PIE ablaut & Renfrew's Celtic Scenario

[Carol F. Justus]

>cjustus at mail.utexas.edu (Carol F. Justus) wrote:

>>The Germanic forms that clearly go back to an IE 'perfect' are the
>>preterite-presents like Gothic wait, witum corresponding to Greek oi~da,
>>i'dmen. The original vowel alternation within a single verb paradigm would
>>seem to be that between the singular and plural as in Gothic wait, witum.
>>In Hittite this vowel alternation is exemplified by verbs like kuenzi,
>>kunnanzi 'strike' or eszi, asanzi 'be' (full grade, zero grade) where the
>>third person plural is zero grade as opposed to other full grade persons.
>>Hittite does not have the e:o alternation, and Germanic too probably did
>>not originally.

>Hittite does have e/a alternation in some verbs of the
>hi-conjugation, with <e> in the plural, <a> in the singular.
>E.g. sakhi, sakti, sakki, [*sekweni], sekteni (sakteni),
>[*sekkanzi] "to know" (similarly in the past tense).

>Miguel Carrasquer Vidal
>mcv at wxs.nl

Quite right. It was the second person plural which had both forms (See the
1981 study: Nr. 7, Lfg. 10 of Kammenhuber's Materialen zu einem
hethitischen Thesaurus which includes all forms attested in texts known). A
major reason for doing exhaustive studies of Hittite 'know' and other verbs
such ak(k)-/ek(k)- 'eat' was to explore the ablaut types in Hittite that
differed from e.g., es-/as- 'be'. At that time I noted (pp. 14-16 on
etymology) just a few etymological relations. The major one was the
comparison with the singular:plural alternation of the old perfect (véda:
vidmá, wait:witum etc.) suggesting that Hittite 'know' might be
grammatically cognate, albeit not formally cognate, with the IE pattern.

An oddity which looked even older was the one with Latin sum, sumus, sunt:
es, est, estis that had a parallel in Hittite uhhi, umeni, uwanzi : auti,
auszi, aummeni, autteni 'see'. Then (as now) there were phonological
explanations for the 'sum' set, but the fact that Hittite 'see' showed such
a similar patterning raised questions in my mind as to whether the
phonological explanation of the Latin forms might not be ad hoc, that these
two (independent?) groupings of a 1sg, 1pl, and 3pl together with zero
grade-like forms as opposed to 2sg, 3sg, and 2pl full grade forms might not
reflect something almost totally lost to us.

In that context the sg:pl alternation looked like a reanalysis or leveling
of an older pattern on a basis (sg vs. pl) that we still understand.
Hittite sak(k)-/sek(k)- forms from texts then known are (1981:4-13):

Prs. Indic.
(variants are not noted unless the vocalism differed)
1s saggahhi,           Imperative / voluntative seggallu
2s sakti, sekti        Imper. sak(i)
3s sakki               Imper. sakdu, sakku
1p sekkueni
2p sakteni, sekteni
3p sek(k)anzi

Preterite Indicative
1s saggahhun
2s sakta
3s sakkis, sekta, sakta
1p sekkuen
2p sakten               Imperative sekten,[s]akten
3p sakkis, sekta, sakta Imperative sekkandu

Participle
SG sekkan-   PL  sakkanta (acc. neuter), sekkandus, sakkandus (acc. common)

The vowel alternation for 'know' ( sak(k)-/sek(k)-) is just about the
opposite of 'be' (es-/as-), barring, of course, the interesting 2pl
variants and the reversal with the participle. From this vantage point, and
the fact that Hittite third plural often contrasts in vowel variant with
all the other persons (esmi, essi, eszi, esmeni, esteni, asanzi and with
kuen- / kunnanzi etc.), when we see that the Greek, Sanskrit, and German
perfect has a regular sg:pl alternation (veda, vettha, veda vs. vidma,
vida, vidur), it looks like a sg:pl leveling, but an old one that got
shared with Germanic (of course we only have one pl form in Germanic, as
Gothic witum is 1p, 2p, and 3p).