From Phil Runkel on 2 Jan 96.
Rick Marken said on 31 December, in re "New Year's
Resolution," that the tracking task is very deep. And to say
that is deep, too, because saying it implies a theory of the
internal hierarchy.
The tracking task seems to take place in a very small world,
a simpler world inhabited by simple creatures who do simple
things. Indeed, actions heavily interlaced with bodily processes
such as the psychophysics of perception, the control of muscular
action, and so on, are highly predictable in psychological
laboratories the world over. Some psychologists, I suppose,
would say that this more physiological kind of behavior is
inherently more predictable and less probabilistic than the more
"mental" kind of behavior--than behavior "mediated in the higher
nervous centers," as they might put it. I think this view
resembles the ancient claim that mind and body function by
different laws. Books on psychology do not put this view into
words any more, but departments of psychology in every university
hint at this view when they establish subdisciplines of
personality psychology, physiological psychology, and so on. I
find the resemblance strengthened when I notice that the research
methods in those subdisciplines differ in ways easy to see (in
the degree of insistence on random sampling, for example) and
when I notice that specialists in one subfield know little about
the work in another and show little desire to know more--when
they seem to believe that knowledge of a colleague's specialty
will not help them in their own. In the departments with which I
have been acquainted, the common attitude seems to be "If those
guys don't interfere with me, I won't interfere with them."
Furthermore, the professional journals of psychology are
specialized by those same subdisciplines. The implication is
strong that most psychologists believe that the laws applying in
other subdisciplines can safely be ignored--that those laws (if
any) have very little application to their own subdisciplines.
Some may say that nature has arranged for one kind of
functioning of the human animal to proceed with a random or
probabilistic admixture and another kind to proceed in a
controlled manner. But to say that, it seems to me, violates not
only the principle of parsimony. It also violates the principle
of evolution that later forms grow from earlier forms, that
nothing grows out of the void, that the new builds on the old.
If the neocortex is an evolutionary development, it should use
the principles that enabled the older brain to work. To claim
that a new principle of functioning can appear in the brain at
some point, a principle not built on an old principle, is to
claim to construct a brain by supernatural intervention. It
seems to me that we should demand of all varieties of psychology
that if a law describes one kind of human behavior, the law
should have its counterpart in other kinds of behavior. We
should not have separate psychologies for walking, eating, naming
colors, gambling, obeying, persuading, teaching, listening to
Beethoven, and so on.
It is parsimonious to theorize that the feedback loop
functions is essentially the same way at all levels of the
hierarchy, as Marken implies. It is as if evolution knew a good
thing when it saw it in the first single cells and has been
elaborating on it ever since. It is fine and necessary to test
whether functioning at the higher levels does indeed work like
always-reliable feedback loops work; Dick Robertson has made a
start with his experiments on self-concept. But let's not make
added assumptions of radical differences in basic functioning
until we need to.
I would sympathize with specialists who choose to limit
their research to the variables popular within their specialty,
and who do not look for ways of functioning that cross boundaries
of specialties, if the principle of control of perception
received no better experimental support than the kind of research
results appearing regularly in journals of personality, clinical
psychology, social psychology, and the like. But the fact is
that the functioning of actual working models of perceptual
control regularly show correlations better than, say, .97 with
actual human functioning. Furthermore, those correlations are
not calculated by taking two (or more) data from this person, two
data from a second person, two from a third, from a fourth, and
so on, in a laboratory where environmental events are kept as
unchanging as possible except for the variables being measured.
Instead, the correlations are made by taking consecutive pairs of
data from human and model; that is, each pair of data consists of
one datum from the person and one from the working model. And
the behavior occurs in an environment in which random changes in
the environment disturb the effects of the acts of the person and
the "acts" of the model. Nevertheless, the models typically
behave almost exactly as the persons being modeled. Since the
models are built on the principle of the control of perception,
the inference is strong that humans function in that same way.
A considerable literature (it is small in comparison with
the mainstream literature, but considerable in comparison with
what is reasonably needed to buttress the conclusions from the
research) now reports the results of modeling the control of
perception. The results, so far, are uniform in their support of
theory. By uniform, I mean that as far as I know, _no single
experimental subject_ has yet been shown to behave with the
experimental apparatus in a way that the model cannot duplicate
to a very high correlation. There are, of course, exploratory
edges of the research that are not yet succeeding as well as the
tracking tasks do. But my statements apply, I think, to all
research that _models_ behavior by computer and compares it with
actual human behavior.
