Essentials of PCT

Continuing the discussion from Where Rick's Chapter 7 on "Social Control" goes off track:

I think that’s a great idea. Perhaps approaching this that way can keep these posts a lot shorter. And since it’s no longer about where my “Social Control” chapter goes off track, I’m moving this discussion to a new thread under the Fundamentals category called “Essentials of PCT”

I agree with most of them. But rather than get into why I disagree with some of them, I will just say that I think none of them, except possibly #2, are essential to a correct understanding of PCT.

Here is my list of what I think are the essential things to understand about PCT . Let’s see if we can agree on these:

  1. The behavior of living systems is a process of control; behavior IS control.

  2. Control is an objective phenomenon that is seen when variables are being maintained in a constant or variable rederence state, protected from the effects of disturbances.

  3. It is essential to identify the variables around which behavior is organizedcontrolled variables – in order to correctly understand it.

  4. Failure to identify controlled variables results in one or another type of behavioral illusion.

  5. Controlled variables are the basic data of behavioral science research from a PCT perspective.

  6. PCT explains the existence of controlled variables as the control of perceptual analogs of these variables, called controlled perceptual variables, that are inside the behaving system.

  7. Since control systems control what they perceive, controlled perceptual variables (inside the system) are exactly equivalent to controlled variables (outside the system).

  8. PCT accounts for complex behavior by postuating a hierarchy of control systems, with systems at ascending levels of the hierarchy controlling more and more complex types of controlled perceptual variables and, hence, more complex types of controlled variables.

  9. Conflict between control systems always reduces or eliminates each system’s ability to control, whether the conflicted systems are in the same or different individuals.

I think these are some of the central facts about PCT that everyone working in this field should be able to agree on. I hope we can agree on them, Bruce, and move on to bigger and better things.

Hi Rick, I agree with all of these… except one of them, number 7. Surely no perceptual analogs of controlled variables are perfect, and neither is control, sometimes due to a partial mismatch between the internal perceptual variable and the external controlled variable?
I actually agree with your statement about conflict too, but it misses any statements regarding how conflict between two systems might nonetheless entail side effects that improve the control of a different system.
Talk to you soon

My essential point #7 was that controlled perceptual variables are exactly equivalent to controlled variables. That’s simply a statement of fact about the PCT model. There is nothing in the PCT model about there ever being a mismatch between controlled perceptual variable and controlled variable. That’s because the controlled perceptual variable is a theoretical construct that is part of the explanation of the existence of controlled variables.

But it is a fact that control is not perfect and this is presumably a result of many differerent factors, one being neural noise. But as I note in Marken & Horth (p. 952) noise doesn’t degrade performance as much as one might think due to the filtering action of the control loop itself. And the controlled perceptual variable is still an exact analog of the controlled variable, the noise being, on average, a small constant offset to variations in the perceptual variable.

It misses them because I don’t think that’s something that is essential to know in order to understand PCT. I can certainly think of situations where conflict between two systems would improve the control of a different system. For example, if A, B and C are living systems vying for the affections of D (another living system) then if A and B get into a fight to the death over D then their conflict has certainly improved C’s ability to control, providing a clear path to victory.

As per my statement about conflict, the conflict between A and B has reduced – actually, eliminated-- their ability to control, which also elimimates a potential source of disturbance to C’s ability to win D’s affections. So just understanding C’s behavior as control and the conflcit as the removal of a source of disturbance to this control is all one needs to know to provide a nice PCT explanation of the situation you describe.


Statement 7 does not seem correct to me either, since perceptual variables are an analog of controlled variables (outside of the system) rather than being an exact equivalent of what is outside the system.


I know. It’s a hard one to get.

I think it might help if you can think of a controlled variable as being a function of environmental variables rather than a variable in the environment. The controlled perceptual variable is the analog of the controlled variable inasmuch as it is the same function of the sensory effects of the environmental variables of which the controlled variable is a function.

Another way to see this is to look at it mathemtically The mathematical definition of the controlled variable (cv) is: cv = f(x.1,x.2…xn) where the x.i are environmental variables. The mathematical definition of the perceptual controlled variable (p) is: p = f(s.1,s.2…sn) where the s.i are the sensory correlates of the x.i of which the cv is a function. The p is the analog of the cv in the sense that the same function, f(), defines both.

