Some disputes about PCT and about the phenomenon of control recur. I wonder if they come from avoidable confusions and misunderstandings.
One likely source is equivocation about one’s point of view. One set of points of view has to do with the roles of observers, observed subjects, experimenters, and experimental subjects. In a related way, our understanding of the word ‘perception’ is modulated according to who is the perceiver and what is perceived. We sometimes slip from one point of view to another in the same discussion–sometimes in the same paragraph or even in the same sentence–and the disputes and arguments that follow come back again and again like Halloween spirits to disturb us.
First, some roles. The observer sees the behavioral outputs of a subject person or other organism, and sees influences that affect concurrent variable conditions in the environment.
The experimenter is an observer who may manipulate disturbances and measure outputs, disturbances, and correlate these with changes in observed environment variables.
The PCT experimenter determines what the subject is controlling. The procedure is called the Test for the Controlled Variable (CV):
- Guess what variable V the subject S is controlling.
- Verify that S can perceive V.
- Verify that S is in fact currently perceiving V.
- Verify that the behavioral outputs of S can influence the state of V.
- Gently disturb the state of V and verify that the behavioral outputs of S negate the effect of those disturbance, so that V is maintained in a preferred state (or a preferred trajectory in state space, or the like).
If any verification fails, return to (1) and guess again. (Phil Runkel specified the Test more carefully in People as Living Things. With Dag’s permission and help, I will be posting that last section of Chapter 7 posted here for reference.)
In this experimental situation, the controlled variable is an aspect of the environment. It has to be in the environment because it is perceived concurrently by E and by S. S controls a perception of it in a preferred state, and E verifies that S is doing this by concurrently controlling in a way that (gently and temporarily) conflicts with S’s control. S and E are each controlling their own respective perceptions of the same variable aspect of the environment.
An immediate source of potential confusion here is that the E has only her own perceptions of environment variables and can only imagine what S is perceiving. In steps 2-4, it is crucial for E to adopt the point of view of S as closely as possible. Success in doing so is demonstrated by success performing the Test. According to the PCT model, S’s perception (a variable that is internal to S) co-varies with the state of affairs in the environment, as does E’s perception of that state of affairs. When E introduces a disturbance (a second variable that E is controlling) and S successfully controls despite that disturbance, then all three values are co-varying: the environment variable, S’s perception of it, and E’s perception of it.
A different potential source of confusion and dispute concerns perceptions, perceivers, and what is perceived.
In our subjective experience, perceptions just are. Indeed, our subjective experience consists of nothing but perceptions in this sense. The tautology appears to be inescapable: perceptions defined as subjective experiences comprise the entirety of our subjective experience. And this meaning is what we almost always have in mind, or assume in the back of our mind, when we talk about perceptions in PCT. If PCT doesn’t ultimately come back to perceptions as we subjectively experience them, what is its significance?
In neuroscience and psychophysics, subjective perceptions are correlated with rates of firing in bundles of nerve axons and dendrites, as regulated at synapses. In the theory, Perceptual Control Theory, the quantity that results from experimental measurement of the CV is taken to be the input quantity, qi, and the value of a perceptual signal (approximately in the psychophysics sense) is defined in proportion to qi. The two variables, the CV and qi, co-vary by a constant multiplier.
The numerical value of p does not have to be an actually measured rate of neural firing. It only has to co-vary proportionally with qi (the quantity resulting from measuring the CV in some relevant units). Proportional co-variation is, essentially, the assertion that when the input quantity and output quantity of a control system model, relative to disturbance quantities, are identical to input, output, and disturbance quantities measured in experiments with living subjects, then control structures hidden within the living subject are revealed by the observable and known structure of the model. In some psychophysics experimental work, the numerical value of neural firing has been shown to be proportional to quantities resulting from measurement of physical environmental variables, but for PCT modeling we do not have to demonstrate that each time; the value of p only has to track the value of the CV in the environment so that they co-vary proportionally.
