intervocalic DEvoicing can also happen / X > Y > X

[Alan R. King]

----------------------------Original message----------------------------
B. Wald says:
 
>It just happened -- because both directions
>are possible (under certain conditions -- certainly NOT ****z > s /V_V).
 
That provoked me to try and think of counterexamples to Benji's latter
assertion here.  I couldn't come up with an example of z > s /V_V, but I
did find one of +voice > -voice /V_V.  Furthermore, if I am right in
thinking that the "devoicing" is an innovation (and thus did happen) rather
than a survival (i.e. unless there was really no change; see the end), then
it provides as good an example of X>Y>X as what Benji called "Central
Spanish" s > z > s.
 
(Okay, I'm going to be a pain and say it: just what is "Central Spanish"
supposed to mean?  The phenomenon in question includes the WHOLE of
Castilian (i.e. Spanish) in its territorial extension.  Are you implicitly
calling Portuguese, Catalan and perhaps Basque "peripheral Spanish"??
Sorry, I couldn't hold that one back.  May I suggest that perhaps Benji
meant something like "Central Ibero-Romance"?)
 
What I have in mind is a dialectal phenomenon within modern Welsh.  First
the relevant background information:
 
In modern Welsh word stress is normally on the penultimate syllable, and
the phenomenon I will describe occurs in the consonant following the stress
when another syllable follows (as it usually does except in stressed
monosyllables, of course), i.e. normally at the boundary before the last
two syllables of a word, for example where the /d/ is in:
 
/'kadair/, /'kader/ or /'kadar/, orthographically <cadair>, 'chair' < Latin
"cat(h)edra" etc.
 
In this post-stress intervocalic position, in standard Welsh, the consonant
may be phonetically simple, as in <cadair> standardly pronounced ['ka:der]
(or in slow speech, ['ka:dajr]) or geminate, as in:
 
/'etto/ <eto> 'yet, again'
 
The distribution of simple and geminate consonants in this environment is
largely predictable (e.g. "voiced" stops are simple, "voiceless" stops are
geminate), as is the distribution of short and long vowels in the stressed
syllable (long preceding simple consonants, short preceding geminate ones),
for which reason the modern orthography reflects neither consonant length
nor vowel quantity in these cases.  But there are some minimal contrasts
too, e.g. <canu> 'sing' versus <cannu> 'whiten'.
 
(Caveat: I have put quotes around "voiced" and "voiceless" because
phonetically all stops in Welsh tend to be voiceless (or at least voicing
is not critical) and the contrast is realized principally in terms of
tenseness and aspiration.  To simplify the exposition I shall henceforth
largely ignore that fact in the transcriptions and terminology used.)
 
That is in STANDARD Welsh.  Now in Northern spoken Welsh, this system has
been altered, in that nearly all consonants in the intervocalic post-stress
position are pronounced geminate (and correspondingly, all stressed vowels
followed by a consonant in non-final syllables are short - but I'm going to
focus on the consonants here).  So corresponding to standard ['ka:dajr] we
will find ['kaddar] in northern Welsh, while ['etto] shows no change since
the consonant is already geminate.  Long vowels in northern Welsh are found
mainly in monosyllabic words (/ka:n/ 'song') and in penultimate syllables
when in hiatus with the following vowel (/'diod/ [di:od] 'drink');
otherwise all vowels are short.  In northern Welsh, then, <canu> and
<cannu> are both ['kannI].
 
Of these two systems, the standard and the northern, the standard one must
represent a diachronically prior stage, for a number of reasons that I
won't go into, but some of which are already fairly obvious.
 
Now: apart from this gemination and shortening, but possibly related to it,
we find some varieties of Welsh in which, corresponding to consonants like
the /d/ in <cadair>, if the consonants in question are "voiced" stops in
standard Welsh, we find "voiceless" stops instead: ['katar] etc.  Assuming
once again that the standard form is diachronically prior, we then have d >
t (or possibly d > dd > t) in post-stress intervocalic position, and
similarly b > p and g > k.  Irritatingly, this phenomenon has been noted in
the south, in an area not contiguous to the northern "geminating" varieties.
 
By the way, we might assume that the same varieties would have had z > s
intervocalically, if only Welsh possessed a /z/ phoneme, which it doesn't;
such is life!  Then again, maybe z > s wouldn't have happened, since the
voiced fricatives /v/ and /D/ do NOT undergo devoicing in this context.
Tant pis!  (Actually, these consonants also seem to resist gemination in
the "geminating" northern dialects, even though the preceding stressed
vowel is STILL short.  That's the reason why I said "nearly all" three
paragraphs back.)
 
As may have been noticed, however, the standard Welsh /d/ which in some
dialects has become /t/ (by the account I just gave) itself derives
ultimately from an earlier /t/.  In the evolution from proto-Celtic (or
Latin, in the case of loanwords) to modern Welsh, intervocalic voiceless
stops were regularly voiced, much as in western Romance languages.  The
modern intervocalic voiceless stops come from original geminate stops -
which explains their gemination in the modern language, as well as the lack
of it in the voiced counterparts in standard Welsh.  So in an earlier stage
of the development of Welsh we have approximately:
 
VttV > VttV
VtV > VdV
 
and similarly for /p/ and /k/.  In northern Welsh ['kaddar] we then have a
further step:
 
'VCV > 'VCCV (hence 'VdV > 'VddV)
 
while in the varieties that now pronounce ['katar] we presumably have a
later reversal of t > d:
 
'VdV > 'VtV
 
Unless, of course, 'VtV in these varieties is actually a SURVIVAL of
Celtic/Latin VtV.  Would anyone more expert than myself like to take up the
discussion from here?
 
NOTE: For anyone who wants to know the detailed diachronic background of
Welsh phonology (in general), I recommend the classic historical grammar by
Morris-Jones.