[From Bill Powers (950105.2000 MST)]
Bruce Abbott (950104.1215 EST) --
Rick: please tell me how you plan to study the properties of human
memory using the Test. Your results should allow you to explain,
for example, how it is that you are able to recognize a familiar
face.
Since Rick would rather prove to you that you're more interested in
preserving the status quo than in advancing PCT, I will answer in his
place.
The problem you pose would more properly be a problem in perception, but
once we admit that a person does perceive faces, we can explore memory
for faces by setting up a control process in which the person acts to
alter a graphic of a face to make it look like the remembered face. I
wouldn't know how to do this myself, but with support (and enough
computer power) one could hire a person competent in graphics
programming and carry out a research program that might prove
interesting. By use of morphing, a continuum of faces could be generated
with variations in several dimensions, the particular face being
embedded in the continuum. By giving a person control over each
dimension, the experimenter could determine how well a person does
remember the target face, how much error is tolerated, and how the
memory changes with time. Disturbances in each dimension would assure
that it is the perception being controlled, and not just the position of
the control handles. From this start a number of related experiments
using the Test could be done, for example determining how much
difference there must be in various dimensions before two faces are
distinguished from each other in memory. Disturbances that are not
resisted reveal irrelevant dimensions. It would also be interesting to
explore "facial expressions" this way -- when does a change amount to a
change of expression of the same face, and when does it cause a change
to a different face?
That would be a big, expensive, lengthy project. There are much simpler
ways to explore memory, more appropriate to our stage of understanding.
In Demo1 there is a control task in which a steady tone is sounded for a
few seconds, and the participant is asked to keep the sound at that same
pitch for one minute while an invisible disturance tends to make the
pitch rise and fall. Obviously (to me) the reference signal during this
control process has to come from a memory of the original tone. The
person is continuously trying to correct deviations of the tone from a
hypothetical reference level. Applying the Test (since we know what the
disturbance is at every instant and can see exactly what corrections are
being added) we can first show that the tone is in fact a controlled
variable; second, we can infer what the reference tone is. By performing
the Test over short segments of the data run, we can detect changes in
the reference tone (the person corrects deviations relative to a shifted
tone), and so get an indication of how accurately the tone is
remembered. By repeating the experiment at intervals but without the
initial presentation of the tone, we can see how stable the memory is
over a longer period of time. This will not work with every person,
because some people can't do this task at all. It would be interesting
to find out why they can't. Saying they are "tone deaf" doesn't exactly
answer the question. Perhaps they can't perceive tones, or perhaps they
can't remember them. Interesting problem, which a smart PCTer should be
able to work out.
Another experiment could be done to see how well people can remember
positions near the center of a computer screen. Flash a dot in some
position, then have the persons move a spot to that position from a
corner of the screen. To show that it is a visual position perception
and not a hand movement that is being remembered, we apply a disturbance
to the spot so that a different movement is required each time to bring
the spot to the correct position.
In another version, we would put a red target on the screen and show a
green cursor 1 cm to the left of it. Then on repeated trials, the person
would move the green cursor from the right edge of the screen to the
same position relative to the red target as in the original
presentation. On each movement, a disturbance would be added to the
green cursor position to make sure that the mouse position is not what
is being remembered. The Test says that if a disturbance alters a
variable significantly, then that variable is not under control. Every
now and then, at the beep that starts a new trial, the target and cursor
colors would be interchanged, to see whether the person is controlling
for "cursor left of target" or "green left of red."
To make sure that the person is not just moving the cursor to a specific
place, we would move the target to a new position at the start of each
trial. This would show us that we are looking at the stability of a
memory for a relative spatial distance, not an absolute position (if
that happens to be true for a given person).
Memory for sequences has probably been investigated enough to satisfy
the demand for the next 500 years, so we can skip that.
We can also test for memory of speeds, both linear and angular. Color
memory can be tested, as well as memory for shapes, sizes, temporal
contours, intensities, and any other variables we can incorporate into a
control task. We can see how memories at different levels of perception
compare with each other; how memories for symbols, for example, compare
with memories for the meanings of the symbols, or how memories for
programs (if-then rules) persist through time, as opposed to memories
for the specific steps in the programs.
In each case, making the variables in question part of a control task
and using the Test enables us to determine just what is being perceived,
and thus just what is being remembered.
This method is highly effective for investigating illusions, too,
allowing quantitative measurement of their magnitude.
Does this answer the question you asked Rick?
···
---------------------------------------------------------------------
Best,
Bill P.