3 Mar 2008

Pre-IE Syncope has an easter-egg surprise for you

In Proto-Indo-European (PIE), it's easy to show that in some stage of Pre-IE, there was once a heavy stress accent (as in English or Italian) that caused unstressed vowels to erode and disappear altogether. In the later PIE stage, tonal accent is reconstructed. Pick any noun that shows accent alternation when declined and we can see the evidence of Syncope plain as day. The animate word *h₂éwis 'bird' (nominative case) is *h₂wei-s 'of the bird' in the genitive case, and likewise *ḱwōn-s 'dog' (nominative case) is *ḱun-ós in the genitive. The only sensible conclusion is that these forms were more regular further into the past. Hence we may infer that before Syncope, there was a vowel between the consonants of *h₂w- in genitive *h₂weis (say, *h₂VwéisV) and that *ḱun- in *ḱunós was originally something along the lines of *kVwVn-ósV[1], thereby matching the nominative and accusative forms which would have been the same shape but with accent on the preceding syllable.

In my theory, the stage preceding Syncope (i.e. Mid IE) appears to have a perfectly predictable accent which is always found on the penultimate syllable unless the word ends in any of a handful of suffixes originating from particles agglutinated sometime around early Mid IE (MIE). Antepenultimate accent is found in the nominative singular in *-sa becoming PIE *-s (from *sa 'the' > PIE *so 'this'), the 3ps verbal ending *-ta(i) becoming PIE *-t(i), (based on *ta 'that' > PIE *to-) and the pronominal inanimate in *-ta becoming *-d (also based on MIE *ta 'that'). I presume by Occam's Razor that all initial clusters in PIE dating to Mid IE were the product of Syncope until such time as it can be proven that some clusters were inherited from before Syncope (and I still haven't found any good reason to believe otherwise). By letting logical simplicity be my guide, this has given me a consistent and simple syllabic structure in MIE that allows only *CV(C) (where C = 'consonant' and V = 'vowel').

So this is all straight-forward and has been concluded by many people before me (such as the previously-mentioned linguist, C.H. Borgstrøm). However I'm dismayed that, whether it be Nostraticists who can't see the trees for the forest or conservative Indo-Europeanists (IEists) who can't see the forest for the trees, no one seems to have noticed what this event of Syncope inevitably implies beyond these more obvious examples.

We are soon confronted with a problem if, by chance, we allow our imagination to dream up a hypothetical word form for the sake of argument, something perhaps that doesn't behave so nicely as we would like. Let's say that there was an MIE verb root like **makéra-. As I've described MIE above, there is absolutely nothing objectionable about such a form. It's only when Syncope operates on it that we start to notice a snag since without any interference, we should expect **mker-. In fact, it's rather silly to expect such a thing because although almost anything is possible in languages around the world, a sequence like */mk-/ in a verb root is unheard of in PIE . We would expect perhaps that *m would remain syllabic somehow but verb roots in PIE simply don't look like this and conform to a stricter structure that disallows many shapes.

Therefore, since all sorts of run-of-the-mill sequences in MIE can lead us very easily to awkward consonant clusters when Syncope is applied, this one example illuminates for us that something more drastic must have happened in pre-IE. This 'something' is phonotactic restructuring which would reestablish sanity in the language's syllabicity despite the murderous deletion of so many unstressed vowels, and this likely occurred through various means depending on the nature of the syllabic violation in question. Metathesis might be one possible tactic for languages undergoing these reductions but in this particular example and knowing what we know of PIE, **mekr- isn't any better than **mker- since we find neither sorts of verbs in PIE at all. If we take our cues from Middle Chinese however, we might expect that the offending onset element was simply deleted. So perhaps we should expect **makéra- to become **ker-, not **mker-, after Syncope. One might also suspect that some phonemes were altered to fit better with sonority hierarchy. Perhaps for example, odd resulting clusters like **ml- might have been fortified to *pl- or *bʰl-[2]. As you can now see, my previously-mentioned suspicion that *mad- 'to be drunk' goes back to maxéd̰a- by way of the vocalization and loss of laryngeal in the onset is implicated in my elaborate network of interconnected thoughts and wasn't just some whim I had at the spur of the moment. (That is, regardless of whether this word is connected to a Proto-Semitic passive participle, I really do suspect that this word once had a laryngeal in the onset because of the root's vocalism.) IEists need to ponder on these riddles more but so far either extremist conservatism or extremist whim are the soups-du-jour. I try to walk the path of balance through this age of bipolar extremisms and I hope that my readers do too.

At any rate, one thing is for certain: Syncope directly implies that a lot more went on than just vowel deletion. So if anyone is searching for a thesis topic, this is an itch just waiting to be scratched.


