë ‘hair. Most of the roots here seem to have involved the PS development *a-, *o- >> *vō-. All the rest involve the cluster *-ps-, though I’m not sure what to make of that fact.

Cluster complications

Another question the paper does not address is how should one analyze heterorganic consonant clusters. Most languages of the world prefer a simple CVCV syllable shape over CVCCV. The latter type is regardless fairly popular in some languages. E.g. my index of the Proto-Sami lexicon contains about 920 roots with clusters, about 600 without. So are clusters to be counted as “medial consonant preceded by a coda”, or as “medial consonant followed by another medial consonant”? Is a word such as PS *tolkē ‘feather’ more or less SPA-compliant than PS *kōlkë ‘hair’? The second does have a neat alternating POA structure; but both the syllable onsets are velars. Which of these is more relevant?

From a preliminary look, it stands out that the relative frequencies of 1st members of clusters resemble quite closely the relative frequencies of single medial consonants — while the relative frequencies of 2nd members of clusters closer resemble the relative frequencies of onset consonants. This would seem to suggest that we should indeed be comparing the first two consonants. But the details could fare differently.

Let’s take a sneak peek at velar/velar combinations for example:

  • *kVkV: severely underrepresented (predicted 18, attested 4)
  • *kVŋV: severely underrepresented (predicted 7, attested 1)
  • *kVkCV: underrepresented (predicted 19, attested 12)
  • *kVŋCV: severely underrepresented (predicted 6, attested 2)
  • *kVCkV: underrepresented (predicted 43, attested 31)
  • *kVCŋV: overrepresented (predicted 3, attested 5)

It seems to be here indeed the case that at least word roots like *kōsŋë- ‘to touch’ are patterning as POA-alternating (= not in violation of SPA). But the underrepresentation of *kVCkV does not fit this hypothesis. Though… the data could also be confounded by one of the most frequent -Ck- clusters being the homorganic *ŋk. I’d need to crunch more numbers here to say for sure.

There’s clearly much to be made of this topic; I am only scratching the surface so far.

[1] They actually use the term “medial”, but I will not, as this seems likely to be confused with “word-medial”.
[2] That is, discounting cases of local assimilation such as np > mp, mt > nt.
[3] I recall Robert Blust covering this topic in his paper __. I seem to have displaced my copy of it, though.
[4] Again, as per Juhani Lehtiranta’s Yhteissaamelainen sanasto (1989/2001).

Tagged with: , , , , , ,
Posted in Commentary, Uncategorized

Consonant clusters in Khanty

My previous example of phonotactic combination analysis was on data that was, despite a few kinks, still largely homogenous. But to showcase how it’s important to have a decent basic hypothesis before going into more fine-grained analysis, here’s a look at a rather different dataset. These are the medial consonants and consonant clusters from the inherited Proto-Khanty lexicon, again per Honti’s data (words with cognates elsewhere in Uralic but absent from Mansi are not included).

Some notes about notation etc. though, before I go on.

  • 1st medial consonants (“C₂”) are listed down. Possible 2nd consonants of a consonant cluster (“C₃”) are listed across.
  • I have analyzed PKh *ə as an epenthetic, non-phonemic segment that is inserted in “difficult” consonant clusters in, roughly speaking, stem-final position. E.g. *peLəm ‘lip’ = underlyingly /peLm/. Without this analysis I would be almost comically short on data.
  • *g and *x mark two segments that only contrast in Western Khanty in back-vocalic roots (as /w/ versus /χ/). Honti conflates both as *ɣ. The contrast is not (directly?) recoverable in front-vocalic roots, nor in words that have been retained only in Eastern Khanty, and seems to have been absent from the C₃ position. I have counted ambiguous cases under *g¹.
  • *L and *Ľ are cover symbols for laterals. PKh had a contrast between a fricative *ɬ and an approximant *l, and might have had even a similar contrast among the palatal laterals, but this is not recoverable in the medial position. (By contrast, the retroflex lateral *ɭ was quite certainly an approximant.)

But without further delay, here is what things look like in this part of the word root — sorted by frequency, again:

Proto-Khanty consonant clusters by consonant frequency

Already one look at this table should tell us though that it would be pointless to compare it against what an assumption of random distribution would predict. Not only are there way too many gaps, there are also several strong correlations apparent. Take for example C₂ *ń and C₃ *ć, which are both found almost solely in the cluster *ńć.

So the first step ought to be determining some basic background rules of phonotactics first. Here is the same data, now sorted by place of articulation instead:

Proto-Khanty consonant clusters by place of articulation

Several qualitative patterns are clear by now.

