poll, questions about higher levels research, reac'n times

[From Dick Robertson] (930409)
  RE: 1) Gary's poll; 2) selection of higher level outputs; 3) problem with
modeling higher orders; 4)request for help in identifying a CEV

1) As far as the poll goes, I don't know how to vote: I wasn't convinced by
the pct advocates because when the debate began I already thought that the
disturbance can't be separated from the output - in any way that could
permit determining the values of the initial disturbance. Since I have to
admit that I couldn't follow all the math, that might get me assigned the
label of a naive, true-believer. But I have played a lot of "test for the
controlled perception" games, especially the three coins game, with
students, and it seems to me that it becomes obvious in an analog fashion
there.

After viewing the debate for what seemed an overlong time I was beginning
to feel concerned about what seemed like a drain of precious time from the
two main modelers in the group, but I'll take Bill Powers' word that it was
worth it in the end (if I read that correctly as his position). I did get
a lot out of Bruce Nevin's linguistic analysis.

2) One spinoff from the debate resurrected a version of a question that I
tried to raise a couple months back-concerning how a control system "knows"
what value to output to lower orders to correct a rising error in itself.
I saw at once that my question was nonsensical in regard to first order
systems, since their output simply increases or decreases to combat error
in the loop. But what about higher order systems, where it would seem that
output "selects" different variables in the system(s) below? For example,
consider a program-level system - like those demonstrated in the dialogue
between Greg Williams and Bill Powers (or for that matter, those between
Rick and Bill and Martin and Allan). Bill and Greg maintained that they
were controlling for definite outcomes in their discussion, I think
principle-level variables, and their programs kept varying word sequences
to match an RS that obviously seemed to be something like, "Ah, now I see
what you mean, you're right of course." In a case like that - how is the
output of the program "choosing" items from the sequence-level repertoire?
Is this simply not covered at all by the theory, or is my understanding of
how it works off the track?

3) I'd like to see more thoughts on the problems of modeling more complex
behavior than tracking. That seems to me at the relationship-control
level, which also seems on the border between "physiological" and
"psychological" phenomena. From that level on down variables seem to be
controlled over periods of time that permit the kind of sampling that is
needed (so far anyway) for computer time-samples. But the levels above
seem to be multiplexed in a way that a variable is brought under control
for a brief moment, then the whole system turns to something else, and
something else, etc. As Phil Runkel pointed out about my self-system study
the disturbance and correction lasted only a couple of seconds, at most,
and then the behavior of each s went on to wholly other things. It doesn't
look like there will ever be enough data points to model behavior at higher
levels in the same way that is done for tracking. Is an entirely different
methodology called for, or can't it be modeled?

Bill Powers' comments on an earlier post of mine on the grade control study
- that I should be questioning each student separately to learn what he or
she was TRYING to do - was a worthwhile admonition. But it opens a whole
host of new questions about finding precise measures of the answers.

4) Request for help identifying a controlled variable (To Rick Marken, esp.
and anybody else who is into reaction times and hierarchy levels)
I have been studying simple reaction times of various sorts and getting
different mean rates for different activities. A group of students in a
lab course all average around .2sec for stopping a sliding ruler
calibrated in msec. They also seem to average around .4sec on a cognitive
rehab task in which s has to wait for a particular display (2 or more
yellow squares) where 3 squares of randomly selected colors appear in
random arrangement on the screen.

My 3rd task is one in which you don't have to act, just observe. My mean
50% correct RT is about .075sec. I can see the difference between the
first two: in #1 you just have to notice movement (2nd order control?) and
press. In #2 you must notice one colored square, then scan to one or two
others to find whether the display meets the criterion (does that seem 3rd
order or higher?)

But in the 3rd task the s sets the speed and then a number (1-9) flashes
somewhere on the screen. After it disappears you have to type in what it
was. So in this one there is no movement command to execute during the
actual task. What is the controlled perception here? I can only pick out
a higher order CP - the desire to see the message, "correct" come on the
screen. Thanks for any help.

Best, Dick Robertson