Sensation, perception, cognition

[From Rick Marken (940922.0850)]

Mary Powers (940921) --

The conventional way of looking at the mind distinguishes sensation,
perception and cognition. These have been considered different abstract
entities, and it is because they have been that you think it
confusing to use one word for all of them. An analogy would be
viewing metabolism, burning, and the turning yellow of last
week's newspaper as different phenomena. There are important
differences, but discovering or inventing or abstracting their
fundamental similarity, the process of oxidation, opens up a lot
of new territory. Calling it all perception (like calling it all
burning) is the first step into that new territory.

This is a wonderful, brilliant observation, Mary. As an ex-conventional
psychologist, I can attest to the importance of the distinction between
sensation, perception and cognition in conventional psychological theories,
especially cognitive-type theories. Your post made me realize that this
distinction is comfortably compatible with the input-output, "assembly line"
type models of behavior that are prevalent in psychology today. Sensations
are the raw materials of this process. They are assembled into usable
"modules", which are perceptions. These perceptions are then assembled,
manipulated, and, in general, used by the cognitive "processes" which are
ultimately aimed at constructing "behavior", such as a "decision" or a
"choice" -- ie. an output.

The PCT model changes one's perspective completely. The "product" of the
control process imagined by PCT is always a perception; specifically, it
is the value of a perceptual signal coming out of a perceptual function. The
outputs produced by the control process are part of the means of
"constructing" (really, controlling) these perceptions. So the products of
controlling are the states of perceptual variables: perceptions. It is in
this sense that "it is all perception". What we are "doing" (according to
PCT) is producing desired states of perceptual variables of different
_types_; we are producing intended inputs, not outputs. In PCT, the
distinction between sensations, perceptions and cognitions is simply a
distinction between different types of products of the control process -- a
distinction that is not nearly fine grained enough; all behaviors (where the
terms "behavior" refers to the intended results of controlling) are
perceptions; they are perceptions of different types.

It does take a while to get used to thinking about behavior this way; but
when you do get used to it, you get a fascinating new perspective on what you
and other people are doing; controlling many different _types_ of
perceptions. These perceptions can be as complicated as programs, principles
and system concepts (which have all, collectively, been bunched
together in conventional psychology under the rubric of "cognition"). The
complexity of the perceptions we can perceive is illustrated nicely by Bill
Powers (940921.1130 MDT) in his reply to Bruce Buchanan (940920.2200) --

Try thinking the thought "I am not thinking this thought."

Whatever it is you "think" when you think this thought (perhaps just the
sound of the words, perhaps some image of yourself not thinking a thought,
etc) it a perception. And it is controllable; you can think that particular
thought (produce that state of the perceptual variable for yourself) again
and again, at will.

Perhaps, instead of calling his book, "Behavior: The control of perception
in order to control values" (as Mary jokingly suggested in a comment about
Locke) Bill should have given it the mellifluous title:

"Behavior: The control of many types of hierarchically related perceptual
variables simultaneously."

Best

Rick

[Martin Taylor 940926 12:15]

Rick Marken (940922.0850) commenting on Mary Powers (940921)

What a pleasure it is to commend one of Rick's postings!

Commenting on Mary's posting, Rick says:

Your post made me realize that this
distinction is comfortably compatible with the input-output, "assembly line"
type models of behavior that are prevalent in psychology today. Sensations
are the raw materials of this process. They are assembled into usable
"modules", which are perceptions. These perceptions are then assembled,
manipulated, and, in general, used by the cognitive "processes" which are
ultimately aimed at constructing "behavior", such as a "decision" or a
"choice" -- ie. an output.

The PCT model changes one's perspective completely.

Yes, and until you wrote this, I hadn't appreciated that the difference
is very much in the direction of seeing "boxes-and-arrows" models as
irrelevant. Of course, a network of ECUs can be interpreted as a set
of boxes and arrows, but each box in a control network does, in principle,
THE SAME THING. It controls its perception, and every perceptual signal
has the same character, whether it be related to what a conventional
psychologist would call sensation, perception, or cognition. In the
"boxes-and-arrows" formulation, each box has a job quite different from
that of any other box. There is a box for syntax analysis, a box for
phonetic labelling, a box for lexical lookup, and so on, with MYSTERIES
inside each box. Each one acts on its complicated input and provides a
complicated output.

Connectionists (neural network) people have long argued against the
boxes-and-arrows view of brain function, but not from the same position
as--shall we call ourselves--connectionist perceptual control people.
In PCT there is room for groups of ECUs that an outside observer would
say act on different kinds of perception, but this only means that the
kinds of CEV they affect in the outer world differ. No ECU is a simple
input-output device like a box in a boxes-and-arrows model.

Great insight, Mary and Rick!

Martin