Corrections and comments to Rick's The Study of LCS

amatic · July 25, 2021, 7:51pm

Lots of mistakes in this overpriced guide from the wannabe leading authority on PCT. I’ll go over and comment the things I feel are not quite correct, or maybe some parts I think are unexpectedly good. I’m writing this as an exercise for myself, something like a journal club. If anyone wants to join the discussion, welcome.

So, let’s see, the first sentence on the cover got me thinking about causes:
“This book is a guide to doing a new kind of psychological research that focuses on the purposes rather than the causes of behavior.”

Yeah, I do want to know the causes of behavior. I want to know causes for changes in behavior, reasons for why people behave as they do. Nothing about control theory prevents learning about the causes of behavior. In a paper from 1980, “The cause of movements …” Rick explicitly names the reference signal as the cause of movements, so the purpose of the movement is the cause of the movement. A bit contradictory to the quoted sentence, and I think both are wrong. The behavior of the loop is the result of interactions of all of its elements, in the organisms and in the environments. Sure, the reference signal is causing things to change, but so is the stimulus and so are the changes in properties of loop functions.

If some element in the feedback loop was different, say the gain was much lower, then the movement would have been different, the gain change would have caused the difference. If anything, control theory promises to find the exact causes of behavior, while ‘conventional’ psychological research finds only correlations.

Nothing wrong with explanations of behavior that use the word “cause”, they are just pretty abstract and coarse-grained, more like a story than a precise model you can plug into a computer, run, and compare to the modeled system. Causal explanations are just stories, or maybe “program-level” explanations, abstractions over events, and relationships between them. The tracking task model is a “lower level” explanation, but we can still talk about how its elements cause some variables to change.

Ok, so my idea of what causes are is completely different from Ricks. Next…

“The conventional methods of psychological research completely ignore the existence of controlled variables”

Maybe. In some papers, I’ve found what I think are controlled variables hidden behind expressions such as “cue” or “salient stimulus”, “reward”, “goal”, “cost” etc. In motor control, some people are explicitly looking for controlled variables, using control theory, etc. But OK, there is certainly a need for different methods of research in psychology.

Rick thinks everything is a behavioral illusion:

“Perceptual Control Theory (PCT) – a theory that explains how living control systems work – shows us that, if organisms are living control systems, then the independent–dependent variable relationships that are found in conventional psychological research are actually side effects of the disturbance-resisting nature of these systems and, therefore, tell us more about the nature of the environment in which these systems do their behaving than about the systems themselves (Powers, 1978)”

Not at all. The 1978 paper talks about the relationship between continuously measured stimulus and behavior variables in a control experiment, not in ‘conventional’ correlational research with samples from populations and all that. It is debatable how useful they are, but the relationships found in ‘conventional’ research don’t say anything about the environment, that is just nonsense.

More later.

(P.S. Since most of my discussions with Rick ended in name-calling and insults, I think I’ll just not reply to any of his comments)

rsmarken · July 27, 2021, 1:36am

Hi Adam

AM: Lots of mistakes in this overpriced guide from the wannabe leading authority on PCT…

RM: I agree that it’s overpriced but I had no control over the pricing decision. I guess I do consider myself an authority on PCT but clearly I’m not an authority on the version of PCT that is being collectively controlled here on Discourse. That version of PCT, which is the virtual reference state of a controlled variable that can be called “PCT”, is not near where it was when Bill was here. So I’m just pulling on my connection to “PCT” by replying to comments that I think are incorrect, in an effort to bring that collectively controlled variable a little closer to my reference for it.

AM: So, let’s see, the first sentence on the cover got me thinking about causes: “This book is a guide to doing a new kind of psychological research that focuses on the purposes rather than the causes of behavior.” Yeah, I do want to know the causes of behavior…"

RM: I describe the book as I do, in the quote above, because it presents an alternative to the approach to research that is taught in every introductory (and advanced) Research Methods course in psychology. Pick up any textbook on Research Methods in Psychology (or related fields, such as education) and you will find a statement declaring that one of the main goals of psychological research is to determine the causes of behavior. For example, here’s one from a very successful text on research methods in psychology: “The principal method for acquiring knowledge and uncovering causes of behavior is research" (Bordens and Abbott (2011) Research design and methods, McGraw-Hill, 11th edition, p7). The approach to research described in my book focuses on the search for the variables organisms control – their purposes – rather than the causes of their behavior.

RM: “The conventional methods of psychological research completely ignore the existence of controlled variables”

AM: Maybe. In some papers, I’ve found what I think are controlled variables hidden behind expressions such as “cue” or “salient stimulus”, “reward”, “goal”, “cost” etc. In motor control, some people are explicitly looking for controlled variables, using control theory, etc. But OK, there is certainly a need for different methods of research in psychology.

RM: I think the controlled variables in conventional research seem hard to find because they are so obvious. And because very few people are actually looking for them, anyway, even in motor control experiments. I wrote the book because I am trying to encourage behavioral researchers to stop looking for the causes of behavior and start looking for the variables around which behavior is organized – controlled variables. And the way to do this is by using methods based on an understanding of behavior as control – the methods described in my book. So I appreciate your recognition that there is a need for different methods of research in psychology.

AM: Rick thinks everything is a behavioral illusion:

RM: “Perceptual Control Theory (PCT) – a theory that explains how living control systems work – shows us that, if organisms are living control systems, then the independent–dependent variable relationships that are found in conventional psychological research are actually side effects of the disturbance-resisting nature of these systems and, therefore, tell us more about the nature of the environment in which these systems do their behaving than about the systems themselves (Powers, 1978)”

AM: Not at all. The 1978 paper talks about the relationship between continuously measured stimulus and behavior variables in a control experiment, not in ‘conventional’ correlational research with samples from populations and all that. It is debatable how useful they are, but the relationships found in ‘conventional’ research don’t say anything about the environment, that is just nonsense.

RM: I borrowed that turn of phrase from Bill. The “nature of the environment” that disturbance-output (independent - dependent variable) relationships tell us about is the nature of the feedback function relating system output to input. But perhaps I should have left out the part about IV-DV relationships telling us more about the environment than about the organism and just said that these relationships don’t tell us much about the organism, period. I think Bill said it the way I say it in the text is because he had used analog computers to measure unknown functional relationships between variables using this “behavioral illusion” – the fact that the relationship between d and o is the inverse of the feedback connection between o and cv.

Best, Rick

amatic · July 27, 2021, 4:36pm

There are lot’s of little things I disagreed with in the introduction, but the larger mistakes are in the diagrams and the equations of control, so I’ll skip the minor ones.

Marken Figure1

AM: The environment part of this diagram is a complete mess. But, really, the problems start with the referenced diagram from Feedback - beyond behaviorism (Powers, 1973):

1973 diagram1

The problem here is the different way of representing functions and variables in the environment and inside the organism ( or System). For some reason, Bill decided to represent system variables as arrows, and system functions as boxes; while the environment variables are represented as circles and environment functions as arrows. Not a very good choice, because there are a few unnamed functions (arrows) in the environment.

In making his diagram, Rick mixes up those two conventions in the environment. The feedback function is in a box, even though it is in the environment, the output quantity is represented by an arrow, but the input quantity is represented by a circle.

These things are a bit pedantic, but they can cause so much confusion - what is a function, what is a variable, etc… And alternatively, when the diagrams follow the verbal and mathematical description, I think they can really help in understanding and clearing things up. Minus points for Bill in this one, too.

Also, I completely disagree with equating the input quantity with the controlled variable. For one, there can be multiple input quantities, but only a single controlled variable per loop. The names are separated for a good reason.

Then the equations… oh boy…

Marken equations

They are incorrect. Try the solve the algebra, it does not work. Try plugging into a simulation, again fail.

