[From Bruce Abbott (951016.1740 EST)]
On Saturday October 14th I initiated a private post to Bill Powers that
resulted in a series of exchanges between us that, in retrospect, could have
and should have appeared on CSG-L, as they represent continuations of the
debate we had been carrying out in public. We have therefore decided to go
ahead and post them here. This posting presents my opening post to Bill and
his reply; I'll present the rest of the exchange in a separate posting.
···
------------------------------------------------------------------------
[Bruce Abbott (951014.1550 EST)] Subject: Getting on with it
Hi Bill,
I hope our recent discussion hasn't made you feel TOO frustrated with me.
What I've tried to do is to present, point blank, some of my views wherein I
perceive we have significant differences of opinion, with the hope that by
bringing these issues to the foreground we can at least try to gain a better
understanding of what precisely we disagree about and what arguments we can
offer to support our position. I even hold out some hope that I can talk
you into coming around to my point of view on at least some of these issues;
by the same token I am genuinely trying to understand your "framework" as I
always hold out the possibility that I may be flat out wrong on a given
position. Of course, before changing my mind I'd have to be convinced of
this. At the same time, I often get the impression that we are actually in
basic agreement about some issues but are arguing about the details. For
example, I agree with you that mere similarity of behavior between members
of different species is no guarantee that the behavior springs from
homologous brain structures, nor is a similarity of brain structure any
guarantee that the two structures serve a common function. Given this
degree of agreement about the basic issue, it would seem that the rest is
almost mere quibbling about the degree to which researchers have been aware
of the problem and done the relevant research before drawing sweeping
conclusions.
I certainly recognize that a multi-layered hierarchical control system can
produce an enormous variety of behavioral output dependent on the
disturbances acting on its various levels and such variables as what step of
a plan is currently being executed. For this reason mere observation of
particular actions is not going to tell you much about the underlying
structure. However, it does seem to me that one can usually do a pretty
good job of inferring the function of a given system by observing its
actions, which will normally be such as to tend to correct for disturbances;
this inference procedure is of course what we call the Test. Furthermore,
although all the brain is active all the time, I believe it is possible
through anatomical, neurochemical, and other such means to identify brain
systems organized to perform specific functions, even though these systems
share the same lower-level control systems with others and many of the same
"higher" perceptual and cognitive structures as well. It is possible, for
example, to identify brain structures specifically organized to orchestrate
all the complex changes we humans experience under the name of "fear." It
is, of course, a control system, and it becomes active when the appropriate
perceptual signal is raised by evaluative processes elsewhere in the brain.
I offer this example, without any supporting evidence, merely to show the
kind of organization I think exists in the brain. Neurobehavioral evidence
supports the hypothesis that such systems exist and that in many cases
homologous structures in humans carry out functions similar to those in the
brains of other mammals, although modified and greatly elaborated during the
explosive growth the the human brain over the past several million years.
You may be worrying that in adopting this view I am subscribing to a form of
S-R view implied by statements such as "this system mediates fear behavior,
and is activated by stimuli identifying potential danger," which is a type
of phrase heard often enough from researchers in this area. Not so. I see
these systems as control systems (or as parts of control systems); the
"fear-eliciting stimuli" are specific types of disturbances to this system,
and the system's actions are to alter references for other, lower-level
systems, alter the gains on some systems, and change the "settings" of
perceptual "filters" so that certain inputs now are more selectively
amplified (attended to). These systems are often named after the behaviors
that commonly follow the disturbance or after their presumed function (e.g.,
"defensive system") rather than in terms of the perceptions being defended,
but this misnaming, while misleading, shouldn't lead one to reject their
existence as cohenent control systems within the brain.
I guess we'll have to put aside some arguments if we're to make progress
in developing PTC applications to behavior. I can see the logic in your
answers to my questions and comments, but can't seem to get across the
framework from which my questions arise.
From what I've stated above, does it seem that I might understand the
framework after all? Or am I still missing something?
I do hope you still plan to read Bowlby, especially Volume 1 (Attachment),
because Bowlby comes about as close as anyone to expressing my view of brain
and behavior (indeed, he's probably where these ideas of mine came from,
although I now have many layers of additional input piled on top, including
B:CP and our many interactions over the past year. In fact, I've started to
reread Bowlby and am positively shocked at just how many ideas I thought I
absorbed elsewhere are in fact exquisitely described in this book. A
caution: Bowlby does talk about "behavioral control systems," a natural
enough thing to do given that this is how even control engineers were
talking about these systems. Keep in mind that B:CP was still four years in
the future. Nevertheless, it is clear (to my mind at least) that Bowlby had
an excellent if nontechnical appreciation of the basic principles and is
clearly aware that references (he calls them "set-goals") themselves are
varied by higher systems and that actions then follow these changes via the
mechanism of negative-feedback error correction.
You sent me some data a few weeks ago. You're going to have to remind me
of where we were in that process and what these data were for. So much
has been going on since then that I've lost the thread, and need a jump-
start. Wait until you're my age and you'll see.
When I'm your age I'm sure I won't do half as well. I'm already starting to
have these problems!
