Re: An Auditory Illusion

[Al Bregman <bregman@xxxxxxxxxxxxxxxxxxxx>]

May 16, 1997

Dear Dick and List

I have another comment on your demonstration of the separate
habituation of left- and right-side words cycling a half-step out of
phase.  I guess you never considered the religious significance of
your finding.  Let me point it out.

I think that the problem of maintaining of two concurrent mental
representations connects with a larger problem in cognition, namely
what the neural basis is for recognition and representation.

The dominant view in cognitive psychology is that the representation
of an abstract entity (a word, for example) is nothing more than a
node (engram, cell assembly) and that recognition is simply the
activation of that representation.


But consider the following two cases:

1.  One sees a red ball on a blue table.

2.  One sees a blue ball on a red table.

According to a "node" theory, in both cases the nodes representing
red, blue, ball, on and table are activated.  What then is the
difference?  The two cases require a different arrangement of the same
ideas, but node-based theories cannot express this.

A computational system that used symbols could arrange them in two
structures as follows.

Structure 1:  (RED colour-of BALL) ON (BLUE colour-of TABLE).

Structure 2:  (BLUE colour-of BALL) ON (RED colour-of TABLE).


Notice that the symbols are ones that were already in the system: red,
blue, on, etc,, but the symbolic structures are new and unique for each
case.

This is easy for a symbol-manipulating system to do, but it is not
clear to me how it is possible for a node-based system to form
momentary structures that combine concepts to represent unique events,
and then dissolve these structures and form new ones as the situation
changes.

There is a similarity to your two concurrent voice situation.
A symbol-manipulating system would have no trouble in forming two
structures:

Structure 3.  (VOICE at LEFT) SAYS X       - where X is a particular word

Structure 4.  (VOICE at RIGHT) SAYS X

However, a node-based recognition system would have to fire the nodes
VOICE, SAYS, LEFT, RIGHT, and X.  According to the preceding argument
there is no node-based mechanism (that I know of) that can say X is
simultaneously at the left and at the right.

Even if the words at the left and right were *different* (X and Y),
there would be a firing of nodes VOICE, LEFT, RIGHT, SAYS, X, AND Y,
with no structure to say which word was on the left and which on the
right.

Node theory has problems because it fails to make a distinction
between types and tokens.  Types are the general concepts, such as
LEFT and VOICE.  Tokens are elements that refer to them but are free
to bind into structures.  There can be as many tokens as you want
referring to a specific type, so you can have the same token more than
once in a given structure; e.g.,

Structure 5.  (RED color-of TABLE) on (GREEN color-of TABLE).

               where TABLE is used twice.


Even if node theory could think up some way of binding the WORD-X node
to *both* LEFT and RIGHT nodes, there would still be only a single
WORD-X node.  I can't imagine how this one node could tire at two
different rates.  On the other hand, two separate symbolic structures
(i.e., descriptions) *could* tire at different rates.

It is generally accepted that node-based recognition systems are the only
ones that are adequate models of the brain's recognition capacity, as
distinct from recognition by computer.  If this is true, your data prove
that the brain as we know it could never form representations that both
minds and computers can do with ease.  This is important for philosophy
and surely for religion.  Computers would have to be allowed into heaven,
but not brains.

Al