4 Jul 2008

A few more words on my new Gemination rule for Pre-IE

I'm getting some great feedback and resistance to my new rule. I love resistance! It keeps me on my toes. As I said before, I'm exploring a better way of explaining what appears to be word-final voicing in PIE based on Jens Rasmussen's published input on an earlier voiced nominative singular *-z and the simultaneous existance of *-d. My view remains that they come from deictic stems *so- and *to-, and I'm sticking to that story. Yet this implies word-final voicing which is exceedingly rare and therefore yucky in a good theory.

Tropylium had got me thinking about a rule based on secondary stress after objecting to the idea that the rule would only occur in unstressed positions, but after pondering on that yesterday (and then again at 3AM tossing and turning), I came back to my senses and realized why this cannot be. As I have it so far, my Reduction rule effectively bifurcates a former schwa (represented as unstressed *a in my MIE notation) to either supershort schwa (the default change) or regular schwa. Since my Syncope rule only deletes supershort schwa but not regular schwa, evidentally the voicing can only have occured before the supershort schwas. This may sound contrived to some but the phonetics are really starting to make sense to me, so let me explain it in detail now.

The reduction or deletion of a word-final vowel with simultaneous gemination of the preceding consonant is commonplace (e.g. Japanese desu /des:/) because this is merely a matter of compensatory lengthening and syncope, both of which are commonplace linguistic processes in languages worldwide. It's important to recognize however that a preceding consonant need not geminate as the following vowel is deleted. In what way the vowel is deleted depends on the idiosyncracies of the language in question and its individual speakers. On the phonetic level, all that's happening in this sort of gemination is a transfer of duration from the vowel to the neighbouring consonant such that the overall duration of the sequence remains the same.

In the rule I'm now proposing for Pre-IE, I suggest that this same gemination occurs, but only in unstressed syllables. Again, this is not objectionable considering the example of Old English wīte "punishment" from wītje (no gemination in accented syllable) versus Old English wēstenne "desert (dat.sg.)" from wēstenje.[1] Compare also wīte < wītje (with long vowel) versus cynne "race (dat.sg.)" < cynje (with short vowel). The importance of the latter comparison will make sense as I explain below. If the lack of stress is some factor in (but mind you, not a *cause* of) this rule, then why and how would this have occurred phonetically?

After pondering, I realized why this might be. In a stressed syllable in MIE, such as in genitives ending in *-ása, I don't expect gemination (and hence later voicing) since the thematic vowel *e in *tesyo "of this" suggests that the *s in the genitive singular was always voiceless (i.e. from early Late IE *təs-ya) in contrast to a voiced *-z in the early nominative singular ending where thematic vowel *o always arose just as it did before other voiced consonants. So clearly in an accented syllable, gemination did not occur, probably because the vowel was stressed and therefore longer in duration. The added duration of the vowel in the stressed syllable[2] would understandably be a credible factor in blocking gemination during Reduction of the word-final supershort schwa (similar to the Old English example above) because a geminated consonant would then compete for time with the accented vowel. The longer stressed vowel naturally won that battle, gemination did not occur, and therefore the genitive remained voiceless. It's brilliant!
In an unstressed position however, there is no such competition for length because there are no tidal effects from the presence of stress. An unstressed vowel is comparatively shorter than a stressed vowel anyways. So as the vowel was reduced at the end, length was then transferred to the preceding consonant to maintain the same overall duration. Thus the unaccented sequence *-Vsa as seen in the MIE nominative singular was free to evolve to *-Vsᵊ (Reduction) > *-Vssᵊ (Gemination) > *-Vzᵊ (Voicing) > *-(V)z (Syncope).

I've also thought of a good reason why secondary stress would be insufficient in explaining this gemination. While secondary stress would have once donned most instances of the agglutinated deictics *-sa and *-ta in MIE, it wouldn't have naturally done so in unstressed MIE *kʷai-ta > *[kʷittᵊ] > PIE *kʷid "what?" yet the inanimate pronominal marker shows voicing nonetheless. Therefore it is the absence, not the presence of stress, that must be a factor in this Gemination rule and this would simply be because a stressed vowel in this language was phonetically long enough to compete against neighbouring consonant gemination in contrast to unstressed vowels which lacked the strength to resist it.

