minimal pairs (was: PIE e/o Ablaut)

[Larry Trask]

Pat Ryan writes:

>>> True minimal pairs, a paltry requirement for phonemicity that would be
>>> undisputed in any other language.

> [SF]

>> Not at all.  Many phonemes are accepted as such *without* minimal pairs
>> even in living languages.

> [PR]

> You assertion by itself does not convince me.

> Would you mind citing an example of any phoneme in any language that is not
> in a minimal pair?

Easy.  English /h/ and [eng] -- the velar nasal, as in 'sing' -- do not
form even a single minimal pair.  They can't, because they are in
complementary distribution: broadly, /h/ only occurs syllable-initially,
while [eng] only occurs syllable-finally.  But we still count them as
two phonemes, and not as allophones of a single phoneme, since their
degree of phonetic similarity is so low.

The existence of minimal pairs may be a *sufficient* condition to
establish a phonemic contrast, but it is not a *necessary* condition.
Even when phones are phonetically similar, we can consider assigning
them to a single phoneme only when their distribution can be stated
by rule.  If we can't state their distribution by rule, then we can't
put them into a single phoneme, even if there are no minimal pairs.

In fact, minimal pairs are very hard to come by for the English [esh] /
[ezh] contrast.  All the minimal pairs I can think of are marginal for
one reason or another.

	obscure, archaic or elevated words:

		'ruche' / 'rouge'
		'leash' / 'liege'

	words of obvious foreign origin:

		'show' / 'zho'

	made-up words:

		'mesher' / 'measure'

	proper names, or derivatives of these:

		'Asher' / 'azure'
		'Aleutian' / 'allusion'
		'Confucian' / 'confusion'
		'shock' / 'Jacques'

There is one minimal pair that works for most of my British students:

	'assure' / 'azure'

But this doesn't work at all in my American accent, since I stress
'azure' on the first syllable -- a pronunciation that invokes giggles
or scowls from my students.

However, even if not one of these marginal pairs existed, it would
not be difficult to show that [esh] and [ezh] are distinct phonemes
in English.  That's because we have near-minimal pairs.  The following
work in my accent, and probably most of them work in all accents:

	'kosher' / 'closure'
	'vacation' / 'occasion'
	'thresher' / 'treasure'
	'fission' / 'vision'
	'nation' / 'equation'
	'masher' / 'azure'
	'pressure' / 'pleasure'
	'condition' / 'precision'
	'contrition' / 'derision'
	'motion' / 'erosion'
	'commotion' / 'corrosion'	
	'fuchsia' / 'fusion'
	'inflation' / 'invasion'
	'solution' / 'delusion'

Clearly, the choice of [esh] or [ezh] cannot be governed by rule.
In fact, there is probably no more economical way to account for
the distribution of these two sounds than to give lists of the
words containing them.  This observation is enough to establish
thet they must be distinct phonemes -- even if we have no minimal
pairs.

> [SF]

>> To show some sound difference is not phonemic
>> you have to show that it occurs in a *strictly* conditioned fashion.  If it
>> is not *uniformly* due to some identifiable set of conditioning factor,
>> then it is left as a phoneme.  This is how it is presented in all of the
>> best texts on phonology.

>  [PR]

>  Could you name a "best text on phonology", and cite a relevant definition of
>  phoneme from it?

I suggest the following:

Francis Katamba (1989), An Introduction to Phonology, London: Longman.
pp. 22-23.

Katamba cites the example of the African language Ewe, in which it is
apparently difficult to find minimal pairs for /f/ and /v/, even though
the two appear in near-minimal pairs with such a distribution that the
choice between them is impossible to state by rule.

> [PR]

> Sorry, I just cannot accept that. If /o/ is an IE phoneme, it should occur
> in true minimal pairs. I have this on the authority of a degreed linguist
> with whom I have consulted on this question. Your reluctance to accept this
> basic method of establishing a phoneme continues to amaze me!

This is *a* method of establishing phonemes.  But it is not *the only*
method of establishing phonemes.  If the distribution of two sounds
cannot be stated by rule, then they can't be assigned to a single
phoneme.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk