The "null hypothesis" is the most
parsimonious one. If morphological conditioning accounts for the observed facts,
why introduce exotic phonemes praeter necessitatem? You talk about the
phonological development of *-os in masculines and *-os in s-stems as if other
things were equal. But in *k^lewos *-os is part of the stem, while in *wlkWos we
have an inflectional ending (and in pre-Slavic the morphological division was
already *wilk-as rather than *wilka-s).
It's not unusual for the same phoneme to
develop differently in grammatical morphemes and in lexical stems.
Let me give you a few characteristic
examples from the documented history of English.
Example 1: In Old English all final
fricatives were voiceless. More recently, originally final [-s] became
voiced in inflectional endings, but not stem-finally or in derivational suffixes
(I ignore special cases like inflections following voiceless consonants). This
is why we have house, palace, kindness and careless with [-s], but horses,
John's, stones and moves (verb or noun) with [-z]. Interestingly, lexicalised
plural or genitival formations in which internal morphological structure
has been obscured through reanalysis, have [-s] (e.g. bodice, else), while
former singulars reanalysed as plurals have [-z] (gallows, bellows). We even
have pairs like once [-s] and one's [-z], both reflecting Old English a:nes,
Gen.sg. of 'one'.
Example 2: Old English initial fricatives
were voiceless as well. In Middle English, the dental fricative /T/ underwent
voicing to /D/ in function words (the, this, that, thou, there, then, thither,
thence, though, thus, etc.), but not in content words (three, thump, thwart,
thigh, thought, etc.).
Example 3: English finger [-Ng-] doesn't
rhyme with singer [-N-] (except locally in NW England), though both derive from
OE words with medial [-Ng-]. Of course it's the morphological boundary in singer
(sing#er) which makes all the difference
in Modern English.
In none of these cases would it make sense
to postulate two original phonemes rather than phonological changes sensitive to
the morphological environment. My claim is that the case of *-os in Slavic is as
clearcut as any of these English examples. Baltic has -as in masculines and
there's little doubt that the Balto-Slavic ending was just that. Then something
happened in pre-Slavic: most likely, in my opinion, the falling together of *-a
(< *-as < *-os) and *-u (< *-ux < *-us) in the Nom.sg. (after
the change *-aN > *uN both paradigms had the same Acc. endings). This
happened only in inflections, not in syllables that belonged to stems, which is
why *slawa(s) 'word' or *neba(s) 'heaven' were not affected.
Nom. Acc.
*-i : *-i
in i-stems
*-u : *-u in u-stems
*-a : *-u
in thematic masculines --> *-u : *-u
As in English (Example 1), one could expect
some thematic masculines to retain the original ending in fossilised
combinations, e.g. with an enclitic. This is precisely what we find in OCS:
rodosI '(this) family', rabotU '(that) servant' < *rada-si,
*arba-tu.
I wonder if there were enough Slavic
s-neuters to account for the generalisation of -o (< pre-Slavic *-a) in
thematic neuters (there are about 20 examples, and not all of them equally
good). The influence of *tod remains a serious possibility. At a certain point,
after the loss of final consonants, pre-Slavic had *ta 'it, that (n.)'
beside neuters like *slawa 'word' and *si:tu 'sieve'. Compare the
following:
sg.
pl.
t-a si:t-u t-a:
si:t-a:
t-a
slava- t-a: slawes-a:
The replacement of *si:t-u by *si:t-a
strengthened the formal symmetry between pronouns and thematic nouns
(already established for masculines and feminines!), but not between thematics
and s-neuters (which retained their *-es- before all case endings). It appears,
therefore, that the influence of the demonstrative pronoun was crucial in this
process.
Piotr
----- Original Message
-----
Sent: Thursday, November 23, 2000 12:44
AM
Subject: Re: [tied] IE *-su and the
Nostratic "equational" marker *-n :)
>These "extra explanations" have been provided by many
authors. The problem has always been of interest to Slavicists. Would you also
say that Slavic *-U (as in the Acc.sg.m.) and *-o (as in the Nom./Acc.sg.n.)
cannot both derive from IE *-om?
Obviously not. The o-stem neuters
in -o can only be explained (from
what I've seen) as deriving from *-od (the
ending taken over from the
pronoun *tod), or from the neuter s-stems
(nom/acc.sg. *-os). The
first solution cannot be ruled out, although it
lacks parallels (*-od
was nowhere else transferred to the neuter
o-stems). The second
solution again leads to the paradox found in the
masc. o-stems: how
can PIE *-os give OCS -U in the o-stems but -o in
the s-stems (c.q.,
by analogy, the neuter o-stems)? Before recurring to
morphological
conditioning factors, one should first investigate the
"null
hypothesis" that the two *-os were not the same phonetically. As
it
happens, there is at least one very good reason to think so: if
we
disregard the thematic nom.sg., the quality of the thematic vowel,
*e
or *o, can be predicted exactly by looking at the consonants
following
it: if it is voiceless (*-e-t (incl. *-e-nt), *-e-s, *e-h1, *-e#),
we
have *e, if it is voiced (*-o-m, *-o-d, *-o-i, *-o-u, *-o-bh), we
have
*o. The major exception is indeed the nom.sg. *-os (and
gen.sg.
*-osyo, if the first *o is a thematic vowel at all). Using
internal
reconstruction, we might thus want to reconstruct this as
**-oz. The
speculation becomes less boundless if we consider the
evidence from
Slavic that nom.sg. *-os is unlikely to have been the same as
the *-os
from the s-stems (with, as far as we know, real voiceless *s).
(But
the question I haven't cracked yet is: was the *-o- really the same
in
both?)