LINGUIST List 10.1880

Tue Dec 7 1999

Disc: What Exactly Are Allophones?

Editor for this issue: Karen Milligan <karenlinguistlist.org>


Directory

  1. benji wald, Re: 10.1867, Disc: What Exactly Are Allophones?
  2. Martin Dr�pela, Re:10.1876, Disc: What Exactly Are Allophones?

Message 1: Re: 10.1867, Disc: What Exactly Are Allophones?

Date: Sun, 5 Dec 1999 22:09:25 -0500
From: benji wald <bwaldHUMnet.UCLA.EDU>
Subject: Re: 10.1867, Disc: What Exactly Are Allophones?

I should have known I would not get off lightly if I strayed from the
sidelines of this debate. Nevertheless, I refuse to plunge into the thick
of it -- but, at the same time, Jorge Guitart asked me some very
interesting and important questions, which deserve answers. I try to
clarify below.

Jorge begins:
>Benji Wald said
> Phoneticians have long been
>> aware that vowels are perceived "continuously", not discretely, even though
>> phonemes are supposed to "always" be discrete (i.e., the phonemes of which
>> phoneticians study the allophones). That means that if there is no gap
>> between the spaces of the /i/ and /e/ phonemes, then there will indeed be
>> an in-between area which speakers will inconsistently identify with either
>> /i/ or /e/.

and then poses the following question:
>Benji, allow me to seize on this:
>If a person who says git and has never said get because she is monolectal
>(and just starting--most probably unwittingly--to be trained in the
>standard) is being taught to say get and when asked repeatedly to say
>[get], says [gIt] instead, what would be wrong with saying that she
>repeatedly perceives (misperceives from the viewpoint of the teacher) [e]
>as [I]? If it is not wrong to say that, how could assigning the incoming
>signal to the slot [I] rather than to the slot [e] be not binary (and
>therefore discrete)?

According to my understanding of the question, it would be wrong to say
that the speaker misperceives /e/ as /i/. Allowing the assumption that a
monolectal speaker might behave as Jorge indicates above, the speaker
demonstrates an incapacity to change a particular pronunciation of a
LEXICAL item, NOT of a phoneme. The same speaker will have no trouble
differentiating "bit" from "bet" or pronouncing them different. In all
known dialects of English these words, and indeed the phonemes (as distinct
diachronic entities) /i/ and /e/ remain distinct. I tried to make clear in
my last message that GBS was not referring to a phonemic merger, but to the
assignment of a particular lexical item (spelled "get") to one "phoneme"
rather than to another ("tra-la-la, you say ee-ther and I say
eye-ther..."). I think Jorge's premiss is possible although not common,
that is, that some dialect speakers might not be able to change their
pronunciation of "git" to "get", and nevertheless have separate phonemes
for /i/ and /e/. I'm not quite sure what the pathology would be called by
neurolinguists and/or if it would be symptomatic of some more serious
neurological disorder. (at least one speaker who has said "nukyalar" all
her life has trouble pronouncing it "nuclear", even under extreme ridicule
- or better yet "possively" for "possibly" but has no trouble
distinguishing "passively" from "passably" < pass-able-ly, and is otherwise
more intelligent than me in all measurable ways.)

>What does it mean to say that a sound that does not exist in your lect is
>perceived "continuously" (your quotation marks) rather than as a unitary
>entity that is then assigned to a given slot in correspondence with the
>learner's native system?

According to my understanding of the phoneticians' finding, the issue is
not about *a* sound, but the CONTINUOUSLY VARIABLE SOUND that can be
sampled throughout the entire vowel space. Presumably, the vowel phonemes
of any language operate within local areas of this space (or of one of
various possible vowel spaces, according to theories that go beyond what
the original phonetic experiments conceived of at the time). As we all
know, Martinet called the area in which the "allophones" (how I tried to
avoid that word!) of a phoneme are distributed (or better,"... in which
they are contained") its "field of dispersion", and the area between any
two such fields, where they could not be pronounced (or manufactured
experimentally) without confusion of the two phonemes contained/constricted
therein, the "margin of security".

Jorge observes:
You have to "put" it someplace! YOu cannot
>assign it to a range that is neither [e] nor [I]. It goes into the
>discrete category /I/.

Well, I'm not so sure that what "you HAVE TO" do is "put everything in
discrete bins", but yes, if you want to have a phonological theory, you
have to do something to have it, so...
According to the discrete theory of (biunique) phonemes, Jorge is quite
right. That is, the vowel of "git" (as a realisation of the lexical item
"get") is a realisation of the phoneme /i/, the same phoneme as in "bit"
and various other words, but not of "bet", or (paradoxically) of the same
"lexical item" "get". Meanwhile, for what Jorge is getting at (or, as they
say today, "where he's going"), we consider next the more pregnant
question.

Now Jorge asks the key question which all phonological theories should be
able to answer, and upon which they should be judged (well, perhaps with
the "in isolation" phrase below as only a part of the larger question):

>What happens when you hear a vowel in isolation that doesn't belong in
>your native lect or language?

Some of them you will hear as very odd sounds indeed, e.g., the Northern
Eurasian "barred i" as in Russian byt', in Japanese an allophone of /u/,
etc. -- not to mention various sounds that occur in (phonetically) distant
dialects of your native language.

To give an example, Jorge describes an experiment in which a particular
area of vowel space is not used distinctively in one language (but is it
used at all? -- under what conditions?) HAS TO be labelled by speakers of
that language, using the symbols they have for areas of vowel space which
are distinctive for them.

