Re: 'Yemls Phonemes

On Tue, 20 Nov 2001 08:27:41 PST, Matthew Pearson
<Matthew.Pearson@...> wrote:

>--- You wrote:
>'Yemls has 4 vowels, distinguished for low vs. high and front vs. back, and
>12 consonants. The syllabary is laid out with a column for each vowel and a
>row for each consonant with an additional row for the null consonant.
>Currently, each vowel and consonant is described as a phoneme. However, in
>actual pronunciation a couple things occur. One is that how some of the
>consonants are pronounced depends on whether the vowel is front or back,
>e.g. /s/ is [s] before /u/ and /o/, but is [S] before /i/ and /e/. Another
>is that /u/ and /i/ are eliminated under certain conditions, so that the
>possibility of minimal pairs occurs distinguished only by [s] vs. [S].
>If I understand correctly, that means that these should be reanalyzed as
>separate phonemes (instead of both being /s/). Is that correct?
>--- end of quote ---
>
>Depends on how abstract you want your representation to be. If you want to
say that the /u/ and /i/ are there underlyingly, and are only deleted on
the surface, then you can get away with treating [s] and [S] as allophones
rather than separate phonemes. If high vowel deletion is a regular,
predicatable process, then I would say that would be the way to go--
especially if the high vowels are recoverable from context. However, if
high vowel deletion is not (fully) predictable, then things get trickier.
Especially if the high vowels are no longer recoverable from context once
they've been deleted, it might be more prudent to treat [s] and [S] as
distinct phonemes.
>
>Matt.
>
Hmmmm. While I haven't pinned down all the rules for vowel deletion, it's
safe to say that the vowel is always recoverable from context (Also, the
diphthongs resulting from CV + V combinations will always be predictable).
That means my current phoneme set is technically adequate. My real problem
is that I want to give a decent idea of how the language is pronounced
without having to notate aspiration, vowel rounding/unrounding, and other
features all the time. That is, phonetic notation is too precise except for
showing a specific utterance of a native speaker (which don't exist), while
phonemic notation isn't good enough.

Jeff