In the experimental point of view, the controlled variable, CV, is in the environment. The reference value is also in the environment as an observed regularity in the value of the CV measured over time. This is why control is an observed phenomenon. The theory, with its postulation of theoretical entities like the perceptual signal p, the reference signal r, and the error signal e, explains the observed phenomenon.
In the theoretical point of view (e.g. in our diagrams of a PCT model) the CV is commonly taken to be the perceptual signal p, and it is sometimes argued that the CV cannot be in the environment, it can only be the perceptual signal in the organism. One source of this may be from thinking of the comparator as the crucial location at which control takes place. It is there that p is subtracted from the reference signal r. But it’s just as crucial that qo is added to d in the environment, for example, and it is absolutely essential that successive variables around the loop co-vary proportionally. Control is a property of the entire control loop.
A second source of this posture that the CV cannot be in the environment may be confusion or equivocation about the nature of the perception that is controlled. As I said earlier, perceptions defined as subjective experiences comprise the entirety of our subjective experience, and this is what we almost always have in mind or tacitly assume when we talk about perceptions in PCT. Our constructivism is sometimes abbreviated to the confrontational maxim “it’s all perception”. If there’s nothing but perceptions, then nothing but perceptions can be controlled, QED. Sounds kind of block-headed when I boil it down to that, but that indeed is what most arguments of this sort boil down to.
More subtly, we agree that there is something Really Real going on in the universe, and we also agree that our perceptions are all that we can know about it. This opens a conceptual window to be able to say, yes, according to the theory we control the internal perceptual signal p in conformity to the internal reference value r, but these are entities of theory that correspond to the environmental measurement qi and the reference value at which we observe that the subject is controlling it. The theory and its elements and relations are themselves only perceptions that bear certain relationships of test and verification (more perceptions) to the universe of subjective experience which is all that we really really have.
Observe carefully here how we have to shift deliberately from one point of view to another in order to talk about this articulately. We should not be blind people standing at just one end of the elephant.
We can go in many directions when we discuss PCT and its ramifications and implications. It is a vastly encompassing theory, and its scope compels us to come at those ramifications and implications from different directions. It is important in all discussions to be clear about these central distinctions of one’s point of view. What role? Subjective perception or perceptual signal p as a theoretical construct?
The test for controlled variables is a controlled conflict; that is, when control has been verified, and the reference value has been identified, the conflict of the disturbance value with the reference value is itself a perception, and the experimenter E is controlling that perception. The conflict itself, and its transience, is controlled by E. Prior to that, while guesses as to the CV are still not accurate, E is controlling a perception of conflict between the disturbance value and what she currently imagines the reference value to be.
This is related to collective control as follows: if the disturbance d succeeds in maintaining the CV at a value slightly different from its reference value, such that the outputs of S cannot overcome d or the loop gain in S does not produce sufficient output (S “doesn’t care”), then the deviant value is collectively controlled by E and S.
Now we come to the reasons for developing hypotheses, testing them, and constructing theories in science which become the organizing framework for developing and testing further hypotheses. It is a systematic way of developing input and output functions and improving our control. But that amounts to nothing if only one person does it. Isaac Newton’s hypothesis of force at a distance as an explanation of regularities observed by Brahe, Kepler, Copernicus, and others, and the mathematics that he invented, would have been as transient as an apple mouldering to compost had he not written and published and had others not read and recognized and developed further consequences of his perceptions. Science inherently involves collectively controlled variables (q.v.). Think of the socially adept Oppenheimer vs. the possibly genius physicist off on a ranch in Montana, unknown and forgotten, as this contrast was portrayed in Malcolm Gladwell’s book Outliers. This is pertinent to the history and current status of PCT.
These last three paragraphs each lend themselves to digression into other topics. The point of this present topic has been that, whatever we are discussing, we need to be mindful of who is doing the perceiving (point of view) and whether we mean perception as we experience it or as we model it.