This reminds me of the Middle English "open
syllable" lengthening (MEOSL). Some time ago Domka Minkova showed that in the
overwhelming majority of cases the lengthening (of [o, a, e]) operated in the
context _C&# (where & = <-e> = schwa), and so could be regarded as
compensatory (the final schwas were weakened and dropped during the Middle
English period). Minkova's explanation predicts that no lengthening should be
expected in words like <shadow>, <body> or <many> (so far
so good) or in _C&R# words, where R = liquid or nasal (since they also have
remained disyllabic). In the latter set, however, we have a good deal of
unconditioned variation, with no length in <kettle>, <throttle>,
<saddle>, <seven> etc. (the modern orthographic geminates are
unetymological), but with length in <raven>, <maple>, <even>,
<over>, <beaver>, <ladle> and a few other words. There is also
historical evidence for long/short variation in a few items (e.g.
<heaven>) in Early Modern English. Now these exceptions are really
problematic for Minkova, since compensatory lengthening (no matter how elegantly
it explains the majority of cases) is ruled out here, and the small
leak can sink Minkova's big ship.
Since the matter was first discussed it has
been pointed out that lengthening very often affects Middle English vowels in
the context _C# (as in <whale> [OE hwæl], <yoke> [OE geoc],
<coal>). This length had been routinely explained as analogical by all
auctores since karl Luick (after the dative <hwale>, etc., where MEOSL
would have applied); this however is another "invincible" (read: unfalsifiable)
explanation that could in principle explain any quantity anywhere but fails to
make the simplest predictions: we still don't know why there is a long vowel in
<coal>, but a short one in <god>. We might just as well (or maybe
better) assume tendential or _sporadic_ lengthening in monosyllabic words before
a single consonant. To sum up, this is how ME words behave (I exclude the
special case of words with "voiced homorganic clusters" like <child> or
<comb>):
CVC etymological length preserved,
very frequent lengthening of [o, a, e]
CVC& length preserved, regular
lengthening of [o, a, e] (dialectally also of [i, u])
CVC&R variable quantity: sporadic
lengthening and shortening (<devil>, <weapon>)
CVCi(j) etymological shortness
preserved, sporadic shortening (<sorry>)
More complex templates (CVCC&,
CVCVC&, etc.) regular shortening, no lengthening
Here is an example of a quantitative change
taking place during the literate phase of the most carefully studied language on
earth. It is clear that the change is prosodically motivated -- factors
such as foot complexity influence the length of the stress vowel. There is
one context (_C&#) in which the change displays Neogrammarian regularity,
buth there are also (structurally similar) contexts (_C#, _C&R#) where
its operation is capricious (rather than abruptly blocked) as if showing the
effects of random lexical diffusion. No subregularities have been discovered so
far and I doubt if there is anything to discover there.
Perhaps we cannot expect more of Winter's
Law.
Piotr
----- Original Message -----
Sent: Monday, January 28, 2002 1:54 AM
Subject: Re: [tied] Slavic voda
... I have no idea at the present moment what subrule should
be added to the conditioning, if any ...