On Mon, 12 May 2003 00:42:43 +0000, Jens Elmegård Rasmussen
<
jer@...> wrote:
>--- In cybalist@yahoogroups.com, Miguel Carrasquer <mcv@...> wrote:
>>
>> With the exception of the dubious 2pl. middle (and we might add the
>> 2sg. athematic imperative in *-dhi), the distribution seems to be:
>> voicing occurs in (pro-)nominal forms, verbal forms have voiceless
>> endings. If verbs were sentence final (SOV), that makes sense.
>
>Nice going, Miguel! I am not sarcastic. While you're ticking, could
>you also get the verbal 1sg *-h2 to be voiced and then really make
>*-o: come from *-o-h2? Could it be done by adding a "primary
>marker", at least that would stop it from being word-final? In the
>external evidence we seem to have a velar ending in the sg. and -m-
>in the du./pl. We need that gap to be bridged. By introducing word
>order you may indeed be on to something.
I see my little excursion on *h1e-g < **h1e-k did not impress you
much. Well, same here. It's not more (or less) credible per se than,
say, my previous attempt at an explanation (emphatic *-ge, added
early). Other than *h1e-g, I have no reason to assume the change *-k
> *-h2 was either earlier or later than the change *-k > *-g in the
circumstances mentioned, so it's possible that we had *-V-k > *-V-x >
*-e-G (gamma) > *-o-G > *-o-h2 in the first person singular. Except I
can't make it work.
Before I go on, some preliminary remarks.
1. We're talking only about _asyllabic_ endings with original
voiceless consonant(s).
For instance, in the nouns and pronouns, we have many desinences
ending in a voiceless consonant (e.g. genitive *-es, *-os, *-(&)s;
instrumental *-eh1 (Hitt. *-et); N.pl. *-es; D/Ab pl. *-bhios, Ipl.
*-bhis; Du.nom. *-eh3, n. *-ih1), but they all contain an original
vowel. Incidentally, the voicelessness of the feminine ending
(thematic *-eh2 > *-a:) proves my point that the underlying form is
*-iq (*-ih2), with inherent vowel.
Schematically:
voiced voiceless
*-p -- --
*-t *-d *-t
*-k *-g (?) *-h2
*-x *-G *-h2
*-s *-z *-s
*-tk *-dh (?) *-th2
We have *-d in the pronominal n.sg. (*tod)
We have *-t in the verbal 3sg. (*h1es-t)
Hypothesis: We have *-g in the pronominal nom.sg. (*h1e-g^)
We have *-h2 in the verbal 1sg. pf. (*-h2[a])
Hypothesis: We have *-G in the nominal n.coll. *-h2
We have *-z in the nominal masc.nom.sg. (thematic *-os)
We have *-s in the verbal 2sg. (thematic *-es)
Hypothesis: we have *-dG > *-dh in the verbal athem. 2sg. imper.
*-dh[i] and in the 2pl. middle *-dhwe[..]
We have *-th2 in the verbal 2sg. pf. *-th2[a]
2. In trying to establish the conditioning factors, we can consider
preceding and following segments. But nothing can be said about
preceding segments, which can be anything. Strictly speaking, I can
only prove that the voiced endings were voiced when preceded by a
thematic vowel, so there is the possibility that they were _only_
voiced when preceded by a thematic vowel. I will assume they were
always voiced (Szemerényi lengthening might be an indication).
Because we're talking about asyllabic consonantal endings, nothing is
likewise all that can be said, in principle, about the influence of
following segments, as there were none. However, even taking into
consideration the shape _before_ the fall of final vowels that I
hypothesize, I could discern no pattern (neuter *-d < **-ta, and 3sg.
*-t < **-ta; nominative *s < **-tu or **-su, 2sg. *-s < *-tu, 3sg.
preterite *-s < **-su or **-tu). Accent is also unhelpful in the case
of asyllabic endings, so the only possible conditioning factor
remaining was position in the sentence, which led to my hypothesis
that sentence-final verbal forms showed voiceless asyllabic endings,
other forms voiced ones.
Now can I make it work for those _verbal_ endings that seem to show
voicing?
The imperative is relatively easy: it has a tendency to be pulled to
the front of the utterance, and need not have been sentence-final at
the relevant stage of PIE.
The middle always has added material after its personal endings
(-r(i), -dh(i), -i/-oi/-o, -m), which then would appear to be of a
different nature than the *-i we see in the present (and the
imperative *-dhi) or the *-e we see in the perfect (*-h2-a, *-th2-a,
*-0-e). Unlike the present and (singular?) perfect endings, the
middle endings would appear to have been word-final, but not
*sentence-final* (the additional sentence-final material was
agglutinated later). An important point is that it has been noted
that Hittite, Tocharian and Germanic (arguably the three most archaic
branches of IE) always have *-o- as the thematic vowel in all persons
of the middle. This would be consistent with voiced endings
throughout, something like:
*-(o)-G-ai
*-(o)-dG-ai [?]
*-(o)-(t)-oi
*-(o)-mV(s)-(dG)-ai
*-(o)-dGw-oi
*-(o)-nt-oi
No problems here in the plural, but the singular is problematic. The
first person sg. is slightly odd, but *-G-ai > *-h2ai and *-o-G-ai >
*-oh2ai > *-a:i (or *-o:i?) are not unacceptable. In the third
person, the *t is missing in the Hitt. hi-stems and in a number of
archaic Sanskrit and Old Irish forms (-e:/-a and -ar), so it's perhaps
not original.
The second person sg., however, does not comply. Hittite clearly has
geminates in 2sg. -a-tta(ri/ti); OIr. passive -ther and imperfect -tha
= Skt. preterit -tha:s, all point to *-th2-. Only Armenian -ar (with
*dh > r as in 2pl. -aruk` < *-a:-dhwes) seems to support *-dh in the
2.sg.
As I said at the beginning, I don't see how thematic present 1p. sg.
*-o: can be from *-o-G, in light of the undoubted voicelessness of the
other strong active forms (primary and secondary): 3sg. *-et(i) [which
can perhaps be explained away as recent], but especially 2sg. *-es(i)
[again Armenian impf. (< opt.) -ir and aorist -er are probably too
little to go on for a putative *-z]. And if an unmarked first person
sigular isn't sentence-final in an SOV language, then what is?
Perhaps Jens can think of way to save that theory (e.g. through the
subjunctive [not sentence final?]), but I'm sticking with my old
explanation of *-o-mW-i > *-o-mW-u (by Umlaut) > *-owu > *-o:u (as in
the u-stem Loc.sg.) > *-o: (except in Tocharian B where it was still
-ew).
Finally another catch: the collective *-h2. If this was really *-G,
the thematic (o-stem) n.pl. should be *-oh2. The traditional
reconstruction is *-eh2 > *-a:. This fits Sanskrit -a:, OCS -a and
Gothic -a, but then those three would also fit -o: < *-oh2. Latin and
Greek have short -a. My theory can be provisionally saved if I repeat
after Beekes: "The short -a of Greek and Latin points to -h2. (The
form, then, is not an o-stem)."
=======================
Miguel Carrasquer Vidal
mcv@...