I would like to get back to these data. The last batch came from Fran
McSweeney, who kindly supplied us with the reinforcement-rate data that go
with the corresponding response rate data that appeared in her recent JEAB
paper. I'll have to dig out the actual schedule values used and their
ordering (which was kept the same across sessions). The data are reported
for successive 5-minute intervals of a 60-minute sessions and represent the
average rates for the final 4(?) sessions under each schedule, which
presumably represented stable performances. The rates fell considerably
over the course of an hour, more so for the faster schedules (e.g. VI 15-s)
than for the slower ones (e.g., VI 120-s). You might want to start by
plotting _cumulative_ data (adding across successive 5-minute intervals) as
a function of time for both reinforcements and responses, for each
individual subject. I think you'll find the plots interesting.
Warmest Regards,
Bruce
-----------------------------------------------------------------------
[From Bill Powers (951015.0945 MDT)] Subject: Re: Bowlby
Hi, Bruce --
I've started reading the Bowlby materials and have skimmed through all
the volumes once, with more detailed reading of Vol 1 on Attachment. I
can see why you recommended it -- he really made an attempt to get away
from the less defensible aspects of Freud's ideas and to substitute
control theory as a different paradigm. It's too bad, in a way, that
Bowlby and I couldn't have got together in the 1950s and 60s when he was
picking up all this stuff about control theory. He really had a pretty
good grasp of the principles, but he didn't know what to do with them
beyond saying that organisms control for outcomes and don't just produce
behavior open-loop. Unfortunately, just as he was getting revved up to
write these books, all this control theory stuff went out of favor. I
don't suppose that Bowlby made much more headway with introducing
control theory after that than I did, although he did a lot of other
work for which he was deservedly recognized.
I'll comment more on the net. For now, my reading confirms the general
impression I've been trying to communicate, which is that the
observations and theorizing in this field are at a different level from
what I would like to see going on in PCT. Just consider one basic
concept, separation anxiety. How do we know what this term means?
"Separation" is easy, but how about "anxiety?" Our only referent for
this term is in our own experience, and we don't understand anxiety in
ourselves any better than we understand it vicariously in someone else.
Bowlby's main concern is not to understand anxiety, but to describe the
conditions under which it occurs and goes away, and the possible
evolutionary reasons for its existence. He never tackles the basic (to
me) question, which is WHAT IS ANXIETY?
In general, this seems to be how all the mental and emotional conditions
of which he speaks are treated. In a PCT approach, I would ask different
questions and look for different answers. The concept of negative
emotions that I have proposed is that the feeling component comes from
error signals in the behavioral hierarchy which reset physiological
reference signals to back up the actions which the same error signals
normally call for. The changes in physiological state that result are
sensed, and in conjuction with the other perceptions that are being
controlled are perceived as what we call an emotional state. The
perceived emotional state has a somatic and a cognitive component. Any
severe emotional state, I have proposed, results from uncorrected error,
produced by an internal or external conflict or some other failure to
control.
When we see anxiety this way, the primary question is no longer how it
feels, because how it feels is simply a consequence of the error
resetting somatic reference signals. The basic question is, what is the
error? In separation anxiety, the error would result from the inability
to prevent the separation, and we can conclude that separation entails a
deviation of some perceptions from their reference levels. And what are
those perceptions? That becomes the question that a PCT approach would
then try to answer.
You'll notice that in this approach it simply doesn't matter WHY these
perceptions are under control; whether because of instincts or learning
or evolution. Those are side-issues. It doesn't matter what previous
experiences prevented learning how to control the perceptions. The
primary issue is what is being controlled here, and what is preventing
success. When we understand that, we will understand what the problem
is, and we will also understand anxiety. We won't have to define
separation anxiety strictly with reference to our own experience of it.
In fact, we will understand our own experience better, too.
This would be my approach to many other phenomena mentioned by Bowlby,
such as imprinting, conscience, guilt, anger, sorrow, grief, and so on.
These are not, to me, basic terms. They appeal to empathy, not to a
model of how we work.
While Bowlby made an attempt to switch to a new conceptual framework, he
was focused not on theoretical understanding but on the middle ground of
psychology, where the aim is to find the circumstances that lead to
distress and also to its ultimate removal, with only a minor interest in
an explanation of the nature of these phenomena. By eliminating the
conditions that usually lead to distress, and encouraging or supplying
those that usually lead to amelioration, the psychologist is able to
help some people to avoid causing or feeling distress, and to reverse
the effects in those already in trouble. Only a fanatic would say that
psychologists should stop doing that because it does not lead to
theoretical understanding, and I am not a fanatic.
Reading at another level I found the work of Bowlby and his colleagues
to show a strong concern for human suffering and a committment to doing
something about it. His observations of the way children and adults
behave under separation and loss touch me and ring true. I think the
findings he describes are of practical use and could easily be applied
to improving the human condition.
However, they do not improve our understanding how HOW behavior works.
That's a different level at which to consider the same phenomena, not to
prevent or cure their effects, but to understand how they work.
Bill