5 Aug 2007

Pre-IE and alternating thematic vowels

I like to ponder about Pre-Indo-European (Pre-IE) from time to time but strangely, despite popular interest, it's a topic that doesn't often get explored seriously by academics, leaving that topic to kooks. (Oh dear Aunt Martha, I hope I'm not one.) Despite the unsaid stigma, long-range comparative linguistics deserves a respected place in academia because questions about where the world's languages ultimately come from are never going to go away on their own.

One thing I've been curious about is the nature of the thematic vowel in PIE and what that says about earlier stages of this reconstructed language. This vowel is everywhere in PIE grammar - in nouns, in verbs, in adjectives. It's everywhere. This vowel often is an unchanging *o (as in the paradigm of *eḱwo- "horse": genitive *eḱwosyo) but every once in a while it throws a curveball and alternates between *e and *o without any immediately apparent pattern, as with thematic verbs. It's a rebel vowel without a cause.

Quirks like this in a language aren't unexpected. In fact, every natural language has some bizarre twist, like English and its irregular conjugation of "to be" (e.g. "am", "is", "are", "was", "were", "been"). All these flaws in this grammatical persian rug can tell us a lot about the older stage of a language. PIE is no different. This is my daring explanation of what took place in Pre-IE to cause this mess in the language. Sit back and pass the popcorn.

Modern thinking now says that the presentive thematic paradigm of *bʰer- (as illustrated above) was in fact only a subjunctive in PIE proper because afterall Anatolian has none of the most basic thematic verbs that are found in abundance in other branches at all. So the famous present thematic paradigm as in the above picture is a slight inaccuracy since this had only developped after the dissolution of the core PIE speech community circa 4000 BCE. Nonetheless, even with updates to the IE model, alternating thematic vowels are here to stay and still show up in PIE presents through the use of suffixes like *-sḱe- (1pp *-sḱo-mes / 3ps *-sḱe-ti) and *-ye- (1pp *-yo-mes / 3ps *-ye-ti). Here too, there is vowel alternation so it must be quite ancient.

Aside from the 1ps, *o is placed before voiced consonants and *e before voiceless ones. I've already explained how the 1ps ending here originates from the perfect ending *-h₂e (see Thoughts on the early Indo-European subjunctive 1ps ending). However, this doesn't in itself resolve why *o is found here unless we surmise that in some stage of Pre-IE the 1ps ending in the subjunctive was once *-o-m, comprised of thematic vowel and 1ps non-indicative *-m which, rather conveniently for me, is used in the post-IE 1ps thematic preterite (e.g. *bʰérom "I carried"). The prior existence of voiced *-m then has produced the *o in *-oh₂ and now the above-mentioned relationship between vocalism and voicing is now consistent and predictable.

However, it begs the question: What caused the change of *e to *o before voiced consonants in the first place? Are there parallels in other languages? It turns out that Modern English can give us an answer. In the English minimal pair sack / sag, the vowel is typically pronounced slightly longer in sag [sæˑg] than in sack [sæk]. It is referred to as half-length and native speakers generally don't realize they're doing it. Even if one were to pronounce final -g as [k] (without any voicing at all), an English speaker will normally still perceive a final voiced stop if the vowel is pronounced with sufficient length. Vowel length alone in these instances has become the perceptual cue to distinguish historical final -t from final -d in English. (See here [pdf] and here for abstracts relating to this voicing effect on English vowels.)

Now we can apply this same phenomenon to Pre-IE by theorizing that there is a single vowel underlying *e/*o alternation. We can write this underlying vowel as (a schwa). At an earlier stage, vowel length first develops on the phonetic level to produce ever-widening variations. The variations then become phonemicized and the bifurcated vowel begins to drift in two different directions of quality depending on the initial length. In the final stage, we obtain the *e/*o pattern via rounding of former *a.

