Re: Reasons (was [tied] Re: Some thoughts...)

On 04-10-08 21:44, Sean Whalen wrote:

> Comparative? This is mainly an internal
> reconstruction.

It was in Saussure's time, when the "sonant coefficients" were used
mainly for explaining minor ablaut patterns, but now it's as comparative
as anything else: you can directly compare things like Anatolian aitches
with Greek prothetic vowels, Indo-Iranian aspiration, failure of
Brugmann's Law, Balto-Slavic intonations etc. It all amounts to good
comparative evidence.

> If complaints that PIE phonology seem
> artificial and not like any real language have helped
> create glottalic reconstructions I thought I'd go for
> the opposite approach (assuming more traditional
> reconstructions are nearly accurate, but incomplete or
> representing a short stage between other more natural
> stages). I made assumptions and tried to find if
> their consequences matched any realities.
>
> For example, assuming f at an early stage, what
> happened to it later? If f>s, does the treatment of s
> appear different in any two environments, and thus can
> be attributed to two underlying sounds? Since I had
> already decided -a'ms>-a'm_>-a:'m to account for
> n/m/s/etc.-stems' nom. sng. but thematic acc. pl. a'ms
> didn't change, probably pl. -f. This further was
> confirmed by -fi>-fu in loc. pl., not explainable by
> an unrounded C (as loc. sng. -(a)wi>-(a)wu).

You ignore the rather solid evidence (defended here eloquently by Jens
Rasmussen) that the nom.sg. *-s was once a voiced segment). The crucial
part of the demonstration is the behavious of the thematic vowel, with
its *-e/o- allomorphy clearly governed by the phonation of the following
segment. If thing are so, the difference was between *z and *s at least
at the time when the *-e/o- rule was operative.

> Similarly, if palatal(ized) nj existed, what did it
> change into? Probably either n or y. Two of the most
> common V affixes are n and y, could they be the same,
> and in what environment would n become either? Also,
> suyus/sunus. I determined laws that work to do
> describe this.

Well, there are arguments for the original identity of some instances of
alternating *n and *r, or some instances of *s and *t, or *n and *t, or
even *h1 and *t -- all of them still tentative at this stage. One
problem for your account is that there is no functional relationship
between *-jo- and *-no- (as opposed e.g. to *-nó- and *-tó-, which both
form quasi-participles, or to *-r/-n- in the heteroclita). I'd also like
to see your detailed explanation of the 'son' words before I believe
that they are phonologically conditioned variants of the same underlying
protoform (rather than related words with different derivational histories).

>>The fact that *h3
>>seems to have had a voicing effect on any preceding
>>stop suggests that
>>it was voiced itself... etc.

> Since all the fricatives are voiceless, it's not
> unlikely one or all had voiced allophones. If not
> before, they were almost certainly voiced when they
> became syllabic.

The whole point about *h3 is that it seems to have been _distinctively_
voiced. Otherwise the voicing would not have any assimilatory effect, as
visible in *h2ap-h3on- > *abon- or *pi-ph3-e-ti > *pibeti. Judging from
their phonetic effects, the three laryngeals were something like *h, *x,
and *G (the latter two in the velar-uvular-pharyngeal range) rather than
anything parallel to *k^, *k, *kW. Even the O-colour of *h3 was not
necessarily a consequence of lip-rounding: note that *kW, *gW and *gWH
don't colour any *e's in PIE!

> I think o/e/0 shifts make it almost certain there
> was one vowel in the past that changed sound for some
> reason, tone seems to do it.

Here I suppose most people (including myself) will agree with you in
principle. The question is if we have any right to reconstruct
distinctive tone values for PIE just in order to account for the basic
ablaut of *e : *o. If we do, we don't really explain anything -- it's a
case of "obscurum per obscurius". If you write that nouns were mostly
marked with one of the tones and verbs with the other, you're only
restating what everyone knows -- that nouns often show the o-grade and
verbs often show the e-grade. I can't see any simplification in your
model. It trades off one kind of complexity against another.