NOTES
[1] I prefer to be more precise than this and reconstruct *kaxʷánasa (nom.)/*kaxʷanása (gen.) in MIE. The reasons for this however are largely beyond the scope of Mid IE and in part related to the rule of Laryngeal Vocalization in late Mid IE which I haven't yet talked about. It suffices for now to observe that PIE's initial consonant clusters are caused by Syncope and placement of stress accent.
[2] I must admit that this particular idea might perhaps be a hasty suggestion considering *mleuh₂- "to speak". At any rate, as you can see, there are many ways in which a language may try to normalize words that violate its syllabicity rules.

11 comments:

  1. There is some evidence of breaking the sonority hierarchy in IE, such as Sanskrit rtá "order of things" and *ndhér- "under".

    ReplyDelete
  2. I was ready for this argument but this is going to be a long answer ;)

    First off, just to be clear for everyone, "under" is *n̥dʰér- with an initial syllabic nasal. However, you probably didn't mean that this was an actual consonant cluster **ndʰ- but rather that you were referring to the seemingly "vowelless" syllable *n̥-.

    I've just read something interesting about how to correctly interpret sonority hierarchy in languages such as Georgian or Klallam where everything seems to go totally haywire. The Syllable in Optimality Theory might be heavy reading but it looks like a worthy book to have in one's library. It discusses the concept of 'semisyllables' to account for what only appear to be violations of sonority hierarchy. For example, it gives Czech srdce which the author analyses as two syllables /sr.dce/ with stress on the first mora on the vowel "r". Thus in Czech, liquids are treated as moraic and both syllables show normal sonority peaks headed by the most sonorant phoneme of the group (i.e. s < r and d < c < e). This is paralleled in Indo-European itself, of course, so that even *streu- can be analysed as /s.treu-/ with an initial semisyllable that cannot bear accent because in PIE a fricative *s is not moraic, only liquids and vowels are. Semisyllables are defined as sequences lacking any onsets that have less sonorancy than the moraic element. So in your example of Sanskrit rtá, r- would be a semisyllable because it's on its own and lacks a less sonorant onset.

    Let's put this another way. In some languages, only vowels are moraic so that if a (semi)syllable doesn't contain a vowel, it's not full syllable. Yet in Klallam, there's crazy crap going on like čɬqə́n̕xʷ cn 'I starve'. So sonorancy rules are dependent on the language and in Klallam, a lateral fricative ɬ represents the peak of the first (semi)syllable čɬ-. Nuxalk is Klallam's evil sister and things are just as bad if not worse (e.g.: Yati slq'ts sk'c tsinu = 'I am glad to see you').

    Anyways, thankfully PIE isn't this crazy (but it's pretty damn close). Nonetheless **ms- is not possible as a cluster since *m will become syllabic and my point is that we don't see verb roots with those kinds of onsets even though a sequence like **masV- cannot possibly be prohibited in the stage before Syncope. So we need to explain why/how these expected onsets vanished in PIE after Syncope.

    ReplyDelete
  3. I'm sorry, Glen, I thought you implied that sequences such as **makéra would become *mkér -- that is, a disyllabic form with an initial syllabic resonant.

    ReplyDelete
  4. Rob: "I'm sorry, Glen, I thought you implied that sequences such as **makéra would become *mkér -- that is, a disyllabic form with an initial syllabic resonant."

    Indeed I did and still do. However, whether this sequence **mker- would have been disyllabic or not (and it would have been because of MIE negative enclitic *na "not" > PIE *n̥- "un-" and MIE *mas "us" > PIE *n̥s) is inconsequential to the fact that PIE verbs never are of such a form. That is, we never see a verb root beginning with a resonant followed by a stop.

    ReplyDelete
  5. Aaah, I just had a flash as to what you may be confused about. According to the terminology used in The Syllable in Optimality Theory that I mentioned above, a mora is used for what the every day guy and gal on the street might confuse with "syllable". In PIE and Sanskrit terminology, a "syllabic resonant" is a resonant (w, y, r, n, l, m) that behaves like a vowel. In Sanskrit rta, r- is a syllabic resonant but it isn't moraic. Instead r- is treated as a "semisyllable" and all of rta is a single mora with two sonority peaks.

    ReplyDelete
  6. It would seem to me that, since we see plenty of evidence of syllabic-resonant + stop sequences in IE, the process(es) by which they were eliminated in verbs cannot be purely phonetic. More than likely, then, the lack of such sequences in *verbs* points to levelling of verbal paradigms.

    ReplyDelete
  7. Rob: "More than likely, then, the lack of such sequences in *verbs* points to levelling of verbal paradigms."

