On 2007-06-06 10:51, tgpedersen wrote:
> How would you state the rule so that the lack of spirantization before
> stops is not an exception to Grimm?
It isn't the lack of spirantisation before stops but the failure of the
*t > *T part of GL to apply AFTER obstruents (so that the second element
of the clusters *sp, *st, *sk{^/W}, *pt, *k{^/W}t is unaffected (the
treatment of *-t-t- is special in its own way). A phonotactic filter
blocking the operation of GL is all that is needed (or an appropriate
ranking of phonotactic constraints in a model that uses constraints
rather than rules such as optimality theory). Proto-Germanic did not
tolerate clusters of fricatives (except for those with *s as the second
element, but *s is notorious for occurring more freely than other
fricatives). A rule-only solution (if you insist on one) is also
possible, but a bit more costly. It would require GL to operate without
exception and then another rule (yrt unnamed, so let's call it Fricative
Dissimilation) to change non-sibilant fricatives back into stops after
other fricatives:
*sp > *sf > *sp
*pt > *fT > *ft
etc.
Expensive, but 100% regular. Something of the sort happened in Old
English, where, in historical times, clusters such as -xs- and -fs- were
"hardened" into -ks-, -ps- (as in <fox> etc.)
Piotr