From Tom Bourbon (930505.1530)
[from Gary Cziko 930505.1925 UTC]
Tom Anastasio is a neurophysiologist at this university studying the
vestibular-ocular reflex (I prefer to call it the retinal image stabilizing
system). ... Here is his preliminary report on Powers 1973. I have already
responded to part of Tom's reaction, but I wanted to share it with the net
in the hope that some PCT veterans like Bill Powers, Rick Marken, Tom
Bourbon, Greg Williams and Wayne Hershberger could also respond (Tom has
given me permission to post this).
[from Tom Anastasio <tstasio@assistant.beckman.uiuc.edu>]
Gary:
I'm really enjoying Powers' book. I'm into chapter 15 (on memory) so I'm
almost finished. So far, I've found lots of ideas which are central to my
area of research, such as vectorized quantities and integration by closed
loops, in addition to control by negative feedback. These ideas are fairly
well accepted by us and have been and continue to be elaborated and
tested by investigators in our field. I would be extremely impressed (and
quite surprised) if Powers himself is the originator of these ideas as they
have been applied to oculomotor neurophysiology. I would be surprised
because I've never seen a reference to Powers in all the papers I've read in
my field concerned with these topics.
Tom A's observations that ideas about control and negative feedback are
prety well established in neurophysiology squares with the sorts of things
Andy Papanicolaou and I find when we look through material in that area. A
fair number of authors seem to have the basic ideas down pretty well, but
many still don't quite have it. The biggest problem usually centers on that
which is controlled -- it is one thing to talk about negative feedback
control, and apparently quite another to recognize that living control
systems control their perceptions, not the variables we so easily view from
the outside. Something Tom A. says a little farther along leaves me with a
question about what he thinks is controlled. (My question is just that -- I
am curious about it and would like to know.)
And his observation that Bill Powers is never cited in that litrature is
similar to ours -- alas!
What's new to me is the idea of hierarchical control in the sense of nested
control loops. This has very interesting implications for the organization
of the brain as a whole. In my opinion, HPCT is a very elegant theory. If
I do move on to higher level parts of the brain in my research, HPCT is an
idea I may well fall back on in trying to evaluate my findings.
Well, PCT as a fallback when nothing else works is better than no PCT at
all, I guess. It would be really nice to see someone who is equipped to do
neurophysiology who caved in and did direct physiological tests of the PCT
model -- tests of the idea that nervous systems control the magnitudes of
their own sensory signals. Think you can keep working on him, Gary?
What I continue to be puzzled by, however, is the apparent persecution
that HPCT adherents seem to suffer. This is brought out in the paper you
sent to which I am responding. Can it be true that scientists who want to
test HPCT are actively restrained from doing so? It cannot be disputed that
science is resistant to change. I think all of us who are subject to peer
review have felt this inertia. I have even spoken out in favor of
reforming the peer review process, to make it more democratic.
No, we have our own PCs now, and we have found ways to look like we are
legitimate scientists, teachers, therapists and the like, while really doing
PCT. So long as no one can tell what we are doing, we seem to get along
pretty well. The problem is not in doing, but in publishing and talking
about what we have done. I am accustomed to the usual songs and dances that
are part of the peer review process, and this is different. The peers (a
majority of them, anyway) are on this net, not on the lists tapped by
editors -- the people they pick are not peers. Seriously. They do not have
an interest in, or knowledge of, the phenomena we study or the use of
generative modeling as a test of hypotheses. As reviewers, they find
themselves looking at foreign material and they treat it as such -- that, or
they imagine it is something they already know and dismiss it as that. (The
collective file of rejections, and the reasons for rejection, would prove that
point.)
But good ideas always manage to generate some support and this allows them
to grow.
And here we are ....
I've got to beleive that in 40 years, HPCT has generated some scientific
results of real merit. In my opinion, the case for HPCT could be made much
more strongly by citing evidence in its favor. This could invlove the use
of HPCT to explain previously mysterious data, or the results of actual
tests designed to directly test the theory. I'm assuming this data is out
there. My suggestion is to heavily incorporate this into Powers' paper.
This is fine -- I agree 100%. And PCT has explained one of the really big
mysteries for me -- the same one I must have been taken by when I discovered
those things that would just wiggle in front of my eyes any time I thought
about them. But control of perception doesn't seem to strike many people as
mysterious or important, and we are nearing the part of Tom A's post that I
think shows how a person who takes a look at PCT and is interested by what
is there can then slip away without knowing it. It is also the point at which
some of the CST popularizers and wordsmiths begin to introduce casual
modifications, the consequences of which they do not imagine or comprehend.
(Not that Tom A. has done that, or will do that. It is just one of the
points where it has happened to other people.)
Include citations and detailed explanations of how HPCT, more than being an
elegant idea, has led to a deeper understanding of SPECIFIC brain processes.
PCT is not just an elegant idea, it is a working model of control. And there
is the idea that brain processes are important, but control of perception --
what happened to it? Doesn't it have a lot to do with specific brain processes?
Again, I know Tom A. has not said that -- not directly -- and he might not
think that at all. But this is the point where I usually begin to wonder,
and to ask a few questions, or float a few ideas to see if they are ignored,
or accepted, or treated as disturbances.
No one will argue against the idea that a driver uses negative feedback to
steer a car down the road.
It is pretty clear that Tom A. never encountered reviewers like ours! Some
of them would even argue that cars and roads are trappings of Western
colonialism and that anyone who attaches importance to data about how people
control cars must be a degenerate capitalistic logical positivist from before
the post-scientific era. And everyone knows positive feedback is more
aesthetically pleasing than the negative kind. (I'm not kidding, there are
such reviewers out there, and they have had their say about PCT.)
But the biggest potential problem in that sentence has to do with a person
steering a car down the road. Yes, perhaps, but not as that process looks
to an observer -- PCT is about the view from the inside, from the driver's
perspective. This sentence might be a casual shorthand version of, "but we
all really know that ...", or it might be an indicator that the writer is
thinking of PCT as a "commands-followed-by-a-servo" model, like those
popular right now in parts of cognitive science.
What will really convince scientists is hard
data, the more neurophysiological the better, that begins to flesh out HPCT.
Sigh. Agreements between a model's predictions and a person's results that
rival the level of precision in parts of physical science, but they do not
count (more logical positivism?), but something really *hard* *would* count.
I agree, completely, that tying PCT more closely to neurophysiology will make
it more respectable to many people, that is one reason I left a tenured
faculty position and moved into my new surroundings. But hard data are
there already -- they just aren't interesting to most folks. (The reviewers
say it, over and over, and so do many of our friends.) Why not say it the
other way: "Gosh! Look at these great behavioral data and these spectacular
predictions. You know, if this model works so well at predicting control
behavior, maybe I should rethink my priorities for physiological research.
Looking to see if nervous systems, endocrine systems and the like are
controlling their sensory signals might be interesting -- and it might pay
off, big time."
We will stay tuned, Gary. Regards to Tom A.
Until later,
Tom Bourbon