  • Almost all of the action goes on in the “edge” cells — those combining peripheral (bilabial/velar) and coronal (dental/alveolar/retroflex/palatal) consonants.
  • Nasal + stop/affricate clusters (highlighted in pink) are easily the most frequent type of homorganic clusters. For bilabials, palatals and velars they are the only attested cases.
  • There is a degree of coronal harmony: dentals/alveolars, retroflexes, and palatals do not combine with one another. [1] For the sibilants, nasals and laterals, this is exceptionless. The rhotic *r and the semivowel *j tolerate some exceptions, perhaps due to how the two lack counterparts at other POAs. One case with *-ćt- is attested, namely *kaćtə- ‘to hit’ — and in Northern Khanty only, actually. This is also one of the clusters that’s demonstrably secondary, as comparison to Mansi *këëćk- indicates that the word is to be segmented as *kać-tə-. Perhaps we can assume that in Proto-Khanty, this cluster still remained impossible.
  • Geminates are uniformly forbidden.

More detailed frequency analysis should probably focus just on the areas that show no obvious restrictions of this kind. And now we can easily pick out a subset of data suited for this:

Coronal + peripheral clusters in Proto-Khanty

Peripheral + coronal clusters in Proto-Khanty

The data’s still a bit scarce, but here the distribution’s at least more randomized. And hence signs of various “minor” historical developments are now able to better stand out. Plus: note that despite my presentation, this is not really two separate datasets — it’s a single, three-dimensional dataset, with cluster order as the 3rd dimension. We can for example note the disproportionally high count of *-x(ə)L- compared to a disproportionally low count of *-L(ə)g-, almost certainly an indication of the regular metathesis of PU *lk and *sk in Khanty.

A full analysis would again be much more work than I am going to just blog out on my free time, though. I have no dout that this general type of methodology, applied to any one given language, could produce a small monograph’s worth of results…

[1] A result very similar to this has been noted already by Eugene Helimski in 2002: an incompatibility of the dentals *n *t vs. retroflexes *ɳ *č in word-initial vs. word-medial position. See: “Eine Regel der Konsonantenkompatibilität im Ostjakischen”, in Veröffentlichungen der Societas Uralo-Altaica 57.
It is obvious that there were no restrictions on initial palatals though, as shown by e.g. *ńoL ‘nose’, *ńoLt- ‘to knead’, *ńeeL- ‘to swallow’, *ńuuɭəm ‘wound’, *ńaLkïï ‘Siberian fir’, *ńaaL ‘arrow’, *ńeLää ‘four’…

Tagged with: , , , , ,
Posted in Methodology

Phonotactics vs. protolanguages

Phonotactic analysis is probably one of the most straightforward tools for statistical etymology. There are others too — but this is an analysis method that will easily bring up a wealth of data that has no real synchronic motivation (arbitraryness of the sign, once again) yet can be assumed to reflect all sorts of historical processes of language development. Usually though in more or less fossilized form, perhaps even quite deeply so.

However, when the object of the analysis is a reconstructed protolanguage, also another option becomes available. This is to take significant quirks as instead suggesting points on which the reconstruction itself could be improved. A reconstruction is not primary data! It is allowed to make argued-for adjustments in just what the reconstruction is in the first place. (Alas, not realizing this is a somewhat common failure mode in studies mixing synchronic analysis methods with reconstructed data.)

For an example of this approach in action, here is a sneak peek at one dataset I am massaging:

OU stats 1

This table shows the co-occurrence of initial consonants and following vowels in the common Ob-Ugric lexicon, as reconstructed by Honti (1982). Since this is for the sake of an example, at this point only some small adjustments in the reconstruction have been added, nothing major. The various non-integer values are due to me splitting most reconstructions that show uncertainty in their reconstruction: e.g. the root listed as *keej-/*kööj- ‘to lek’ has been tabulated as 0.5 *kee-, 0.5 *köö-. An exception to this though is the correspondence type marked by Honti as “uu/ïï” which actually outnumbers several allegedly regular vowel correspondences, and seems to deserve a line of its own.

“B”, “BB” and “FF” moreover indicate correspondences that are sufficiently irregular that Honti has only dared to report if the data points towards a back or front vowel, and a long or short vowel.

So the question is: might we be able to determine if there is anything odd going on here? For just one example, while roots with zero onset are quite abundant, there seems to be an absence of roots beginning with *o-. But then again, random holes occur elsewhere in the table as well. So is this a sign of something being wrong with the reconstruction? a reflection of some earlier soundlaw in the development of Ob-Ugric? or perhaps, of nothing at all? Hard to say using only qualitative tools.

Forming some simple quantitative predictions from this type of data is however not hard. For a first approximation, say we assumed a fully random distribution of roots, with no interdependences in the occurrence of consonants vs. vowels. In this situation, the expected number of roots beginning with a given *CV- sequence could be calculated from just the total vowel and consonant frequencies. For example *-ää- occurs in 44/724 ≈ 6.1% of the roots; *ɬ- occurs in 53/724 ≈ 7.3% of the roots; their predicted co-occurrence is thus 0.061·0.073 ≈ 0.44% of the roots, i.e. the expectation value of roots beginning with *ɬää- is about 3.2.

