Return of the minimal pairs

[Larry Trask]

--On Sunday, May 20, 2001 1:58 pm -0400 "Paul S. Cohen" <pausyl at AOL.COM>
wrote:

[on non-initial [eng] in English]

> Actually, there's another possible view of this situation that's
> completely consistent with classical phonemics:  since [[eng]] and [h]
> are in complementary distribution in English (yes, I know about the
> Trager&Smith tradition of talking about [h] and [[schwa]] being
> allophones), and since they are, in some sense, phonetically similar
> (they comprise a class that can be characterized as "the back, continuant
> consonants"), why not put them together as allophones of /[eng]/ or /h/
> (whichever you like)? Occam's Razor would seem to demand doing so.  This,
> I would hope, shows off some off the inadequacies of classical phonemics.

Perhaps not necessarily.  Putting [eng] and [h] into a single phoneme was
proposed by somebody in print several decades ago, perhaps not entirely
seriously.  Apart from the violence this analysis does to native-speaker
intuitions, it founders on the rocks of phonetic similarity: [eng] and [h]
have no phonetic features in common apart from those few shared with all
consonants.

And phonetic similarity is certainly a necessary condition in classical
phonemics, and it was recognized as such: see, for example, page 108 of
Hockett's classic 1958 textbook.  In English, using a following apostrophe
to mark aspiration, we find that *each* of [p'], [t'] and [k'] is in
complementary distribution with *each* of unaspirated [p], [t] and [k].  On
distributional grounds alone, then, we have a free choice: we can group
[p'] with any one of [p], [t] and [k], and similarly for the other cases.
Even the most rigorous classical phonemicist never believed that these
several analyses were equally plausible, and therefore a criterion of
phonetic similarity is essential.  Of course, before the advent of
distinctive features, the notion of phonetic similarity could not be
defined in a principled way.  But features allow us to build a principled
criterion of phonetic similarity -- and any version of this criterion I can
think of rules out the assignment of [eng] and [h] to a single phoneme.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk

Tel: (01273)-678693 (from UK); +44-1273-678693 (from abroad)
Fax: (01273)-671320 (from UK); +44-1273-671320 (from abroad)