David Peterson wrote:

> <<snip everything>>
> I think you'd run into some problems, no matter what the transcription
> system.  Because, essentially, you wouldn't need one.  All you'd need
> is to know the word.  In fact, you could do it with pictures.  You'd
> have to have a different picture for each language, if you were
> looking for a particular vowel, or you'd have to specify, "How you'd
> pronounce the English word for [picture]".  Even there, though, you'll
> run into trouble.  For example, I can envision a British English
> speaker coming up with this kind of a strategy for his/her conlang,
> and while trying to convey the vowel [Q] (low, back, rounded) in his
> dialect, he might say, "Pronounce this the same way as you would the
> vowel in 'car'."  For him, that's [Q].

'car' is [kA:], the same as American, but without an [r\].  [kQ] would
be thoroughly weird.

> For an American English speaker?  It might come out as something like
> [Ar\].  So a word like /kana/ could become [kAr\nAr\].  IMO, confusion
> *always* ensues if you bring natlang examples in to try to describe
> how a sound is pronounced (though it's sometimes useful).

[Ar\]  isn't a vowel, though.  The vowel is [A].  Now, we have [A:] vs
[Ar\].  However, there is no extra distinction, simply a different
phonetic realisation.  So, we can stick with [Ar\]=ar, say.  The only
real problem is that (southeastern) English English uses [A:] in three
places - 'car', 'father', and 'grass'(NOT 'cat' as some people insist).
These represent three different sounds in other ones:  [Ar\], [A:], and
[&], in most of them, only one of them being the same as British
English.  This means that we need to distinguish between the three.
We'll call the first 'ar', the second 'aa', and the third 'ae', along
with simple [&] - 'a'.  That means we've dealt with all the short 'a'
sounds.  There is the place with the most cros-dialectal problems,
actually.  I think everything else would be easy.  And 'kana' would be
[log in to unmask]