> Also, they're not
> pre-encoded, there are rules to delete a tone
> immediately followed by an affix with a tone. This
> allows one underlying form for each noun, verb, etc.,
> with various surface forms.

That doesn't change the fact that you ascribe the fundamental contrast
between *e and *o to an underlying tonal contrast without an independent
motivation.

> See some of the derivations I sent for the
> different cases of "sister" and "dog". If swa`fa'r-
> is the underlying stem, gen. affix -a's deletes tone
> to get swa`far- and the toneless a will be affected by
> later a-deletion rules. When simplification causes
> only one tone per word: swe'sros. The tone of the
> final syllable is the only tone for kuxva'n-, so:
> kuxvno's. By analogy with words with only final tone
> (common), morphemes like -su'/-su (and all tone-marked
> plural affixes) are simplified to their most common
> form, thus swesrsu' instead of swe'srsu.

The same forms are adequately accounted for in some derivational models
(discussed here more than once in the past) that only have recourse to
independently required lexical accent (surfacing as stress) and don't
employ primary tonal contrasts. They are superior to your tonal solution
by being more parsimonious.

> Also, I don't rely on pre-marking to account for
> e/o variation in pHyo:, pHyeis, pHyei, pHyomen, etc.

You mean the thematic vowel? I should think its ablaut has already been
explained ;-)

> So, in Greek there was definitely analogy to change
> *tetHo:- to *tethe:-, but no possibility that any
> analogy would create o:/e: on model of o/e?

If anything, analogy is likely to level out paradigmatically related
forms, leaving isolated derivatives untouched. What would <tho:mos> be
analogous to? But in the case of the verb <titHe:mi> (aor. <etHe:ka>,
etc.) there was surely enough ground for analogical levelling in favour
of <e:>. In general, the Greek perfects of *CEH- and *HeC- verbs
(because of their erratic phonology after the loss of the laryngeals)
were subject to so much innovative restructuring that it would be naive
to derive them directly from PIE protoforms. The more irregular
alternations we get from phonological development, the more likely is
the elimination of residual alternations. See what happened to Vernerian
alternations in Germanic strong-verb conjugation (while non-paradigmatic
effects of Verner's Law have survived unscathed).

> What's
> the explanation for fe:ci:? Do you mean the perfect
> didn't always have o?

Do you mean Latin perfects simply continue PIE perfects?

> Even if there was lack of coloring in certain
> environments, that isn't just a consequence of my
> theory (*-oh2 but *-ah2a, turned to Greek -o: and -e:
> (then reanalyzed and becoming -ome:)).

Apart from the questionable derivation of *-o: from PIE *-o-h2 -- as far
as I can see, only underlying //e// could be coloured to *a by an
adjacent *h2, which means that the A-colouring effect is younger than
qualitative ablaut.

> No, str- doesn't need to be a special case; sr- is.
> Even if str- and sr- once existed side by side (since
> a-deletion could turn stara'- to stra'- it's hard for
> me to determine) there would be a time with no sr-.
>
> Since sr->str- (and only this, not sn->stn-, etc.)
> there is no sr- in PIE until f>s (so, fr>sr) and
> probably some cases of sara'>sra' will create sr-,
> too, unless sr->str- is a lasting rule and still in
> effect.

Other than your desire to find a job for *f, is there any independent
argument against PIE *sr being simply what it seems to be? You could
make your case by showing that some extended derivatives of a root
reconstructible as *ser(X)- have initial *str- already in PIE (and not,
say, in Germanic, Slavic and Thracian, as *sreu- does, for example); but
that doesn't seem to be the case. Alas, you have just immunised your
theory against this kind of verification with the "probably some cases"
clause above, which could be used to disqualify any counterevidence as
an exception (if not, circularly, as an instance of *f).

Piotr