[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE/ an auditory illusion


Your proof that Dick's auditory illusion constitutes the deciding vote
for sending computers, rather than brains, to Heaven, may be to the
point but I have to remind you that such a view seems to depart from some
earlier ones you expressed in your classic 1991 opus. It appears
pretty clear that, when both Voice A and Voice B say (simultaneously) "X",
we have a Greek chorus (of two) which, through a single loudspeaker at
least, should form a single stream because both voices share the same
temporal structure which must prompt the system to group all components
together. So, all this is a matter of grouping -- i.e., segregation.

Which brings me to the following structure consisting of two
intermingled events:

        Talker A says "X", while
        Talker B says "Y".

If the (male) listener happens to perceive the structure * (asterisk as
in linguistics)

        Talker A says "Y", while
        Talker B says "X",

is he considered as having segregated the two events?
Before taking a general vote on this, let me just forcefully express my
negative opinion. No, because Talker A is the listener's wife and "X"
was the utterance "Come here, love", while Talker B was another, definitely
attractive female unknown to the wife. Utterance "Y" is irrelevant --
if the listener responds as if he obtained the above (incorrect) percept,
the next thing he receives will be the divorce papers.

So, symbol-manipulating systems can save your marriage while node-based
ones can, at best, serve you with an alibi. Forget heaven.

It was thanks to recognizing the subjective utility of discriminating between
the two above structures that I proposed at the 1995 Mohonk meeting,
and have continued to do ever since, that it is essential to associate
a particular **value** (e.g., blue, red) with a particular attribute
(=type, e.g., ball, table) for a scene to be correctly analyzed.
In other words, correctly recognizing the mere PRESENCE of
types and values does not lead to segregation of streams/events UNLESS
the type-value association is correct.

The positive fallout of all this is that segregation can be easily measured
with a psychophysical paradigm (e.g., 2AFC) and can be learned through feedback.

        Pierre Divenyi

Pierre Divenyi             Speech and Hearing Research (151)
                           V.A. Medical Center, Martinez, CA 94553, USA
Phone: (510) 370-6745;     Fax: (510) 228-5738
E-mail :                   PDivenyi@ucdavis.edu