    No, not quite. I agree that it's something more than phonetic but I think it relates to general morphology at that stage. Ugh, I have so much to say on this!! Bear with me as I ramble.

    From what I observe, nouns seem to add a prothetic vowel in certain cases which I call a-Epenthesis just before Syncope and was surely phonotactically motivated in order to restructure awkward or excessive syllables into something more pronounceable. (I argued with Jens Rasmussen on Cybalist about reinterpreting his O-Fix rule as a-Epenthesis many times but while he could admit that he could offer no parallel for his epenthesis of /o/ in other languages, he also wouldn't accept that my solution was a stronger one. To my shock just today, I found out that Serbo-Croatian has an eerily similar rule to mine and it goes by the same name! But I digress.)

    Point is, PIE verbs don't seem to use a-Epenthesis at all and I've noticed years ago that early Late IE (eLIE) seems to have banned 'strong forms' of roots that start with a zero-graded syllable. (Strong forms are the forms of stems used in nominative & accusative cases, or in indicative singular verbs, while weak forms are their zerograde counterparts when accent shifts to the suffix.) So for example, an eLIE noun like *kunás (> PIE *ḱunós 'of the dog') can only be a declined noun based on its form (because of its zerograded root *kun-) and at this stage, no derivative of "dog" can start with *kun- in the nominative or accusative cases either. All derived stems with *kwan- as their head will retain *kwan- in their strong forms.

    If my "strong/weak stem" rule is accurate, then naturally **m̥kér- would be banned by virtue of simply beginning with an initial zerograded syllable which would have led speakers to hear a 'weak stem' instead of a strong one. On the other hand, even though **mékr̥- is a valid strong form on the surface, we can only get to it by added metathesis. But even if metathesis had operated on it, its final resonant creates problems for word syllabicity when consonant-initial affixes are applied. Imagine for example this root extended with *-s- like **mēkr̥sti and you can start to see the problem with that.

    So... we see that offending verbs like these did not survive intact and yet were not resolved by metathesis or epenthesis. Cluster simplification through deletion then looks like a strong possibility, as I've already suggested for *mad- 'be drunk' < eLIE *mäd- < *m'xed'- < MIE *maxéda-.

    ReplyDelete
  8. Just for the record here (I suppose you'll agree and just left it unsaid?), one other argument for PIE's initial clusters to have been analogy-borne would be how almost all adjacent language families / proto-languages have just about none of them - including Nostratic applicants.

    My question, tho.. probably more of something to ponder than something to expect any quick answer to... but: if you explain basically with the sonority hierarchy which clusters remained and which eroded, how do you arrive at complete gaps for dental + /l/, or the even less regular set of stop + nasal clusters?

    ReplyDelete
  9. Tropylium: "[...] one other argument for PIE's initial clusters to have been analogy-borne would be how almost all adjacent language families / proto-languages have just about none of them - including Nostratic applicants."

    Yes, I agree, however an argument based on internal considerations carries more weight than an appeal to conjectures about Proto-Nostratic. One need not look to external languages to realize that PIE once had a simpler syllabism.

    Tropylium: "[...] if you explain basically with the sonority hierarchy which clusters remained and which eroded, how do you arrive at complete gaps for dental + /l/, or the even less regular set of stop + nasal clusters?"

    Sonority hierarchy would have nothing to do with that. However, come to think of it, it seems like you're hinting to me that there's a better solution to my original proposal that PSem *θalāθu 'three' became MIE *tareisa before becoming PIE *treis. The idea of a revised MIE *taleisa (with *l at first instead of *r) becoming *treis due to some aversion to initial *tl- during Syncope might indeed be something to ponder. Nifty!

    ReplyDelete
  10. I decided to do another one of my Borg-intensive Google searches in order to attempt to disprove my current views. How exhilarating! In Mirco Ghini's Asymmetries in the Phonology of Miogliola (2001), it's confirmed on page 42 that "sequences of coronal+lateral have a strong tendency to be avoided by the majority of languages, whether they have constraints on double occurrences of coronals or not."

    Ergo, there is reason to revise my reconstruction for MIE 'three' to *taleisa which then has an even stronger correlation with the Semitic word *θalāθu with lateral. The superior optimality of this new view titulates me in an unnatural way. Thank you, Tropylium. ;)

    ReplyDelete
  11. Yes, coronal+lateral isn't a popular cluster type. Lateral affricates (as appearing in various Caucasian language families of the neighborhood) would be another possible route of decay, but there's no clear PIE outcome for those.

    It might be worth of note that /l/ seems less stable than /r/ in individual IE branches too (Romance, Indo-Iranic).

    ReplyDelete