"centum"/"satem" "exceptions" [was Re: Northwest IE attributes]

[Miguel Carrasquer Vidal]

"Richard M. Alderson III" <alderson at netcom.com> wrote:

>On 11 Jan 2000, Rick Mc Callister wrote:

>> Why the exception kravih?
>> Is it because of the vowel or a combination of /r/ and non-palatal vowel?

>> Is pekus exceptional because of the velar vowel?

>These and similar "exceptions" to so-called "satemization" were seen by the
>Neogrammarians as evidence for a third series of dorsal obstruents in PIE,
>that is, that along side the palatovelar *k{^y} and the labiovelar *k{^w},
>there was a "plain" velar (or possibly back velar) *k.

>Since then, it has been argued that PIE had either (1) palatals and velars,
>and labiovelars are a late development, or (2) plain velars and labiovelars,
>and palatals are a late development.  Today's consensus view seems to be a
>third alternative, that PIE had palatals and labiovelars, and plain velars are
>an odd development of one or both.

In my opinion, the likeliest possibility is that there were
indeed three series (plain, palatalized and labialized), which
developed out of a Pre-PIE single velar series when in contact
with different vowels (which we might notate as **a, **i and
**u).  The allophones became phonemic when these vowels were
either lost or transformed.  So far, that makes sense.  Figuring
out the details is a different story.

One thing that, in this view, seems quite certain is that
stressed **a > *e and unstressed **a > zero (also: **ai > *ei ~
*i, **au > *eu ~ *u).  We can also imagine that unstressed *i and
*u were lost, leaving behind only palatalization [which I'll
notate as ^] or labialization [which I'll notate as w].  So
*h1'ek^u-os "horse" would be **h1'akiw-as and *nekwt- "night" <
**n'akut-.  The problem is the fate of stressed **i and **u, for
which we can hypothesize spontaneous diphthongization to *ei, *eu
(unlikely, I'd say, but an interesting possibility to account for
possible Pre-PIE long *i: and *u:), or loss as in the case of the
Slavic jers (with, as in Slavic, occasional retention to avoid
excessive consonant clusters, e.g. **CiC > *C^C, but **CiCC >
*C^eCC, likewise for **CuC > *CwC, **CuCC > *CweCC).
Additionally, we should also consider the possibility of rising
diphthongs **ia, **ua > *^e, *we.

I do believe there is evidence for palatalization and
labialization outside of the velar series.  Borrowing an idea of
Jens Rasmussen, I have reconstructed the 2p.sg. active verbal
ending *-s as **-tw (< ***-tu), where labialized *tw is subject
to specific soundlaws: it becomes *s initially and finally (or
rather, it merges with *sw in these positions), but becomes *t
between vowels, as in the 2pl. active ending *-te- (< **-tw-a-).
The special development of final -**tw can alse be seen in the
PIE nominal plural *-(e)s (< **-atw), where Armenian shows the
special development *-sw > -k`, and the locative suffix *-i
becomes -u after the labialized consonant (except in Greek, where
*-sw-i > -si).  Slavic has -*s^U, with a shibilant that is
normally explained as analogical from the i- and u-stems, but
might well reflect the normal development of original non-final
*sw in Slavic (as in s^estI "6" < *swek^tis, or possibly the 2sg.
thematic ending -es^I ~ -es^i < *-e-sw-(e)i).  A special rule of
dissimilation of two adjacent labialized sibilants (*swVsw >
*usw) would account for *(i)us "you" (pl.) < **swesw < **tu-atu
(but Hittite sumes < *suwes), and Armenian vec` < *uswec` <
*sweswek^s "six" (cf. also OPr. uschts "sixth").  Here PIE *sw
reflects borrowed Semitic s^ (Akk. s^es^s^e-t- "six").

Other cases of *t/*s alternation can also be explained as
original **-tw > *-s and *-tw- > *-t- (e.g., pace Szemere'nyi
[Myc. is not Pre-Greek]), the pf.act.ptc. masc.nom. *-wo:ts
(Greek -o:s), obl. *-w(o)t- (Greek -ot-), n. *-wos (Grk. -os),
fem. *-ws + ih2 (Greek -uia) [so the fem. suffix -ih2 (like -i in
the present verbs) was only added after the working of the law
*-tw > *-s(w)]).

The active verbal endings also contain evidence for a labialized
*mw, in 1pl. -*mwen ~ *-mwes ( < *mu-an, *mu-atu), -men in Greek
but -wen in Hittite, and o-vocalism *-mos < **-mwes in Pre-Latin.
Similarly, the independent 1pl. pronoun with m- and w- reflexes
in the IE languages: ***mu-atu > **mwesw > *mes or *wes (or
*wei).  A lexical item like *mel- "grind" (Latin Vollstufe mol-,
Greek Nullstufe mul-) can be best interpreted as original *mwel-.

There is also evidence for palatalized non-velar consonants, as
in the well-known doublet *nem- ~ *iem- "to take" (< **n^em-), or
possibly the -i/-n- heteroclitics (Skt. asthi, asthnah "bone") if
from **-n^/-n^-, cf. non-palatalized *-r./-n- < **-n/-n-.
Palatalized **l^ might also have given *i, but the only example I
can remember at this moment is my rather wild idea to link the
Hittite 1sg. imp. -lit (-lut, -allu(t)) as in <eslit> "may I be,
let me be!" with the IE optative in *-ieh1- (<**-l^et-, with -t >
-h1 as in the ins.sg. Hitt -it, (non-Anatolian)IE *-eh1).

See further Gamqrelidze and Ivanov, which I don't have handy
right now, on their hypothesized *s^, *sw, *tw, *dw.

It is clear that a system with both palatalization and
labialization, as I advocate for (Pre-)PIE, is not stable.  Old
Irish (as the result of a much later and independent development)
had both i-coloured (palatalized, slender) and u-coloured
consonants, but the u-colouring quickly became marginal and was
lost before Middle Irish. The loss of jers in Slavic has led to a
full series of palatalized consonants, but not to a parallel set
of labialized/velarized consonants (not at the phonological level
at least).  By the time of common PIE, the plain-palatalized-
labialized distinction had only been retained in the velar stops,
and there too the system was in the process of breaking down,
with confusions between palatalized-plain or plain-labialized,
which account for the hesitations we find even among the ranks of
the two main blocks ("satem" and "centum").  The generalization
of front *e as the "default" vowel may have favoured the
palatals, while certain positions may have had specific effects
(e.g. de-palatalization before *r [*k^r > *kr], or
de-labiaization before *u [*kwu > *ku]).  But that still leaves
plenty of non-palatal, non-labiovelar "plain velars" which cannot
be explained away.

=======================
Miguel Carrasquer Vidal
mcv at wxs.nl