[From Bill Powers (990711.0617 MDT)]
Rick Marken (990710.1940)--
But there is a fairly large base of research on at least
one kind of control process: manual tracking. Here is one
possible experiment; it tests the notion that what is stored
in memory is a perceptual signal, not an output signal
(B:CP Fig 15.2).
Have a subject keep the cursor aligned with a sequence of
4 cursor positions. Call the positions 1, 2, 3 and 4. On each
trial the target occurs for a couple seconds in a different
sequence of positions: 1,2,3,4 or 1,3,4,2 or 1,2,1,3 or
whatever; the subject is to keep the cursor under each target
position that appears. The cursor is being subject to continuous
disturbance while this is being done. The subject must move a
mouse (output) appropriately to keep the cursor (perception)
controlled (aligned under each target position).
etc.
Excellent suggestion. It's funny how much more interesting the conversation
gets when real experiments that someone could actually do are proposed. I
have another suggestion: write this up as a stand-alone PC or Mac program
(or both) and pass it around to people who promise to run the experiment
with at least two subjects and send you the results.
I did another experiment with some similar elements, not concerning
sequences but definitely concerning whether people learn to produce outputs
or to control inputs. This one was published in
Levine, R. L. and Fitzgerald, H. E., Editors (1992) Analysis of dynamic
psychological systems, Vol. 2; Chapter 13: A cognitive control system. p.
327-340.(New York: Plenum Press)
This experiment involves using a mouse to move a cursor to point to the
right answer to an arithmetic problem. Numbers from 0 to 22 are listed down
the right side of the screen. On the left, a problem such as "7 -3 = ?" is
presented. The subject moves the cursor up or down to the answer (here 4)
and holds it there until the computer beeps to indicate the correct answer
(after a delay of 1 second). A series of 50 problems is presented, once
during the first experimental condition, and again during the second
experimental condition.
In the first experimental condition, the mouse position directly affects
the cursor position. Thus whenever the participant makes the cursor point
to a particular number, the mouse and the arm that moves it come to the
same position. Under these conditions it could be said that the subject
learns a motor act that provides the right answer to each question, just as
if the answer were being "emitted" using writing or speaking muscles.
Learning how to answer the question could be viewed as learning to emit the
right response to the stimulus provided by the written problem.
The results of the first experiment appear to bear out this interpretation.
The first figure is a plot of the handle position that corresponds to the
answer pointed to by the cursor for each trial. The plotted points lie
exactly along a straight diagonal line, showing that the mouse position
corresponds exactly to the right answer on each trial. The correlation of
mouse position with the correct answer is 1.00.
In the second experiment, the conditions are exactly the same except that a
disturbance is applied to the cursor position along with the effect of the
mouse position. Now the position of the cursor that is used to indicate the
right answer is determined by the sum of the mouse position and the
magnitude of a disturbance.
The program is set up so that for each problem, a random disturbance with a
magnitude between 0 and 22 is generated first; then an arithmetic problem
is constructed such that the mouse position required to indicate the right
answer will correspond to this randomly selected number.
Results: After 50 such problems have been generated and the answer
indicated, a second plot of handle position versus right number is drawn.
Now, of course, the plot is a random scatter, exactly as random as the
series of random disturbances generated for each problem. In a very long
series of problems, the correlation of handle position with right answer
would be guaranteed to approach 0.00.
This proves without a doubt that a person does not learn to emit a series
of actions as answers to arithmetic problems, but learns to control
perceptions of the right answers, altering actions exactly as required to
make the perception match the right answer.
I'm going to post this PC program on my FTP page with a link from my Web
page (www.frontier.net/~powers_w). This version has only 20 trials per
position, to save the participant from terminal boredom. The name of the
self-extracting zipped program file will be lrn.exe. Download to its own
directory, then run it to expand it. The runnable program will be
plenum.exe (named after the publisher).
P.S. Important note: after each plot, tap the space bar to go on.
PPS: The web page and ftp page are updated.
Best,
Bill P.