Ambiguous figures: DynaNeck 3

[From Bill Powers (2004.05.10.1546 MST)]

More work on the dynamic Necker cube.

1. As suggested by Martin Taylor, the plot of the angle errors in x and y
now has the average subtracted out so you'll see a plot no matterf what
orientation you choose for the cube.

2. As I more or less promised, you can now set the frame rate to make the
blanking bars less noticeable. This is a temporary fix until I find out how
to synchronize the display with the vertical retrace interval. Use the
slider bar (frame rate shown in label). When ready to start the run, click
on Start Run. That panel will disappear. Quickly put the cursor in the
center of the figure to start, then start controlling.

3. As suggested by Rick Marken, the output text file now includes more data:
   b. Each following line (numbers separated by commas, lines end in cr, lf) :
       x, x error
       y, y error
       dx, x disturbance
       dy, y disturbance
       mx, x mouse position
       my, y mouse position

       NOTE: x = dx + mx, y = dy + my.

4. The size of the disturbances has been increased so the mouse has to move
by an appreciable amount to keep control. I supposed this ought to be
another adjustment.

I think that a sudden drop of the error (x or y) to nearly zero is a good
indicator of when the person makes the internal switch required to restore
negative feedback.

5. Oh, yeah -- I renamed the executable file to DynaNeck.exe for you.

Best,

Bill P.

DynaNeck2.ZIP (209 KB)

···

a: First line, number of data points. (selected to make runs last 1 minute)

[From Rick Marken (2004.05.10.2200)]

Bill Powers (2004.05.10.1546 MST)

More work on the dynamic Necker cube.

I ignored the variations around the vertical axis and just tried to
control the horizontal alignment of the pairs of vertical lines that
make up the cube. When I did this I got very clear plots of the
exponential runaway when the cube perceptually reversed. This runaway
could be seen in the top error plot (the one of x or horizontal error),
which was pretty flat during when I had the horizontal alignment of the
vertical lines under control. When the perceptual reversal occurred
the error curve did a brief, exponential increase -- how long I don't
know because I haven't done the data analysis yet -- that ends abruptly
when a higher level system is able to re-adjust the polarity of the
connection of output to the new perception.

A very elegant demonstration of hierarchical control, I think.

Best regards

Rick

···

---
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

[From Bill Powers (2004.05.11.0245 MST)]

Rick Marken (2004.05.10.2200)--

I ignored the variations around the vertical axis and just tried to
control the horizontal alignment of the pairs of vertical lines that
make up the cube. When I did this I got very clear plots of the
exponential runaway when the cube perceptually reversed.

That suggests a button that selects x, y, or both types of disturbance.

Best,

Bill P.

[From Rick Marken (2004.05.11.0945)]

Bill Powers (2004.05.11.0245 MST)]

Rick Marken (2004.05.10.2200)--

I ignored the variations around the vertical axis and just tried to
control the horizontal alignment of the pairs of vertical lines that
make up the cube. When I did this I got very clear plots of the
exponential runaway when the cube perceptually reversed.

That suggests a button that selects x, y, or both types of disturbance.

I would rather have a button that selects x, y or both types of effect of
the mouse. I think the y disturbance is good because it contributes, I think
to the figure reversals. These reversals seem to happen at the point when
the vertical lines become perfectly alighed, but I think I can check this in
the data.

What is the time scale of the data, by the wqy, in points per second? You
probably said it but I missed it.

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

[From Bill Powers (2004.05.11.1419 MST)]

Rick Marken (2004.05.11.0945)--

I would rather have a button that selects x, y or both types of effect of
the mouse. I think the y disturbance is good because it contributes, I think
to the figure reversals. These reversals seem to happen at the point when
the vertical lines become perfectly alighed, but I think I can check this in
the data.

What is the time scale of the data, by the wqy, in points per second? You
probably said it but I missed it.

Have you been able to compile the source code using Delphi? Just do (after
starting Delphi) a File | Open | ... | project2.dpr and you'll get a
message asking whether to create a missing resource file, to which you
answer yes. Then type F9 to compile and run, or control-F9 just to compile
for errors. The Pascal code of interest is in Unit1.pas, probably hiding
behind the Form1 window.

The time scale is whatever you set the frame rate to, in iterations per
second. The first line in the data file shows how many iterations there are
in one minute, so you can get it that way, too.

Best,

Bill P.

[From Rick Marken (2004.05.11.1420)]

Bill Powers (2004.05.11.1419 MST)--

Have you been able to compile the source code using Delphi?

I have Turbo but not Delphi. It's OK. The program as it exists give me data
I can analyze using Excel. I'll eventually work up a java version as a demo.

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400