[From Rick Marken (951231.1900)]
Bill Powers (951230.0500 MST)
One doesn't simply convert a rule of language directly into
output; instead, one starts producing a sentence, perceives what has
been produced so far in terms of fit to some rule or principle, and
selects the next word, in part, to correct incipient departures from the
rule or principle.
Bill Powers (951231.0930 MST) --
The PCT concept of control says that what we really control are
perceptions.
After reading Bill's lovely posts on control of language and speech I
realized that anyone who understood what was going on in the basic
compensatory tracking task could have come up with these analyses.
But you do have be able see the generality of the principle illustrated
by these experiments; in particular you have to be able to see that the
perception of distance between cursor and target in the compensatory
tracking task is the same, in principle, as the perception of the
grammatical structure or phonemic content of an utterance.
Once you understand the principle of control of perception you can see
that all behavior -- simple behavior, like controlling the position of a
cursor, and complex behavior, like controlling your the position of a
preposition-- is control of perception.
A control system controls perceptual signals; these signals must be
derived from physical variables or lower level perceptions via a
perceptual function. The perceptual signal that comes out of this
perceptual function can be an electrical current, the mechanical
displacement of a coil or a series of neural impulses. But whatever it's
physical nature, it is this _perceptual signal_ that is actually controlled
by the control system; and it is the _perceptual function_ that
determines what aspect of the system's environment is controlled
when the perceptual signal is controlled.
Much has been made of the fact that PCT has only been demonstrated
in "simple" or "trivial" situations, like the compensatory tracking task.
But the "simplicity" or "triviality" of the compensatory tracking task is
as illusory as the "simplicity" or "triviality" of Galileo's inclined
plane experiments. The compensatory tracking task shows that, when
people control, it is a perceptual representation of some physical state
of affairs that is controlled. This is not simple; it's very _deep_.
The results of the compensatory tracking task show how control works --
regardless of what aspect of the environment is under control. Any
aspect of the environment that can be controlled will be controlled in
the same way; it will always be the output of a perceptual function that
is controlled.
Once PCT has been shown to account for control of simple perceptual
variables, we know that it also must account, in principle, for control of
complex perceptions. The argument is simple; PCT says that a perceptual
signal, p, is controlled and that all perceptual signals are a function
of environmental variables, q.i. So p = f(q.i). This statement is true
whether p is a simple function of q.i (as in compensatory tracking) or a
complex function of q.i (as in language and speech).
The compensatory tracking task also shows that control (the stabilized
position of the cursor) is not the result of stimulus or informational
guidance, selection by consequences, or output generation. Indeed,
the "simple" compensatory tracking task shows that none of the
exlanations of behavior that have been advanced in the behavioral
sciences can account for the subject's ability to control line position.
Much has also been made of the fact that very little research has been
done on the PCT model of behavior. This is true; much more research
has to be done to examine the details of control and the control model.
But the research on PCT that has already been done shows that
behavior is, indeed, the control of perception. This means that _all_
conventional models of behavior are rejected by the results of the
simple compensatory tracking experiment that have already been
reported in the literature (eg. Powers, 1978; Marken, 1992; Bourbon &
Powers, 1993).
Much has also been made of the fact that there is a hugh amount of
behavioral data already out there and that PCT cannot and will not be
taken seriously until it accounts for a lot of this data. But the results
of compensatory tracking experiments also show why PCT cannot be
expected to account for any randomly observed relationship between
an environmental and a behavioral variable; it can't be expected to do
this when there is no way for the PCT theorist to know what variable
was under control -- which is almost always.
There are many potentially interesting and reliable relationships that
might be observed between environmental variables (such as the
number of people in a room) and behavioral variables (like the
probability of the heater going on) when studying a control system, such
as a thermostat. But if you have no idea that the heater is the output of
a system controlling perceived temperature, then finding a PCT explanation
of a result (like "the number of people in the room is a disturbance to
controlled temperature") is just a matter of luck.
The results of simple, compensatory tracking experiments show that
1) all control involves the control of the output of perceptual functions
2) conventional cause-effect models of behavior are, therefore, incorrect
and 3) PCT cannot, in general, be expected to account for non-PCT
behavioral research.
Because I am writing this on the eve of the New Year, I suggest that we
all resolve to apply, in 1996, the lessons about control that we should
have learned in 1978 (when Bill Powers published the results of some
very simple and elegant compensatory tracking experiments in his
Psych Review paper): one of those lessons is "now that we know that
behavior is the control of perception, let's stop trying to understand
behavior as though it were being produced by a cause-effect system and
let's start trying to understand behavior as though it were being
produced by a living control system."
I hope 1996 will be the year when we finally increment the number of
PCT researchers by one, or two, or ...
Happy New Year
Rick