Graziano et. al

[From Rick Marken (2002.06.10.0930)]

Bill Powers (2002.06.06.1224 MDT) --

Concerning the Graziano et. al. paper, check out the reference from

Ranson

and Clark starting at the bottom of p. 121 in B:CP. Looks as if Ward

did a

very similar experiment prior to 1947.

I'm afraid I forgot to take a look at it this this weekend. But I did
think of something that _might_ be worth doing. Couldn't you reproduce
the Graziano et. al. results relatively easily in your "Little Man"
program? It might be neat to have a version of the Little Man where the
"user" could inject signals of various durations. These signals are the
equivalent of the electrical signals used by Graziano et al (and Ward,
too, I suppose). The signals (depending on their magnitude) specify the
x,y,z target location for the little man's fingertip.

The "surprising" finding in Graziano et. al. is that longer duration
signals lead to very different results when stimulating the same
cortical neurons than shorter duration signals. This variant of the
Little Man demo could show that this is precisely what results when the
stimulus signal is setting the reference for a perceptual input (eg.,
x,y,z finger at position). With short signals all the Little Man would
do is produce apparently purposeless "jerky" movements. With a longer
signal we should see what Graziano et. al. saw; smooth, purposeful
movements to a goal position.

I would do this myself but 1) you are a heck of a lot more familiar with
the Little Man code than I am and 2) I am actually working on another
modeling project (not baseball; something else that I think I'll
describe at the meeting) where I am successfully reproducing some very
interesting data using a simple two level PCT model.

Best regards

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

[From Bill Powers (2002.06.10.1753 MDT)]

Rick Marken (2002.06.10.0930)--

>Couldn't you reproduce

the Graziano et. al. results relatively easily in your "Little Man"
program? It might be neat to have a version of the Little Man where the
"user" could inject signals of various durations. These signals are the
equivalent of the electrical signals used by Graziano et al (and Ward,
too, I suppose). The signals (depending on their magnitude) specify the
x,y,z target location for the little man's fingertip.

The Graziano et. al. results suggest that position reference levels are
determined not by magnitude of a signal but by where in a perceptual map a
reference signal is applied. The basic closed-loop architecture still
applies, but the detailed organization is not the same as in the simple PCT
model. Another way to look at this is that there is a separate control
system for every final position that the monkey's hand can reach! To do
that in a simple way requires a clever organization which I have yet to
figure out.

This leaves me wondering many things. What would have happened if the
experimenters had simply turned on the stimulating signal and left it on
for ten or twenty seconds? One hypothesis is that in one of the cases,
where the hand closed, turned, and moved to the mouth, it was not simple
end-point control but control of an _event_ of the sort I proposed for the
fifth order. Would the event have repeated as long as the signal remained
on? That would tell us a lot about the nature of the control system
receiving the signal.

What would have happened if the signal strength (frequency) had been
changed from one episode to the next? There is confusion between the
dimensions of behavior associated with position in a map and other possible
dimensions associated with signal strength.

There were some experiments where the hand simply moved to a spatial
position rather the carrying out a complex set of control processes in
parallel. Did these loci lie between the ones evoking complex patterns and
the ones at a lower level near the central sulcus?

?The "surprising" finding in Graziano et. al. is that longer duration

signals lead to very different results when stimulating the same
cortical neurons than shorter duration signals. This variant of the
Little Man demo could show that this is precisely what results when the
stimulus signal is setting the reference for a perceptual input (eg.,
x,y,z finger at position). With short signals all the Little Man would
do is produce apparently purposeless "jerky" movements. With a longer
signal we should see what Graziano et. al. saw; smooth, purposeful
movements to a goal position.

I'm not optimistic about any of these people suddenly seeing how the analog
control model works. I haven't been able to get anyone in these fields to
look at or discuss the little man model or talk about the concept of
continuous control. What good would it do to set a position reference
signal in the LM model and show the difference between turning it on for a
short time and a long time? People who won't look at the basic model won't
look at that variation on it either. I'll be happy to let someone else tilt
at that windmill.

Best,

Bill P.

···

I would do this myself but 1) you are a heck of a lot more familiar with
the Little Man code than I am and 2) I am actually working on another
modeling project (not baseball; something else that I think I'll
describe at the meeting) where I am successfully reproducing some very
interesting data using a simple two level PCT model.

Best regards

Rick
--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org