Peter T. Daniels scripsit:

> Do we define "phoneme" no more generally than "a phoneme of a given
> language is ..."? If we did that, how would we be able to identify the
> phonemes of a previously unanalyzed language?

From what I understand, linguists do not deal in phonemes any more,
except in the restricted sense in which physicists deal in Newtonian
physics: a useful approximation of pedagogical value that is known to
be inadequate to explain the phenomena.

Nor do I see how phonemes (as opposed to "the phoneme") can be defined
without reference to a particular language.

> Note that, mutatis mutandis, such a definition of "phoneme of a given
> language" wouldn't even work.

No more it would. But then writing, unlike speaking, is consciously
learned behavior; learned, furthermore, with reference to a specific
tradition. When the tradition is lost, all such judgments become
correspondingly uncertain, taking on the status of reconstructions
rather than primary data.

If you don't like my pirating of the term "grapheme" for this purpose,
I invite you to provide a substitute term.

--
Híggledy-pìggledy / XML programmers John Cowan
Try to escape those / I-eighteen-N woes; http://www.ccil.org/~cowan
Incontrovertibly / What we need more of is http://www.reutershealth.com
Unicode weenies and / François Yergeaus. jcowan@...