Don't touch my phonemes (was: minimal pairs ex: PIE e/o Ablaut)

[Robert Whiting]

On Thu, 16 Nov 2000 David L. White <dlwhite at texas.net> wrote:

>        Look, either we admit that English speakers could coin
>/lith/ vs /lidh/, whereas Japanese speakers could not coin /biri/
>versus /bili/ or we (somewhat obstructionistically) do not.

I say we admit it.  The question then becomes "what does it
prove?"

>If we admit it, there must be an explanation.

I'm glad to see that you subscribe to the principle that patterns
must be accounted for by rules.

>The view of "the phonemic principle" that I and several others
>have been pushing provides this, while the view that Mr. Whiting
>has been pushing does not.  I personally prefer to construe the
>concept of "phoneme" in such a way that meaningful and accurate
>predictions, which yield an increase in understanding of
>linguistic behavior (even hypothetical), can be made.

This is fine so long as you don't try to foist logical fallacies
off on your adoring public.  To come back to my original question
of "what does this prove?", the answers are easy (even if those
not trained in logic can't see them).  The fact that Japanese
speakers could not coin /biri/ vs. /bili/ proves that [r] and [l]
are not phonemes in Japanese.  The fact that English speakers
could coin /liT/ vs. /liD/ proves that [T] and [D] either are
phonemes in English or that they could become phonemes in
English.  The problem is that "are" and "could become" seem to
mean the same thing in your idiolect.  If you are over 35 and
a native-born citizen of the U.S. then you could become President
of the U.S.  Does this mean that you are the President of the
U.S.?  I don't think so (although it is possible, after all
somebody has to be and nobody seems to know who :>).

The logical fallacy is in assuming the converse of a universal
(A-type) statement to be true.  Being distinguishable sounds is a
necessary condition for phonemes.  Therefore the statement "all
phonemes are distinguishable sounds" is valid.  But the simple
converse of this, "all distinguishable sounds are phonemes" is
not a valid inference.  If it were true, then being
distinguishable sounds would be a sufficient condition for
phonemes.  It is not.

To prove that this is not true, we could trot out the example of
German [c,] and [x] again, but there is a more direct way
involving the segments we have been discussing.  Let us go back
to the point where intervocalic voiceless fricatives were voiced
in English.  This change produced voiced allophones of the
voiceless fricatives.  Could these sounds be distinguished by the
speakers of the language?  Pretty certainly, since the Irish
monks who introduced the Latin alphabet into England heard them
and provided signs to write each of them.  Could the speakers
have coined words in which these sounds contrasted?  Pretty
certainly, since if speakers can hear the difference between two
sounds they should be able to contrast them if they had wanted
to.  Does this mean that these sounds were phonemes from the
moment of their creation?  No. Nobody (except Mr. White) disputes
that these sounds remained allophones throughout Old English.

But according to your "phonemic principle" these sounds were
phonemes from the moment they were created.  I don't believe it.
It took quite some time before these sounds began to be used as
phonemes.  Now the fact that they were distinguishable sounds
provided the potential for them to become phonemes.  But
potential is not reality and despite this potential, they aren't
phonemes until they are used as phonemes.

Psychologists tell us that everybody is capable of murder.  Since
everyone is a potential murderer, then, according to your brand of
logic, everyone is a murderer.  Don't tell the Governor of Texas
this or he will have his Rangers round up everyone and start
executing them.  Every man who does not suffer from erectile
dysfunctionality (not to be confused by speakers of Japanese with
the electile dysfunctionality in Florida :>) is a potential
rapist.  According to your brand of logic, that means that every
man is a rapist.  Potential is just not reality, and saying that
it is doesn't make it so.

Now there is nothing wrong with potential.  Bookies make a lot of
money out of it.  For the odds offered before a race or a game
simply reflect a numerical evaluation of the potential of each
entry or team to win.  But at the end of the day, it is not the
entry or the team with the highest potential that wins, but the
one that crosses the finish line first or that scores the most
points.  Potential may help sell tickets, but it is reality that
goes into the record books.

Similarly, there is nothing wrong with identifying potential
phonemes.  But potential phonemes are just potential phonemes
until they cross the finish line.

>        The sort of distributional standards that are commonly
>used to determine phonemes can fail to predict what speakers are
>sensitive to and able to control.

