Powers' Model of a PCT-Based Research Program

Hi Bruce

BN: The general path for emergence of higher levels (both species evolution and individual development ) is probably by repurposing neural structures of the same kind as are used for simpler means at lower levels. Frans has proposed that the obvious development of human cognitive capacities beyond emergence of the Systems level at about week 75 has its explanation in the proportionally larger size of the human cerebellum relative to cerebral cortex as compared to our evolutionary cousins. He proposes that cerebellar structures of the sort that control configurations are repurposed to control more ‘abstract’ perceptions such as concepts.

RM: I don’t believe Frans ever proposed this. The evidence for hierarchical control, much of it collected by Frans, suggests that there is no repurposing of lower level control systems to produce higher level ones. I see the evidence as supporting the idea that the higher level control systems that emerge during development use the existing lower level control systems to achieve their higher level goals. The existing neural structures involved in the control of configurations, for example, still control configurations when a higher level of control emerges (the configuration control systems are not “repurposed”); these now lower level control systems are used by the new, higher level control systems as the means of controlling the more abstract variables (relationships, sequences, programs, etc) that these higher level system control.

BN: You talk of strategies of war as programs. Clausewitz, Sunzi, Macchiavelli (yes, he also wrote about war) and others are prescriptive as well as descriptive. We use language and logic (which is a specialized use of language and dependent upon it) to review what actually has happened and tidy it up…

RM: I think you are questioning whether war actually involves control of what Powers calls program perceptions. And I agree that it may not. I was just using Bill’s names for possible perceptual types to make a point. But I am not married to the current names of the perceptual types. I think they are probably basically correct but Bill’s proposed research program would be aimed at determining whether there are perceptual types and, if so, what those types are and what might be the most appropriate labels for them.

RM: I think there is evidence for the existence of several of the types of perceptions Powers named. I describe some of this evidence in my “Hierarchical behavior of perception” chapter in More Mind Readings. The evidence is based on studies measuring the fastest rate at which a variable can be perceived and/or controlled. One example is that the fastest rate at which a sequence of tones can be perceived/controlled is the same as that for a sequence of visual shapes or a sequence of finger taps. This suggests that one type of perception that people control is a sequence, regardless of what the components of the sequence are, or their modality.

RM: Another way I thought might work as a way to see if controlled variables fall into categories or types may be similar to the method used to classify predation strategies in the paper that you cited:

  • Stephen D. J. Lang, Damien R. Farine (2017). A multidimensional framework for studying social predation strategies. Nature Ecology & Evolution, 22 August 2017.

RM: This looks like it might be a really interesting paper. It sounds like they are classifying predation strategies in terms of where each strategy sits on several different dimensions. I thought one way to classify controlled variables might be to use a form of multidimensional analysis called non-metric multidimensional scaling, which is based on expert ratings of the similarity of controlled variables to each other. If there are discrete types of controlled variables then these should show up as clusters of controlled variables in different parts of the multidimensional space into which the similarity ratings are fit.

BN: Here’s a public-facing overview (August 22, 2017):On the hunt: describing group predation across the animal kingdom. Researchers outline a new way to define and classify how groups of animals hunt together.(On the hunt: describing group predation across the animal kingdom | Max-Planck-Gesellschaft)

RM: One particular statement in that summary of the article caught my attention:

But, after separately scoring the different sub-populations, the result was unexpected: different populations of killer whales use different combinations of strategies. These differences largely depend on the type of prey that they hunt. Fish-eating killer whales extensively use communication when hunting, whilst those that eat seals and other marine mammals hunt silently, instead relying on specialised roles to force their prey into ambushes.

RM: The most interesting part is what I bolded. Since the prey is a disturbance to one of the main variables the whales control – something like “getting food into the mouth” – the actions they take to get this variable under control would be expected to differ depending on the type of prey; different type of prey disturb this variable in different ways which would require different actions to keep this variable under control. This is consistent with the fact that control systems control by producing output that compensates for disturbances to the controlled variable.

RM: So all killer whales probably control the same variable – amount eaten – using a cooperative predation strategy; they just have to use different ways of cooperating – different predation strategies – to compensate for the different types of disturbance produced by the different prey in order to control for eating it. My guess is that if a group of killer whales who prey on fish were moved to a region where the prey was seals and a group of killer whales who prey on seals were moved to the region where the prey was fish, both groups would quickly learn to predate the way the other group had been doing it.

RM: The Lang and Farine framework for classifying predation strategies suffers, I believe, from the fact that it is a classification of this behavior in terms of emitted output rather than controlled input. But I think there would be much to learn from it based on their observations of the whales and the attempts to classify their behavior. So thanks for the reference, Bruce.

