Bruce Abbott on E. coli

[From Bruce Nevin (980204.0931 PST)]

Rick Marken (980130.1450) --

Rick commented on

the latest edition of Bruce Abbott's Psych Methods text

Unfortunately, Bruce describes our model in a way that
makes it seem like E. coli navigates to a target by responding
to sensed input:


Anyway, I think Bruce A. had a great opportuninty to sneak a
brief description of a control of perception model (our E. coli
model) into his methods text but (I think) he missed it.

I can see that descriptions of what control systems do need to be written
very carefully because of readers' demonstrated skill at interpreting them
in terms of linear causation. I can see that Bruce Abbott has not done this
in the quoted passage. I can imagine that this could be distressing, when
seen as a missed opportunity. I can see how one might take this as an
indicator that Bruce himself was interpreting the E. coli model in terms of
linear causation. (One's grasp of PCT can grow over time. It could reflect
his misunderstanding at the time of writing the text, or his not taking the
above-mentioned care in writing it at the time, i.e. he might write it
differently now.)

Bruce, do you disagree?

Here's the rest of Rick's quote and comment for reference:

[...] I quote:

"...Marken and Powers (1989) suggested that E. coli's nutrient-
seeking mechanism chould be modeled as a control system in which
the concentration of nutrients around E. coli affects the time
between tumbling" (p. 489).

This statement is basically true (though it is really the
_difference_ between the actual and the _reference_ nutrient
concentation that affects time between tumbling) but it neglects
the fact that time between tumbles also affects the concentration
of nutrients around E. coli. So _perceived_ concentration of
nutrients affects behavior (time between tumbles) and behavior
affects perceived concentration of nutrients. There is a closed-
loop of cause and effect; perceived concetration of nutrients
is _both_ a cause and an effect of behavior _at the same

The model doesn't get to the target because behavior is affected
by concentration gradient; it gets to the target because it is
acting to make a perception of nutient concentration match a
secularly adjustable reference for that perception. If the reference
is set for a lower level of perceived gradient, the model produces
that lower level of perceived gradient -- and there is nothing the
gradient itself can do about it. To an observer, this "lower
reference" E; coli will appear to have the goal of circling (rather
than going to) the target. It is not the gradient outside the
organism that causess this behavior (circling); it is the model
acting to keep the perceived gradient at the (lower) reference
that results in this behavior as a side effect.

  Bruce Nevin