Intervocal and final are probably not the same date. Final -s shows up in
the ogam inscc., but I don't think intervocalic ever does. As far as
relative dating, hence inferential absolute dating, perhaps, a very
similar treatment of intervocal simplex -s- occurs in Brittonic. The river
name Tay appears as Taua within the Roman period, and that would go back
to *Tausa (I'm not sure whether that is attested or not in Tacitus,
Ptolemy, or some such place). Often the peculiar weakening of s is
regarded as uniquely Insular Celtic, but there are a few likely cases in
Gaulish: indas < *sinda:s in the Larzac tablet (c. AD 90) and SVIOREBE (if
from IE *swesor- 'sister') on an earlier isnc. on stone. There may be a
few more. The tendency must be quite early, at least. In theory, you might
have s as a phoneme alternating with h (to sometimes nil) in external
sandhi. Internally, though, once it was gone throughout a given paradigm
(that is, where it was intervocal in all forms) the element would change
phonemically. In Brittonic, suffixes, like the superlative, retain an
initial h- from s-; so morpheme boundaries seem to be a factor as well.
With part (b) of the question, I'd expect , e.g., *ti:reas or *ti:reah,
before total loss of the final. I doubt that Patrick ever heard
*ti:resos, even from the oldest druid in Ulster.
ta' bro'n orm ach ni'l cuidu' mo'r sa fhreagair seo.
J T Koch
On Fri, 9 Jan 1998, Richard Coates wrote:
> Dear friends,
> Would anyone care to suggest an authoritative absolute date-range for
> (a) the loss of sibilance, (b) the complete loss of Goidelic
> intervocalic and final /s/? Specifically, how late is it likely that
> there was any phonetic sibilance at all in reflexes of the _s_-stem
> genitive singular desinence _-esos_?
> I realize I am probably asking for the moon, but maybe someone has it
> to give.
> Bui'ochas mo'r,
> Richard Coates
> University of Sussex