Best

Rick

I agree with you (and with Frans) that the emergence of eleven levels of perception and control evidenced by studies of mothers and infants appears to be genetically programmed. I’m not sure why it is that difficult to see that my phrase

does not refer to the development of the perceptual hierarchy in infants up to the (proposed) Systems level, but that it refers instead to the immediately following phrase

I bet Frans would be willing to send you a PDF of his chapter or the section of it in which he discusses the relation of cerebellar to cortical functions. I haven’t sought his permission to post it here.

It appears that there may be regular ‘regression periods’ associated with the emergence of cognitive capacities after a child is about a year and a half old, and that these episodes are also marked by insecurity, liability to depression and illness, and so on. This doesn’t appear to be a hot topic of investigation. I don’t have anything very substantial i.e. peer reviewed to point to. Here’s some research that got popularized in the 1970s:

** Roger Gould, M.D.** [… i]n his book Transformations … presents his view that adult psychological development consists of the “dismantling of the illusions of safety developed in childhood” [and] suggests that these illusions are confronted in a time-sensitive sequence as one progresses through the life-cycle. His research was used extensively in Passages , by Gail Sheehy.[1]

Sheehy’s popular books are more widely known. Gould accused her of plagiarising his research (he published his book two years after hers), and she settled giving him 10% of royalties from Passages (1976). Her several related books published in the 1990s were apparently free of that.

Though research appears to be thin, I think few would deny the existence of developmental stages and significant transitions in life. These episodes, or some of them, are probably grounded in physiological developments. Evidence for this would support my speculation about the development of new perceptual input functions by reorganization and the recruitment of existing types of control loops to control these new perceptions, including the possibility of new types of perception; both learning and evolution are inherently open-ended. Such learning might be persisted by social demonstration and practice among social animals (more formally in human culture). The input functions and control loops might over sufficient time even come under genetic control. To my knowledge, this has not been studied from our perspective, but what I have seen suggests that some degree of regular timing persists throughout the human life cycle. Are these evidence of higher levels, or elaboration of new input functions creating inputs for the levels established by week 70, or development of specialized partitions within the existing hierarchy (as e.g. the specializations for language), or some combination with a nascent higher level emergent?–such questions are unanswerable in the present state of research, but that does not preclude asking them.

We need to bear this open-ended developmental context in mind as we look to identify and categorize types of perceptions and their means of control, which after all is the focus of this topic.

Hi Bruce

BN: I’m not sure why it is that difficult to see that my phrase

does not refer to the development of the perceptual hierarchy in infants up to the (proposed) Systems level, but that it refers instead to the immediately following phrase

RM: I didn’t have any problem with these statements. The only part of your very useful post with which I had a problem was when you said “[Frans] proposes that cerebellar structures of the sort that control configurations are repurposed to control more ‘abstract’ perceptions such as concepts”. [empahses mine]. Maybe I didn’t understand what you meant by “repurposed”. What I thought you were saying was that development of the ability to control more abstract perceptions – such as concepts – involves using the neural structures that produce lower level perceptions – such as configurations – for a new purpose; to produce the more abstract higher level perceptions. If that’s not what you meant by “repurposing” then, as Roseanne Roseannadanna would say, “never mind”:wink:

Best

Rick

It involves using the same kinds of structures, in the same part of the brain, but these structures are repurposed to receive in their perceptual input functions perceptual signals of a higher order than the signals that are received by such structures that are not ‘repurposed’ in this way. This is accomplished by reorganization, either by evolution and then developmental or by learning. The perceptions that they control are more ‘abstract’ because they are created from perceptual signals of a higher order.
which create more ‘abstract’ perceptions to be controlled.

For example, Relationship perceptions may be more abstract forms of Configuration perceptions. Tom Bourbon’s demonstration of two cursors and one target did not distinguish whether the subject was controlling relationship perceptions or a configuration perception.

That we can see the alignments of cursors and target as relationships or as configurations suggests that relationship control is a more ‘abstract’ form of configuration control. Henry’s figure showing the changed rate of firing in a nerve bundle as a monkey flexed its arm was taken as an illustration of configuration control, i.e. the configuration of the arm around the elbow. Might it as reasonably be taken as an illustration of the monkey controlling a relationship of proximity of the lower arm and hand to the upper arm? Or if it is understood as a relationship of distance between hand and shoulder along a vector that is simultaneously determined by shoulder angle, isn’t that more appropriate than configuration for describing reaching movements?

The H of HPCT may not be tidily stratified across the board. Nature does whatever works.

