Andreas Johansson:
> Paul Roser wrote:
> >How clicks are counted, or even how phonemes are counted, is another
> >question entirely, and I know that And has certain views on this topic
> >(with
> >which I am in sympathy) that one ought to specify the environment one is
> >counting phonemes in, eg prevocal/word-/syllable-initial or syllable-final
> If I'm getting And's point correctly, one 'd get multiple sets of "phonemes"
> for each language, none of which's size is directly comparable to a
> traditional phoneme count. So, AFAICT, accepting And's idea would leave the
> statement "!Ui has over 140 phonemes" vacuous, rather than factually wrong
> and to be replaced with a statement "!Ui has [a certain number] phonemes"
> Am I missing something?

That's right. "My idea" is to boost the size of the phoneme inventory
while simultaneously doing away with local phonotactic constraints.

But above and beyond that, and not especially associated with me, is the
idea that it is better to see the inventory of elemental phonological
units as subsegmental rather than segmental. For most phoneme inventories,
not a lot is gained (at a superficial level) by treating segments as
complexes of subsegmental elements. But big inventories are usually big
by virtue of systematic combinations of subsegmental elements -- e.g.
the Khoisan lg in UPSID has all vowels in long and short and oral and
nasal versions, and each basic click has several different versions
according to the secondary articulation. It's more insightful, in
analysing the fundamental building blocks of this phonology, to
separate out length and nasalization as discrete elements, and
similarly for clicks and their secondary articulations. This move would
shrink the inventory of phonological elements down to a fairly ordinary

It's certainly misleading to describe a language with the basic 5
vowels, in long/short, oral/nasal, breathy/creaky/modal versions
as having a lot of vowels (viz 5 * 2 * 2 * 3 = 60). And similarly
for consonants. We need some different criterion to measure the
phonetic richness of a language (e.g. what proportion of the IPA
chart systematically corresponds to phonetic realizations of
phonological forms in the language in question). I guess that
the languages that win on that score are found in the Caucasus
(and much beloved by Bfowol).