However, there may be a third alternation that this earlier schwa produced, that of *i. Stems in *-o- sometimes show up as *-i- stems, particularly in compounds, without discernible difference in meaning. In reduplicated verb forms, we find reduplication with *e (*bʰe-bʰor-) but also reduplication in *i (*di-deh₃-). This may be the effects of schwa when positioned behind word accent, rather than after it. In these instances, short schwa would become *e while longer schwa rose to *i instead of becoming *a (> *o) as we find when positioned after the accent. So *dideh₃- would be from earlier *[dəˑ'dehʷ]- while *bʰebʰor- from *[bʰə'bʰar]-. In compounds, accent would have originally been on the last component of the word and the "thematic schwa" of preceding stems would be in an open syllable, making it longer. Thus *-o- is reduced to *-i- in these circumstances. Over a few generations, it would have become fashionable to treat these compounding *i-stems as independent stems, as an alternative to *o-stems without difference in meaning.

If someone can think of a better explanation for all this, I'm all ears. My head hurts and I'm taking two ibuprofen.


  1. Wow, I'll have to think about this a bit more.

    But there's at least a small mistake you made.

    his may be the effects of schwa when positioned behind word accent, rather than after it.

    shouldn't behind be before


    So far everything looks about right.

    so you say that *e will arise only after unvoiced stops. and *o after voiced and voiced aspirated stops?

    But when *o is found before the stress it becomes an *i.

    Why exactly do you take *a as the original sound? Do you base this on the Anatolian languages having a instead of o?

    By the way, I wouldn't mind arguing that the genitive of *eḱwo- is actually *eḱwos, just like the nominative, and that the *-syo ending was taken from the pronouns, based on Hittite that is. But if you don't count Hittite, I guess the reconstruction with *syo is fine. :D

  2. phoenix: "shouldn't behind be before"

    Excuse my hobbit dialect ;) To me, it means the same, but I admit I talk weirdly-like ;)

    "Why exactly do you take *a as the original sound?"

    Tonnes of reasons! It explains the rarity of PIE phonemic *a, nicely deconstructs *e/*o ablaut into a simpler alternation of height, exposes the relationship of PIE *o with *a in the cognate stems of other groups like Aegean and Uralic, and makes it possible to neatly explain thematic alternation by way of the most trivial yet well-documented phonetic phenomena.

    "Do you base this on the Anatolian languages having a instead of o?"

    No, Anatolian did merge PIE *o with *a, however this doesn't mean that it wasn't encouraged to do so by an *o-less para-IE dialect already in Turkey, a dialect that, say, escaped the "Late IE Vowel Shift" wave ;) Just a conjecture (but a very intriguing one that just won't get out of my head).

    "By the way, I wouldn't mind arguing that the genitive of *eḱwo- is actually *eḱwos,[...]"

    I would leave *eḱwosyo as is. I think it's fine. However... I might have already mentioned on the ''Cybalist'' forum (which I don't participate in anymore) my theory that the earlier form used a genitive like *ekwə-s-ya literally meaning "which (is) of the horse", created at the time when the thematic animate would have been first created. Here, *ya (> *yo-s) was semantically redundant but was a simple tactic to disambiguate it from an otherwise identical nominative (*ekwə-s). Of course, if that were correct that would open up Pandora's box some more and further hint that thematic nouns were... [drumroll, please]... misanalysed genitival derivatives. Whoops! ;)

  3. Interesting thoughts. I'm not familiar enough with the PIE corpus to evaluate whether the e~o alternation is likely conditioned by voicing of the following consonant -- and the controversy over whether the traditional plain voiced/voiced aspirated distinction should be interpreted as a distinction between voiceless ejectives and plain voiced consonants would clearly have consequences for your hypothesis.

    Based on the data you give in your entry, I think there might be an alternative explanation, again phonetically based, albeit slightly more complex.

