Re: [tied] PIE dorsals

On Fri, 8 Dec 2000 20:36:39 +0100, "Piotr Gasiorowski"
<gpiotr@...> wrote:

>>What Thracian and Phrygian (and the second hand information on Illyrian: centum or satem?) seem to show is that the palato-velars were maintained as such (without assibilating) until classical times, judging by spellings such as Gord- and Zord- (/g^ord-/?).
>
>I don't think Thracian and Phrygian (or Armenian and Phrygian) were closely related. Judging from onomastic evidence, Thracian was fully Satemic (it also shared some features with Albanian and the languages of Northern Europe, e.g. *sr- > *str-). Examples of Phrygian sibilants from IE dorsals are found before front vowels (zeuman < *g^heu-mn 'fountain', etc.) but not elsewhere, which shows that Phrygian was no more Satemic than French is.

It's my understanding that a variant form <Mane-zordum> (besides
<Mane-gordum>, <Gordium>) is known in Phrygian.

>>I'm not so sure *K's are as rare as all that. Just as an experiment, I counted almost half the Sanskrit entries starting with k- (the ka- and ka:-'s, as a matter of fact) in the index to IEW, for a total of 175 words and morphemes. 19 of them were only referred to from outside the *k/*k^ or *kw pages. 30 were referred to from the *kw pages, 126 from the *k/*k^ pages. That's four times as many *k-'s as *kw-'s. I counted 117 Sanskrit forms starting with s'a-/s'a:-.
>
>>Even if the *k-set contains more loanwords, onomatopoeia and Pokornian mistakes than the *k^ and *kw sets (which might be true), all it shows is that the unnatural "gap" in the unmarked member (if due to a phonological event in pre-PIE) tended to be filled with borrowings and onomatopoeic formations (I'll leave Pokorny out of this), just as we
>would expect (the same happened with the *b gap).
>
>Certainly. And the process went on for three thousand years or more, generating new *k-words and strengthening the incidence of the unmarked phoneme until it rose to the level observable in Sanskrit. I'm sure you realise that counting Sanskrit entries tells you something about the relative frequency of "new" and "old" k in (Old) Indic but is not a good approach to estimating the relative frequency of dorsal types in PIE. Note also the methodological bias: any Sanskrit word referred to from the *kW pages is likely to be securely attested (*kW guarantees that a Kentum cognate has been identified), while a large proportion of the *k etymologies are likely to be less rigorously controlled if not outright *kuku.

There's of course a small bias in that /k/-words attested only in
satem languages may be assigned to either *kw or *k, but that number
is relatively small, or IE dictionaries like Pokorny's wouldn't have
separate entries for *k and *kw. I don't agree with your last
sentence: a large number of the *k etymologies have centum cognates
too: those cognates simply fail to show evidence of labialization.
It may have been more interesting to count only those words (the ones
that have *k in centum *and* satem languages), but I don't have the
time for that. My experiment shows, however, that we have plenty of
satem *k-words to choose from.

>The PIE state of things can be assessed more reliably by inspecting more-or-less complete inventories of uncontroversial etymologies compiled for some common semantic fields, preferably in a work less antiquarian that Pokorny's dictionary. Many such lists can be found e.g. in Mallory and Adams (EIEC). If numbers and animals are not enough, take anything else, e.g. body parts:
>
>(a) *k^reh2- 'head', *g^enu 'knee', *g^Hesr 'hand', *bHa:g^u- 'arm' *h1orgHi- 'testicle', *dng^Huh2- 'tongue', *k^erd- 'heart', *smek^-/*smok^wr 'chin/beard', ...
>
>(b) *(h)okW- 'eye', *jekWr(t) 'liver' *(h)nogWH- 'nail' (fewer examples but all of good quality).
>
>(c) *K is only found in *krep- 'body' (in a "blocking" environment), *pokso- 'side' (which seems to have actually been *pog^-so-, judging from some plausible cognates cited in EIEC), *kok^-so/sah2- 'hollow of a major joint' (dissimilation possible) and a handful of obscure items like *gutr (Lat. 'throat', Hitt. 'nape'), *twek- (?) 'skin' (Gk., Skt., Hitt.?) and *gHelun-ah2- 'lip' (or rather Gk. 'lip', ON 'jaws', Arm. 'palate'). You could say that they outnumber the *KW-words, but to say so is to ignore the fact that they are also far less reliable.
>
>As usual, *k^ is robustly attested, *kW is significantly rarer but securely evidenced, and *k is by and large restricted to "fringe" vocabulary.

And yet there remains an irreducible number of solid etymologies,
which are not fringe words, are not explainable by phonetic
environment, and that contain *K. We can argue about which ones they
are precisely, or if their number is comparable to or lower than that
of words with *Kw. I too used to believe there were only two dorsal
series, but in the end I just gave up making excuses.

>>What is Meillet's (or Kortlandt's) explanation for the position before non-syllabic *u (*k^u- vs. *ku- [vs. *kw-])? Without an original *k^ ~ *k opposition, the reflexes in the different languages become totally incomprehensible (they're difficult enough *with* *k ~ *k^).
>
>Why not *kw- : *kWw- (reduced *kWeu-) > *k^w- : *kw-?

(I'll use your transcription here on) Because where we find ablaut
variations, the words with *kw alternate with *keu, while *k^w
alternates with *k^eu. I'm not aware of a single example of ablaut
alternation *kw ~ *kWeu-.

>In some cases we may be dealing with a split. Slavic has the mysterious alternation *sw-/*gw- (*svist-/*gvizd- 'whistle', *svEtja/*gvEzda 'candle/star'), which suggests that expressive voicing was incompatible with satemisation (East and South Slavic *zv- in such words is secondary and dates back to the "second palatalisation"). The Satem opposition *k^w-/*kw- is rare and each example had better be considered on its own right.

It's not just a satem opposition. In Greek, *k^w > pp/tt, *kw > k and
*kW > p/t. In Latin *kw sometimes becomes w (vapor), while *kW, *k^w

> qu.

The existence of separate reflexes of *kw vs. *k^w, both in satem and
centun langugaes, seems to me to be undeniable evidence that
explaining *k ~ *k^ away as positional variants simply doesn't work.
The environment is special (and rare), but it was identical for both
*k and *k^.


=======================
Miguel Carrasquer Vidal
mcv@...