As hominids developed into early humans their brains got larger, specifically, the cortex and cerebellum. The cerebellum is 10% of the volume of the brain but holds 50% of the neurons in the brain, and in its purview are configurations, transitions, and relationships, not only physical-motor but also more abstract. I must relocate this topic to the Neuroscience category so as not to digress too much from the topic at hand (and I will post references there), but the relevance here is that because of anatomical parallels and co-location there are likely to be important analogies between the ‘concrete’ perceptions that are easier to subject to test/experiment and the more ‘abstract’ perceptions that are often more difficult for an experimenter to control. This may be very useful for experimental design, and if true is surely relevant to the categorization exercise proposed in this topic.

Hi Bruce

RM: Maybe I didn’t understand what you meant by “repurposed”. What I thought you were saying was that development of the ability to control more abstract perceptions – such as concepts – involves using the neural structures that produce lower level perceptions – such as configurations – for a new purpose; to produce the more abstract higher level perceptions.

BN: It involves using the same kinds of structures, in the same part of the brain, but these structures are repurposed to receive in their perceptual input functions perceptual signals of a higher order than the signals that are received by such structures that are not ‘repurposed’ in this way.

RM: I think I would understand this better if you would describe the data that led to the formulation of this theory. This is, after all, meant to be a thread dedicated to discussion of how to do research based on PCT. So what are the observations that led to this repurposing theory and how does the theory account for them?

BN: For example, Relationship perceptions may be more abstract forms of Configuration perceptions. Tom Bourbon’s demonstration of two cursors and one target did not distinguish whether the subject was controlling relationship perceptions or a configuration perception.

RM: Of course Tom’s research didn’t distinguish whether the subject was controlling relationship or configuration perceptions because “relationship” and “configuration” are just names Bill came up with to describe what he thought were two different types of perceptual variables. Tom’s experiment showed that you can get coordinated behavior (the dance to which you are invited) as a side effect of two independent control systems controlling two different variables (the distance of two different cursors from a target) when the outputs of each system is a disturbance to the variable the other system is controlling.

BN: That we can see the alignments of cursors and target as relationships or as configurations suggests that relationship control is a more ‘abstract’ form of configuration control.

RM: It shows that we can see the same thing in two different ways but it doesn’t say anything about the relationship between those two ways of perceiving the situation.

BN: Henry’s figure showing the changed rate of firing in a nerve bundle as a monkey flexed its arm was taken as an illustration of configuration control, i.e. the configuration of the arm around the elbow. Might it as reasonably be taken as an illustration of the monkey controlling a relationship of proximity of the lower arm and hand to the upper arm? Or if it is understood as a relationship of distance between hand and shoulder along a vector that is simultaneously determined by shoulder angle, isn’t that more appropriate than configuration for describing reaching movements?

RM: Right. Henry chose to call the angle of the arm at the elbow a “relationship”. But I think the main import of Henry’s work was that he found a hierarchical relationship between the firing of the nerve bundle that corresponds to elbow angle and the firing of nerves bundles descending from higher levels of the nervous system.

RM: None of the PCT research that has been done to date can tell you the type of variable an organism is controlling – whether it is what Bill called a configuration, relationship, sequence, program, etc – nor can it whether there even are types of controlled variables.

BN: The H of HPCT may not be tidily stratified across the board. Nature does whatever works.

RM: You bet! We really don’t even know if the organization of control systems is hierarchical, heterarchical, some kind of network or something else. One aim of this thread is to see if we can think of research that would help figure this out.

BN: I must relocate this topic to the Neuroscience category so as not to digress too much from the topic at hand (and I will post references there),

RM: No need to do that. Neuroscience research is more than welcome in this thread. The only thing that is not welcome is theoretical speculation based on intuition rather than data.

BN: but the relevance here is that because of anatomical parallels and co-location there are likely to be important analogies between the ‘concrete’ perceptions that are easier to subject to test/experiment and the more ‘abstract’ perceptions that are often more difficult for an experimenter to control. This may be very useful for experimental design, and if true is surely relevant to the categorization exercise proposed in this topic.

RM: You bet. That’s why this neurophysiological discussion should stay here (or be posted to both this thread and a thread in the Neuroscience category). If you could give a nice, clear description of some of the results from Henry’s lab that would be great!

Best

Rick

Included thus by reference to the appropriate topic.

I have asked Henry for comment but I have no indication that “the results from Henry’s lab” include specific investigations of the anterior and posterior cerebellum. I will shortly be asking Frans if he would kindly review my references to our conversation and his chapter in the Handbook. Maybe not right away, it’s a busy day here.

