[from Jeff Vancouver 961125.09:55 EST]
Well I tried to program the spiral experiment using Visual Basic. I was
completely unsuccessful. I stopped when I could not figure out how to
record the X and Y coordinates of the mouse. Instead I compromised with
a less than adequate solution, but it approaches what I seek.
Below are the data from several different type of runs. I edited the
"visible" word to label the condition. Recall that in the blanked
condition one still has kinesthetic and vision feedback of the mouse, as
well as feedback from the screen when the blacked out cursor passes over
the edges of the target. I had initially suggested blanking the entire
screen, but doing it during runs not as a run. So, I closed my eyes to
simulate blanking the screen (I still get the kinesthetic feedback, but
that is it. I tried three variations on this theme. First, I alternated
eyes open, shut, open for 10s (called Alt 10s). Time was measured using a
paced "one and two and . . . " counting to myself. Then I tried eyes open
and shut alternating every 2s (called Alt 2s). Finally, I tried eyes open
for 3s and closed for 1s (called 3,1,3). I discuss the results below.
jeff
6.31 visible
6.21 visible
6.01 visible
117.72 Alt 10s
6.44 visible
115.10 Alt 10s
28.38 Alt 2s
36.67 Alt 2s
32.29 Alt 2s
29.16 Alt 2s
11.66 3,1,3
10.41 3,1,3
11.87 3,1,3
11.38 3,1,3
32.79 3,1,3 Mouse got stuck
12.19 3,1,3
On the second to last run the mouse ball got stuck on some dirt on about
the last 4s of the run so it did not move.
It is perhaps interesting to note that I did substantially better in the
Alt 2s then the Alt 10s even though my eyes were closed half the time in
Alt 2s and only a third in Alt 10s, the ability to correct more often once
my eyes were opened accounts for the better performance (given that when I
opened my eyes I was always making a substantial correction -- esp. in
the Alt 10s condition).
Otherwise, the data clearly show the advantage of closed-loop control.
I think that the data (mine and others) clearly demonstrate that we use
closed-loop control. If there are any open-loop only advocates out
there, I invite a different interpretation.
Now that this point has been made, I would like to get on to the point I
was trying to make. Unfortunately, I am at the mercy of the Pascal
programmers.
The point I was trying to make is that control with the mouse, screen, or
whatever blanked (i.e., model-based control), is better than random,
static (no movement), or linear trajectory performance. Thus, as a
comparison, the average RMS from those models needs to be added to the
output.
I am not looking for an "excuse" here, I am looking for a reconciliation.
I think that reconciliation is in the nature of the transfer function _we
all think_ is created to track the spiral. That given sufficient time to
create such a function, it can create a controllable perception of the
movement of the target even without the target being visible. This
function will not be as good as it would with on-line data, but it will be
better than if no function were there (as indicated by comparing to
random, no movement, or linear/trajectory movement). Any takers?
Later
Jeff
···
_________________________________________________________________________
Jeffrey B. Vancouver
Assistant Professor Phone: (212)998-7816
Department of Psychology Fax: (212)995-4018
New York University e-mail: jeffv@psych.nyu.edu
6 Washington Pl., Rm 578
New York, NY 10003