In our computer models of perceptual control we often use the same equation to represent both cv and p. So when I say that cv and p are equivalent, I am speaking from the perspective of the model. In reality, p is at best linearly and at worst monotonically related to cv. But PCT assumes that the functions that define p’s are exactly equivalent to the functions that define the corresponding cv’s.

Yes. Your list addresses some fundamentals of PCT, which has been mostly concerned with behavior of individuals.

I provided a list of some points of agreement, or at least not contested, within the discussion of multi-individual interactions, with the proposal that we set those aside as relatively firm context and discuss other items as matters for which PCT does not presently have good answers.

You’re welcome to abandon that discussion and start this one. In general, I agree with the intentions that I perceive in your list of ‘essentials’. I have some suggestions about how to express them.

Might be best to omit the word ‘objective’, and say just that control is an observed, experienced phenomenon.

Bill did challenge the notion that a phenomenon (a perception) can be ‘actually objective’ in some way:

Objectivity as I understand it is a perception (what else could it be?) that depends upon collective control — current, remembered, or imagined. Scientific consensus is its most disciplined form.

But I don’t want to make too much of Bill rejecting ‘objectivity’, his point in 1993 was “PCT is about experience”. What Bill was driving at was not the objectivity of the phenomenon of control, but rather that it is a straightforward, undeniable perception like rotation once you know how to recognize it, not at all abstract or theoretical. The slight lag and the asymptotic draw toward zero error gives an unmistakable feeling to disturbance-resisting, as when you gently pushed the arm projecting from that little box of his when we met in your LA hometown, and felt it resist your pressure in that unmistakable, slightly pulsating way — the feeling of life. As Dag was driving Bill someplace, his visceral experience of resisting Bill’s variable pressure on the steering wheel from the passenger seat. Aha! That’s what control is.

This distinction between CVs and CPVs (?) is a new way to put it. Last I knew the perceptual variable is the CV, and it is a function of a set of physical variables perceived to be in the environment when you wear ‘physical science glasses’. When you don’t wear ‘physical science glasses’ you experience the CV as though it were what is ‘present in the environment’ — the taste of lemonade, the color purple. The only way to put on ‘physical science glasses’ is by participating in the scientific consensus of the physical sciences (wrt the matter at hand), which for us generally means accepting statements of such consensus as true. Yes, ma’am Ms. physicist or Mr. textbook author, you’re the expert, not I. So we rely upon variables collectively controlled by physical scientists as feedback paths for controlling variables that matter to us for PCT.

From the point of view of experimenters, their experience is of controlling something in the environment and having their control actions resisted by the subject. The experimenter is ‘actually’ controlling a perception, of course (per the PCT model), which is an analog of effects of physical variables {v1, v2, … vn} on sensors (per the physical science models), but all anyone has to go on is experiences. “PCT is about experience” is one of the essentials of PCT. When we forget that we risk treating theoretical entities as realities. Sometimes even ‘projecting’ them into the environment as though they were experiences like the taste of lemonade, the color purple, and the phenomenon of control.

Hard to defend ‘exact equivalence’ for an individual control hierarchy. In addition to Warren’s concerns this formulation covertly smuggles in an assumption that something like the taste of lemonade is present as such in the environment. Not just hard to defend but outright indefensible for collective control. At the risk of again boring you so much that you stop reading … Say you’ve confirmed that a variable that you perceive is consistently restored to a reference value when you disturb it. You’ve even identified a living control system (LCS) whose actions as you perceive them seem to be responsible for resisting your disturbances. How do you know that is the only LCS controlling that variable? You have identified a virtual controller which might or might not have only one LCS contributing to it, the limiting case to which PCT has mostly limited itself. The answer, how you know, is of course that you consider influences from other LCSs to be confounding variables which your experimental setup is designed to eliminate. Sorry. Only works for artificial lab situations.

So am I going off topic and reverting to the Social Control topic that Kent started? If these problems are not also essential to PCT, then PCT so constrained is irrelevant to a great proportion of the experiences that interest people. And that is a problem which is essential to PCT in another way.

