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

Robert Whiting whiting at cc.helsinki.fi
Tue Nov 14 16:36:18 UTC 2000


On Sun, 05 Nov 2000 CONNOLLY at LATTE.MEMPHIS.EDU wrote:

>>On Thu, 2 Nov 2000, David L. White wrote:

>><snip>

>>> For English /th/ vs/ /dh/, the question to ask is something
>>> like "In a story where two brothers were named "Lith" and "Lidh",
>>> would an audience be able to keep them apart?"  The answer is
>>> clearly yes, regardless of the various other consideration that
>>> some have noted, therefore the distinction is (not surprisingly)
>>> phonemic.

>Bob Whiting replied:

>>No, the answer is:  if there were two brothers named Lith and
>>Lidh then the distinction would be phonemic.  As long as there
>>aren't, there is no phonemic distinction.  It doesn't matter that
>>the two sounds are capable of being distinct phonemes.  Until
>>they are used as phonemes by the speakers of the language, they
>>aren't phonemes for all that they are recognizably different
>>sounds.  When speakers start coining unrelated words where [th]
>>contrasts with [dh] then they will be phonemes.  The fact that
>>you could coin words where they contrast is not sufficient.

>This is simply wrong.

Which means that opinions differ. :)

>We could, theoretically, coin words where aspirated [ph]
>contrasted with plain [p] -- but if we did, people would have to
>learn to control, and pay attention to, a hitherto redundant
>feature.

This is simply wrong. :)  If we started using aspirated p [p']
and unaspirated p [p] contrastively, the real problem would be
distinguishing unaspirated [p] from [b].  People can easily
distinguish [p'] from [p] as in Chinese (a billion Chinese can't
be wrong :>).  When pinyin transcription was introduced in 1979,
most people seemed to get the impression that the pronunciation
of Chinese had changed:  Peking was now called Beijing.
Linguists know (or should know) that Chinese pronunciation didn't
change a bit.  The new system was just considered closer to the
actual pronunciation (particularly for English speakers) than the
Wade-Giles system that it replaced.  The <P> in Peking and the
<B> in Beijing both represent the same Chinese sound, unaspirated
[p] (Chinese does not have a voiced/voiceless distinction).
English [b] is simply closer to the sound than English [p'] is in
this position (where it is always aspirated).

A speaker of Finnish would have no difficulty getting the correct
pronunciation from Peking (except for the final nasal velar which
does not appear in this position in Finnish) because Finnish has
only unaspirated [p] and no /b/ phoneme.  Indeed, as a native
speaker of English, my English-trained ear almost always hears
Finnish initial [p] (both word and syllable) as [b] because it is
not used to hearing unaspirated [p] in this position and the
nearest equivalent to it is [b] (the puff of air that constitutes
the aspiration of [p'] delays the onset of voicing for the
following vowel; without this puff of air, voicing begins earlier
with the result that the consonant sounds voiced.  Not enough to
make it sound like a full-bodied [b], but enough to make it sound
more like [b] than [p], especially in an unexpected environment.

So you can't just start shifting contrasts around one at a time.
You have to consider the effect on rest of the phonological
system.  If we wanted to use [p] and [p'] contrastively, we would
probably have to give up /b/ as a phoneme.  And the same is true
of other stops.  English only gets away with using unaspirated
stops after [s] ('pin' vs. 'spin', 'top' vs. 'stop', 'cot' vs
'Scot') because English phonotactics does not permit [sb], [sg],
or [sd] (except at morpheme boundaries:  'disburse', 'disguise',
'disgrace', 'disdain', 'jurisdiction').  Using [p], [p'], and
[b], etc. all contrastively is cutting distinctions very fine.
Consider the minimal (if any) difference in pronunciation between
'discussed' and 'disgust' in contrast to the clear(er)
distinction between 'cussed' and 'gust'.  I suspect that there
are very few languages that actually do this.  I suspect that
most languages will either have a contrast between voiceless [p]
and voiced [b] or between unaspirated [p] and aspirated [p'], but
not a three-way contrast between [p], [p'], and [b].  Perhaps
some typologist could confirm or refute this.  Stefan?

>But people already hear, and pay attention to, the voicing
>distinction in the interdental fricatives, just as they do in the
>other English fricatives and affricates.

And I still say that the fact that a person can hear and
recognize the difference between two sounds does not ipso facto
make those sounds phonemes of his language.  I can hear the
difference between bright l and dark l.  You might say that I can
only do this because I have studied Russian and Arabic.  But I
imagine that most English speakers could hear the difference
between bright l and dark l.  Does this mean that bright l and
dark l are phonemes in English?  No, because they are not used
contrastively.  Does this mean that they aren't phonemes in
Russian?  No, because there they are used contrastively.  So
whether two sounds are phonemes or not does not depend on whether
they can be be recognized as different sounds or not, it depends
on whether they are used contrastively by the speakers of the
language or not.