Apparently, the goal was to simplify the equations, so one option is to use ‘steady-state’ conditions, meaning when the variables stop changing, they are all stable, the controlled variable reached the reference, and the disturbance is not changing. In that case the equations should loose the (t) because this is the steady state - they are not changing in time:

a1. qo = O * (r - p)
a2. qi = F * qo + d
a3. p = I * qi

(this is similar to B:CP, p. 286)

Alternatively, to include time in the analysis, the equations need to show how the current state of the variables depends on the previous state. After LCSIII, p. 82

b1. qi(t) = F * qo(t) + d(t)
b2. p(t) = qi (t - tau) // assuming I = 1, tau is the transport delay
b3. e(t) = r(t) - p(t)
b4. qo(t + dt) = qo(t) + [ O * e(t) - Kd * qo(t) ] * dt

The output function here is a leaky integrator, not just a multiplier, so that the output quantity depends on previous values of the output quantity. It could also work without the transport delay, but not without the integrator in the output.

Then, of course, there is this:

There is no way to get from 1.1 and 1.2 to 1.3 and 1.4. What Rick call’s here “making the appropriate assumptions about the values of I, O and F” is a bluff. It doesn’t work. Also, while 1.4 is correct, 1.3 is not correct.

Here is a fix:

a1. qo = O(r - p)
a2. qi = Fqo + d

Plugging a1 into a2, assuming I = 1 and p = qi
p = F( Or - Op ) + d
p = FOr - FOp + d
p + FOp = FOr + d
and finaly
p = ( FO / (1 + FO)) r + d / (1 + FO)

Here, taking O to be a large number like 1000, and F = 1, we get
p = 1000/1001 * r + d/1001 (compare to eq. 7 in BCP, p. 288)

the first fraction is near 1 and the second near zero, so we get:
p ≈ r

Similarly, the equation relating qo with r and d should go:
qo = O (r - (F qo + d))
qo = O(r - Fqo - d)
qo = Or - FOqo - Od
qo + FOqo = Or - Od
qo ( 1+FO) = O ( r - d)

and finaly
qo = O / (1 + FO) * (r - d)

In this case, just like in the first one, we can’t see much without assuming large O, and some value for F, let’s say 2.
so

qo = 1000/ (1+2*1000) * ( r - d)
qo = 1000/2001 ( r - d)
qo ≈ 1/F * (r - d) (compare to eq. 9 in BCP, p. 289, asuming ke and ki are one)

Note the difference from Rick’s 1.3. qo = r - 1/F * d

Namely, according to my solution, the output qo will be the inverse feedback function (1/F) of the disturbance input only if the reference r is zero.
qo ≈ - 1/F * d

This is the behavioral illusion equation, by the way.

It is a long tradition that scientists make bets instead of bickering. So, Rick, I won’t argue with you. If you want to put some money where your mouth is, I’ll take your bet.

MartinT · July 27, 2021, 4:45pm

I suspect that Bill would be rather disappointed if he were to come back and find no progress in the development of PCT since he died. However, what is mostly discussed here (except by RM, whose vision seems to change from day to day depending on who he is arguing against) is firmly based on PCT as described by Bill.

Using the TCV as a basis for trying to understand the problem of why RM makes his complaint, I conclude that he controls two main perceptual variables by his contributions to the Forum in response to disturbances cause by the contributions of others: (1) The level of conflict between himself and other contributors, with a reference level of “high”, and (2) the degree of understanding of Perceptual Control and its implications by others, with a reference level of “low”, both of which he controls by filling so many threads with ever-changing descriptions of “authoritative” PCT that contradict what someone else contributed to the thread…

Martin

rsmarken · July 27, 2021, 10:26pm

Hi Adam

AM: There are lot’s of little things I disagreed with in the introduction, but the larger mistakes are in the diagrams and the equations of control, so I’ll skip the minor ones.

RM: Thanks for the detailed analysis. I, too, disagree with lots of the things you say but I do appreciate you taking the time to do such a fine grained evaluation of my book.

RM: Your two main criticisms concern my Figure 1.1 and my derivation of the equations for input and output from my equations 1.1 and 1.2. I think Figure 1.1 is fine. I also think my derivations of the input and output equations are fine as well; at least we both end up with the same results.

