("(Pierre Divenyi)" )

Subject: 
From:    "(Pierre Divenyi)"  <marva4!pdivenyi(at)UCDAVIS.EDU>
Date:    Wed, 23 Sep 1992 16:57:04 PDT

ucdavis!auditory(at)vm1.mcgill.ca

Re: Jont Allen's comment.

Plomp once defined speech as "a waveform slowly varying in intensity
and frequency". I have yet to find a more succint and to-the-point
characterization of speech: it implies the AM and FM nature of
the signal and sets ("slowly") some limits to the modulation
frequency.

The sinewave analog experiments (with due respect to Bob Remez
and others) sins against one of the criteria -- the AM quality.
Thus, despite any degree of speechiness of the percept,
sinusoids as formants will never satisfy any proponent of
various versions of the duplex theory, as an adequately controlled
analogy. (Doug Whalen, are you listening?) As a matter of fact,
I think that the comparisons can even mislead the uninitiated,
especially when the issue is perceptual separation of multiple,
simultaneous speech samples.

Jont, your comment is most pertinent to some of us who, during the past
few years, have been cooking up experiments in the hope of finding
a handle on explaining the "cocktail-party effect", and ways to
overcome its loss. As Sid Bacon and ourselves (independently) have
found it out, the effect may well be related to simultanously tracking the
envelopes of the multiple speech samples. If this is truly the case,
the sinewave analog controversy may be no more than a tempest in
a teapot.
	Pierre Divenyi