Algebraically, the formula for this expectation value comes out as C·V/A, where C is the attested count of the onset, V the attested count of the vowel, and A the number of roots altogether.

The actual number of attested roots beginning with *ɬää- happens to be indeed 3 (*ɬääpət ‘7’, *ɬäärəɣ ‘ruffe’, *ɬäärɣət ‘hard’). So in this case the prediction is spot on! Many of the other CV combinations seem to work this well too, “off” by about 1 at most. But larger deviations also can be found. Here is the full table of differences between the attested and expectation values, with some color-coding applied:

OU stats 2

As an initial observation, note the gradual accumulation of random holes and peaks: a lesser number of roots are off by about 2, even fewer off by about 3, etc. Also unsurprizingly, bigger deviations are mainly found towards the upper left, where the data is denser.

At this point we could continue quantitative analysis. Making various starting assumptions about expected variance in the vocabulary and then doing a whole bunch of math would probably be able to tell us if the general patterning of the data shows statistically significant deviations or not. But… this seems like a bit too much work. For one, parts of the table would end up having to be recalculated if we were to adjust the underlying reconstruction even just a bit (e.g. by splitting a given proto-vowel in two). And for two, it is not at all obvious what should be our default hypothesis! It is already known that languages tend to prefer some phoneme combinations over others. And yet, AFAIK, a universal typology of this has yet to be developed even qualitatively. Applying detailed rigorous methodology while relying on guesstimated background assumptions would be a waste of effort.

Instead, I think at this point a qualitative human intervention can already tell us how likely is it that there is anything interesting going on here at all. Rather than aiming for assessing every single entry, let’s check out just the lowest-hanging fruit. The 5 most aberrant *CV- sequences in the data are:

  1. *wuu: +9.0
  2. *kuu: +7.4
  3. *ää: +6,9
  4. *mee: +5,2
  5. *kää: -5.0

Since my initial point is to demonstrate that calculating phoneme co-occurrence rates among a proto-language’s lexicon can reveal evidence for adjusting the reconstruction, then surely this sort of evidence should be found in this end of the data, if at all.

And indeed, it looks like that at least the first case is not an accident. In part it probably reflects the fact that the contrast between *uu- and *wuu- is not very clearly indicated in the data at all. Most Ob-Ugric varieties have lost *w before rounded vowels; and some others like Pelymka Mansi and Kazym Khanty have by contrast introduced an epenthetic *w before some rounded vowels. In other words, we may already suspect that having as many as nine roots “too many” indicates that some of Honti’s *wuu- roots here should be actually reconstructed with plain *uu- instead.

A look at Southern Mansi suggests a few good candidates. These are the words where Honti assumes shortening *uu > *u in Mansi (although this is a change he does not really present any conditioning for):

  • #668 *wuuj- >> SMs oj- ‘to swim’ (~ Pelymka wuj-, Kazym wooś-)
  • #682 *wuulɜ >> SMs olā ‘pole’ (~ Pelymka wula, Kazym wooɭ)
  • #689 *wuunč- >> SMs onš- ‘to run over’ (~ Pelymka wunš-, Kazym wuš-)
  • #708 *wuur >> SMs or ‘edge’ (~ Kazym wur)
  • but: #706 *wuur >> SMs wor ‘possibility, way’ (~ Kazym wur)

This looks like Southern Mansi may actually have maintained a contrast between *w and zero in this environment. And, better yet: Honti also fails to list any examples beginning with (zero onset plus) *uu that would have any potentially incriminating reflexes at Pelymka, Kazym, or other similar dialects. So there seems to be no obstacle to adjusting the reconstructions to *uuj- ‘to swim’, *uulɜ ‘pole’, *uunč- ‘ to run over’, *uur ‘edge’. In the case of ‘to swim’ we can even verify this with external evidence! Consider Permic *uj- ‘to swim’. Normally Permic should retain evidence of *w even before rounded vowels (as in Finnish uusi, Hungarian új ~ Komi выль /vɨlʲ/ ‘new’), but no such thing appears here.

Recognizing w-epenthesis also allows cleaning up #701 *wuupɜ ‘older sister’, where *w seems to have again been posited only on the basis of Pelymka wuup. The Khanty reflexes like Tremjugan oopïï, Kazym opi, Obdorsk apii do not support positing *w- at all. Neither does the Proto-Samoyedic cognate *apå. By external evidence, #688 *wuunč ‘nelma’ (a type of salmon) similarly seems to be a case of secondary *w: contrast Proto-Samoyedic *ånčɜ, Komi удж /udž/.