RM: For the input equation we both get p ≈ r (although I put in the time variable, t, but just to show that these are variables that vary over time, as noted in footnote 3 of Chapter 1). For the output equation, however, you get a result that looks different than mine. Your result is:

qo ≈ 1/F * (r - d) (1)

RM: and I get

qo = r - 1/F * d (2)

RM: You go on to note that your result is consistent with eq. 9 in B:CP, p. 289, and it is, in the second edition. My result is consistent with eq. 14 in B:CP , p. 277 in the first edition. Bill got there by including the “environment function”, k.e, which is the effect of a disturbance on the controlled variable. So Bill’s equation comes out to be

qo = r - k.e/F * d

RM: I am taking k.e to be 1. So your equation (1) and mine (2) are both right after making the appropriate assumptions about the the functions O, F and k.e. But surprisingly, your equation leads you to a conclusion that is inconsistent with the way living control systems behave. You say:

AM: according to my solution, the output qo will be the inverse feedback function (1/F) of the disturbance input only if the reference r is zero.
qo ≈ - 1/F * d
This is the behavioral illusion equation, by the way.

RM: My equation, on the other hand, leads me to the conclusion that the behavioral illusion will be seen even if the reference r is non-zero; it occurs as long as r is a constant. And this is in fact observed. You can demonstrate this to yourself by doing the behavioral illusion demo, first doing a whole run keeping the spider “in the box” and then another run keeping the spider slightly to the left (or right) of the box (there is not enough range in the output that would allow you to keep the spider too far to the left or right of the box).

RM: Of course, you only get a good picture of the illusion if k.e is constant also. But, both equations (1) and (2) can be seen as describing a behavioral illusion, if the illusion is seen as simply being the appearance that stimulus variations, d, are the cause of response variations, qo, via the organism.

Anyway, thanks for the comments and I look forward to hearing what you have to say about the rest of the book.

Best, Rick

rsmarken · July 28, 2021, 7:06pm

Hi Martin

MT: I suspect that Bill would be rather disappointed if he were to come back and find no progress in the development of PCT since he died.

RM: Of course he would, although I think your idea of what constitutes “progress in the development of PCT” would be quite different than his.

MT: However, what is mostly discussed here (except by RM, whose vision seems to change from day to day depending on who he is arguing against) is firmly based on PCT as described by Bill.

RM: It’s certainly based on PCT as described by Powers, but not all that firmly. For example, your concept of a controlled environmental variable (CEV) is particularly problematic. It is either redundant with the concept of controlled variable (CV) and, thus, unnecessary, or it is simply wrong, which it is when conceived of as a variable in the environment that is not equivalent to the corresponding perceptual variable. As Bill said in one of his replies to you: “The CEV can never be defined independently of the perceptual signal…” [Bill Powers (940831.0610 MDT)]. That’s because the perceptual signal is a theoretical variable that explains the existence of what you call the CEV (and what Bill and I call the controlled variable or controlled quantity).

Best, Rick

MartinT · July 29, 2021, 4:29pm

Since this last message by Rick is irrelevant to Adam’s comments, as was mine to which Rick replies, I belatedly moved my next contribution to [Real Reality CEV and Perceived Reality CV]. My hope is that this thread will continue to be about Adam’s comments on Rick’s book.

amatic · July 30, 2021, 9:13pm

More on causality

I keep reading that somehow stimuli don’t cause responses. “PCT looks for purposes not causes”. As if “purpose” and “cause” are opposing concepts.

For example:

RM, page 7: The difference is in the way control theory is mapped to the behaving system, which results from a difference in the way behavior is viewed. MCT views behavior as output caused by stimulus input, whereas PCT views behavior as the control of input."

This is so silly, and I’m only on page 7. In PCT, as defined in Bill’s publications, behavior is caused by stimuli. Let’s see the Science paper already referenced for a plot (Feedback: beyond behaviorism, Powers, 1973).

It starts:

BP: “The basis of scientific psychology is a cause-effect model in which stimuli act on organisms to produce responses. It hardly seems possible that such a simple and venerable model could be in error, but I believe it is. Feedback theory shows in what way the model fails, and what must be done to correct our concepts of organized behavior.”

