Delays in SD experiments

[From Bill Powers (950309.1210 MST)]

Rick Marken (950309.1020)--

     If you have my "Mind Readings" book, look at the paper called
     "Levels of intention in behavior". Bill and I found that it takes
     about 1/2 second to recover from a polarity reversal (which is
     "signaled" by exponentially increasing error). It also takes about
     1/2 second to change the relationship (tracking vs mirroring)
     between target and cursor movement after an auditory signal. In
     your case it takes about 1/2 second to change the reference
     position of the cursor after a signal.

I think we're seeing a slightly longer time delay than appeared in the
reversals experiment. Bruce got about 500 milliseconds, and I got about
610; you'd better fire up the experiment and see what you get (the delay
has to be adjusted by hand right now, but I'll try to come up with an
automated determination if Bruce doesn't beat me to it). As I recall,
you found a little less than 500 milliseconds for reversals, didn't you?
Maybe not enough less to make a difference.

This very long delay after a change in a low-level variable (sensation
level?) suggests that the control process is (as you suggest) a high-
level one, and that the low-level signal is actually passing through all
the intermediate levels. It would be nice to have a modification of the
experiment in which the mouse is used to keep the color constant; the
well-practiced reaction time should be very much less, more like 10/60
sec. Or your perceptual experiment might tell us the same thing.

···

---------------------------------------
I'm wondering about a "target designation" process. In the real human
visual system, positions are probably handled in terms of place in a 2-D
map. In a field of targets, we somehow select one of them, and
immediately, or when desired, the control system brings the place where
the cursor is over to the place where the target is.

I had suggested that one control system might bring the target to the
center of vision, after which another control system brings the cursor
to the target that was selected this way. We would really need apparatus
for measuring direction of gaze to do this properly. The eye movements
wouldn't necessarily have to achieve a perfect registration of the
designated target in the center of vision, because I suspect that
control of the distance between target and cursor could be accomplished
even with the eyes fixated on one point. But we need to think about how
a person can pick out one object in a field of objects and control
relative to that one rather than any other. This is a way of setting a
reference signal that doesn't lend itself well to our scalar-signal
model -- even though the scalar model is probably equivalent to a map-
based model in its characteristics, or close enough.

It seems that every time we start getting into modeling a phenomenon,
the questions multiply. What we really need is a lab and the time for
exploring all the ramifications that show themselves whenever we get
serious. Wonder how Fred Good's search for support is doing.

And have you had a chance to try compiling the Pascal models on your
Mac?
--------------------------------------------------------------------
Best,

Bill P.

[From Rick Marken (950309.1300)]

Bill Powers (950309.1210 MST) --

I think we're seeing a slightly longer time delay than appeared in the
reversals experiment.

Actually, the reversal was about 500 msec; it was the switch from tracking to
mirroring (and vice versa) based on a siganl tone that was only 420 msec. But
the reversal time is probably comparable to the tracking/mirroring time
because it must take some time for the person to notice (based on increasing
error or whatever the perception might be) that the reversal has occurred. So
I suspect we're looking at 420 msec for the switch in both cases. Of course,
this estimate of switching time is based on my data and I'm one quick
study;-)

Bruce got about 500 milliseconds, and I got about 610; you'd better fire up
the experiment and see what you get

Aw, do I hafta? That means collecting data and all that stuff, doesn't
it? Why do I have to go through all that trouble when I already know how it
works;-)

What we really need is a lab and the time for exploring all the
ramifications that show themselves whenever we get serious. Wonder how Fred
Good's search for support is doing.

Me too.

And have you had a chance to try compiling the Pascal models on your Mac?

Yes. I can see that it will not be trivial (at least, not for me) but I
think I can have Mac versions of the PC Turbo programs before the next CSG
meeting.

Best

Rick