On Wed, 17 Jul 2002, Miguel Carrasquer wrote:
> >
[Jens:]
>I believe the underlying forms are here based on nothing
> but wishful >thinking. I see no evidence that a:-stems (made from
> thematic stems) and >"i:-stems" (made from athematic stems) are
> identical i underlying >structure.
>
[Miguel:]
> In my view, they have. The feminine suffix was **-ih2. The pertinent
> soundlaws are:
>
> 1) [stressed]
> **í > in open syllable: *yé
> > in closed syllable: *íC1#, *íC1C2 (at least for C1 = *h2)
[JER:] But accented í stays that way; it is the stem part of the vrki:s
inflection; it is accented in Caland forms of the type Skt. rjí-s'van-,
Gk. argí-pous where the compositional variant of -ró- is -í-.
> 2) [unstressed]
> **i > *y
[JER:] OK
>
> 3) ["svarita-lengthened"]
> **i: > *ye:
[JER:] I know of no evidence. The type ma:terie:s must have the suffix
*-yeH2- with regular lengthening before the nominative marker, this
yielding IE *-ye:H2-s with absence of coloration. I agree with Mayrhofer
that the reason it did not become **-yek-s (or *-yak-s) is that laryngeal
hardening (before the nominative sibilant) did not work after a long
vowel.
>
> 4) [absorption by thematic vowel]
> **%'i > *% > *e ~ *o, according to the voicing of the next segment.
[JER:] Assuming % means e/o, again I see no evidence. Are you talking
about *-e/o- + *-iH2- yielding *-e-H2- (> long a-stems)? Other y's do not
do this, thus the postthematic form of the optative whose full-grade form
is /-yeH1-/, i.e. also /-yeH-/, is *-o-yH1- + endings. The ntr.du. og
o-stems is *-o-yH1. There is very little support from within the language
itself.
> For athematic nouns, we have proterodynamic (de:vi:-type):
>
> N.sg.
> A) <light root> : **'-i:h2 > *'-ye:h2
> B) <heavy root> : **'-ih2 > *'-yh2
>
> cf. Gothic band-i (heavy root), sun-ja (light root), Lith. -e: and
> Lat. -e:s feminines. Elsewhere unstressed *-ye:h2 was reduced to
> *-yeh2 and further to *-ih2 > *-i: (i.e. the same development as for
> stressed *-íh2) or *-ja: (*a:/*ja: as *o/*jo).
[JER:] A rule differentiating posttonic /-iH2/ into *-i:H2 and *-iH2
depending on the complexity of the preceding root segment is without any
foundation that I know of. In my reading, bandi/sunja reflects Sievers (in
pre-Germanic terms, i.e. *bhondh-ia:- vs. *sn.t-yá:-, processed before
syllabic n became /un/): The old nom. in Gmc. *-i: was retained in bandi
because it was supported by i-forms in the rest of the paradigm, while the
inflection of tha short type offered no such support. After the levelling,
the words begin with /bandi-/ or /sunj-/ throughout their paradigms. It
hurts a bit to say so, but most examples really do not support Pedersen's
analysis of Lith. -e: as covering *-eH- of any kind. Even alu-de:
'Bierfass', allegedly from *-dheH1-, could just as well be *-dhH1-iaH2.
Still z^va~ke. : face:s and la~pe: : volpe:s look good. But Lith. z^e~me:
'earth' on which Pedersen based his type, lends itself to the unforced
analysis as "the low one", being to z^e~mas 'low' what se~ne: 'old woman'
is to se~nas 'old'; and since se~nis 'old man' is definitely a(n) *-(i)yo-
stem, the corresponding feminine should be derived from a stem in
*-(i)ya:-. The nom.sg. of Gk. fem. u-stems, -eîa reflecting *-ew-iH2,
lends very poor support for your rule.
> For both types, oblique **-íh2-a:s > *-yáh2os > *-ya:s, etc.