A native speaker of English who hears a Chinese pronounce the
initial unaspirated [p] of <Beijing> will probably hear it as
[b].  If he then hears the same Chinese pronounce an initial
aspirated [p'], he will probably be able to hear the difference.
But he will probably think that the contrast is between voiced
[b] and unvoiced [p(')] because that is the contrast in his own
language.  The fact that he can hear the difference between [p]
and [p'] does not mean that these two sounds are phonemes in his
language (although they are in Chinese).  He will just assume
that the contrast is the same as in his own language (nearest
equivalent).

>And there are those minimal pairs, such as ether:either and
>thigh:thy, which cannot be dismissed by *synchronically*
>attributing them to some morphological conditioning or to a
>native:foreign dichotomy.

Part of the problem is that these words were there in the
language before there was any question of [th] and [dh] being
separate phonemes.  For a long time 'thigh' and 'thy'
were pronounced the same.  Then, with the lenition of the initial
[th] of unstressed function words, 'thigh' and 'thy' suddenly have
different initial sounds and constitute that holy grail of the
phonologists, a "minimal pair."  Golly, says everyone, this must
mean that [th] and [dh] are different phonemes because they make
it possible to distinguish 'thigh' and 'thy' as different words.
The problem is that 'thigh' and 'thy' were always different
words.  So the question is, for those who insist that differences
in meaning must mean the presence of different phonemes, what
phonemes differentiated 'thigh' and 'thy' just before this event?
Answer:  none.

The lenition of initial [th] in function words is simply a change
that affected a class of words (function words) and didn't affect
the complementary class (content words).  Whatever the reason for
this change may have been, I am fairly confident that it didn't
happen so that it would be possible to distinguish 'thigh' from
'thy'.  The two words just happen to be in different classes and
so end up with different initial sounds.  Since these sounds are
determined by the class that the word belongs to, and not by the
fact that they are different words, I find it difficult to accept
that this is a phonemic distinction.

>Let me drag in a notorious German example.

Please do, because I think that this example clearly illustrates
exactly what I have been talking about.

>[x] and palatal [c,] are very nearly in complementary
>distribution, with [x] found only immediately after back vowels
>a(:) o(:) u(:) au.  [c,] is found after front vowels and
>diphthongs, all consonants, and (for some speakers) initially.
>But the diminutive suffix -chen has only [c,] even after back
>vowels -- a rare occurrence, since -chen normally fronts
>("umlauts") the preceding vowel.  So one finds alleged minimal
>pairs such as Kuchen [khu.c, at n] 'cake' and baby-talk Kuhchen
>[khu:c, at n] 'little cow', which some claim demonstrate that [x]
>and [c,] must be assigned to different phonemes.  And
>furthermore, many (most?) German speakers hear the difference
>quite clearly.  But are they separate phonemes? No, as is shown
>by the fact that Kuchen : Kuhchen form a subminimal pair: the
>vowel of the latter is clearly (but non-distinctively) longer
>than the former, indicating that it was in root-final position,
>before suffix -chen.  Then, on synchronic grounds along, we may
>write the alleged minimal pair as /ku:x at n/ 'cake' but /ku:+x at n/
>'little cow': the boundary /+/ conditions both the extra vowel
>length and the palatal realization of /x/.

But this is precisely my point.  Juncture (or morpheme boundary)
is not a phonological condition but a morphological one.
Creating a silent phoneme /+/ and calling it juncture doesn't
change this fact.  It just eases the phonologist's conscience.
The justification for considering juncture (morpheme boundary)
a phonological condition seems to be axiomatic.  Phonologists
start with the axiom that only phonological conditions can affect
phonology.  Since morphological conditions (morpheme boundary,
word boundary) often affect phonology, these must be phonological
conditions because, axiomatically, only phonological conditions
can affect phonology.  Doesn't anyone see this argument as even
a little bit circular?  Apparently the phonologists don't because
they then import these morphological conditions into phonology as
symbols (+ and #) and then proceed to bandy them about as if they
were really phonological conditions.

If, whenever you have good reason to believe that two apparently
contrasting sounds aren't really phonemes, you invent a silent
phoneme that has some morphological (or morphophonemic or
morpho-lexical or lexico-historic) basis, call it a segmental
phoneme, and insert it into the phonemic analysis so that the two
sounds no longer contrast, then you are just pandering to those
who say that only phonological conditions can affect phonology.

