What is attention?

[From Kent McClelland (960428.1700 CDT)]

Having just got back from a three-day Midwest Faculty Seminar (attended by
a bunch of cognitive psychologists and philosophers) on "Mind and Brain,"
I've been reading with some interest the collection of recent posts on
consciousness and awareness. Here's my take on these problems. I'd like
to focus on what "attention" is and how it might be produced.

Suppose for convenience we imagine the HPCT hierarchy as split into several
perceptual "modalities," each a making up somewhat separate neural
hierarchy. One modality would be vision, another hearing, and a third
perhaps the combination of taste and smell. Perhaps the fourth might be
labeled "body-receptors external" and would include perceptions of touch,
heat, and the proprioceptive system which tells joint position and tensions
on tendons. A fifth modality might be "body-receptors internal" combining
perceptions of heart rate, breathing rate, stomach acidity, all those gut
feelings we associate with emotions.

Within any of these modalities there would be a limit on the number of
perceptions we can stabilize (control at a given unchanging reference
standard) during some patch of time. Trying to stabilize two perceptions
of the same perceptual order within the same modality at the same time
probably results in conflicts (like trying to say two words
simultaneously--it doesn't work), so the hierarchy won't be able to
independently stabilize two same-order perceptions within the same modality
at once.

And we probably can't stabilize two perceptions of different orders within
the same modality, either. In order to stabilize any given perception, the
lower-order perceptions which contribute to it must be able to vary freely,
so that any disturbances to the stabilized perception can be countered.
And if a stabilized perception contributes to an active higher-order
perception within the modality, this will have to be temporarily put on
idle (perhaps with its gain lowered but ready to resume control immediately
at the current reference standard) for the lower-order perception to remain
fixed. Since the higher-order perceptual systems operate on a slower time
scale (and often intermittently in terms of the time scale of lower-order
systems) having them temporarily "on hold" might not be a problem.

Pretty often, I would guess, we operate in a "quiet" enough environment
that some of our lower-order perceptions are stabilized without any effort
on our part. I look at a wall and note the color. I look a moment later
and see the same color, not because my actions have countered any
disturbances in the meantime but just because the lighting hasn't
significantly changed. As Bill Powers puts it in B:CP, such systems can
operate in "observation mode." When we don't have to actively stabilize
the physical perceptions in a given modality, our perceptual hierarchies
are then at liberty to stabilize other perceptions within that modality in
"imagination mode." In other words, we can carry on the interior talking
and listening or viewing of minds-eye imagery or summoning of imagined
feelings that we call "thinking."

With all these processes going on in several perceptual modalities, what if
there were an "observer" system within the brain that got "reports" on the
currently stabilized perceptions (whatever their perceptual order) from
each of the perceptual modalities? Perhaps the function of that system
would be to resolve possible conflicts between the perceptions being
stabilized in different modalities by picking one modality to "pay
attention to" at any given moment, so that any conflicts which arose would
be resolved in its favor. What I'm groping to describe is an "attention
system" that moves the camera of attention across all the possible inputs
and fixes it from moment to moment on one perception or another.

If this observer system had a memory, it could lay down a "memory track"
that associated the simultaneously stabilized perceptions from different
modalities. Perhaps the memories would be somehow indexed by elapsed time.
Then playing this track back in imagination would be equivalent to our
short term memory--the way we reconstruct what was happening to us a moment
ago or five minutes ago or yesterday.

Perhaps the same "observer" system or a closely associated system could be
getting simultaneous "reports" on the large uncorrected perceptual errors
(if any) that are currently occurring in each of the modalities. The
decision about which modality currently needs attention could then be
somehow resolved by comparing errors and seeing which is the greatest.

From here on, my proposal, I think, starts to tie in with Bill Power's

ideas about the "reorganization system" and learning.

I could spin this fantasy out a little further but don't have the time
right now. Let me know if anyone can see some plausibility to this

By the way, my conclusion from listening this past week to the latest
reports on current brain research is that those guys are still pretty
nearly clueless, though some are doing interesting work. Without the PCT
framework to put things in perspective, several of the high-powered
researchers who talked to us seemed (to me) needlessly confused by their
own results.