--- In phoNet@yahoogroups.com, "Jean-Paul G. POTET" <potetjp@w...>
> "For example /b/ is a phoneme in English. It is realised as [b] in
> <scribe>, but as [p] in <*scription> of <subscription>" Jean-Paul
> "I don't think so. You're importing a Latin rule neutralising /p/
> v. /b/ into English. English gets close with the "b" (I think
> better than {b}, if we have to avoid <b> lest it embolden the
> following text.) However, the "p" is definitely pronounced
> differently to the "b", so I don't agree. The best counterexample
> can think of is surnames like 'Hobson' and 'Gibson', to compare
> with 'gypsum'." Richard WORDINGHAM, ENGLAND
> No. It's not a rule that concerns romance languages, but,
apparently all
> languages. When a voiced consonant is followed by an unvoiced
consonant the
> universal reaction is to devoice the former.

Not so. Lachmann's law in Latin, that the vowel before a voiced
consonant is lengthened before the -t- of the past paticiple,
requires that such assimilation hasd been suspended. This law is
not found outside Italic. There was a long discussion of this
starting at http://groups.yahoo.com/group/cybalist/message/14125 .

Another counterexample is _progressive_ assimilation in Sanskrit, at
least for voiced breathy + voiceless combinations. I gave my
summary understanding of what was going on at
http://groups.yahoo.com/group/cybalist/message/14173 .

Finally, there is the infamous English word 'Aztec'. The opposition
of /zt/ and /st/ remains.