[From Rick Marken (941104.2100)]

Bruce Abbott (941104.2020 EST) --

I have consulted with the aliens, and they refuse to accept your

analysis.

I have looked over the code of ECOLI4 more carefully and it looks like

the aliens are right; the program does seem to instantiate the three

term contingency. The probability of tumbling in the presense of S+ is

decreased when a tumble results in reinforcement (S+) and increased

when a tumble results in "punishment" (S-); similarly, the probability

of tumbling in the presense of S- is decreased when a tumble results in

reinforcement (S+) and increased when a tumble results in

"punishment"(S-). If this program results in E. coli moving up the

gradient to the target, then I have to admit that behavior is indeed

controlled by its consequences -- the aliens are right!

I suppose I will write the program up tommorrow and sadly watch as E.

coli moves steadily toward the target, guided at every step by its

consequences. I don't understand how this could happen -- after all,

the probability of a reinforcement following a tumble in the presense of

S+ is the same as the probability of a reinforcement following a tumble

in the presense of S-. So, in the long run, the probability of tumbling

in the presense of S+ should be about the same as the probability of

tumbling in the presense of S-. So the behavior of E. coli should be a

random walk. But apparently there is a magic here that I don't

understand. You say that the behavior of this model is not a random

walk. If you are right, then control theory is wrong. Control theory says

that behavior is the control of perceptual consequences; you have

demonstrated the behavior is controlled by perceptual consequences.

Gee, and it seemed like such a good theory.

Ah well. Back to behaviorism;-)

Best

Rick