[From Bill Powers (951010.0130 MDT)]
Bruce Abbott (951011.2350 EST) --
Bill Leach said it well but I want to repeat the idea. Why do
psychologists measure things like distances and speeds and directions,
and then use them to explain metaphors like imprinting and attachment?
Output variables in chicks: peeping, movement. Peeping affects
proximity through the action of the mother hen's control system,
movement by closing the distance between visual representation of
mother and (I believe) her sounds; pathways: visual, auditory.
Additional inputs: sensation of warmth, physical contact with
mother or others in brood (speeds restoration of error to zero):
That's what you would say now. Were those observations reported when the
study was published? Were any measurements made of the relationships
among these variables? Were disturbances applied in all possible
directions to identify these controlled variables, to prove that they
were actually under control? Or are you just offering plausible general-
As I recall, the study examined the following behavior of chicks or
ducklings or goslings (I don't remember which) that had been
imprinted on an object whose motion could be readily controlled by
the investigator. (It was mounted on an arm extending from a
central pivot, which moved the object in a circular path through a
circular corridore.) The imprinted birds followed the object as
they would have followed their natural mother under normal
Well, that's interesting. Was this study really done with multiple
chicks, or with one chick at a time? The way chicks follow their natural
mother, as I have observed many times (including pictures of goslings
following Konrad Lorenz), is in a line. That is, the first chick follows
the mother, the second chick follows the first chick, and so on, as in
the CROWD demo called "oneline." So it looks as if only the first chick
need imprint on the object; the others imprint on another chick. The
observer, however, who is hung up on the idea that the chicks imprint on
the mother, may have seen only the relationship between each chick and
the mother, and may not have tested to see what, in fact, each chick was
following. If all the chicks were following the mother, they would
cluster around at roughly the same distance from the mother (as in the
"guru" demo). If they were simply following a familiar moving object,
they would follow in a line. One way to test this would be to pull one
chick aside slightly with a thread; if the chick behind it followed the
deflected chick, the chick behind was following another chick, not the
Control of distance implies obstacle avoidance as well. Were the chicks
controlling for preventing collisions as well as for maintaining
distances? Were distances or velocities under control? Was direction
under control as well?
If chicks imprint on any large moving object that is seen at the right
time, then presumably they will thereafter follow "any large moving
object." In other words, the controlled perception is the distance from
some large object that is moving. So the next question is whether a
different large moving object can be substituted. If not, the controlled
quantity has been misidentified: the chicks are maintaining something
else constant, not just the largeness or the relative distance.
Incidentally, if the chicks follow at a fixed distance from a moving
object, the object is not moving relative to the chicks. So perhaps
"moving object" is not the right concept in the first place. The object
moves mainly from the observer's point of view.
If imprinting is really specific to one moving object, then clearly more
perceptions than size and motion are being controlled; perhaps there are
also color, odor, markings, shape, and so forth. Each one of these
should be investigated, until it can be determined what the chick is
actually controlling. And when that is done, there will be no need to
invoke a metaphorical term like "imprinting" or "attachment." The chick
is not literally attached to the object; that is only a false appearance
created by the chick's control actions. And there is certainly no
evidence of "imprinting" of anything on anything else. If the chick
learns to control certain perceptual variables, all that is observed
will follow from that alone. What do the ideas of imprinting and
attachment add to that?
However, the situation was then changed so that when the birds ran
toward the object, it retreated from them faster than they could
approach, and when they ran away from the object, it approached
them faster than they were retreating. After some experience with
this arrangement, the birds learned to move in the direction that
would keep them within a specific distance from the object.
This was Wayne Hershberger's thesis experiment. He found that chicks had
a great deal of trouble solving this problem, and they did it mainly by
showing very abnormal behaviors. You seem to be talking about a
replication of Wayne's experiment, which found different results
(control of distance from the object, with no apparent difficulties).
This experiment shows another phenomenon, the ability to reverse the
sign of the output function when the sign of the environmental feedback
function reverses. The next logical experiment would be to switch back
and forth from the normal relationship to the reversed one, to see how
rapidly the chicks could change the internal sign of the control system.
This would give some information about a higher level of control system,
acting on the parameters of a lower-level system (as in Rick Marken's
experiments with reversals).
Dagc Forssell pointed out over the phone today that this reversal
phenomenon is a perfect proof that it is perception, not output, that is
The problem with these studies is that the object was to find out
something about the metaphorical phenomena of imprinting and attachment.
As a result, the REAL phenomenon, which is the learning of control
processes, was barely investigated. A great deal more could have been
learned if the experimenters had just abandoned those two vague concepts
and concentrated on studying controlled variables.
That's what I mean by starting from scratch. Abandon all the old
categories and start over by studying control processes. The more you
learn about control processes, the less use you will have for terms like
imprinting and attachment (except to communicate with people who still
believe they mean something real). You can still investigate chicks
following things, if that is what you want to investigate.
Even Copernicus didn't demand that astronomers throw away their
laboriously and meticulusly assembled maps of the heavens and start
from scratch. Their conceptions, yes; the observations, no.
He did, however, demand that they throw away all their laboriously
assembled diagrams showing epicycles and deferents and all the concepts
they thought they were studying. The observations were simply records of
where the planets were relative to the stars at successive times. These
Copernicus kept. The rest went in the garbage can.
I recommend that we retain the records of the chicks' velocities and
positions relative to the moving objects at successive times. And throw
the rest in the garbage can. If we accidentally throw something useful
away, we will reinvent it anyway, only this time on solid grounds.
It seems to me that the phenomena of attachment and loss, extending
as they do from lowly chickens clear across to the human child and
parent, have already demonstrated their dependence on specialized
control systems designed by evolution for the purpose of keeping
the child safely within its parents' protective influence.
I don't think that terms like attachment and loss have any place in a
scientific study of behavior. They are not phenomena; they are metaphors
and interpretations by an observer who chooses to see the same thing in
different behaviors. If they belong anywhere, it is at the END of the
study, where you try to interpret the actual findings for the layman who
knows only a non-technical vocabulary. You may actually find surprising
similarities between the control behavior of chicks and the control
behavior of human children. The idea of a scientific investigation is to
let such similarities come as a surprise, not to jump to conclusions
based on the human habit of classing different things together and
calling them by a single name.
I also don't think that evolution designs anything for any purpose.
"Keeping the child safely within its parents' protective influence" is
not a goal of a chicken or a chick, much less of evolution. You're
describing how a human being would see the _consequences_ of controlling
as the chicks do, but the chicks know nothing of those consequences.
They control only what they can perceive, and that does not include
verbal generalizations as far as I know.
The benefits of this analysis are already being widely felt in
clinical and other applied settings. It absolutely astonishes--and
saddens--me to hear you say that you think such research is
antithetical to PCT.
As I said, I'll take the observations, if they are actually adequate for
a study of control processes. Whether you consider the applications of
the analysis to be beneficial depends on how you interpret the effects.
I have never heard any psychological researcher claim that the
applications of his ideas to clinical settings were harmful or
irrelevant. You could just as well be talking about the enthusiasm with
which psychologists typically lap up the latest fad. Maybe there have
been benefits: I wouldn't know. But I would rather have someone other
than the originators of the ideas or the psychologists applying the
ideas evaluate the results.
I suggest you read Bowlby's (1969) book and see whether I doesn't
change your opinion.
OK, I'll get it through interlibrary loan. That will take a while.