>
> Hysterodynamic (vrkí:s-type):
> N.sg. **-íh2-z > **-yéh2-s > *-íh2s
> Obl. **-ih2-ás > *-yh2ás > *-(y)yás
[JER:] But the general principle of IE ablaut is the opposite of this:
Vowels are *retained* when accented, not inserted in accented position
(except by analogy, this explaining most Schwebeablaut cases). Therefore
also the alleged distribution, posttonic *-ye:H2 vs. accented *-íH2, is
against what little solid evidence the language itself puts at our
disposal for the analysis of its phonological processes.
>
> Thematic:
> N.sg. **-%'-ih2 > **-%'h2 > *-eh2 > *-a:
> but V. **'-%-ih2 > *-oi(h2)
>
> Oblique:
> A) **-%'-ih2-a:s > *-éh2os > *-á:s
> B) **-%-íh2-a:s > *-oyéh2os > *-oyá:s
> C) **-%-ih2-ás > *-oyh2ás > *-oyyás
>
> Type (B) underlies the Sanskrit oblique forms:
> G. **-oya:s > -a:ya:s
> DL. **-oya:i > -a:yai (L. -a:ya:m)
>
> Type (C) is found mainly in the end-stressed instrumental (Arm. also
> dative?):
> Sanskrit:
> I. **-o-yh2-át > *-oyh2áh1 > *-oyya: > -aya: (no Brugmann lengthening)
> Slavic:
> I. **-oyh2áh1 + -m > *-oj(j)a:m > -ojo~
> Armenian:
> fem. obl. *-oyyV(:)(C) > -oj^
>
> Elsewhere we have type (A).
[JER:] You are inventing rules and types of PIE just to suit a single
language, and only a single inflectional type of that language. That falls
flat on its face if the -y-'s of the IIr. a:-inflection are regarded as
analogical, i.e. as part of endings that were adjusted to the other
feminine type, which had gen. in -ya:s, dat. in -yai, instr. in -iya:. If
you add that to -aH- (or *-a:-) you get precisely -a:ya:s, -a:yai, -aya:.
The Arm. form in -oj^ is locative, commonly explained as a sandhi variant
of the locative particle *-dhi; it does not turn up in the inflection of
old a-stems, but has its place in the "ea"-stems which are properly old
neuter io-stems.
> [Jens:]
> >Give me a rule that explains the coming and going of -p- in IE,
> >not Sanskrit alone, and I'll take it under advisement.
>
[Miguel:]
> The causative suffix is in origin the verb *ey-e-ti (Hitt. iyami "ich
> mache", píjami "schicke hin", uijami "sicke her"). I don't know why
> Sanskrit chose the variant with preverb *p(e)- to make the causative
> of roots ending in a laryngeal (and a few others).
[JER:] But 'make; send' must be the verb corresponding to Gk. hí:e:mi,
i.e. IE *H1yeH1-. Now, using that to explain a /y/ of a causative suffix
as in *mon-éye-ti just reak havoc in the elucidation of PIE. This looks
like something Bopp could have written before he found any phonetic
rules.
> [Jens:]
> >Greek épion is no
> >more secondary than Greek aorists at large, for the 3pl would have been
> >*pH3i-ént with that structure in any case.
>
[Miguel:]
> The point is that é in *pH3i-ént is not the thematic vowel, which I
> believe would have absorbed the *y.
[JER:] the type is found in, e.g., Skt. syáti 'binds', dyáti 'binds',
chyáti 'cuts', all without that absorbtion. Do you deny the existence of
io-stems? They must be very embarrassing to you, for they look very much
like thematic derivatives from thematic stems, i.e. "-o- + -o-" > -io-,
accented *-ío-, as in Ved. mitrá- => mitría-. That does not support either
absorption or a change of í to yé very well. It does support, on the other
hand, the time-honoured rule of reduction of the thematic vowel to /i/
when not accented, followed by my rule of initial accent: *meytló- + -ó- >
*mytlió- > *mytlío- > *mitlío-, the last steps of which are younger than
the Schwundablaut that put the unaccented root into the zero-grade.
Jens