On Thu, 27 Feb 2003 03:27:20 +0100 (MET), Jens Elmegaard Rasmussen
<
jer@...> wrote:
>It may be note that PIE has other cases of an interchange of /m/ and /H2/.
>One could cite the roots *gWem- and *gWeH2- 'come, go' (combined to a
>paradigm in Greek baino:, eba:n); there are also *drem-/*dreH2- 'run' and
>another *drem-/*dreH2- 'sleep'. Could the form with /H2/ be pausa
>variants? The alternation would not be *very* different from Old Norse
>springa, prt. sprakk (with -kk < *-nk < *-ng). Still, it would demand
>something like a labiovelar nasal, and thus does not look particularly
>appealing.
This *m ~ *h2 alternation has been on my to-do list for a long time,
but so far I haven't been able to come up with a solution that made
sense. Jens' remarks have turned my attention to the problem again,
so let's have another try at interpreting this mysterious alternation.
In principle, there are a number of ways in which we can view the
problem:
1) There are two roots *gWem- and *gWah2- (etc.), and they are
etymologically unrelated. The similarity in meaning and shape is
accidental.
This possibility is hard to disprove, but I can reject it by an appeal
to "gut feeling": I "just know" that *gWem- and *gWah2- are related.
On the other hand, I can accept this possibility for now in its
"agnostic formulation": Until a good solution is found for the
alternation *m ~ *h2, the two roots *gWem- and *gWah2- (etc.), must be
treated provisionally as etymologically unrelated.
2) There was a root *gWeX-, which split into *gWem- and *gWeh2-
depending on position/context. This is Jens' speculation above, and
the area where I have been searching on and off for a solution, in
vain. Jens' suggestion of something like a labiovelar nasal (*ngW ?)
might work, but has its problems. In itself, I have no trouble
accepting the presence of labiovelar *ng in (pre-)PIE, and my personal
theories about pre-PIE phonology would make the existence of a
labialized variant *ngW unavoidable (I believe there was stage when
_all_ pre-PIE consonants had labialized variants). If *nW gave *m in
certain contexts, as I believe, then *ngW could also have given *m.
Unfortunately, I cannot find a pathway from *ngW to *h2. If we
reconstruct the root for "blood" with a labiovelar nasal as
*h1ésh2angW-, then the result in final position is *h1ésh2r.gW (Skt.
asr.k), not +h1ésh2&2.
There is a variant of possibility (2) where the proto-phoneme *X is
replaced by a consonant cluster:
2bis) *gWem- and *gWeh2- derive from a root *gWeXY (where *X is
perhaps *m, and *Y perhaps *h2), which gave sometimes *gWem-,
sometimes *gWeh2-.
But this (at least for a cluster *mh2) simply does not work. We can
explain the loss of *m before *h2 in certain contexts (see below), but
then we would expect an alternation *gweh2- ~ *gWemh2- which is simply
not what we see. There is no +gWemh2-
Some more possibilities:
3) The root is *gWe-, and *-m and *-h2 are suffixes (root extensions).
4) The root is *gWem- and *-h2 is a suffix / root extension.
5) The root is *gWeh2- and *-m is a suffix / root extension.
Possibility number three is somewhat similar to number one. It is the
best "agnostic" solution if your gut feeling tells you thet *gWem- and
*gWeh2- _are_ related, but you do not want to get tangled up in idle
speculation. Perhapds that's why I find it unsatisfactory. I prefer
to get tangled up in idle speculation.
For the time being, I'd like to propose a solution along the lines of
possibility number 4. The root is *gWem- and the variant *gwah2- has
a suffix *-h2. Here my provisional speculations about PIE *a come in
handy. A form *gWe:mh2 or *gWo:mh2 (pre-PIE **gWa:mh2), possibly with
Szemerényi lengthening due to *-h2, would have been nasalized to
*gWa~h2, giving *gWah2. Conveniently, we cannot tell the difference
between *gWeh2- and *gWah2-, so why not? The unextended root remains
as *gWem-. Now all I have to figure out is what the function of this
*-h2 was. I can find nothing particularly "passive" about *gWah2- as
opposed to *gWem- (except perhaps Arm. ka- "stand" vs. e-ki, e-kn
"come" [aor.]), so linking it to the Sanskrit aorist passive in -i (<
*-h2 ?) is not obvious (and the Skt. aorist passive form of the root
gam- (*gWem-) is a-ga:m-i anyway).
Of course my solution above divorces the PIE *m ~ *h2 alternation in
PIE *gWem-/*gWah2-, *drem-/*drah2- from the problem of first person
marking in Nostratic. I would keep the suffixes *-m (active) and *-h2
< **-k (stative) strictly separate, as they are in Uralic and
Eskimo-Aleut. We can reconstruct stative *-k vs. active *-m, which
also neatly fits the Afro-Asiatic evidence (stative *-k(u) vs. active
*-ni/*na-, with /n/ from /m/ before front vowel [as in Hausa singular
<ni> "I" vs. plural <mu> "we"]). There is a possibility to unite the
two, if we take into account second person active *-t versus stative
*-th2 (< *-tk), and we analyze the latter as 2nd person marker /t/ +
"stative marker" (possessive marker?) /k/. In that case, perhaps 1st
person stative /k/ may ultimately be analyzable as 1st person /m/ +
"stative marker" /k/, with the /m/ eliminated by a very ancient
(pre-)Nostratic soundlaw.
=======================
Miguel Carrasquer Vidal
mcv@...