Hi Rick, I guess I just have an aversion to skyhooks. How does the correct mathematical function based on all the necessary inputs spring into being and work perfectly? Doesn’t making this assumption ignore the important principle of reorganisation in reducing error in control units that is not reduced from closed loops that operate at their current organisation?

In an earlier post you said “Surely no perceptual analogs of controlled variables are perfect …due to a partial mismatch between the internal perceptual variable and the external controlled variable?” [emphasis mine]. In PCT, perception is not a “matching” process; it is a constructing process. The perceptual function defines the aspect of the environment that is constructed as a perception. So there is no issue of how well a perceptual function matches the function that defines the controlled variable. The controlled variable is not something in the environment to be perceived correctly; it is the aspect of the environment, defined by the perceptual function, that IS perceived, .

When I described the controlled variable as cv = f(x1,x2…xn) I was not describing what the perceptual function should perceive; I was describing what the perceptual function does perceive. In PCT, the perceptual function, f(), defines the aspect of the environment that an observer sees as the cv. The controlled variable, cv, can be considered the observer’s perception that corresponds to the perception, p, that the control system is controlling.

This is actually the best way to look at what is going on in the test for the controlled variable (TCV). The experimenter is trying to perceive (as the cv) the same aspect of the environment that the control system is perceiving (as p) and controlling. The TCV is successful when the observer has determined that cv = p.

I should add that, if the test is successful then cv = p even if the control system under study is a human who denies that you have identified the cv correctly (see. B:CP, 1st Ed., p. 235-236; 2nd Ed., p. 237-238).


Thank you for the clarification. I understand.

A suggestion though. Rather than using the term “controlled variable (cv)”, using the term “perceived control variable (pcv)” might lead to less misunderstanding or confusion by others.


The problem with that is that it could suggest that there is a controlled variable “out there” to be perceived. My point is that the controlled variable IS a perception, just as it is (per PCT) in the system whose controlling is being observed.

As observers we don’t think of the controlled variable as a perception, nor do we need to. For example, in the compensatory tracking task we can see that the distance between cursor (c) and target (t) is being controlled; we can see (because it is protected from disturbance) that the distance between c and t is the controlled variable, we see that the variable being controlled is c - t. I don’t think it’s necessary to add that that distance is a perception.

So I will stick with calling the controlled variable the controlled variable.

That’s all you need to know in PCT; how control works in individuals. Applying PCT to groups (collectives) of control systems just involves applying PCT to all the individuals in the group. Powers demonstrated this long ago with the CROWD program. I demonstrated it in Ch. 7 of The Study of Living Control Systems.

I wasn’t using “objective” in that sense. I was using it in the sense that control can be considered objective because two or more people can agree that it is happening.

Yes, and I agree with that, of course. Bill was just saying what I’ve been saying about the controlled variable; it is not an objective entity in the environment; it is a perception (an experience).

The difference is that the CV is the observer’s perception of the system’s CPV (otherwise known as p).

I don’t believe anyone experiences CVs any other way then as “present in the environment”. If someone is experiencing CVs, such as the distance between cursor and target, as a perception in their head then they are either on something or should seek treatment.

I think I have shown why Warren’s concerns are unwarrented (no pun intended – well, maybe a little;-) If you can read what I said as covertly smuggling in an assumption that “something like the taste of lemonade” (a CV) is present in the environment then I have really failed to explain my point. I’ll try again: The CV is an observer’s perception of the perceptual variable that the system is controlling. It’s all perception, but neither the experimenter nor the control system treat it that way. However, competent PCT observer/modelers assume that the CVs they see being controlled are perceptual variables for both the system under study and themselves…

Don’t you just open your eyes and look?

I think you are describing a situation that never happens. I will not even try to answer this until you give me a concrete example of what you are talking about. And be sure to include an example of a virtual controller made up of one LCS. I’d really like to see what that is.

The problems you describe are not essential to PCT because, as far as I can tell, they exist only in your imagination. PCT explains control phenomena (or, if you prefer, experiences) that are seen in the behavior of an individual or groups of individuals. I have not seen even one concrete description of the “great proportion of experiences that interest people” that you think my (and Powers’) version of PCT doesn’t explain. I think you are bluffing. I see you and call you.