I'm not trying to say that juncture doesn't exist, or even that
it doesn't affect pronunciation (it often effects things like
pitch, duration, and pause), or that it doesn't affect meaning
(consider the difference between 'I see you've got an ice cream'
and 'I see you've got a nice scream').  All I'm saying is that
juncture is not a phonological condition, it is a morphological
one, and making a segmental phoneme out of it and throwing it in
among the segmental phonemes and saying "see, the sounds don't
contrast any more" may not be the best methodology.

Admittedly, doing this makes it easier to understand some things
(for example, analyzing 'unclean' as /@n+kli:n/ in contrast to
'uncle' /@ngk at l/ makes it clear why the n of 'unclean' doesn't
assimilate to the velar), but if there is a "phoneme" that is
only trotted out when it is needed and ignored otherwise, one
begins to wonder about its reality quotient (as a phoneme, not as
an event).

>The fact that all such "minimal pairs" contain precisely the
>diminutive suffix -chen confirms this analysis.

Actually, it makes it seem more like a self-fulfilling prophecy,
since '-chen', as a bound morpheme, specifically a suffix, will
always be preceded by juncture, whereas in words where the <ch>
is part of the root, <chen> can't be preceded by juncture.

I'm not saying that the analysis is wrong.  Quite the contrary,
I agree with the conclusion completely:  [c,] and [x] are not
different phonemes in German despite the apparent contrast in
words like 'Kuchen' and 'Kuhchen'.  I'm just saying that I would
have gotten there by a method other than making a phoneme out of
a morphological feature.

I would have said that contrasts that can be accounted for by
rules are not phonemic contrasts.  The allophonic distribution
rules say that [x] appears after back vowels and [c,] after front
vowels.  Therefore the pronunciation of [x] in 'Kuchen' is
required by rule.  There is another rule, obviously of higher
precedence, that says that the diminutive suffix '-chen' is
always pronounced with [c,] regardless of its environment.
Therefore the pronunciation of [c,] in 'Kuhchen' is required by
rule.  Since both pronunciations are required by rule, the
contrast is not phonemic.  QED.  Now some people will say that
the rule that says that '-chen' is always pronounced with [c,]
is not a phonological rule, but a morpho-lexical one and that one
can only evaluate phonemic contrasts through phonological rules.
These people will then take a morphological feature, make it
into a segmental phoneme and then show that it blocks the
contrast.  Occam's Razor gives my method the edge, because it
does not need the additional entity of a silent phoneme.

>It is therefore unlikely, though possible, that German could
>coin new words in which [x] and [c,] would (appear to) contrast
>in some other environment, just as it is unlikely that English
>would coin new words contrasting precisely in aspired [ph] vs.
>plain [p].

This is an interesting spin that you have put on this.  It makes
it seem like your analysis supports what David L. White said
while in fact it demolishes it.  I would have said that it is
possible, but unlikely, rather than the other way around.
But either way, clearly the Germans recognize [c,] and [x] as two
different sounds (otherwise they wouldn't appear to contrast in
'Kuhchen' and 'Kuchen'), but they don't use them contrastively
(as phonemes).  We agree on this.

So let's modify what Dr. David said just a little bit.  "In a
German story where two brothers were named 'Luchen' [lu:x at n] and
'Luhchen' [lu:c, at n], would an audience be able to keep them
apart?"  The answer is clearly yes ('Kuchen' and 'Kuhchen' prove
this).  Would [c,] and [x] then be phonemes?  Well, yes (as long
as you can convince the audience that 'Luhchen' doesn't mean
"little Luh") because the sounds are used contrastively in a way
that can't be predicted by rule.  Does this mean that [c,] and
[x] are already phonemes in German?  Most emphatically not.  You
have proved that they are not your way and I have proved it mine.

The fact that two sounds are potentially contrastive does not
prove that they are phonemes in any particular language.  The
mere fact that the same two sounds can be phonemes in one
language but allophones of the same phoneme in another language
should be adequate proof of this.

>But English does have a few minimal pairs for the interdentals,
>ones which cannot be explained on phonologucal grounds.

But you haven't explained German [c,] and [x] on phonological
grounds.  You have explained them using morpho-lexical
information.  The fact that you have made a morphological
condition into a phonological one (or rather used one that the
phonologists already had waiting to be used for purposes like
this) doesn't make this a phonological explanation for anyone who
knows the difference.

>True, in my pronunciation, the stressed vowel of _either_ is
>longer than that of _ether_ -- but this is in line with vowel
>length before other voices vs. voiceless consonants.  And there's
>no difference in _thigh_ vs. _thy_.

