I was curious which arguments and examples Cowgill gives
exactly in favour of the proposed soundlaw -eti > -ei, as in
the 3rd. person sg. of the thematic verb, (Warren Cowgill,
"The personal endings of thematic verbs in Indo-European",
in: Grammatische Kategorien, Funktion und Geschichte,
Wiresbaden, 1985).
The exact formulation of the soundlaw is:
"voiceless (? relevant material for *d actually seems to be
lacking) dental stops, aspirated or not, preceded by a
short, unaccented, non-high, non-nasal vowel, and followed
by word-final -i disappeared in the prehistory of all Greek
dialects early enough that the resulting hiatus had
contracted already by Mycenaean times"
This explains the 3sg. -ei of the thematic verb (provided
the Greek retraction of the stress in verbal forms preceeded
the proposed soundlaw).
Most of the space devoted in the article to discussing the
soundlaw is concerned with the counterexamples, which result
in the string of restrictions in the formulation above:
- the vowel must be short because of athematic forms like
(Doric) hista:ti, dido:ti, tithe:ti.
- the vowel must be unaccented, presumably because of the
numerous words like thésis, básis, etc., which have -V'ti (>
-V'si) [the point is not discussed by Cowgill, probably
because it's too obvious]
- the vowel must not be high because of pérusi "last year"
- the vowel must not be nasal because of eíkosi/wí:kati "20"
Other counterxamples are explained as analogical:
- imperatives like tétlathi, héstathi
- locatives like oíkothi
Taht leaves (apart from the thematic 3rd. person) only three
positive examples to support the soundlaw, teh first of
which I do not understand:
- Epic 3sg. thematic conjunctive -e:isi. Like 1sg. Epic
-o:mi, this is a conjunctive form apparently reinforced by
an athematic ending (-si, -mi). We would expect *-e:ti with
retention of /t/ after a long vowel, and perhaps Cowgill is
right that -e:isi continues this *-e:ti (> *-e:si), but I
fail to see how this supports a development -eti > -ei.
- The imperative dídoi, found in Pindar, and in early
Boeotian and Corinthian inscriptions, proposedly from
*dídothi
- The conjunction kai "and", besides Cypriot kas, from
unstressed *kati and stressed *káti.
This last example is similar to a number of other words
which Cowgill doesn't mention: protì ~ prós, potì ~ pos,
where the form with -s is explained as sandhi variant before
vowel *p(r)oty > p(r)os. The loss of -t- is unique to
<kai>, but that makes me suspicious as the reality of a
soundlaw here: in taht case we would also expect *proi and
*poi. As it is, the thing depends on the etymology of <kai>
and its relation to Cypriot kas, ka(t), a question which is
uncertain.
In the case of dídoi, I think Cowgill may be right in
deriving it from *dídodhi, but that hardly constitutes solid
evidence for the proposed soundlaw. It might just be the
elimiation of a dental stop by dissimilation in this
particular dental-stop-rich word form.
In sum, I see no convincing argument in favour of a phonetic
development -eti > -ei. The solution must be found
elsewhere, and in my opinion that is in the phenomenon
Cowgill discusses elsewhwre in the same volume in relation
to Old Irish, where the present tematic third person is
reconstructed as *-et (without -i), a phenomenon that is
also found in Balto-Slavic, and perhaps also in Greek.
=======================
Miguel Carrasquer Vidal
mcv@...