Re: Allophones Question

At 9:44 AM +0100 2/19/03, Christophe Grandsire wrote:
>En réponse à Angel Rivera <mktvr@...>:
>
>>
>> I'm readin Pike's _Phonemics_ lately [looks old, but like the only text
>> on it my
>> library has], and apparently:
>>
>>     << It is advisable to consider as the norm that segment which has
>> the least
>> limitation in distribution in the language, and appears to be the least
>> affected
>> by surrounding sounds.  >>
>>
>> ...which is why I considered it safe to use the Spanish example.
>>
>
>And you would be wrong to do so. After all, what is the distribution of the two
>allophones of voiced stops in Castillian Spanish? Simple: voiced fricative in
>intervocalic position, and voiced stop *everywhere* else! If you look at
>limitation of distribution, the fricative allophone is much more limited in
>appearance than the voiced stop. Now because of Spanish's simple syllable
>structure, you get often intervocalic consonants. But it doesn't that if you
>reason in terms of environment, the voiced stop allophone is freer to appear
>than the voiced fricative one.
I seem to remember that the distribution is more complicated than you describe.
Ah; here it is. Joan Mascaró has this to say: "(Basque and) Spanish have the
stop after pause and nasal, and for [d] after [l]. The fricative appears after
vowel, glide, and nonnasal consonants (except, as noted, after [l] in the case
of [d])." Both environments seem equally complex, and Mascaró later notes that
"there is no compelling argument for having underlying stops or underlying
fricatives." (Quotes are from his 1984 paper "Continuant Spreading in Basque,
Catalan, and Spanish" which appeared in _Language Sound Structure_, a
festschrift for Morris Halle.) So it is an open question for Mascaró.

>Moreover, even if this argument you gave was in your favour, it wouldn't be
>enough, because this is *not* the main criterion for deciding what a phoneme is
>and how to label it. Why? Simply because of the definition of phonemes itself:
>phonemes are abstractions manipulated by the mind of the *speakers*! And as
>such the main criterion to decide what's phonemic is the speakers' intuition,
>and any phonemic analysis that disagrees with the speakers' intuition is
>probably flawed. And that is valid for how to label phonemes too.
Ah, but whose phoneme are you talking about? the structuralist phoneme was
discerned by investigating the distributional patterns of sounds in a corpus of
data; Pike's heuristic cited above by Angel is firmly within that tradition.
The intuitions of the native speakers had nothing to do with it in principle,
even if field linguists used their consultants' intuitions to help them sort
out troublesome data.

Generative grammar discarded the structuralist phoneme in favor of underlying segments
which occupied a position roughly equivalent to the structuralist
morphophoneme. The claim was made of these underlying segments that they were
psychologically relevant to speakers of a language; entries in the mental
lexicon were encoded in terms of these underlying segments. While the status of
native speaker intuition *as data* was elevated in generative syntax, there is
no evidence in the literature or in practice that linguists considered
phon-etic/-ological intuitions to be important in the same way in the analysis
of a language's sound patterns.

The choice of /d/ as the phoneme in Spanish seems to be as much a convention based
on history and spelling as on synchronic distributional facts.

>So now let's
>look at the Castillian speakers' intuition. Well, if you ask them, they will
>tell you that /d/ is indeed a stop, or that they would put /g/ with /k/ rather
>than with /x/. That's already a strong sign. Not only that, but if you
>pronounce "ciudad" ("city") as [Tju"da] instead of [Tju"Da], the Castillian
>speaker will think you have an accent, but won't be able to explain exactly
>why, and won't misunderstand you. While if you pronounce [Dar] instead of
>[dar], the Castillian speaker will surely interpret this as */Tar/ *"zar" and
>thus won't understand you at all (I give this last example because it happened
>to me. A slip of the tongue it was), and will ask you to repeat. That's an
>extremely strong sign that [d] is considered to be the main sound and [D] is
>just unconsciously pronounced between vowels.
But if [D] is only considered an allophone of a more basic /d/, then there should
be no confusion of [D] with [T], which belongs to a separate phoneme. That is,
there is no version of /T/ which is pronounced [D]; [D] can only be /d/. So
your example casts doubt on your analysis of underlying /d/.

>So calling this phoneme /D/ is
>certainly wrong, /d/ is better as it fits with the speakers' intuition (Occam's
>razor here. Choose the simplest way to explain all the data).
You're using two different standards. Is the speaker's intuition the relevant
criterion, or is simplicity the relevant criterion? They are not necessarily
the same. I might agree with you that /d/ is underlying, but only because
distribtutional evidence may favor this as the simplest solution. I don't think
that speakers' intuitions are reliable at all in these matters. This is
reflected in my own work; I have never relied on speakers' intuitions. I may
use them occasionally to probe the facts, but I certainly wouldn't build an
argument based on them; they are simply too unreliable.

Dirk
--
Dirk Elzinga                                               Dirk_Elzinga@byu.edu

"It is important not to let one's aesthetics interfere with the appreciation of
fact." - Stephen Anderson

Reply