Newsgroups: sci.lang
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!newsfeed.internetmci.com!in2.uu.net!news.sprintlink.net!new-news.sprintlink.net!eskimo!rickw
From: rickw@eskimo.com (Richard Wojcik)
Subject: Re: Phoneme list needed!
X-Nntp-Posting-Host: eskimo.com
Message-ID: <Dq08AF.KIE@eskimo.com>
Sender: news@eskimo.com (News User Id)
Organization: Eskimo North (206) For-Ever
References: <31681010.5699096@news.cdsnet.net> <316beef6.5010819@news.cdsnet.net> <316df7b5.2747293@news.cdsnet.net> <4klvs3$j9t@news.acns.nwu.edu>
Date: Wed, 17 Apr 1996 11:37:25 GMT
Lines: 41

In article <4klvs3$j9t@news.acns.nwu.edu>,
Sean M. Burke <sburke@babel.ling.nwu.edu> wrote:
[snip]
>Phonemes are an abstraction from phones, and you're not solving
>anything by bringing in with feature analysis, because that's one step
>/further/ abstracted.  Yes, that means it's even more language
>specific.

Sean, I think I agree with your understanding of what phonemes and phones
are, but don't you think it would be better to say that phones are an
abstraction of phonemes?  Note, particularly, your claim about phones
being "continuous" in the following:

>
>At the level of phones (in the sense you're looking for it), language
>is a /continuous/ phenomena; asking for all the phones in language is
>like asking how many clouds can be in the sky; /first/ you have to get
>pictures of clouds, then you have to draw the lines between the clouds
>(no small task if they're kind of brushing against eachother), then
>you have to impose some system for classifying them.  The result
>generally becomes an artifact of your classification system, not of
>the data.

The point is that phonemes might be considered an abstraction of phones
only in the sense that they have to be discovered by some kind of phonetic
analysis.  The native speaker of a language, however, perceives the
phonemes as discrete, clear elements and has to figure out how to pronounce
them or map them onto auditory streams.  Linguists are used to thinking of
phonemes as being abstractions of phones because, historically speaking,
phonemic analysis has been treated as a discovery procedure based on data
that consists of a discrete phonetic segmentation of an acoustic stream.
One ought to avoid defining phonemes, however, in terms of the way
linguists (or language learners) come to discover what they are for a
language.  They should have a definition that describes their functional
role in language.  For example, one might define phonemes as the phonetic
segments that speakers associate with the morphemes of a language.  Other
phonetic segments can exist in the abstract, but they can't normally be
associated with the morphemes of a language.
-- 
Rick Wojcik  rickw@eskimo.com     Seattle (for locals: Bellevue), WA
             http://www.eskimo.com/~rickw/