— Moreover the above type of scenario is not the only possible kind of explanation for why a particular sound sequence might be non-randomly overrepresented. A different issue seems to concern the following two words:

  • #659 *wuuč ‘town’
  • #660 *wuučəm ‘weir’

Wider Uralic etymological references generally consider these words to be based on one and the same root. Cognates such as Northern Sami oahci ‘barrier, obstacle, reef’ or Tundra Nenets ва” /waːʔ/ ‘fence’ seem to point to the original basic meaning having been simply ‘fence, obstacle’, from which the two attested meanings are easily derivable. Perhaps also #657: *wuuč- ‘to fish’ is a part of the same bundle. Honti indeed even includes small footnotes in the lexicon commenting on the possible relationship of these three words. It’s not clear to me why he regardless lists them separately.

Altogether at least eight of the roots where Honti reconstructs *wuu- seem to be superfluous in some sense. A pretty good catch for such a simple statistical tool, so far.

I’ve only taken a more casual look at the other top-5 cases, but some instances of *kuu- also might be illusory. More briefly:

  • #229 *kuuďmɜ ‘ashes': according to a recent proposal from Ante Aikio, this would be a derivative of the root listed by Honti as #227 *kuuď-/*kïïď- ‘to disappear’.
  • #261 *kuulpɜ ‘net’ is generally considered an old derivative of #245 *kuul ‘fish’.

Some less directly apparent phenomena may also have shaped the data. For one, I have here only charted out the co-occurrences of initial consonants + initial vowels. Perhaps a look at medial consonants, or the few stem vowels that are found in the data, would turn up other results. In theory it is even possible that some initial *CV- effects are the secondary product of sound changes involving medials instead. Suppose initial X had some interaction with medial Z, and this then had some interaction with vowel Y; this would already suffice to generate a correlation in some direction between X and Y. Hence, with this mode of analysis, it seems efficient to attack the data from multiple directions. Take a couple of snapshots from different angles, look thru the biggest problems that come up, recalculate the results after any adjustments… and see if this then brings to highlight any new issues.

Tagged with: , , , ,
Posted in Methodology

Primary vs. secondary *ë

I claimed in my post “Two Lemmata” that the reconstruction of Proto-Uralic *ë rests on quite firm ground by now. Regardless, it is still not too rare to see studies which fail to recognize the idea. [1] Apparently the existence of this proto-vowel cannot be yet considered to have reached the status of general consensus. Why is this?

Assuming that the relevant literature has simply gone unread might be a bit too uncharitable. I believe a better reason for why doubts persist would be that no single unified source discussing the reconstruction of this vowel is available; the information needs to be pieced together from disparate sources. I hope to have previously provided a brief overview, though, and in this post I will explore some additional complications.

Probably one obstacle has been that the evidence for *ë is not trivial. For all other PU vowels, the evidence of Finnic, which has been presumed highly archaic, can generally be taken as direct: PF *a < PU *a, PF *o < PU *o, PF *ü < PU *ü, etc. (with only minor conditional shunts). The PF vowels also generally remain intact in the descendants. And only in Finnic does the contrast *a/*ë seem to be irrecoverably lost. Hence, one necessary precondition for accepting PU *ë is to accept that the Finnic vowel system does too contain innovations, even major ones.

(You’d think there should be no need to explicitly spell out something this basic, but alas, long-outdated ideas about “key languages” have persisted for long in Uralic studies. Better safe than sorry…)

The direct evidence in East Uralic

The best evidence for the reconstruction of *ë comes instead from the quite distinct reflexes in the easternmost branches: Mansi (*ë > *ëë), Khanty (*ë > *ïï) and Samoyedic (*ë > *ë, *ï). Hungarian *ï, though it has in the modern language merged with the front vowels i ~ í, is also quite distinct in its refusal to adhere to vowel harmony. However, in general the vowel systems of these groups have been subject to much innovation, and it takes care to wring out evidence from here.

The single most important observation, I believe, is to look beyond individual details and to note that among all these four branches — i.e. across the East Uralic group in entirety — the general categories of non-open unrounded back vowels appear cognate to each other. Thus we can find correspondence sets such as the following:

  • H ín (: ina-) ~ Ms *tëën ~ Kh *ɬaan ~ Smy *čën ‘vein, sinew’
  • H nyíl (: nyila-) ~ Ms *ńëël ~ Kh *ńaaL ~ Smy *ńëj ‘arrow’
  • H nyír (: nyira-) ~ Ms *ńëërəɣ ~ Kh *ńaarəɣ ~ Smy *ńër ‘cartilage’
  • Ms *ëët ~ Kh *aapət/ɔɔpət ~ Smy *ëptə ‘hair’
  • H al- ~ Ms *jal- ~ Kh *ïïL ~ Smy *ïlə ‘under’
  • H máj ~ Ms *mëëjt ~ Kh *muukəL ~ Smy *mïtə ‘liver’
  • Ms *tëët ~ Kh *ɬïïkəL ~ Smy *tïtə ‘Swiss pine (Pinus cembra)’
  • Kh *ïïkət- ~ Smy *ïtå- ‘to put up (e.g. a net)’

