Re: Optimum number of symbols

Tom Wier wrote:


>Quoting Roger Mills <romilly@...>:
>> Ray Brown wrote:
>> >At 10:44 pm -0500 21/5/02, Thomas R. Wier wrote:
>> >[snip]
>> >>(TW)Actually, I was referring to different forms:  <Rad> and <Rat>
>> >>are homophonous in the singular, and yet are phonologically
>> >>distinct in the plural:  <Räder> /RE:d@/ and <Räter> /RE:t@/.
>> >>This does bear on the criticism of a true phonemic system, since
>> >>a true phonemic system will fail to capture phonological
>> >>neutralizations like that in the German data I presented.
>> >>That is, /Rat/ is really two distinct words: /Rat/-1 and /Rat/-2,
>>
>> (RM)Quite so.
>> >(RB)Surely, phonemically they are /ra:d/ and /ra:t/ respectively, tho
both
>> >pronounced [ra:t], since in syllable final position /d/ is pronounce
[t].
>>
>> (RM)The first person to suggest that to the early phonemicists was
drummed out
>> of the corps for introducing an "abstraction", I think.  The generative
>> idea of underlying (abstract) forms, operated on by a series of ordered
>> rules, was IMO a great improvement, even though Chomsky and Halle carried
>> it a little too far.
>
>(TW)I see that I have confused, rather than enlightened, my intended
>audience.  My point in using phonological notation rather than
>phonetic was to reinforce the point that /d/ and /t/ constitute
>a salient distinction in German phonology generally, outside this
>word, and that when this distinction is neutralized, it is
>neutralized  to one of these two phonemes, but not both.
No no no, not confusing at all.  I, and Ray (I'm sure), are well aware of
the demands of the theory.  My own problem with that brand of phonemics was
that, by relying only on the surface forms of a language, it rather
willfully ignored a lot of what speakers know about their language.  It
always struck me as counter-intuitive to posit /Rat/ = /Rat/-1 and /Rat/-2,
(and all the many other words with voiced/voiceless alternants), and then
simply list the various allomorphs-- /rad-/ in such-and-such environment
e.g.-- as if linguistic information is stored in some sort of gigantic
telephone book or computer file that one has to search every time one opens
one's mouth.  Surely native speakers of German "know" that there is a word
{rad} and a word {rat} and that {rad} (_by rule_)> [rat] when it stands
alone/in the nominative case, or however we want to say it.

(Other aspects of one's language may very well be stored in a "list"-- Engl.
irregular verbs or plurals, for ex., where forms are not predictable.)

It's certainly true that neutralization-- even in such a relatively
transparent case as final-devoicing-- was a problem for phonemics.  (Along
with such things as French [gra~, gra~d]  et al.)

 >I was not
>refering to a kind of Chomskyan underlying specification per se,
>which would surely posit something like /Rad/ and /Rat/.
I realize that, but hope you agree that some sort of underlying level is an
improvement...  Of course, Chomsky/Halle and Classical Phonemics are extreme
and opposite viewpoints,  and the answer, if there is one, probably lies
somewhere between the two....:-)

Replies