[From Bill Powers (2004.05.13.0913 MST)]
David M. Goldstein (2004.05.12.0950 EDT)--
I ran the program. It is interesting. I was able to keep one view of the
cube close to me. I was able to see the other cube at the same time.
Your phone call yesterday was most interesting -- I can't even imagine what
it would be like to see both orientations of the cube at the same time. You
were going to send me the data files, called NeckData.txt. I can't seem to
drop the habit of using 8.3 file names.
I'm still working on DynaNeck, or perhaps we should say DynaCube since Neck
doesn't sound like what is meant. I'm going to make it read as well as
store the data in floating-point comma-delimited files to preserve
accuracy, so we can trade data files easily over the internet. Also I'm
going to use Delphi's handy dialogues for opening and saving files (the
normal Windows procedures). If someone has a specification for how a
standard comma-delimited text file should look, I'd appreciate getting it.
I'll force a standard ".nck" file extension on the file names, which you
can read either using the DynaCube program or, after renaming the extension
to ",txt", any text file reader.
Also, I'll put in some check boxes to specify which variables are to be
plotted, and two buttons for saving data and reading it to plot. You'll
then be able to examine other people's data files and save your own under
different file names. All plots will have the average value removed (but
printed beside each plot) so the traces will appear in predictable positions.
>
The two lines that the program produces, can you please go over what each
one is. One is green and one is blue.
What kind of graphics card are you using ??? On my computer, one is green
and the other is red! The red (upper) trace shows angular deviation of the
cube about a vertical axis from the average orientation; the green (lower)
trace shows angular deviations from average about a horizontal axis.
To orient the cube properly it's important that you move the cursor to the
middle of the figure before starting to track. I'll place the start button
in the right position to facilitate this (it will disappear when the run
starts).
I hope that others are experimenting with this program. Aside from the
perceptual effects of interest, there are some other aspects of the
experiment that have theoretical importance. The cube, in any of the
possible perceived orientations, has attributes which are continuously
variable under feedback control, which do not change the cubeness of the
figure but nonetheless can be controlled over a large range of values --
which repeat at different positions of the mouse. The components of the
cube -- lines, angles, distances, shapes of surfaces -- can be seen to
change systematically, without changing the perception of a cube. If these
changes were related any differently we would not see a cube but an object
changing its form. We can obviously learn a lot about configuration
perception by playing with this experiment. So who is going to do it?
The theoretical problem is, of course, what kind of computation could start
with a two-dimensional display of lines on a surface and come up with a
perceptual signal invariant with respect to the transform being used?
I'm replanting my drought-destroyed back yard with gramma grass at the rate
of about 60 or 70 square feet per day, on my hands and knees, and it's time
to do my stint.For this I went to electronics school and college?
Best,
Bill P.