The alignment is not perfect, but it’s far better than we’d expect to happen randomly. It’d take some odd coincidences to end up with this situation from an original system containing no “ë-type” vowels. [2] I suppose there is the theoretical possibility of proposing *ë to have been an East Uralic innovation, or proposing a set of similar but not identical parallel innovations in the four groups, but I have not seen this done convincingly. [3]

The individual details of course still need examination as well. A 1st-degree correction factor is to note the mainly stem-vowel conditioned split developments in Hungarian (*ë-ə > i ~ í vs. *ë-a > a ~ á) and Khanty (*ë-ə > *aa vs. *ë-a > *ïï). There is very little direct evidence for the original stem vowels in any of the Ugric languages, and the Samoyedic evidence has its limitations as well, but their western relatives help here: cf. e.g. Finnish suoni, nuoli, hapsi vs. ala-, maksa, ahtaa. You may also notice that the H and Kh splits run in largely opposite directions, and indeed I do not think any examples are known where H í or i would correspond to Kh *ïï. There are moreover also some apparent exception cases with *ë-a > *aa in Khanty, though, so the exact analysis of this split may require further fine-tuning.

Secondary *ë in Hungarian and Mansi

As 2nd-degree corrections, it also seems to be the case that East Uralic *ë-type vowels can regardless in some cases represent conditional developments from different PU vowels altogether.

One prominent source of secondary *ë is cheshirization in Mansi. In what seems likely to be a late change, expected Proto-Mansi *oo followed by a velar consonant develops to *ëë followed by a labialized velar. Typical examples include *čaŋa- > *čooŋk- > *čëëŋkʷ- ‘to hit'; *ńoxə-lə- > *ńooɣl- > *ńëëwl- ‘to follow’. (Contrast Samoyedic *čåŋå-, *ńo-.) This is a fairly self-evident change on account of being one of the only regular sources of labiovelars in Mansi (together with similar effects triggered by other labial vowels). It has previously even inspired claims that perhaps all cases of *ëë in Mansi are similarly secondary — say, in Erkki Itkonen’s mid-1900s model of Finno-Ugric vocalism. [4] However the other cases resist explanation by similarly simple conditioning. “Redistributionary” splits, which do not lead to the creation of any new phonemes or even allophones, do happen! Being able to condition the appearence of a sound in one environment is not sufficient evidence for concluding that its appearence in other positions would therefore have to be conditioned by something as well.

And indeed, we can find even contrasts (near-minimal pairs) between primary and secondary *ëë in Mansi. Consider e.g. *këŋkə- ‘to climb’ > Ms *këëŋk-; but *aŋa- ‘to open’ > Ms *ëëŋkʷ- ‘to undress’. As the shift *oo > *ëë / _K has normally left a trace in the form of the labialization of the following velar consonant, then roots like the first could only be accommodated into the system by abandoning regularity and switching to a much weaker model running on “sporadic” sound changes.

 Another sound law responsible for secondary *ë-type vowels also seems to be identifiable. This is a type of “illabiality assimilation”:  *o > *ë / _jC.

This development has long been recognized for Hungarian. E.g.:

  • *kojə-ma > *kojmV > *këjmV > hím ‘male’ (cf. Skolt Sami kuõjj ‘husband’ < PS *kōjë)
  • *pojə-ka > *pojɣV > *pëjɣV(-w) > fiú ‘boy’ (cf. Finnish poika)
  • *kojɜ-ta- > *kojðV- > *këjðV- > hízik (hízo-) ‘to become fat’ (cf. Mordvinic *kuja ‘fat’)
  • *tojə-ntV > *tojdV > *tëjdV(-w) > tidó ‘birch bark’ (cf. unsuffixed Udmurt /tuj/, Komi /toj/) [5]

The first two cases are well-known and relatively clear. I am not sure if the latter two have been previously noted, but they seem to work equally well. A fifth case might additionally be *kojə-ra > *kojrV > *këjrV > here ‘drone; testicle’ (cf. Finnish koiras ‘male’) — though it is unclear why we get here a mid vowel e, instead of the expected i ~ í. [6] It’s also interesting how hím (hime-) and here follow vowel harmony; yet the shift *k- > h- still indicates them descending from back-vocalic originals.

It is also fairly clear that the change only occurred in closed syllables: this is shown by e.g. *kojɜ > háj ‘fat’, *pojə > faj ‘species’ (though the semantic development here seems questionable), *śojə > zaj ‘noise’.