Marken also said that PCT should not be expected to
"explain" research done with the mainstream theory of linear
causation from one environmental condition to another, such as
the correlation between a boss giving orders to underlings and
the actions of the underling after hearing the orders (my example
here, not Marken's). That's manifestly, unquestionably,
absolutely right. The chief reason that expectation should not
be cherished is that almost all mainstream psychological research
(except maybe psychophysics and some experimentation by Karl U.
Smith et alii) consists of reports of correlations over persons
between environmental events. (I gave examples in my "Casting
Nets and Testing Specimens" on pages 11-12 and 26-29 and in
chapters 4, 5, and 7.) Correlational research on actions cannot
reveal how the actors function. It can show only what phenomena
cluster together with a certain class of action. As an example:
why do more fires, on the average, occur in cities with the
larger fire departments? There is no end to collecting research
reports of the form: "I found a correlation between X and Y." I
once did a study showing a correlation between the mean score on
an academic aptitude test in a high school and the distance from
the high school to the nearest coal mine.
The correlation between frequency of fires and the size of
the city tells you nothing about how city councils operate or how
fire departments fight fires. But the way the city council
operates and the way the fire department fights fires produce the
underlying actions that bring about that correlation. The
correlation occurs as a by-product of the actions citizens and
councils and firemen take in controlling the perceptions they
hold dear. When, however, the firemen start spending more time
persuading the city council to pass laws that will reduce fire
hazards, the correlation will weaken. That does not mean that
human nature changes, or city politics, or fire departments. I
means merely that firemen and others are choosing new features of
their environment through which to take actions to control their
perceptions.
PCT explains why and when the actions living creatures take
will produce reliable patterns in the environment (that is,
reliable consequences) and therefore associations or correlations
among classes of environmental events. PCT explains how it can
come about that you hear something you recognize as Beethoven's
"Fur Elisa" when you see someone sitting at a piano and looking
at that music. But PCT does not predict whether that is going to
happen with any random person at any random time. It does not
claim that sitting at a piano and looking at the music produces
the sound of "Fur Elisa." Indeed, PCT counsels that you should
_give up_ trying to find high correlations between such events as
your seeing subjects "stimulated" by the sight of a piano with
the music for "Fur Elise" on it and your hearing the "response"
of the subjects moving fingers in such a way that you recognize
"Fur Elise." PCT cannot predict a high correlation (or low,
either) between sitting and looking at the music, on the one
hand, and making the sounds of "Fur Elise," on the other hand,
because the correlation does not depend only on that possible
connection between what the environment offers and what the
player chooses to do. The correlation also depends on whether
music teachers and publishers bring "Fur Elise" to the attention
of players, on the proportion of piano students who stick with it
long enough to acquire the necessary skill, on the proportion on
players who choose to put their energy into "Fur Elise" in
preference to other music, and on other environmental
opportunities and obstacles that PCT has nothing to say about.
And for that matter, no other theory, either, will succeed in
predicting what any random piano payer will do in the face of all
those environmental uncertainties. PCT can never explain or
predict that a correlation will occur between two particular
kinds of action or, indeed, between some internal state and some
particular kind of action in the environment, because it can
never predict the state of the environment (unless the
environment is reduced to such simple predictability as the
Skinner box). It is the glory of living creatures that they can
stabilize the perceptions they want to stabilize despite the
disturbances from the turbulent and unpredictable environment.
The parallel Marken made with Galileo's experiments with
balls rolling down inclined planes is exact. I can hear
Galileo's critics: "What in the world do these toys have to do
with the impetus of a cannon ball?" Galileo could have said, "I
won't try to explain impetus. I will urge you to _give up_ doing
research on impetus. Cannon balls do act like these little
balls." And I can hear his critics grumble that until Galileo
can explain things people care about, like how impetus works, he
isn't going to get much attention. But despite my pleasure in
Marken's example, I don't want to go too far with analogies
between research on PCT and research on physics, because the ways
living creatures behave very differently from cannon balls.
I hope I have said something helpful to somebody--and nothing
confusing to anybody.