On 16 November 2012 09:40, J. 'Mach' Wust <[log in to unmask]>wrote:

> On Fri, 16 Nov 2012 00:19:41 +0100, Christophe Grandsire-Koevoets wrote:
> >[...] French people make a strong distinction between the
> >vowels /u/, /y/ and /i/, and the approximants /w/, /ɥ/ and /j/.
> Does ist? Of course, prescriptivist French calls for a difference between
> _loua_ [lu.a] and _loi_ [lwa]. But whether that is only prescriptivism or
> not, I don't know.

It's definitely prescriptivism in this case, and _loua_ can indeed be
pronounced [lwa] (although that'd be uncommon, given that it doesn't exist
in Spoken French at all :P ). But that doesn't change the fact that we can
*hear* the difference.

> Then again, my command of spoken French ist not that
> certain, either. I think, however, that the most convincing argument for
> distinguishing between half vowel phonemes /i/, /y/, /u/ and approximant
> phonemes /j/, /ɥ/, /w/ would be the existence of sequences such as /wu/ or
> /ji/ (such as in English _woo, yield_), because a distinction between
> [lu.a] and [lwa] can be analyzed as a morphological boundary effect
> (something like /lu.a/—/lua/) without the need of distinguishing two
> phonemes /u/—/w/.

I can't think of actual words that have those sequences (well, actually, I
can think of one: _yin_ in French is definitely pronounced [jin]). I can
tell you however that any French person is capable of pronouncing those
without a problem (barring speech impediments). A phonemic distinction
needn't happen in *actual* words to exist anyway. The idea that only those
distinctions happening in extant vocabulary can be considered phonemic has
always sounded preposterous to me.

> I think a case could be made for analyzing [wa] and [ɥi] as diphthongs
> since they pattern like simple vowels, and not like CV combinations. Both
> occur in simple morphemes after complex onsets such as [tʁ], e.g. _froid,
> fruit_. This gives yet another possibility of analyzing the [lu.a]—[lwa]
> difference without recurring to different /u/—/w/ phonemes: /lua/—/lu͡a/
> (the latter should show up as _ua_ with the tie bar).
I disagree. I just analyse those by saying that a complex onset can be of
the form CCA.
Christophe Grandsire-Koevoets.