Interestingly there seems to be evidence of this change having extended to Mansi as well. At least three promising and two potential examples can be found:

  • *kojə-ra > *kojrV > *këjrV > *këër ‘male animal’ (cf. Fi. koiras)
  • *kojwV-lV > *kojlV > *këjlV > *këëĺ ‘birch’ (cf. Fi. koivu)
  • *soja-tV > *sojtV > *sëjtV > *tëëjt ‘sleeve’ (cf. Skolt Sami suäjj < PS *soajē; unsuffixed *soja > ujj in Hungarian)
  • ? *poskə > *poɣɬV > *pojɬV > *pëjɬV > *pëëjt ‘cheek’ (cf. Fi. poski)
  • ? *ńojta > *ńëjtV > *ńëëjt > *ńääjt ‘shaman’ (cf. Fi. noita)

The 4th has a kind of a chicken-and-egg problem: after primary *ë there is some evidence for a shift *ɣ > *j (e.g. *mëksa > *mëëjt ‘liver'; *wëlka- > *wëëɣl- ~ *wajt- ‘to rise’) [7], but we obviously cannot use both *ëë to condition the *j and *j to condition to the *ëë. A possible ad hoc solution would be to reconstruct something like #pojsəkə, but let’s not.

The 5th requires a shift from *ëë to *ää, seemingly due to the influence of two flanking palatal/ized consonants. It is not clear though if this should be dated to the Proto-Mansi level, or perhaps later. Northern Mansi /ńaajt/ and Southern Mansi /näjt/ could actually regularly reflect PMs ńëëjt as well: the former thru the regular lowering *ëë > *aa, the latter thru the regular fronting *ëë > *ee adjacent to palatalized consonants + vowel shortening to /ä/. For these changes a perfect parallel is PMs *ńëëraa > *ńeerää > SMs /ńärää/ ‘legwear'; [8] a word not of Uralic inheritance, but here the regular back vowel is still found in Eastern Mansi /ńëërə/, Northern Mansi /ńaara/. It is only the Eastern and Western reflexes of ‘shaman’ that point to older *ää specifically.

It’s moreover possible that the 2nd case actually indicates instead a fairly similar change: *o > *ë / _ĺ. In this light two further interesting words are PU *śod₁ka > *soĺɣV > ? *sëĺɣV > Ms *sëëĺ ‘goldeneye’ (cf. Finnish sotka); and Ms *këëĺt- ‘to peel (e.g. hamp)’, which has been compared to Mari *kŭðaša-, Komi /kuĺ-/, Udmurt /kɨĺ-/ ‘to undress’, and behind which a PU root *kod₂V- could be reconstructed. [9]

There is no clear evidence on how *-od₂- is reflected in Hungarian — this has not been a frequent sound sequence. However, one old lexical comparison (that the UEW rejects) might be rehabilitable if we assumed that also this change occurred in Hungarian: *śod₂a ‘war, fight’ (cf. Finnish sota ‘war'; Mari *šuðala- ‘to scold’) > *śod₂a-nta- > *soĺdV- > *sëĺdV- > szid ‘to scold’? A cluster simplification *ĺd (? > *ɟd) > *d would also have to be assumed though.

However, even though these changes are highly similar, there is a strange complication that seems to preclude an analysis as a common Hungarian-Mansi innovation. In most words where Hungarian points to this kind of a secondary *ë, the Mansi development differs — we see a loss of *-j- instead:

  • *kojə-ma >> *kum ‘man’
  • *kojə-ta- >> *kaat- ‘to become fat’
  • *pojə-ka >> *piw ~ NMs /piɣ/ ‘boy’
  • *tojə-ntV >> NMs /toont/ ‘birch bark’

At least the 2nd and 3rd of these are clearly irregular: *-jt- is a perfectly valid consonant cluster in Mansi (cf. ‘sleeve’ and ‘shaman’ abov), and there are no parallels for a vowel development from *o (or for that matter, any other back vowel) to Ms *i. The 1st brings to mind the developments *kojə > *kuj ‘male’, *śojə > *suj ‘sound’. Was ‘man’ perhaps derived in Mansi from a vowel-stem variant *kojəma > *kujəmV > *kujm?

Perhaps it is relevant that the irregular loss of *-j- in these words extends also to Khanty: *kaatLə- ‘to become fat’, *pak ‘son’, *tontəɣ ‘birch bark’. A fourth example of this is also known, the word for ‘louse': Ms *tääkəm, Kh *teeɣtəm (also Hungarian tetű); contrasting with Finnic *täi, Udmurt /tej/, Komi /toj/. [10] We could perhaps suppose a loss of *j before a consonant cluster to explain the last two… Though *-ktV is not really a typical Uralic noun formant, and so I also wonder if the Ugric words for ‘louse’ are not perhaps instead somehow related to the quite similar root *tikte found in Tungusic.

