[_delirium]

Ah yeah, that makes sense to me. Even if you treat 'd' as a kind of modifier in French (it turns the 'j' from a sibilant into an affricative) it's clearly still having an effect on the pronunciation. I was somewhat objecting to whether the the 'd' itself is being pronounced, but that does get pretty hair-splitting, especially since /d/ is a stop. The transcription /dʒ/ sure makes it look like it's pronounced, but if IPA used a single character to transcribe the affricative, the 'd' would look like it disappeared in French too. But, granted, rather than having no effect (as in English), in French it'd still modulate the pronunciation of the 'j' even in that analysis.

Incidentally, I wonder if there are any languages where /d͡ʒ/ != /dʒ/ or if it's just impossible to produce that sequence.

[jtauber]

Polish apparently contrasts /t͡ʂ/ and /tʂ/

See http://en.wikipedia.org/wiki/Affricate_consonant#Affricates_...

[jtauber]

I'm trying to think of other modifier examples. Not a perfect analogy but I don't think you'd say that, in English, the <h> in <sh> is silent.

I wonder about Italian, though. Would it be correct to say the <g> is silent in <gli>?

Seems a much tougher call in Italian. I would say /d͡ʒ/ has a much closer relationship to /dʒ/ than /ʎ/ does to /gli/ for example.

[jtauber]

Also /d/ + /ʒ/ does become /d͡ʒ/ in English (and potentially all languages that have /d/ and /ʒ/) so even if you had a single symbol for /d͡ʒ/ that was unrelated to that for /d/, the relationship between /dʒ/ and /d͡ʒ/ is still close.