    I notice that in the data you give, /o/ appears before three kinds of consonants:

    1) h2 (the "a-coloring" laryngeal, often hypothesized to be a pharyngeal)
    2) r
    3) nasals m and n

    It so happens that all three of these classes of consonants often have the effect of lowering adjacent vowels cross-linguistically. Pharyngeal H and 9 in Arabic tend to give an "a-coloring", i.e. a lower and retracted position, to adjacent vowels: a (normally a mid to low front vowel) is pronounced low and back; /i/ often lowers to a more [e]-like vowel and /u/ lowers to some extent to a more [o]-like pronunciation. The exact details vary, of course, by dialect. If the hypothesis that h2 is a pharyngeal is correct, a lowering effect on PIE schwa would be a straightforward phonetic consequence of an adjacent pharyngeal.

    /r/ sounds often involve a slight pharyngeal component involving the retraction of the tongue root. This is probably behind the common change from coronal (alveolar) trills to uvular trills in various languages (NW Europe, beginning with French in the 1600s as well as some Arabic dialects). This pharyngeal component is also plausibly behind the common phenomenon of mid-low front vowels lowering before (or occasionally after) a coronal /r/ trill. We find this in Arabic again, where /r/ behaves like a pharyngealized "emphatic" consonant, so that /a/, which is realized as a front low vowel with most consonants, is pronounced lower and farther back next to /r/. We find a similar effect with /r/ in English: look a the /a/ in fat and fad vs far, and it also pops up in the lowering of /e/ before /r/ in some French loans, e.g. clerk (cf. the surname Clark) and sergeant.

    Finally, PIE schwa moving to /a/ and/or /o/ before nasals is a plausible result of lowering accompanying nasalization, something we find in French, for example, where nasal vowels are lowered with respect to their oral counterparts (and compared to Canadian French, this process is even more advanced in standard European French). Compare the vowel of "vent" (wind) in Occitan and Catalan: /ben/, Canadian French: /vae~/ or /va~/, and standard European French: /vA~/ or /vO~/, where I use [ae] for a low front vowel, [a] is the same as the IPA value, and [A] and [O] respectively stand for a low back vowel and a low to mid-back rounded vowel.

    I don't have references to any studies on hand to support all this, but I have read literature describing all these phenomena in depth and it should be easy to track down if you want to use it...

    I may be off track here but it seems to me that what these three environments have in common is the fact that modifications of the vocal tract brought about by pharyngealization and nasalization both have the effect of expanding the size of the resonance chamber thus lowering the frequency of the vowel. Again, there are references to this somewhere in the phonetic literature.

    It should be useful to follow up this hypothesis and compare it to your original voicing-based hypothesis. I think this is an interesting lead to follow; good luck with it!

  4. Holy cow! :) Thanks a lot for the in-depth comment, Chris!

    To respond adequately to what you've said, I need to write a new blog entry. It doesn't seem fair to stuff this all in the murky shadows of a commentbox.

    So stay tuned. For now, contemplate the origins of *tesyo "of this" (from stem *to-) and Szemerenyi's Law.

  5. Ok, that was very interesting, but it still doesn't answer the question that I've been trying to find an answer to for the better part of 2 weeks now.

    So in here you're saying that the -o is the result of the original schwa being next to a voiced consonant and the -e the the result of being next to a voiceless.

    If that is the case, then how do we come up with different words in Latin and Greek for words like 'tooth' deriving from -e and -o stems respectively? I'm not questioning you're theory here, I really just want to understand cuz this is driving me crazy...

    I read earlier that this alternation between -e and -o stems had to do with grammatical case. You gave 'pods' vs. 'pedos' as an example, pardon the lack of accents. So, what's going on here, because I'm seeing the vowel preceding a voiced consonant in both cases, and what are the determination criterion for which nouns will derive from -e or -o stems in daughter languages?

  6. "If that is the case, then how do we come up with different words in Latin and Greek for words like 'tooth' deriving from -e and -o stems respectively?"

    Thematic vowels only. The root vowel of nominative *pōds or the *e of genitive *pedós are not thematic vowels.