 In Mansi, further examples of apparent secondary *ë can still be found as well. The residue includes e.g. Fi. os-ta- ‘to buy’ ~ Ms *wëëtaa ‘ware'; Fi. otta- ‘to take’ ~ Ms *wëët- ‘to pluck’. [11] Itkonen in his critique has claimed that *ëë would be even the most frequent correspondence of West Uralic *o, and this seems to still hold up pretty well even once we remove the words showing Finnic *oo (< *a/*ë via Lehtinen’s Law) from the count. It might still be possible that there has indeed been a default development *o > *ë in Mansi, only one bled by several conditional developments. — Regardless: this type of secondary *ë must still be distinguished from primary *ë, which is instead normally reflected as *a in West Uralic, and is further supported by the Samoyedic evidence.

[1] For just one example, no mention of this result appears in what I belive is the newest overview of Hungarian historical phonology available: the fifty-odd page appendix in Andras, Róna-Tás & Árpád, Berta (2011): West Old Turkic: Turkic Loanwords in Hungarian. Wien: Harassowitz.
[2] This can be contrasted with the Western end of the family. “Ë-type” vowels are not at all unknown here either. However, these show no relation to each other. E.g. Ter Sami has the vowels /ï/ and /ïë/, from Proto-Samic *ō and *oa < PU *a and *o. Skolt Sami has õ [ɘ] and â [ɜ] plus the long versions, under various conditions from PS *ë < PU *i, *ü, *e-ə. And the various languages of the Southern Finnic areal have õ [ɤ ~ ɨ], mostly from *e, though in some cases from *o.
[3] At least Reshetnikov & Zhivlov (2011; see Bibliography) have attempted an analysis to this effect, but they do not analyze Hungarian or Khanty, and they exclude some material previously reconstructed with original *o that turns out to be quite relevant. A recent follow-up in Zhivlov (2014) has abandoned the idea.
[4] He has presented some detailed critique against the reconstruction of *ë (“Vokaaliston kysymyksiä”, 1988, Virittäjä 92 pp. 325–329), though it seems this never led to much further discussion of the matter, and after Itkonen’s death in 1992 no one else seems to have had much interest in defending his system of vowel reconstruction.
[5] An alternate reconstruction *tejɜ- would also work for the 1st syllable vocalism, but this would predict a vowel-harmony-compliant **tidVw > ˣtüdő in Hungarian.
[6] It would be possible to hypothetize e.g. that inherited *ë had already been split to Old Hungarian *i vs. *a at this date, and that *oj first yilded not *ëj, but rather *ej, which was later assimilated to *i; and that in ‘drone’, *j was then lost early, leaving a mid vowel. The Mansi evidence seems to support an earlier shift specifically via *ë, though.
[7] There are other words as well with a more limited distribution; cf. Honti 1982: 29–30. These words mostly feature an alternation between a base form with *-ëëɣ- and an oblique stem with *-aj-. I would assume that this *j was later generalized to the nominative in the body part terms ‘liver’ and ‘cheek’, which will only rarely occur as subjects.
[8] On a slightly off-topic note, I am not sure if the Southern Mansi long open stem vowels should be taken as original. They don’t seem to contrast with the corresponding short full vowels, and indeed, they correspond to short stem vowels in the other Mansi dialects. They also regularly condition shortening of 1st syllable vowels. I suspect some sort of a prosodic effect here: e.g. ˈV₁-V₂ > V₁-ˈV₂ when V₂ was a full vowel, followed by lengthening of the newly stressed V₂, and if applicable, shortening of the newly unstressed V₁.
[9] The shift *u > /ɨ/ seems to be regular in Udmurt before coda /ĺ/. Other examples include *kad₂a- > PP *koĺ- > *kuĺ- > kɨĺ- ‘to stay’ (cf. Komi koĺ-); *kod₂ka > PP *kuĺ > kɨĺ ‘disease, evil spirit’ (cf. Komi kuĺ); *neljä > PP *ńoĺ > *ńuĺ > ńɨĺ ‘4’ (cf. Komi ńoĺ). Contrast though retention before intervocalic /ĺ/ in muĺɨ ‘berry’, tuĺɨm ‘topmost yearly growth of tree’.
[10] Mari *ti is ambiguous: this could also derive from e.g. *täkV or *tikV. Samic *tikē is though probably an unrelated loan from Germanic (or perhaps from the same pre-Indo-European source as the Germanic words).
[11] These two might suggest a dissimilation *wo- >> *wëë- at first glance, but a counterexample is *woča > Fi. ota-va ‘fish trap’ ~ Ms *wooš ‘weir; fence; city’.

Tagged with: , , , , , , , ,
Posted in Reconstruction

Etymologically opaque Votic words

For later reference, here’s a collection of etymologically opaque (to me) Eastern Votic words harvested from my new dictionary. I will not attempt any detailed analysis yet. (Presumably some investigation into Russian, Ingrian, Estonian, maybe even Latvian & German could turn up known cognates for many of these.)