There! Now is this new theoretical account good enough to please the populus?


NOTES
[1] Van der Hulst, Word Prosodic Systems in the Languages of Europe (1999), p.342. (see link).
[2] Blevins, Evolutionary Phonology: The Emergence of Sound Patterns (2004), p.173 (see link): "In many languages, a stressed syllable is longer in duration than a segmentally identical unstressed syllable."

UPDATES
(July 04 2008) I corrected "Since my Syncope rule only reduces supershort schwa but not regular schwa [...]" to "Since my Syncope rule only deletes supershort schwa but not regular schwa [...]". Sorry for the potentially confusing choice of words.

7 comments:

  1. It pleases me. Just have to nag at this tiny thing. If the gemination is compensatory lengthening (which I'd expect it to be), reduction and gemination were in fact simultaneous. After all, if first the vowel reduced, there would be no sense of 'need for lengthening'.

    So I'd call the stage 'Reduction with Compensatory lengthening of preceding consonant'. Rather than presenting them as two individual steps. But of course it doesn't really matter much. It just makes the phonetic credibility of the lengthening clearer if you label it as simultaneous and as compensatory lengthening.

    ReplyDelete
  2. Unless I can think of a better way of presenting the rules, all I can do is put Gemination under Reduction with indentation, a kind of "sub-rule". I think what I'll do is save up all these changes I'm making and issue the third draft of this document on July 15 with my latest version of the Etruscan dictionary. I've already made numerous additions and changes to the Pre-IE file offline as I piece together the full scope of my theory and work out lingering contradictions.

    ReplyDelete
  3. A huge problem I see here is, elsewhere you have posited vowel reduction causing compensatory lengthening of the previous vowel, so it's unclear why here it should suddendly lengthen the consonant insted.

    The thematic *o under scrutiny could actually be explained just as well either way, but a long vowel doesn't help with the voicing in *-d...

    There's also a smaller issue, that consonant lengthening requires fairly specific constraints for prosodic weight. Consider: When the change is from (C)VC to (C)VV, bimoraicity is preserved. But siccing apocope on (C)VCV, (C)VC should be an acceptable outcome, if coda consonants count as moraic. If they don't, while long vowels do, (C)VVC will work (typical example: the English "magic E"). But getting (C)VCC requires either onset consonants - or coda consonant clusters, but not single consonants - to be moraic, and this to be preferred over vowel length.

    I've never seen Japanese described as an example of this kind of compensatory gemination (after all, the coda consonant /N/ is moraic), but rather as an example of vowel devoicing.

    I have some counterarguments on the Old English too, but this is for another comment...

    -BTW, should eSnips be requiring registration to download your pdfs? Nowhere does it seem to say that it should, and yet it does.

    ReplyDelete
  4. Tropylium: "A huge problem I see here is, elsewhere you have posited vowel reduction causing compensatory lengthening of the previous vowel, so it's unclear why here it should suddenly lengthen the consonant instead."

    As I explained, the lengthening of the preceding consonant is the by-product of the shrinking of the word-final schwa to a supershort schwa before disappearing altogether. This is simple, garden-variety compensatory lengthening.

    I think you're exaggerating the "problem" (if it really is one) just a tad. If we take a Mid IE nominative singular form like *CVCa-sa, then gemination is free to occur to produce *CVCᵊ-ssᵊ (and then *CVCᵊ-zᵊ via Voicing). Thus the final two syllables still have the same duration as before and vocalic length merely transfers to the previous intervocalic consonant.

    So once we have voicing of the "clipped" deictic, Syncope is triggered due to heavy stress accent and so the duration of the remnant supershort schwa word-finally contributes to adding length to the surviving vowel in the now monosyllabic word: *CV:C-z. Again, compensatory lengthening.

    What rule does that violate? What am I missing?