He describes the experiment:
 I ran an experiment with Spanish dominant
>Puerto Rican kids. As part of it a native speaker of American English who
>knew linguistics pronounced [I] (the vowel of, say,PIT) in isolation. We
>asked the children to write down the
>sound they heard (they were literate--seventh graders). Some kids heard it
>as Spanish /i/ (they wrote the letter I) and some heard it as Spanish /e/
>(they wrote the letter E).English [I] has features in common with both
>Spanish [i] and Spanish [e], so this was no surprise. Now, no kid failed
>to decide either for I or for E There were only two discrete slots, and
>they put it in one or the other! How would you characterize these results
>in a framework that denies
>discreteness?

As I anticipated above, speakers will not perceive every vowel they hear as
a realisation of some phoneme in their own language. However, the
experiment demonstrates that speakers can be FORCED to assign some
particular sounds, which may very well sound WEIRD to them, to one or
another phoneme in their own language. This may even be easier with vowels
than with consonants since vowel sounds are perceived continuously
according to previously mentioned phonetic experiments. (The fun one with
consonants is to record a labio-lingual stop and then play it for speakers
of any number of languages and see whether they choose "p" or "t", and
further to watch them scratch their heads as they hear a sound that sounds
"peculiar" to them.)

Getting back to the last thing I said about vowels, I am not supposing that
ALL possible vowel phonemes (i.e., local areas of any vowel space in any
language) can be assigned to some vowel phoneme or other in ANY language,
but Jorge's experiment demonstrably represents a subset for which it IS
possible. Indeed, perhaps English speakers can even assign mid front
rounded vowels to particular English phonemes (I would expect them to vary
between front unrounded and central vowels, according to their particular
dialects -- I, for example, would take the French vowel of "peur" and if
forced, assign it to the English phoneme represented in "bird", "word",
etc. -- nevertheless, I and countless others who I could get to do the same
assignment as I, would consider such a pronunciation of "bird" to be very
odd. Now for French "peu" I would choose my vowel of "book" as closest,
even though that vowel is less rounded for me than for most varieties of
English, etc.) In sum, as Jorge says, English "i", as in "pit", "git",
etc. has some affinity to both Spanish /i/ and /e/. I assume a Northern US
or Southern British variety of "i", which is lax and thus lower than
Spanish /i/ but not (usually) as low as Spanish /e/ in such varieties.
Thus, if Spanish speakers contrast the sound with Spanish /i/ and are
forced to pick a "letter" (= phoneme) to APPROXIMATE it, then, finding it
suitably FRONT but LOWER, they will choose "e", but if they contrast the
sound with Spanish /e/, then finding it suitably front but HIGHER they will
choose "i". Meanwhile, they realise that the English sound is NEITHER, and
the community as a whole, and probably every member thereof, is
INCONSISTENT on which of the two Spanish phonemes to assign it to. The
experiment, performed as a one-time pass, cannot check on individual
consistency, but I would not expect speakers confronted with such a task to
be consistent in the strategy they use (i.e., comparison with the closest
higher vs. lower vowel), since I do not think the task relates directly to
any task they more normally or naturally do in using their language (or
English, to the extent that any of them use it).

There is more that can be said about phonemic theory, and the notion of
"discrete" as an all-or-nothing proposition, when confronted with imminent
mergers and near mergers in particular dialects, and about cross-dialectal
and cross-language "perception" and "labelling" (the latter being the
object of Jorge's experiment, NOT perception). Anyway, for the moment,
even though I am not adverse to theories of discrete phonemes, WHERE
APPROPRIATE, I think care should be taken to distinguish in experiments
between what speakers CAN do and what they ACTUALLY do (or DO do, or MUST
do, or whatever). That is, in principle, if we ask speakers to perform
discrete operations and they show that they CAN do them, that does not mean
that that's what they are inclined to do when left to their own devices.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re:10.1876, Disc: What Exactly Are Allophones?

Date: Tue, 7 Dec 1999 13:39:49 +0100
From: Martin Dr�pela <martin.drapelafpf.slu.cz>
Subject: Re:10.1876, Disc: What Exactly Are Allophones?

Dear all,

the discussion on the nature of (allo)phones is really fascinating, but
I think that Hugh Buckingham has hit the nail on the head:

> Allophone Enthusiasts:
> First, tell the students to look up the nice little prefix "allo-" in
> the dictionary.

Being one and doing so I think that David Crystal's {1997) definition of
'allophone'
is fairly uncontroversial and satisfactory. He begins with

"allo- A prefix used generally in LINGUISTICS to refer to any noticeable
variation in FORM of a linguistic UNIT which does not affect that unit's
FUNCTIONAL identity in the language."
and further below continues:
"The relationship between allophones and phonemes is one of REALIZATION:
a phoneme is 'realized' by its allophones."

Just to add to this, in his definition of 'phone' we will also find that
"from the viewpoint of segmental P H O N O L O G Y, phones are
the physical REALIZATION of PHONEMES: phonic varieties of a phoneme are
referred to
as ALLOphones.", PHONOLOGY double-spaced here by myself.

Now, do we still have a problem here?

Sincerely,
Martin Drapela


Crystal, D., (1997), A Dictionary of Linguistics and Phonetics, 4th ed.,
Blackwell Publishers Ltd.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue