NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/bit/listserv/csgl/1851

Text File | 1992-12-14 | 9.3 KB | 186 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!paladin.american.edu!auvm!VAXF.COLORADO.EDU!POWERS_W
X-Envelope-to: [e-mail]
X-VMS-To: @CSG
MIME-version: 1.0
Content-transfer-encoding: 7BIT
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.csg-l
Date:         Mon, 14 Dec 1992 15:53:32 -0700
Sender:       "Control Systems Group Network (CSGnet)" <[e-mail]>
From:         "William T. Powers" <[e-mail]>
Subject:      Neurological research
Lines: 171

[From Bill Powers (921214.1045)]

Mark Olson (921214.1010) --

You may have less frustration in sending if you set your right
margin to 65 and be sure you have a hard carriage return at the
end of each line. If you send a long string without a return you
can overrun a buffer in your mainframe, which will cause the
transmission to be rejected. Remember, mainframe software was not
made to serve people; people were made to serve mainframe
software.
-------------------------------------------------------
Your point about using information about the brain is a good one.
I did a good deal of that in writing BCP. In fact when I went to
work at the VA Research Hospital in Chicago, I was full of
ambitions, with the Northwestern University Medical School (and
library) right across the street. I thought good, I can just go
through all the neurological literature and look up what the
various parts of the brain do, and build the model around that. I
was soon disillusioned. Maybe the information was there
somewhere, but if so it was like a card index that someone had
dumped on the floor, stirred, and put back at random. You'll find
some neurological references in BCP, but not nearly as many of
them as I had hoped to accumulate. I did make a note about Hubel
and Weisel. Basically, however, my ambitious neurological project
was a bust.

As you say, brain researchers have found a good deal of
interesting material without using PCT, more in recent years with
improvements of technology. The problem is not with the
technology, however. It's with the concepts of behavior against
which neurological findings are compared.

When you stop to think about it, neurological findings are ALWAYS
based on SOME theory of behavior. Without any theory of behavior,
all you have is a record of lesions in various parts of the brain
and some recordings of spikes and potentials from electrodes in
anatomically, but not functionally, known areas. You see the
theory of behavior not in the findings about the brain, but in
the descriptions of external correlates of brain activity.

This problem has not escaped brain researchers. Mary brought home
a book on lesion research recently: Damasio, H. & Damasio, A. R.;
_Lesion Analysis in Neurophysiology_ (New York: Oxford University
Press,1989). In the introduction, D&D say

  "Naturally, the lesion method can only be as good as the finest
level of cognitive characterization and anatomical resolution it
uses. In other words, the method's yield is limited by:
   1. The sophistication of the neurophysiological testing or
experimentation with which anatomical lesions are correlated.
   2. The sophistication of the theoretical constructs and
hypotheses being tested by the lesion probes.
   3. The degree of sophistication with which the nervous tissue
is conceptualized ...   4. The anatomical resolution of the methods used." (p.
 9)

From a modeler's point of view, the sophistication of
neurophysiological testing and experimentation is not very high.
In fact, evaluations of what is wrong with the behavior of a
person with a brain lesion tends to rely on subjective and rather
crude classification of symptoms rather than models of brain
function.

A large part of testing for neurological deficits consists of
presenting stimuli to patients and recording how they respond. If
you hold up a card with a picture, or point to an object, can the
subject utter its name? If you tell the subject to point to your
finger, then the subject's own nose, then the finger again and so
forth, do the movements seem normal, and are the end-points
accurately located? Do the movements seem retarded or uneven? Are
there tremors or oscillations? Can the patient repeat back a
spoken sentence, read a written sentence, follow spoken or
written directions to do something? Can the subject name colors
correctly, in all parts of the visual field? Can the patient
speak the correct name of a seen person? Can the patient sing
when so commanded? Is reading speed impaired? Is grammar correct?
Can the patient understand and/or generate sentences of normal
length? In general, is the response competent and appropriate to
the stimulus?

And so on and so on. What's going on is nothing more than an
informal assessment of superficial aspects of behavior to see if
the patient can do all the things that normal people do, and in
the familiar way. An atmosphere of formality is generated by
using Latin terms -- alexia for inability to read, prosopagnosia
for inability to recognize persons, constructional apraxia for
inability to make well-formed sentences (I think). If you strip
away the Latin, what you have left is just a subjective
description of what the person can't accomplish that normal
people can. The main thing the Latin terms do is to allow you to
say that the person "has" alexia, "has" prosopagnosia, etc.

These deficits give some clues as to what various parts of the
brain accomplish. But they don't tell us what those parts of the
brain DO -- that is, what functions are carried out in these
particular networks that normally result in the externally
visible consequences that we call reading, pointing, naming, and
so forth.

Such reports of what's wrong are analogous to the report a
technologically naive person makes to an auto mechanic: "it makes
a funny noise sometimes; it pulls to the left; the acceleration
is sluggish above 30 miles per hour; the steering wheel shakes."
When the mechanic sets out to fix the problem, he doesn't look
for a funny noise or a pull to the left or a center of
sluggishness or a steering-wheel shake.  Those are just the
symptoms, outcomes, consequences. The mechanic understands how
the car works, so he looks for a hole in the exhaust pipe, a
tight wheel bearing, a malfunctioning carburetor, or anunbalanced tire. He
 doesn't say, "Oh, you car has odd-noisia, or
dextromobilia, or accelerotardia, or manipulo-oscillia" and go
look up the recommended treatment in a big thick book. He reasons
out what might underlie the symptoms on the basis of the theory
of operation of an automobile, and that theory tells him what is
REALLY wrong with the car. That's what a good theory of behavior
does for you, when it's tied to the actual functions of the
device. It lets you reason your way to the layers of organization
that underly superficial symptoms.

Even the interpretation of neural connections themselves is
conditioned by the background theory:

"In the new approach, subjects' behavioral responses are not just
linked to the stimuli that eventually triggered them, but are
connected to mind processes and representations that handled the
stimulus and generated the responses according to some mechanism.
The investigations no longer shy away from formulating hypotheses
about those mechanisms and attempting to test their validity,
indirectly, by measuring external responses." (p. 14).

The brain is a maze of connections. If you are under the
impression that stimuli cause responses, you will select a path
through this maze that connects inputs to outputs, and ignore all
other connections (even while noting, in passing, that they
exist). You'll stick an electrode somewhere the in the middle of
this network and present a stimulus to the senses and look for a
response from the electrode. Then you'll try to stimulate that
spot and see what response results. If you get a response you'll
say that you've traced out an S-R path from input to output. If a
reference signal entered into this path somewhere, you'd never
find it.

There's no way NOT to apply a model when you're doing brain
research. The normal accepted model is an S-R model. This way of
thinking about behavior limits what brain research can discover
-- or at least which of its discoveries it can purport to make
some sense of.

A neurologist who understood PCT would do different kinds of
investigations and see different meanings in the findings. Most
certainly, such an investigator would use different methods of
assessing deficits. The investigator would be trying to find out
not just what this person can control and can't control, but how
the parameters of control have changed from those shown in normal
people. Quantitative experiments would replace informal
subjective classifications of symptoms. Perception and action
would not be considered separately, but as a closed-loop system.
The whole approach to diagnosis would change. Instead of just
going through lists of behaviors that a normal person should be
able to produce, the researcher would be building up a picture of
the functions that enable such superficially-observable behaviors
to be seen. The specific quantitative deficits would point to
specific underlying brain functions like perception, comparison,
and output that would produce such symptoms if operating in a
particular defective way.
I think that PCT would revolutionize neuropsychological research.
-----------------------------------------------
Best,

Bill P.