    Tropylium: "The thematic *o under scrutiny could actually be explained just as well either way, but a long vowel doesn't help with the voicing in *-d..."

    No, but I'm unsure what sort of counterexample you're thinking of. So see the above example.

    Tropylium: "But getting (C)VCC requires either onset consonants - or coda consonant clusters, but not single consonants - to be moraic, and this to be preferred over vowel length."

    Tropylium: "BTW, should eSnips be requiring registration to download your pdfs? Nowhere does it seem to say that it should, and yet it does."

    Hmmm. I'll have to test it to see what's happening there. Thanks for telling me.

    ReplyDelete
  5. Wait, I realize now having misunderstood somewhat; you're not trying to trigger gemination by apocope, but by reduction. A different issue comes into play here - namely that you're dealing with a language without a phonemic length contrast. If you have any real-life examples of lengthening caused by an entirely subphonemic pronunciation change, I'm all ears. Alternately, if your normal/supershort shwa distinction is not supposed to be allophonic, I'd also be very interested in seeing an example of a language distinguishing length only in reduced vowels. I've actually been thinking of your "normal / supershort shwa" distinction as qualitativ, with the former still by this stage basically [a], or at least closer to [a] than [ə].

    This comment
    once we have voicing of the "clipped" deictic, Syncope is triggered due to heavy stress accent and so the duration of the remnant supershort schwa word-finally contributes to adding length to the surviving vowel in the now monosyllabic word
    additionally seems to propose getting compensatory lengthening twice out of the loss of one segment, which is pretty implausible. I was under the impression that it was the 1st vowel lost you were crediting for cases where the root vowel lengthens?

    Additionally, if you explain *CVCa-sa > *CVCᵊ-ssᵊ rather than > **CVCCᵊ-ssᵊ by the root vowel being stressed and thus long, how do you reconcile this with moments later compensatorily lengthening the root vowel anyway? Is it long, or is it not?

    An idea that intrigues me would be non-compensatory lengthening based on foot type considerations, similar to the theory of gradation by fortition proposed for Finno-Samic. The problem with something of this bredth is, I'd expect to see more remnants of it. What we're dealing with here seems much too specific.

    ---

    Now for that other point I mentioned previously. An inconvenience about taking your precedents from Germanic is that the origins of stød & gemination are far from clear-cut, and intertwined with PIE itself. I'll be very surprized if you, as a glottalic theory proponent (of sorts) have not seen before the argument about stød actually representing a retention of glottalization? And the Old English gemination you bring up is part of a pattern where gemination occurs before a variety of sonorants, which are not necessarily lost in the process ("apple" etc.) so I do not find necessary to consider it a case of compensatory lengthening (also briefly covered under the previous article).

    My previous suggestion of insertion of a glottal stop, BTW, has the plus side that glottal stops exist beforehand in the language, while geminates do not (again, if I haven't lost track of all of your ideas). The objection about needlessly adding another phonation type is unfounded, because we may well be operating before the change from ejectivs to creaky voice. I'm finding it difficult to motivate this insertion, tho.

    ReplyDelete
  6. -BTW, should eSnips be requiring registration to download your pdfs? Nowhere does it seem to say that it should, and yet it does.

    I've never had to register. So this is really odd.

    ReplyDelete
  7. Tropylium: "Wait, I realize now having misunderstood somewhat; you're not trying to trigger gemination by apocope, but by reduction."

    Precisely.

    Tropylium: "If you have any real-life examples of lengthening caused by an entirely subphonemic pronunciation change, I'm all ears."

    Hmm, that's a good challenge and it's led me to the demise of my argument considering a poignant quote from International Journal of Slavic Linguistics and Poetics, vol.42 (1998), p.111 (see link): "In typological terms it does appear that languages which do not already possess quantitative distinctions do not develop phonemically long vowels in response to the loss of a following segment (cf. de Chene and Anderson 1979: 508). However, one should note that it is likely that the lengthening of the vowels e and o occurred not at the time when the weak jers were lost, but actually began earlier, at the time of the reduction of the jers."