If you stop reading here, at the first paragraph, it might sound as if Bill is saying that stimuli are not causing the responses. But you’d be wrong. In the next paragraph he explains that this view is correct, but incomplete. Responses do depend on stimuli, and then the next stimulus (the proximal one, ie the CV) depends on the response. If these were continuous variables, changing over time, there would be stimulus(t) is causing response(t + dt) and response(t+dt) is causing stimulus(t+ 2*dt), or whatever, there would some time passing for the signal to travel around the loop (unlike in Rick’s formulas from my previous post). Bill even writes it explicitly: “Stimuli do cause responses.” (page 355 of the same paper, 3rd paragraph in the left column)

So, when the manual control theorists ‘see behavior as caused by stimulus input’, there is nothing wrong with this view, and, other than the naming of the signals, nothing different from PCT. Although, there are other differences.

Otherwise, it is pretty rich seeing an author who just failed at both writing and solving basic equations that describe a control loop, and also draws some awful diagrams, bash on people who seem to actually know some control theory, and draw fairly nice diagrams.

What, the reference signal is in the environment? What a heresy. But what about the “reference level”? The reference signal is a neural signal specifying the amount of the perceptual signal, but we could also use a quantity Bill calls the ‘reference level of the input quantity’, qi*, and make a diagram where everything is in the environment. Like so:

amatic · July 31, 2021, 9:22am

This one got me thinking, something seemed pretty off, and I couldn’t quite understand why:

RM (page 10). Since perceptual variables are assumed to be analogs of controlled variables, the same function that defines the perceptual variable, p(t), also defines the controlled variable, qi(t). It is in this sense that both qi(t) and p(t) are perceptual variables. Both are functions of the sensory effects of physical variables. While the functions that produce p(t) are the input functions, I, in the organism under study, the functions that produce qi(t) are the input functions of the human who is observing the organism’s behavior.

I mean, sure, if the observer can see the variable controlled by the (human) subject, the cv is the “perception of the observer”. But if the observer records the perceptual signal p, then both the p and cv are the perception of the observer. The quoted paragraph is mixing different meanings of the term ‘perception’. All the variables in the control loop are theoretically observable, and therefore could be in the observers ‘perception’, meaning experience. The observer can’t see the subject’s experience, but he can record signals and quantities in what he sees as the organism ‘inside’ and the organism’s environment, including p and cv. Following the same logic, the subject or organism doesn’t see his perception signal p, he is experiencing the controlled variable in the same way as the observer, as the cv in his environment.

So, no, the cv is not the analog of p in the sense that one is in the subject and the other is in the observer. Both are in the observer’s experience, as variables in a model, while the subject has the experience only of the cv, and theoretically, there is an analog of his experience that could be measured in neural signals.

Also, again, the controlled variable is not always the input quantity. And why use the shortened name of a different expression? If you want to shorten “the input quantity”, use qi, if you want to shorten the controlled variable, use cv. I don’t get it.

Next… drumroll…

The behavioral illusion

The code for this experiment, available on Rick’s website, has the correct formulas, so the experiment was performed correctly, but the explanation, the algebra, and the diagram are all wrong, and even the bloody graph does not have units on either axis.

Let’s see the diagram first. The dashed line is supposed to be the feedback function, but it says CV = kS, defining the controlled variable as a multiple of the stimulus? The function inside the ‘head’ defines the reaction as R = 1/k * S? On the good side, it doesn’t have the reference signal (because it is assumed to be constant? zero?), and all the variables are in the environment, similar to my diagram above, showing you don’t really need the environment-organism limit every time. Here is my correction, following the code:

In the experiment, the image of the spider is increasing in size when it goes to the left or right, but it is the smallest in the center. The participant tries to keep it in the center by moving the mouse. Standard compensatory tracking.

The controlled variable (CV) is the size of the spider on the screen. The O is the organism function, R is the position of the mouse, k is the feedback factor, k * R is the feedback quantity, S is the stimulus or disturbance. The controlled variable is defined as CV = S + k * R.