Intentionally so, because everything that speakers are sensitive
to and able to control is not necessarily a phoneme according to
the present definition.  This may be in the background as an
unstated (but fairly obvious) necessary condition for phonemes,
but it does not form any part of the definition of a phoneme even
if it is implied by the "contrastive" part of the definition.

>For example there is in one African language (I forget which,
>but I think it is in the Nuer Dinka area) which has predictable
>long-range nasalization that speakers are sensitive to.  The
>long-range is the key, because the nasalization, though
>abstractly predictable, is not predictable from phonetic
>implementation.  (Speakers do not really feel compelled to open
>their velums the whole way, as opposed to the way that English
>speakers feel compelled to aspirate under certain circumstances.)
>In a case like this, the question of whether nasalization is or
>is not phonemic depends on what we want the word to capture.  I
>prefer it to capture speakers abilities, enabling us to predict
>(among other things) what coinages are possible and what are not.

Ah, so you are the one who wants to change the definition of
phoneme.  I'm afraid that phoneme as it is now defined is based
on the actual usage of sounds in a particular language, not on
what sounds the speakers of the language can distinguish.
Potential use as a phoneme may, as you say, be useful linguistic
information, but it doesn't fit into the phonemic description of
a language as it is presently constituted.

>Mr. Whiting evidently prefers it to capture distributional
>truths, among other things.  As I said before, perhaps
>over-charitably, we will have to agree to differ.

Capturing truth can be a tricky business as Pontius Pilate once
pointed out.  But charity has nothing to do with agreeing to
differ.  "Superciliously" might be a more accurate term.  It is
possible for people to have honest differences of opinion over
facts (such as whether /ch/ is one phoneme or two), but the rules
of logic are not subject to different interpretations.  And I
don't need your permission to disagree with a logical fallacy.
So I will continue to disagree as long as you continue to claim
that any sounds that the speakers of a language can distinguish
are phonemes in that language.  Until you can get the definition
of phoneme changed, phonemes will continue to be the sounds that
are actually used contrastively in a language to effect
differences in meaning.

Now if you want to say that sounds that can be distinguished and
could be used contrastively by the speakers of a language are
either already phonemes or could become phonemes in that
language, then I would have no objection.  There is nothing
logically inconsistent in such a statement.  Whether there is an
inevitability about such potential phonemes eventually becoming
actual phonemes that makes this a powerful tool of prediction or
not, I don't know, but it might be interesting to investigate.

>        It should be noted that the matter of boundaries, how
>many types there are and how (other than elegance of description)
>they are to be detected, is not so clear as some would have us
>think.  For example, it is fairly common in Old English for what
>we think of as the first element of a compound to be written as a
>separate word.  Nor is it clear that the sort of boundaries we
>have been talking about are properly described as
>"morphological", as any account of phonetics and phonology
>described in their own terms would at the very least have to
>include a recognition of word boundaries.  But if we admit that
>at least there are stronger "word-like" boundaries on the one
>hand and things quite a lot weaker than that on the other, then
>the matter of "kuchen" can be handled by treating the diminutive
>as "ku-chen" or even, more radically, "ku chen", with "chen" as a
>somewhat exceptional word (given a looser definition of the word
>"word", which does not after all have a revealed meaning).  The
>pronunciation referred to does indicate the presence of a strong
>boundary of a fundamentally different type, general rather than
>merely morphological, compared to the boundary in something like
>Latin "paribus" (however we divide that).

All very interesting and it certainly has a bearing on the
gramaticalization of independent words into derivational
suffixes, but I'm not sure what its relevance is to the present
discussion.  I'm sure that everyone knows that whether a phrase
is considered one lexical item or two does not depend on whether
it is written as one word, with a hyphen, or as two words
(someone is certain to remind us that writing is allegedly
secondary to spoken language) but on whether it is considered as
one stress unit or two.  Consider for example English 'bedroom'
and 'living room'.  Although the first is written as one word and
the second as two, both are single lexical items (i.e., they are
both a single stress unit).  'Living room' as two stress units
implies something different (a room with an animus of its own).
So it is not a question of whether 'kuhchen' is written
<kuhchen>, <kuh-chen>, or <kuh chen>; it is a question of when
any putative *'kuh chen' became a single stress unit.  That is
when the juncture ceases being a word boundary and becomes a
morpheme boundary.


Bob Whiting
whiting at cc.helsinki.fi