    This is an intriguing quote. So then should I propose something similar for Pre-IE such that phonemic vowel length arose *before* Syncope?

    Tropylium: "Alternately, if your normal/supershort shwa distinction is not supposed to be allophonic, I'd also be very interested in seeing an example of a language distinguishing length only in reduced vowels."

    I'm merely calling them schwas without concern for their exact phonetics right now. If you think a qualitative contrast of [a]/[ə] is a more natural contrast than a quantitative one, by all means. It might very well be a more optimal account. Of course, if long vowels arose earlier (as per my musing in the above paragraph), then perhaps schwa had length contrasts along with all the other vowels!

    Tropylium: "I was under the impression that it was the 1st vowel lost you were crediting for cases where the root vowel lengthens?"

    Again: Both monosyllabification and the Clipping rule affecting suffixed deictics produce a lengthened vowel in the syllable preceding the deleted schwa.

    You keep saying that I'm getting length twice from the reduction/deletion of the supershort vowel but I just don't see what the problem is because I've explained how this can be. If Gemination developed at the same time as the unaccented vowels were being reduced, then the growing length of the consonant is connected to the shrinking length of the neighbouring vowel. So far, so good.

    Once we have a geminated consonant neighbouring a surviving vowel, there remains length in that vowel. There is at this point a contrast between word-final *-s# and *-sᵊ# which needs to be accounted for. Obviously, the difference between the two is *still* that of length. So once Syncope occurs, the length has to go somewhere, so why on earth not to the preceding vowel in the nominative singular?

    Tropylium: "Additionally, if you explain *CVCa-sa > *CVCᵊ-ssᵊ rather than > **CVCCᵊ-ssᵊ by the root vowel being stressed and thus long, how do you reconcile this with moments later compensatorily lengthening the root vowel anyway? Is it long, or is it not?"

    It's important to note that we're discussing both phonetic and phonemic length in the same breath and this is potentially confusing for both of us. In MIE, I don't believe there was a *phonemic* length contrast. However, considering that the language used a stress accent, it would be natural that stressed syllables were longer than unstressed ones, a matter of *phonetic* length, not phonemic. The root vowel in our example is phonetically longer in MIE but with the help of lost schwas, it becomes phonemically longer as well. The same vowel is not affected in the accusative singular because the case ending *-m does not derive from a deictic with word-final supershort schwa (but rather from *-am with a different syllable structure). So, there is nothing to create the same long vowel that exists in the nominative singular due to Clipping.

    Tropylium: "An idea that intrigues me would be non-compensatory lengthening based on foot type considerations, similar to the theory of gradation by fortition proposed for Finno-Samic."

    I think I'm going to have to bone up on these feet. I admit I'm a little hazy on them and this theory of mine is starting to get complex.

    Tropylium: "I'll be very surprized if you, as a glottalic theory proponent (of sorts) have not seen before the argument about stød actually representing a retention of glottalization?"

    Yes, of course I've read this but I'm not so sure.

    Tropylium: "And the Old English gemination you bring up is part of a pattern where gemination occurs before a variety of sonorants, which are not necessarily lost in the process ("apple" etc.)[...]"

    Maybe so, but the example shows that gemination may occur in unstressed positions while stressed positions may avoid it. I'll try to find more clear-cut examples though.

    Tropylium: "My previous suggestion of insertion of a glottal stop, BTW, has the plus side that glottal stops exist beforehand in the language, while geminates do not [...]"

    Yes, I thought of this. However, I have a sneaking suspicion that ejectives had already eroded to creaky-voiced stops before contact with Proto-Semitic. I seem to recall phonetic correspondences that seem to suggest that Mid IE couldn't have had ejectives at that time. (I'll have to look up examples to show you what I mean though. Bear with me.)

    So far, I'm thinking that gemination was merely an allophonic variation of non-geminated phonemes. In other words, [ss] is an allophone of *s and [tt] is an allophone of *t until Voicing turns these sounds into different phonemes, *z and creaky *d̰. I'm not adding extra phonemes or phonation types at all, only allophone variants.

    ReplyDelete