The problem is how to get from:

R = O (CV)
CV = S + k*R

to

R ≈ -1/k * S

In other words, how is it possible that the stimulus-response law measured in this experiment reflects the inverse (in this case reciprocal, or multiplicative inverse) of the feedback function?

Rick gets very creative in the text. He refers to the CV both as the “controlled variable” and the “effect of the response on the controlled variable”, in the same paragraph, and then writes “CV=0.5 * R” and “CV=2 * R”, and references the incorrect formula 1.3, and magically, the results are “precisely what is expected”.

To get from 1 and 2 to 3, the first thing we need to assume is that the organism function O is, as usual, a high gain amplifier, and also that the reference is zero. Then we say: in this control system the CV will be approximately equal to the reference, so CV ≈ 0. Then we have, from eq. 2:

S + k * R ≈ 0
k * R ≈ -S
R ≈ -1/k * S

And that’s it. If the control is good, the feedback quantity (k * R) is the opposite of the stimulus (S), and for this to be true, the response needs to vary as -1/k * S.

Well, disaster. Or as Rick said, “correct with respect to the purpose of the book”. What, as a guide to doing research? Looks pretty wrong with respect to that purpose.

The TCV

In comparison with the intro, the chapter on the TCV seems pretty good. The coin game, car following, an interesting anecdote from Feynman… Not bad.

The pseudocode for a model of pursuit tracking could be improved a bit. p. 31. If two variables are always equal, just use one. Also, you need to store the calculated values of c, so:

could be:

 For i = 1 to N:
    p = t[i] - c[i]    //or p= arcsine[(t[i] - c[i])/s]
    c[i+1] = c[i] + slow * [ gain * (r - p) - c[i] ]
 next i

Also, there could be an explicit time-step and time constant instead of the “slow”, but OK.

rsmarken · July 31, 2021, 8:15pm

Hi Adam

RM, page 7: The difference is in the way control theory is mapped to the behaving system, which results from a difference in the way behavior is viewed. MCT views behavior as output caused by stimulus input, whereas PCT views behavior as the control of input."

AM: This is so silly, and I’m only on page 7. In PCT, as defined in Bill’s publications, behavior is caused by stimuli. Let’s see the Science paper already referenced for a plot (Feedback: beyond behaviorism, Powers, 1973).

AM: It starts:

BP: “The basis of scientific psychology is a cause-effect model in which stimuli act on organisms to produce responses. It hardly seems possible that such a simple and venerable model could be in error, but I believe it is. Feedback theory shows in what way the model fails, and what must be done to correct our concepts of organized behavior.”

AM: If you stop reading here, at the first paragraph, it might sound as if Bill is saying that stimuli are not causing the responses. But you’d be wrong.

RM: Oh my!

AM: In the next paragraph he explains that this view is correct, but incomplete.

RM: I scanned through the article and can’t seem to find where he says that this view (that stimuli are not causing the responses) is correct, but incomplete.

AM: …Bill even writes it explicitly: “Stimuli do cause responses.” (page 355 of the same paper, 3rd paragraph in the left column)

RM: And if you stop reading there it might sound as if Bill is saying that stimuli do cause responses". But if you read on after that statement you will see that Bill is saying that that statement is correct, but incomplete:

BP: "If one knew the controlled quantity associated with a given stimulus-response pair, one would see more regularity in the relationship, not less. In fact one would see an exact quantitative relationship, for the effects of the response on the controlled quantity must come close to canceling the effects of the stimulus on that same quantity, and both these effects are mediated through the environment, where the detailed physical relationships can be seen. That implies, of course, that given knowledge of the controlled quantity one can deduce the form of stimulus-response relationships from physical, not behavioral laws. [emphasis mine]

RM: What Bill is saying is that there is a causal connection between stimulus and response in a control loop but observed relationships between stimulus and response reflect properties of the environment – the physical laws relating the response to the controlled quantity – not of the organism – the behavioral laws relating stimulus to response. In other words, the idea that observed S-R relationships reflect the causal path from S-R via the organism is an illusion – what came to be called the behavioral illusion. Some evidence that this is what Bill meant can be seen in his reply to one of the two behaviorists who commented on his Science paper:

BP: In the section on controlled quantities in my article, there appears an approximation , g (d) ≈ - h(o), which says that the cause-effect relationships that can be observed between stimulus events and consequences of nervous system outputs – responses – are expressible wholly in terms of the physics of the local environment, containing almost no information about the behaving system at all. I see no way in which behaviorism [which views behavior as caused by stimulation – RM] can survive a full understanding of the derivation and significance of this harmless expression. If control-system theory does indeed describe correctly the relationship between organisms and their environments, behaviorism has been in the grip of a powerful illusion since its conceptual bases were laid. [emphasis again mine]

RM: By the way, in that paper Bill plays somewhat fast and loose with the word “stimulus”; sometimes it refers to what we would call a disturbance variable and sometimes it refers to what we now called the controlled variable (which Bill called the controlled quantity in that paper) . At the beginning of the paper he uses “stimulus” to mean “controlled variable”, as evidenced by this quote:

BP: “But it is equally true that stimuli depend on responses according to the current organization of the environment and the body in which the nervous system resides”.

RM: This shows that, here, “stimulus” means controlled variable because disturbances don’t depend on responses but controlled variables do. Once Bill introduces the concept of the controlled quantity later in the paper he uses “stimulus” only to mean disturbance.

AM: So, when the manual control theorists ‘see behavior as caused by stimulus input’, there is nothing wrong with this view, and, other than the naming of the signals,

RM: Yes, this is exactly the way Bill (and I) see it as well:

BP: If one’s primary purpose is to keep pilots from flying airplanes into the ground or to make sure that a gunner hits a target with the shell, that is, if one’s purposes concern objectivized side effects of control behavior, the man-machine blunder [committed by manual control theorists – RM] amounts to nothing worse than a few mislabelings having no practical consequences. [emphasis mine]

AM: nothing different from PCT. Although, there are other differences.

RM: And these differences make all the difference. The main difference is that manual (or engineering psychology) control theory is about determining how well a person can control a particular variable while PCT is about determining what variables are being controlled when a person is performing well. My book is aimed at describing how to do the latter.

Best, Rick

rsmarken · July 31, 2021, 8:37pm

Hi Adam

AM: This one got me thinking, something seemed pretty off, and I couldn’t quite understand why:

RM (page 10). Since perceptual variables are assumed to be analogs of controlled variables, the same function that defines the perceptual variable, p(t), also defines the controlled variable, qi(t). It is in this sense that both qi(t) and p(t) are perceptual variables. Both are functions of the sensory effects of physical variables. While the functions that produce p(t) are the input functions, I, in the organism under study, the functions that produce qi(t) are the input functions of the human who is observing the organism’s behavior.

AM: I mean, sure, if the observer can see the variable controlled by the (human) subject, the cv is the “perception of the observer”. But if the observer records the perceptual signal p, then both the p and cv are the perception of the observer. The quoted paragraph is mixing different meanings of the term ‘perception’.

RM: I’m sorry if it’s confusing to you. It seems rather clear to me. In a simple tracking task, the distance between cursor and target is the perceptual variable controlled by the subject – the CV from the subject’s perspective. It is also a variable I (and/or the computer) can perceive – the CV from the observer’s perspective. If the observer were able to record the perceptual signal that is an analog of the CV from the subject’s perspective, the observer would have wonderful confirmation of the PCT model, seeing that her perception of the neural firing rate, p, corresponds to her perception of the CV.

AM: Next… drumroll…
The behavioral illusion

AM: The code for this experiment, available on Rick’s website, has the correct formulas, so the experiment was performed correctly, but the explanation, the algebra, and the diagram are all wrong, and even the bloody graph does not have units on either axis…

AM: …Well, disaster. Or as Rick said, “correct with respect to the purpose of the book”. What, as a guide to doing research? Looks pretty wrong with respect to that purpose.

RM: There is an unfortunate error here. The formula under the dashed line should be CV = kR, not CV = kS. The text has it right but the Figure is wrong. Thanks for pointing it out.

Best, Rick