Re: An Auditory Illusion

[Al Bregman <bregman@xxxxxxxxxxxxxxxxxxxx>]

On Mon, 28 Apr 1997, Richard M Warren wrote:

> The question I have concerns a puzzling variant of verbal transformations
> that occurs when each ear hears the same repeated word dichotically with
> the delay of 1/2 the repetition period separating the repeating
> statements in each ear -- the illusory changes occur independently on
> each side, so that different forms can be heard simultaneously at each
> ear. ...  the question is:  If there is a single speech processor, how
> can the same stimulus be heard asynchronously as two forms that are
> sometimes dramatically different?


Dear Dick

The phenomenon implies that at least some part of the "speech processor"
is not serial.  It seems that the stage responsible for extracting
phonetic information from the acoustic signal can carry this out in
parallel on two or more signals.

However, since all the energy enters the ears mixed together, what is a
"signal"?  I imagine that a "signal" is whatever the auditory system has
grouped into a single stream.  That is, extraction of phonetic categories
goes on in parallel in concurrent streams.  This parallel extraction can
be done on the basis of spatial location, as in the demonstration
described in your e-mail.  But I would also expect, since pitch is a good
basis for the segregation of speech streams, that if you replicated the
experiment with a man's and a woman's voice both coming from the same
location, you would also see independence in the verbal transformation
effect (VTE) in the two voices. Perhaps low-pass filtering and high-pass
filtering of a man's voice would also create a pair of signals that, if
looped asynchronously, would display independence of VTE.

There is presumably a serial (capacity-limited) process that operates on
the interpreted phonemes, and possibly on other interpretations that have
been extracted in parallel.  The serial process may be the one that
synthesizes a final meaning based on the parallel-processed features.
Evidence for a serial stage is that we cannot understand two concurrent
conversations in any detail.

What about the counter-argument that a SERIAL processor could keep track
of two streams of speech by rapidly alternating back and forth between
them?  I think that this could be ruled out as a mechanism.  The single
speech processor would have to not ONLY alternate back and forth, but
would have to "get tired"  of phonetic interpretations independently for
the two signals.  This seems unlikely.

I think, then, that your experiments on VTE force us to the conclusion
that the speech processor is not a homogeneous serial mechanism.  It seems
likely that there is a stage that is preattentive and operates on
concurrent auditory streams in parallel.

Regards,

Al

----------------------------------------------------------------------
Albert S. Bregman,  Professor,  Dept of Psychology,  McGill University
1205  Docteur Penfield Avenue,   Montreal,  Quebec,  Canada   H3A 1B1.
Phone: +1 514-398-6103  Fax: -4896 Email: bregman@hebb.psych.mcgill.ca
Lab Web Page: http://www.psych.mcgill.ca/labs/auditory/laboratory.html
----------------------------------------------------------------------