On the Indo-European Forum (e.g.
http://groups.yahoo.com/group/cybalist/message/14181 and
http://groups.yahoo.com/group/cybalist/message/14407 , Guto Rhys has
been asking about the forms that resisted Brythonic sound changes /s/
> /h/ and /st/ > /s/ in the context #_V. Before these changes,
Brythonic did not have /h/. The exceptional forms do not show any
immediate phonetic conditioning, though it is plausible that
dissimilation may have helped resist the change /s/ > /h/. A
possible complication is the influence of Latin while these sound
changes were occurring, though I am not sure that they did not
precede its arrival on the scene.

I find it hard to view the change of /s/ to /h/ as being a sporadic
change that occurred frequently enough to be viewed as the regular
development, for that would imply an unmotivated, unconditioned split
of one phoneme /s/ into two, /s/ and a new phoneme /h/. (I do not
believe this /h/ would have been identified with the allophone of /k/
before plosives, including geminates.)

I find it easier to contemplate a phonemic re-analysis whereby
initial /s/ before vowels was assigned to a different phoneme /h/
(perhaps phonetically [S]) from other occurrences, e.g. in initial
clusters /sp/, /st/ and /sk/. It then seems plausible that /st/
might then be realised as [s] or [st] in free variation, and then
variants in [s] might be internally lexicalised as beginning /s/
rather than /st/. This would be a sporadic change of initial /st/
to /s/ which ultimately spread through almost the entire lexicon.

It then seems quite plausible that once the change /st/ to /s/ was
underway, forms that preserved the original initial /s/ might be
borrowed from dialects that had not [yet] undergone the change /s/
to /h/, yielding exceptional forms such as Welsh <saith> 'seven',
instead of the regular *haith.

However, dialect mixture seems less plausible for the retention
of /st/, e.g. in Breton <sterenn> but Welsh <seren> 'star', or Welsh
<ystum>, Breton <stumm> 'bend'. This explanation claims that a
dialect that had initial /h/, /sp/, /sk/ and /s/ then added an
initial /st/ to its repertoire although rendering such a word
with /s/ would be unlikely to cause confusion.

I think it may be relevant that, in English at least, [st] acts like
a 'pseudo-phoneme'. Teachers have observed that English children
have to be explicitly taught, at least at the start of a word, that
<s> + <t> is /st/; it often (always?) failed to shorten a preceding
long vowel as Old English became Modern English; and 'aste' is the
only group in which a final 'e' lengthens a vowel preceded by two
consonants. (It is just possible the massive variation in the IE
partial reduplication of verbs in st- is relevant here. I don't
think there is such variation for plosive plus liquid, but there is
significant variation for 'laryngeal' plus resonant) A more remote
example of the combination's special nature is Classical Hebrew
s^tayim 'two (f.)', which is one of the very few Classical Hebrew
words to start with two consonants.

Is my description of the mechanisms for the sporadic failures of the
sound laws s- > h- and st- > s- plausible? Are there better
explanations, either in the case of Brythonic, or generally for such
phenomena?

I have read the discussions of the similar combination of the Satem
changes /k^/ > /c/ and /kW/ > /k/ on the Indo-European forum, but I
don't think I've overlooked any pertinent parallels from that case.

Richard.