Phonemic split

[Henry M. Hoenigswald]

Traditionalists, including, via Jakobson, generativists, want their
segments bite-sized ([allo]'phones' Æ 'phonemes') and claim a privileged
existence in nature for that particular kind of  segment. Some also believe
they are painting a scenario of just what goes on phonetically and
sociolinguistically along putative lines of descent (such is Janda's
commendable but not necessarily feasible concern   [1999;  I thank the
author for an advance copy] and operate with two, and only two,  separable
events, one merger and one split,  when in fact, and despite what is said,
there are only homonyms and non-homonyms are to be handled.  This is  the
stuff of which the 19th century+  formulation of sound-change and the
(highly successful)  'comparative'  method is made where sporadic sound
change is excluded, not for any possible empirical reason but by
definition.  Notational preferences, such as the choice of some one
segmentation, do not matter for that purpose, homonyms being homonyms under
any notation.   Some think it elegant to imply, say,  that the Indo-Iranian
merging of IE /o/ with  /e/ after  /t/  is 'the same (rule?)' as the
merging of /o/ with /e/ after /kw/.   If, however,  we elect to make our
statements about  ..CV..  sequences rather than about   C..  and about
..V   separately,   /te/ and /to/ merge while  /kwe/  and  /kwo/ don't.
None of this directly describes events, let alone events  with a
chronology.-   See Hoenigswald 1999 [Lehmann Festschrift]  (where it would
have been better to speak noncommittally of 'entities' rather than
'phonemes' so as to forestall undesirable implications;  Bhat's critique
[Hoenigswald loc. cit., fn. 5]  is  instructive).  The question is what the
sociolinguistic welter has to be like in order to qualify for profitable
interpretation in the grand old framework of descent, replacement, merger..
and all the rest.

Henry M. Hoenigswald
908 Westdale Avenue
Swarthmore PA 19081-1804
<henryh at babel.ling.upenn.edu>
Tel: 1-610  543-8086