[From Bruce Abbott (951019.1720 EST)]
Bill Powers (951018.1530 MDT) --
Bruce Abbott (951018.1530 EST)
Rick must have read my previous couple of posts, wherein I argued this
point (that much observable behavior is not a side-effect of control, as
here defined). I can only say I'm delighted to see that he--and you--
agree with me on this.
Don't leap to conclusions. My thesis is still that all behavior, all of it
with no exceptions, is produced for the purpose of controlling some
perception of the behaving organism. If this is true, then it follows that if
you think you see a behavior that does not have this purpose, you are
misconstruing your observations and describing them in terms of irrelevant
effects of the organism's actions. Every action can be seen as having many
different effects; in this case, you have simply picked the wrong effect as
being significant. The way to prove this is to find a controlled variable and
show that some other effect of the same actions is actually controlling that
variable.
I must apologize for having inadvertently left out a critical word in the
quoted paragraph above: irrelevant. So let me restate:
Rick must have read my previous couple of posts, wherein I argued this
point (that much observable behavior is not an irrelevant side-effect
of control, as here defined). I can only say I'm delighted to see that
he--and you--agree with me on this.
If you go back and reread those posts, you will see that the corrected
version is indeed what I had argued. I think it says something that you
failed to notice the discrepancy. With this correction, my statement is
entirely consistent with the thesis that all observable behavior is produced
for the purpose of controlling perceptions, and what you say following is
entirely consistent with my viewpoint. I do have, however, a few comments
to make, as stated below.
For example, the apparent fact that baby chicks behave in such a way as to
move in parallel with the mother hen is a misinterpretation, from the PCT
point of view, of what the baby chicks are actually doing. While it would be
necessary to carry out formal tests of this hypothesis, one description a
PCTer might want to try out is that the chicks are maintaining the mother hen
at a certain distance and direction relative to themselves. In other words,
the _relevant_ dimensions of the chick's actions are the changes in position
and walking direction that continually correct errors between the mother's
perceived direction and position and some reference direction and position,
both measured from the chick's point of view. The parallelism of the chick's
movement path and the mother's, while it can certainly be observed, is a real
but irrelevant side-effect of the chick's actions, because the chick is not
in a position to perceive that parallelism; the parallelism appears only in
the observer's perceptions. What the observer perceives is irrelevant to the
chick's behavior.
I agree with you here; however, I must point out that the actual observers
(unlike the observer in your hypothetical example) seem to have had no
problem identifying the relevant controlled perception (distance from mother
hen or last-chick-in-line-behind-her), and did so by conducting appropriate
tests. Such tests are of course necessary if one is to be sure that the
correct variable has indeed been identified.
Once a verifiable way of describing the relevant aspects of the chick's
behavior at this level of organization has been found, one can then treat
this whole control system as the output function of a higher level of
control. What effects of controlling relative direction and distance of the
mother, among all those we can see, are the ones that the chick is concerned
with maintaining at reference levels? Again, many possibilities may come to
mind, and each one has to be tested to see if the assumed variable is
actually under control by the chick. In the course of this investigation, you
might find other lower-level control systems that are also used to control
the same higher-level variable.
Again I agree, but must note that one may develop an hypothesis about
higher-level control, based on a functional analysis of the system. That
is, one can ask, almost immediately, "what function or functions does this
system serve?" In this example it is fairly easy to see how the observed
behaviors would aid in the survival of the chick, by keeping the chick under
the protective wing, so to speak, of its mother. Many of the observed
behaviors, both of the chick and of its mother with respect to the chick,
make good sense when viewed in this light. This hypothesis then serves to
guide the search for other controlled variables in addition to perceived
proximity to mom. Thus, previous research on this phenomenon can provide
the starting-point for a concerted effort to elucidate the detailed
structure of the perceptual control systems involved, at all levels.
In all of this there is never any need to look for analogies between the
chick's behavior and experience and a human being's. Even doing this with a
human being does not rely on analogies with the way other human beings behave
(except as one source of hypotheses to test). If there are parallels and
valid analogies, they can be pointed out later, and will be all the more
significant for not having been assumed at the start.
Yes, anthropomorphizing can be dangerous. But what I'm talking about is
different. What I'm talking about is called scientific inference--always
dangerous, but unfortunately the only game in town. Given the obvious
parallels between the chick's behaviors and those of rats, cats, dogs,
monkeys, chimpanzees, and humans, and given that they all appear to serve a
similar set of functions in the life of the infant, one can legitimately
infer the possiblitiy that similar control systems may be involved, that
they may have emerged long ago in a common ancestor and may still share a
common basic structure (though no doubt elaborated in different ways in
different species). Inferences, however logical, can be wrong; they should
not be taken as proven fact, but rather as hypotheses to be subjected to
test. Meanwhile, they suggest specific questions to ask and answer; they
guide research.
There are many, many well-established behavioral phenomena that can serve as
a starting-point for a PCT analysis. If PCT is to become accepted science,
it must be able to demonstrate its power to supply coherent explanations for
those phenomena that fall within its domain, and objective, reliable
evidence to back them up. Ethology and psychology have in many cases done a
good job of describing important aspects of behavior and the circumstances
under which it tends to occur; what they have often failed to do is identify
the correct underlying mechanisms by which these actions come about. PCT
will be valued for its ability to supply the missing mechanisms and thereby
provide a unified account of disparate phenomena. In doing so it will
establish strong cross-disciplinary connections. If you don't want this to
happen, I recommend that you maintain this assertion that the contributions
of other fields of behavioral research are irrelevant to a science based on PCT.
Regards,
Bruce