>#1 says: "The changes go from less sonorant to more sonorant." As it
>stands, it's demonstrably false. Sonorisation is common enough to be
>expected only in some contexts (e.g. in intervocalical consonants, where
>it can be regarded as a kind of assimilation and so falls under #0,
>making #1 redundant!); it's reverse is common enough to be regarded as
>natural in other contexts (e.g. various forms of initial or
>stress-conditioned fortition, as in Romance, Hindi [not to mention
>English], word-final devoicing in hundreds of languages, etc.).
>Likewise, deletion is more common than epenthesis, but both _may_
>happen, as Richard has pointed out.
Isn't it so that it _must_ happen?
I would think the true "axioms" of sound change are:
1) there are no "sinks"
2) there are no "sources"
Every phoneme /x/ must have at least one out-transition (/x/ -> /non-x/)
and at least one in-transition (/non-x/ -> /x/). Otherwise, the sources
would have long ceased to be possible phonemes, and sound-change would have
stopped altogether as soon as all phonemes had been attracted to the sinks.
Now, since we know that for the null phoneme, sound changes /x/ -> /0/
exist, there must also be transitions /0/ -> /x/ (something from nothing).
To be sure, epenthesis is not the main mechanism (which I think is word
composition, operating at the lexical/semantic level, above the
Miguel Carrasquer Vidal mcv@...