Re: Allophone Problem

Joseph Fatula wrote:
> >   - [nef] > [neva]
> >   - [niv] > [niva]
> >
> > Among words with the "-a" suffix, this [e] vs. [i] distinction is the
> > only thing showing the difference between words like [neva] and
> > [niva].  Are these minimal pairs?  Are [e] and [i] separate phonemes?
> >
>
> Looking at everybody's responses so far, it seems as though most of you
> think these are two separate phonemes.  Disagreement seems to be
> centered around the idea of distinct underlying phonemes that happen to
> have the same phonetic realization.  I'm not sure whether I like this
> idea or not.
This is the old Bi-uniqueness Problem of classical Phonemics:  1) given a
phonemic form, the phonetic form must be predictable, and 2) given a
phonetic form, the phonemic form must be predictable. Thus the problem with
e.g. German [bUnt] =? bunt 'colored' or Bund 'association'; and oddities
like Engl. house ~houses [haUz@z], vs. most (all?) other words with a /-s/
that doesn't voice in the plural. Same for /-f ~ -v-/ -- of course the
_spelling_ differences give it away; there are historical reasons.

In the mini-ex you gave, it is in fact possible to predict the phonemic
forms from the phonetic, _provided_ that the vowel [e] occurs in just that
one environment-- if ones hears a form with [e + vd.fric +V-suffix] one
_knows_ that it must have a vl.fric in the un-suffixed form, since that is
the only possible source of [e]. (One also knows that [niva] _can't_ derive
from */nif/.) How you deal with that is another matter.

The cop-out :-) analysis is simply to say: there's this vowel [e] that's
phonemic (or contrastive) only under two conditions: presence of a vowel
suffix, and intervocalic C-voicing.

That ought to raise red flags-- conditions usually indicate that something
predictable is going on, and the forms are not a true minimal pair.
(Similarly, if houses::kisses [haUz@z : kIs@z] was our best evidence for the
s:z contrast in Engl. it would be suspicious for similar reasons --1)
morpheme boundary 2) presence of apparently irreg. voicing in 'houses', as
evidenced by the larger number of [-s# ~ -s+@s] forms.

It's true that such forms were a quandary for phonemics; but using inflected
or derived forms of the same word to establish minimal pairs was always a
little suspect. Toward the end of the classical period (as generative ideas
were beginning), a few people took up the idea of "ordered rules", so that
e.g. you could have _phonemic_ Germ. /bund/ + a rule devoicing finals. But
the idea of really abstract underlying forms came later.
>
> The examples I gave you were not the actual ones I've been looking at.
> I disguised it like this because I didn't want to start Yet Another
> English Pronunciation Thread, since I'm actually trying to diagnose
> something in the dialect of English I speak.  By this rendering of
> things, I'm using way too many vowel sounds, far more than I'd ever put
> in any conlang.
>
Hmm, it would be interesting to know what the question is........