On 2006-08-06 23:30, Sean Whalen wrote:
> In PIE the desiderative affix was -s. (and -z. after
> voiced obstruents). It is usually reconstructed as
> -h2s but no outcome of h2. > a is seen in any language
> but in Celtic and Italic. Why would didh3-h2s- >
> dits- in Skt not *diti:s.-? Why h3omh2.sos > óomos in
> Gk not *ómaos? If retroflex C's existed it makes more
> sense for 0>a between C and retroflex C in Celtic and
> Italic (sim. to Ir. retroflex-C y/w > i/u retroflex-C
> y/w word-initially). The lengthened V/C in Indo-Aryan
> can be explained by the differing effects of s. vs s
> on the preceding.
In handbooks, the desiderative suffix is usually reconstructed as
*-hs-/*-s-, with the *-s- variant found after obstruents. A popular
theory about its origin claims that the laryngeal (whose index is
usually left unspecified) was attracted into the suffix through the
reanalysis of roots with a final laryngeal (such as *telh2-, kWelh1-,
*gWerh3-, etc., with *telh2-s- > *tel-h2s-). This theory makes no sense
to me, since it still fails to explain why the laryngeal should be
invisible after obstruents if it was "analogically" added after roots
like *k^leu- or *kWer-.
I don't think we can avoid reconstruct the suffix as *-hs- (most likely
*-h1s-, quite possibly related to the verb *h1es-), assuming early but
hardly surprising simplification of *-Ch1s- to *-Cs- (where C = any
obstruent, including PIE laryngeals). Desideratives in *-(h1)s-e/o- are,
I believe, subjunctives of a once-athematic compound formation in
*-h1es-, where the first member had either a lengthened vowel of
reduplication, just as in ordinary thematic presents, also presumed to
be former subjunctives. The reduplicated type yields the Indo-Iranian
desideratives like Skt. cíki:rs.ati < *kWí-kWr.-h1s-e/o- 'wants to make'
and Old Irish futures like didma 'will tame' < *didma:se/o- <
*dí-dm.h2-(h1)s-e/o-.
piotr