  • aimo ‘carbon monoxide’
  • alëtsë ‘mitten’
  • hilkeä ‘ugly’ — if this is not a hypercorrect cognate of Finnish ilkeä ‘evil’.
  • hulkkuag ‘to travel’
  • hülpeä ‘disobedient’
  • ikolookka ‘rainbow’ — a compound based on lookka ‘bow, curve’, but the 1st element is unclear.
  • jahsaag ‘to take off shoes’ — does not seem related to Finnish jaksaa ‘to have energy for’.
  • kaaliag ‘to lick’
  • kaputta ‘sock’
  • kineri ‘melted fat’
  • koltši ‘old-fashioned ladle’
  • kosma ‘hair’
  • lainatag ‘to swallow’ — does not seem related to Finnish lainata ‘to borrow’.
  • lautta ‘cowshed’ — does not seem related to Finnish lautta ‘raft’.
  • liblo ‘oat awn’
  • linnaasëd ‘malt’
  • lohko ‘soup’
  • lühtši : lühdže- ‘pail’
  • läntü ‘milk’
  • mauttši ‘intestine’
  • naka ‘cask spigot’
  • nakliska ‘some part in a sleigh’ (“the informant is unable to explain what exactly”)
  • nëikko ‘rockable cradle’
  • nättšelikko ‘burdock’
  • nätši ‘uncooked (of bread)’
  • nättü ‘rag’
  • ootava ‘cheap’
  • pallo ‘pigeon’
  • pelssimed ‘loom’
  • peltta ‘leftovers of threshing’
  • pihta ‘shoulder’
  • pilpa ‘dandruff’
  • pärähmä ‘fathom, armful’
  • raaka ‘twig’
  • ramitsaag ‘to limp’
  • ratiz ‘granary’
  • rehnüüz ‘entrance hall’
  • rehtilä ‘griddle’
  • ringuttaag ‘to stretch’
  • ripa ‘footwraps’
  • ripila ‘fireplace poker’
  • rooppa ‘porridge’
  • rootšiag ‘to dig, to rummage’
  • ruttaag ‘to hurry’
  • śalko ‘foal’
  • servä ‘edge’
  • sippelikko ‘ant’
  • sisava ‘nightingale’
  • sultsiag ‘to wash’
  • surmukaz ‘relative’ — probably not derived from surma ‘death’?
  • säblä ‘kitchen hook’
  • šitinka ‘bristle’
  • šlotta ‘slush’
  • taari ‘ale’
  • tahtši : tahdžë- (!) ‘chaff’
  • tauttaag ‘to take’
  • tiheh ‘mosquito’
  • turvaz : turpaa- ‘ladder’
  • tuutikko ‘washbundle’
  • türü ‘food comprising breadcrumbs mixed with milk or water’
  • tšiutarë ‘coldroom’
  • tšiutto ‘shirt’
  • tšäppeä ‘beautiful’
  • uhër ‘auger’
  • unka ‘wooden cup’
  • upa ‘bean’ = Est. uba.
  • ursi : urtë- ‘bed curtain’
  • vaattaag ‘to look’
  • valo ‘dung’
  • varo ‘hoop’
  • veelatag ‘to soak’ — compound with vete- : vee- ‘water’?
  • vokki ‘spindle’
  • väitšiäg ‘to call’
  • ördžähtäässäg ‘to wake’
Tagged with: , , ,
Posted in Etymology

A potential Turkic-Yukaghir loanword

A project I am working on and off is compiling lexical parallels that have been proposed in connection to various proposed external relationships of Uralic. Occasionally this kind of work turns up nice new etymological insights.

One of the best-retained — and also one of the more specific — verbs of motion reconstructible for Proto-Uralic is *kälä- ‘to wade': reflected in e.g. Northern Sami gállit ‘to wade’, Finnish kahlata ‘to wade’ (an old loan from Samic), and Hungarian kel ‘to rise’. (The meaning ‘to rise’ is found also in Mansi and Khanty; the latter also has ‘to step up on land’.)  This has been compared with the Yukaghir verb *kel- ‘to come’. The pairing is phonetically OK, but semantically it does not seem impressive. It might be acceptable if a relationship between Uralic and Yukaghir were already established, but it offers hardly any evidence for a relationship in the first place.

Interestingly enough, the same Uralic verb has also been compared with Turkic *gel- ‘to come’ — with the exact same semantics and an equally compatible phonetic shape! (E.g. already Björn Collinder in Fenno-Ugric Vocabulary, 1955/1977, reports both comparisons.) Probably the first step here should be to analyze the Yukaghir word as a loan from Siberian Turkic, and worry about any possible Uralic relationships later.

I would predict that pitting the Uralo-Yukaghir and Ural-Altaic hypotheses against each other may turn up further cases like this where a straightforward loan etymology is available. It’s already been noted by Rédei in his “Zu den uralisch-jukagirischen Sprachkontakten” (1999, in FUF 55) that many of the Uralic-Yukaghir lexical parallels extend to some of the “Altaic” languages as well…

Tagged with: , , , , ,
Posted in Etymology

Get every new post delivered to your Inbox.