Re: [tied] PIE dorsals

----- Original Message -----
From: Miguel Carrasquer Vidal
To: cybalist@egroups.com

Sent: Thursday, December 07, 2000 2:54 AM
Subject: Re: [tied] PIE dorsals

>Why should the 3-way

distinction have had to be preserved *somewhere* for it to have been real?

Because IE is a large family with numerous branches. As I said, the Satem group is pretty close-knit (either as a genetic taxon or as a very old Sprachbund) and may arguably be treated together, but for the other branches (Hellenic, Germanic, Italic, Celtic, "Illyroid", Tocharian and Anatolian) you'd have to propose several independent mergers of *K and *K^ -- a non-Satemic conspiracy. In evolutionary biology, outgroup evidence is used to distinguish a shared innovation from a shared retention. Since Anatolian can be regarded as an outgroup with respect to the other IE languages, it's likely that the Kentum dorsals are original (I mean PIE), since this is what we find in Anatolian. The Luwian counterevidence is meagre, doubtful and inconsistent. It's symptomatic that only little-known and hard-to-analyse languages are alleged to have this three-way contrast (Albanian some time ago, Luwian at present, Thracian or Phrygian at any time).

>>(Re: the relative rarity of *K) I’ve never seen a convincing

counterargument.

>Actually, your (3) is one. "There is a bias in

the Brugmannian reconstruction: anything that has a *$ reflex in any Satem language is automatically assigned to the *K^ set."

But it's a counterargument that cuts both ways. The origin of the *$ reflexes is problematic if they don't go back to *K^. Secondary palatalisations usually produce different results in Satem languages. But even if you turn a blind eye to this problem and assign the doubtful cases to the *K set, the other dorsals are still several times more frequent.

>Undoubtedly, in some environments (in some languages), the opposition

betwen *K and *K^ is neutralized. So is, in some environments (in some langages), but to a lesser degree, the opposition between *K and *Kw, but that doesn't mean we should consider them positional allophones of a single phoneme (although it's of course easy, perhaps
unavoidable, to think that, historically, *K and *Kw (and *K^) must all have a common origin).

>I

do not deny that in the satem area the boundary between *K and *K^ becomes fuzzy. But it does not follow that this reflects a split of *K into *K^ (normal development) and *K (in "blocking environments" etc.). It can also mean a near-merger between *K^ and *K (with *K^ swallowing up large portions of earlier *K, esp. in I-I).

>The number of cases where, even in I-I, one

finds "unmotivated" *K (no *r following, etc.) is high enough for me to prefer the latter option. Hell, I've even gone so far as claiming that the loss of the distinction between *K, *K^ and *Kw is nothing but the final stage of a process whereby a three-way split in quality of *all* consonants was
gradually lost in pre-PIE.

>[...] Just in case I wasn't clear enough, I'll state

again that these "blocking environments" are far too ill-defined ("loose") for my taste.

I wasn't too precise about those "blockers", and forgot to mention one environment specified by Meillet, namely the position after PIE *u (*jugom, *leuk-, *dHugh2te:r [where *g occurs in two overlapping blocking environments]). In this environment Sanskrit and Balto-Slavic show *K quite consistently, but Armenian has *$ (Armenian is in general the most consistently "satemising" language). For the sake of completeness I can add that in onomatopoeic words *K would have been restored to prevent the loss of imitative value (a cuckoo doesn't go "shoo-shoo" ot "tsoo-tsoo"), and that dissimilation could affect words with more than one dorsal: *k..k- > *k^..k- rather than *k^..k^, e.g. in *konk- > *k^onk-'hang'.

The "dispalatalising" environments are studied is the following article:

Kortlandt, Frederik. 1978. "IE palatovelars before resonants in Balto-Slavic". In: Jacek Fisiak (ed.). Recent Developments in Historical Phonology. The Hague: Mouton.

Kortlandt compares the Balto-Slavic distribution of the reflexes of "*K^" with those in Indo-Iranian, Albanian and Armenian with all the scrupulousness of a Leiden scholar, and his findings corroborate Meillet's approach. I find his analysis convincing. The only point where I disagree with Kortlandt is his insistence that the two PIE articulations were *k^ and *kW. This would require independent and hard-to-motivate shifts in the Kentum branches.

As for the number of *K words, I'd say that it's amazingly low for what is apparently the unmarked subset of a natural class. If loanwords and onomatopoeic words are excluded, the are very few good cases left. The rarity of "unmotivated" *K is striking in any lexical field. I mentioned PIE numerals in my posting, but the same is true of body parts, elements of nature, names of animals etc. Look at this:

*K^ -- *k^uo:n 'dog', *k^asos 'hare', *(h1)ek^wos 'horse', *pork^os 'pig(let)', *(h1)elk^is 'elk', *pek^u 'livestock', *g^Hwe:r 'wild animal', *g^Hans 'goose', *h2rtk^os 'bear', *h1eg^His 'hedgehog', *dHg^Huhs 'fish'... (nothing recherché about them)

*KW -- *wlkWos 'wolf', *gWo:us 'cow', h2agWnos 'lamb', ...

*K -- *gerh2- 'crane' (how much longer can you make this inventory)?

If we exclude a handful of "North-Western" words that are found in Balto-Slavic but not elsewhere in Satem and have Germanic, Italic and/or Celtic cognates (*dHrougHos 'companion', *gHostis 'guest', etc.), there remain several problematic cases like *legH- 'lie' or *dlhgHos 'long', but not too many of them. If the shift *K > *$ began somewhere within the Satem area and spread through lexical diffusion, it would be hardly surprising if we found, say, 80% of consistently shifted items, 15% of forms that vary between *K and *$, and a 5% residue of *K words.

Brugmannian *K^ is also significantly more frequent than *KW, which squares well with the interpretation of *K^ (= my *K) as the unmarked member of the opposition *K : *KW. This is what we find in Kentum. During the Satem dorsal shift the markedness values were reversed and a new *K (reflecting the merger of PIE *KW with unshifted old *K) became the unmarked ("plain") dorsal.

>And I'll add that a three-way distinction between palatalized, plain and labialized/velarized consonants is in no way typologically implausible, although it would seem to be diachronically instable (and thus not terribly common). A good example is Old Irish, where besides the well-known "broad" and "slender" (palatalized) consonant qualities, there was also "u-colouring" (labio-velarization), caused by (formerly) following *u or *u:. Even in the oldest Old Irish glosses, "u-quality" was losing ground to "neutral quality" (maybe it was more of a merger, given the velarized quality of the Gaelic "broad" consonants). The same may have happened in Slavic, where the loss of the yers <I> and <U> may have initially resulted in (the phonologization of) three consonant qualities, later resolved into two (palatalized vs. plain [though often labio-velarized, especially /l/]) or even back to one in modern Slavic languages.

Such a contrast is indeed possible, though typological considerations would make one expect some place-dependent asymmetry in the system. All kinds of secondary articulation go very well with dorsals and almost equally well with coronals; "labial" is typologically the least frequently modified articulation, and labials with extra lip rounding as a distinctive feature seem to occur only in very special circumstances (e.g. in systems with no contrastive vowel roundness, as in Aranda).

As for your Slavic scenario, if you can show some evidence (preferably a set of minimal pairs) for a ternary contrast like *t : *t' : *tW, we can discuss it.

Piotr