I have explained my reasons for not considering 'thigh' and 'thy'
a minimal pair above.  My reasons for not considering 'ether' /
'either' a minimal pair are similar (ultimately relying on the
fact that the pronunciations are required by rules), but if you
prefer, I could invent a silent phoneme that blocks the voicing
of intervocalic [th] in loanwords (there is no need for this
since it wouldn't voice anyway [there is no indication of one
having voiced], but it does provide a "phonological" basis for
the lack of contrast).

Everyone seems to object to my using these rules to determine
whether the sounds involved are phonemes or not, saying "you
can't use these rules because they aren't phonological rules."
To which I could say (if I really wanted to) "Of course they are
phonological rules because they are rules about how these words
are pronounced.  And what is phonology about except how things
are pronounced."  But I don't want to say this (yet) because
there are other aspects of phonology that I don't want to get
involved in.

Everyone also seems to be under the impression that I am trying
to prove that [th] and [dh] are not separate phonemes.  This is
not the case.  It would not bother (nor surprise) me if [th] and
[dh] could be shown to be separate phonemes.  All I am saying is
that the evidence that [th] and [dh] are phonemes is inadequate.
As I have said before, there is no reason why [th] and [dh]
shouldn't or couldn't be phonemes.  Indeed, when one looks at the
evidence, they should be phonemes based on the parallel
development of /z/ and /v/.  All I say is "where is the evidence?"
Where are all the contrasts that would show unequivocally that
that [th] and [dh] are phonemes as there are contrasts that show
that /z/ and /v/ are phonemes.  And I don't consider 'thigh' and
'thy' or 'ether' and 'either' any more indicative of the
phonemicity of [th] and [dh] than 'Kuhchen' and 'Kuchen' are
indicative of the phonemicity of German [c,] and [x].

>So the fact that new words contrasting only in having [T] vs.
>[D] could be formed without any new feature becoming distinctive
>confirms that the two are searate phonemes, and have ben so for
>some time.

And I still say that when new words with this contrast are
formed, then they will be phonemes.  They might be phonemes
now, as you claim, but if so, they are unused phonemes -- just
sitting in the closet with their wrapping and ribbons still on.
But the argumentation used here seems considerably less than
convincing.  For instance, one might say that [x] could be used
as a phoneme in English by forming new words where it contrasts
with /k/ or /g/.  Since these words could be formed without any
new feature becoming distinctive (as a voiceless velar fricative
the contrast between /x/ and /k/ or /g/ would be the same as
the contrast between /f/ and /p/ or /b/), this confirms that [x]
is already a phoneme in English and has been for some time.
I don't think so.

>>And as you unintentionally point out, when they become phonemes,
>>[dh] will have to be written <dh>, otherwise you couldn't
>>distinguish between <Lith> and <Lith>.

>Well, it should be so written already, but we ignore all sorts of
>other distinctions, and it doesn't bother us.

It's not so much that we ignore other distinctions as it is that
the graphic system's ways of distinguishing them are not always
consistent or intuitively transparent.  But once you get the hang
of it, the writing system is actually quite well adapted to the
language, and the lack of transparency is generally because the
writing has preserved etymological information that has
disappeared from the spoken language.  Perhaps you should read
Chomsky and Halle.  Except for some word final distinctions
between -<th> and -<the>, no attempt is made to distinguish
between [th] and [dh].  It's a good thing the pronunciation is
predictable.

One thing you can say for sure though is that <dh> never
represents [dh] (except in the name of the character 'edh').
Otherwise it is regularly realized as [d] ('dhow', 'dharma',
'jodhpurs') and generally indicates a foreign word (except, of
course, in native compounds where it represents two sounds [e.g.,
'madhouse', 'deadhead', etc.]).  Incidentally, I'm still looking
for an explanation of why words from languages with phonemic /dh/
don't come into English with [dh].  It seems curious, what with
[dh] being a phoneme in English and all, that phonemic /dh/ comes
into English as [d].  Many people have hastened to assure me that
[th] and [dh] are indeed phonemes in English, but everyone seems
reluctant to address this point.

><s> stands for /s z S Z/ (the latter two as in _pressure_ and
>_pleasure_ -- in fact, we don't evan have an official way to
>spell /Z/, though <zh> would seem obvious.  And don't even ask
>about the vowels: treating <y w> as doublets of <i u>, we have
>effective five vowel symbols for a very large phonemic inventory.
>I'd be glad to see ,dh. in use, but it really has nothing to do
>with whether [T D] are separate phonemes.

Maybe not, but you don't get any help from the orthography on how
to pronounce them.  And although there are only five vowel
symbols in English, there are actually quite a few more, because
the five are combined in various (but not always consistent) ways
to provide much more than five representations for various vowel
sounds.

Bob Whiting
whiting at cc.helsinki.fi



More information about the Indo-european mailing list