23 Nov 2008

Laryngeal abuse - Phonemes caught in the reconstructive crossfire

Today I'm here to warn you about the tragedy of laryngeal abuse. This is when you see a long vowel in some proto-language and a devilish thought comes to mind like "Gee, I wonder if that long vowel is underlyingly the result of a vowel-plus-laryngeal combo?" And then before you know it, you've gone and rearranged the entire proto-language according to your laryngeal-obsessed whims. Laryngeals are fun, but we have to keep a level head too.

I remain convinced that Bhadriraju Krishnamurti's version of Proto-Dravidian is one such example of this laryngeal abuse at work, and it seems to me that this can be quickly resolved by examining what a mess he makes out of the pronominal system of this language. As far as I'm concerned it's supposed to look like this (i.e. how most Dravidianists reconstruct it):

singularplural
1p*yān
(obl. *yan-)
*yām / *nām
(obl. *yam- /*nam-)
2p*nīn
(obl. *nin-)
*nīm
(obl. *nim-)
3p (reflexive)*tān
(obl. *tan-)
*tām
(obl. *tam-)

However, Krishnamurti has proposed the following[1]:

singularplural
1p*yaHn*yaHm / *ñaHm
2p*niHn*niHm
3p (reflexive)*taHn*taHm

He then suggests that the laryngeals disappear in oblique forms. However the process by which this happens is obscure and left unexplained. In contrast, the idea that long vowels are reduced when used in oblique cases or when preposed to another noun is natural and commonplace. For example, we may take note of French moi "me" versus enclitic m(e) "me, myself", the latter being preposed to verbs as the object (e.g. Elles m'aident. "They help me."; Vous me dérangez "You disturb me."). Since we know where French comes from (i.e. Latin, of course!), we know how absurd and off-track it would be to reconstruct Proto-Latin **me(H) "me" in ignorance of attested Latin, placing a laryngeal in there that appears and disappears conveniently like the Cheshire Cat without rhyme or reason.

I have to say that I just don't buy Krishnamurti's reworking of the pronominal system. Whether Dravidian ultimately has a few laryngeals lurking about is, to be fair, a seperate issue that may still hold true, but these pronouns surely don't contain any. To add them here makes analysis more difficult rather than less.


NOTES
[1] Krishnamurti, Comparative Dravidian Linguistics (2001), p.336 (see link).

5 comments:

  1. Uralic might be another case of Laryngeal Abuse. The PU laryngeal occurs in very restricted environments and from what I can tell occurs in complementary distribution with *p, *k, *l, *r, and *j.

    ReplyDelete
  2. Mordrigar: "Uralic might be another case of Laryngeal Abuse. The PU laryngeal occurs in very restricted environments and from what I can tell occurs in complementary distribution with *p, *k, *l, *r, and *j."

    I'd like to argue on that point with you. From what I see, in the specific case of Uralic, it would actually help by explaining away long vowels and reducing the language to CV syllables. Unlike with Proto-Dravidian, it appears to me that laryngeals would aid in its analysis rather than hinder it. So we'd have *näxi "woman", *toxi- "to bring" and *puxï "tree" without causing problems like they do in Dravidian.

    However, do you have a counterargument that I've overlooked? What do you mean by "complemtary distribution with *p, *k, *l, *r and *j"? Can you provide sources or evidence to support this?

    ReplyDelete
  3. One thing that's important to note is that *x is always followed by *i or by a consonant followed by *i. Intervocally *p and *k appear before *a, but not *i (though there are exceptions in Finno-Ugric).

    Internal clusters of Cl are rare. Only *p, *k, *r, *w, and *x followed by *l are attested. The combination *rl appears in one word with an uncertain following vowel. This word is limited to Hungarian and Mansi. So much is unclear about it. The combination *wl appears before *i and therefore cannot be a source of *xl. While *pl and *kl appear in few words they always appear before *a, and not *i. So *pl and *kl (and possibly *rl) are in complementary distribution with *xl.

    Other combinations are even rarer. These are *xm, *xn, *xs, *xs', *xr, *xd, *xt, *xN (velar nasal)and *xd'. The combination *jm appears before *a or an indeterminate vowel, but never demonstratbly before *i. The one example of *jn is before *a. *jr appears before *a but not *i (but there are some indeterminate cases as well). *lt appears before *a, but the only example before *i is Finno-Ugric.

    I'm still doing some investigation of these, but so far I haven't found any counter-examples. Note: I'm using Starostin's database which has a buggy search feature and is inconsistent about showing the laryngeal in the Sammallahti reconstructions.

    I'm currently checking through PIE to see if there are any possible pieces of evidence (which presumes a genetic relationship between the languages). Some promising examples:

    PIE * pekW, to cook
    PU *pexi, to cook, boil

    PIE *g'ep, mouth
    PU *käxli, tongue

    PIE *wegH', to carry, drag
    PU *wixi, to bring, carry, convey

    PIE *megH', to give, sell
    PU *mexi, to give, sell

    My discovery of all this (prompted by Janhunen's suggestion that intervocalic *k followed by *i would yield *x has forced me to reconsider much of my earlier work on Indo-Uralic. But that's progress for you.

    ReplyDelete
  4. "From what I see, in the specific case of Uralic, it would actually help by explaining away long vowels and reducing the language to CV syllables."

    Oops, forgot to address this point. PU has plenty of well established internal clusters so I'm not sure what the advantage is of assuming only CV syllables. If we assume these arose from some kind of syncope on the second syllable then we should expects all manner of internal clusters, yet only a few combinations are found.

    ReplyDelete
  5. If I may intercede again, aren't you two arguing different issues? Original long vowels (not deriving from consonants) vs. derivation of laryngeals from other consonants before their later demise?

    And yes, proposed IE cognates for Uralic *x are many, but usually proposed as loans. I won't strain Glen's comments section for this topic tho, you probably plan on raising it back at Uralica too anyway?
    (presumably you're "Etherman23"? I'm J.V.)

    ReplyDelete