On 2007-06-07 16:41, tgpedersen wrote:
>> No, "classical" GL is not a rule with an exception. No linguists
>> uses the word "exception" in this way.
>
> Aha. So whoever calls it an exception is not a linguist.
The term "exception" is reserved for isolated words that don't seem to
obey any regularities. In other words an exception in this sense is "a
case that does not conform to a rule or generalization":
http://www.bartleby.com/61/27/E0262700.html
Here is an example: Middle English /O:/ has become a diphthong ([oU],
[&U] or something similar) in mainstream accents of English (including
RP), in words such as <bone, home, boat, load, loaf, oats, oak, moan,
stone, road, rode, drove, old, bold, only>. The diphthongisation was
suppressed (blocked or reversed) before /r/, as in <oar, boar, lore,
roar, lord, more, sore, hoar, glory>. This lexical set represents not an
exception (they all conform to a simple generalisation) but a
subregularity. _Exceptions_ are quirky, unpredictable cases like <one,
once, sorry, gone, shone, cloth, broad>, which have the "wrong" vowel
instead of the one expected in a given context.
>> The _blocking environment_ is part of the original formulation of
>> GL, in itself fully regular, so the outcome of GL remains
>> predictable.
>
> That's right, GL has two in-built exceptions.
GL includes special clauses referring to lexical sets containing certain
clusters:
(1) *tt > *ss
(2) *t remains unchanged after PIE *p, *k(W), *s
ELSEWHERE,
(3) *p, *t, *k(W) > *f, *T, *x(W)
A true exception would be a word in which, e.g. PIE *t > PGmc. *t
word-initially, or, say, *pt > *st. Actually, it was once believed that
many unpredictable deviations from the pattern of Grimm's Law should be
admitted, but since the time when those deviations were shown (e.g. by
Verner) to follow regular patterns of their own, they have been known as
"apparent exceptions".
Piotr