My recollection is that he and Bill called it the configuration of the arm. When I say that we can see it either as configuration or as relationship I mean we as armchair analysts. What level is actually controlled when I take a sip of tea is surely not equivocal, so this purely verbal ambivalence (configuration? relationship?) is a failure to refer to empirical data. I believe Henry’s data suggests it is the lower of those two levels. But which of them is lower, is it always the lower or sometimes the upper of the two levels in hierarchical control, and how do we know? Phenomenologically, I’d say we control the configuration as means of controlling the relationship, and I believe this is what Bill and Henry said. This is confirmed by the illustration of stimulating the monkey’s nerve bundle, because no relationship is involved (there is no target that the monkey is reaching for, and the probe provides the reference signal that would come from a system controlling the relationship of hand to a target).

Henry replied

I think it’s reasonable. Not too controversial either.

Hi Bruce

RM:…If you could give a nice, clear description of some of the results from Henry’s lab that would be great!

BN: Included thus by reference to the appropriate topic.

RM: I’ve read some of this stuff. My interpretation of it is quite different than theirs. But I didn’t ask to see work on the cerebellum. I wanted to see a nice, clear description of Henry’s work, which is the kind of neurophysiology that is directly relevant to testing the PCT model. How about giving it another try.

BN: I have asked Henry for comment but I have no indication that “the results from Henry’s lab” include specific investigations of the anterior and posterior cerebellum.

RM: Still, from what I understand of Henry’s work it is far more pertinent to PCT research than the stuff on the cerebellum that you posted.

BN: My recollection is that he and Bill called it the configuration of the arm. When I say that we can see it either as configuration or as relationship I mean we as armchair analysts. What level is actually controlled when I take a sip of tea is surely not equivocal, so this purely verbal ambivalence (configuration? relationship?) is a failure to refer to empirical data.

RM: I currently know of no empirical data that will tell you what type of perception is being controlled or even whether their are types of perception. Powers’ model for a PCT based research program – the program I am trying to develop – is all about figuring out how to find out whether controlled perceptions fall into different types and, if so, what those types are and whether they are hierarchically related.

BN: I believe Henry’s data suggests it is the lower of those two levels.

RM: That is my understanding as well. The research in Henry’s lab shows that there is a hierarchical relationship between structures in the nervous system associated with the relationship between the variables involved in control. They named the variables controlled at different levels of the nervous system using the names Powers used to identify what he thought were the different types of perceptions at different levels. These names are nice descriptions of the variables involved in Henry’s studies but we still don’t know whether these names describe actual different types of controlled variables, as predicted by the hierarchical PCT model.

RM: The work in Henry’s lab (as I understand it) shows that the apparent hierarchical relationship between the behavioral variables involved in control is reflected in the hierarchical relationship between the structures in the NS that are involved in control. So it would be great if you could provide a nice, clear description of Henry’s work because I think it is very relevant to Powers’ vision of a PCT-based research program. The neurophysiological research you posted about the cerebellum provides much to coarse a picture of the operation of this component of the nervous system to be of much use use in terms for evaluating models of hierarchical control.

Best

Rick

Hi Bruce

BN: Henry replied

HY: I think it’s reasonable. Not too controversial either.

RM: Henry is commenting on the cerebellar stuff you posted. I agree that it is non-controversial. And I suppose you could call it reasonable. I just find that the interpretations of this research seem to go way beyond the data – data which, as I said, is way to coarse to inform any conclusions about how control works.

RM: The research on the cerebellum seems to be at the level of quality as the ablation studies of Bizzi et al. who cut the spinal (and possibly also higher level) afferents in monkeys and found that they could still walk and climb (although not nearly as skillfully as before the operation and only after a long period of learning). So they concluded that sensory input was unnecessary for skilled behavior – walking and climbing could be produced by central neural effect generators. In other words, they concluded based on this neurological evidence that behavior is the control of output. I think the conclusions about what the cerebellum does, based on people who are born sans cerebellums (fortunately they weren’t ablating these structures) are of the same quality.

RM: But, again, what I want to hear about in this thread is the research in Henry’s lab, which is very high quality work on the neurophysiological basis of control. A nice, clear, simple explanation for non-neurophysiologists would be nice.

Best

Rick

My link to the neurophysiological discussion was my response to the first sentence (which you left out).

As to the second sentence (which I should have separated, to avoid inviting your misinterpretation), you will have to ask Henry for descriptions of his research. All I have is his statement that what I was proposing was reasonable and not controversial. Maybe my reasonable and uncontroversial speculation about a PCT understanding of CogSci research into the role of the cerebellum aligns with some of his current research interests, or maybe he will turn to the cerebellum in the future. The fact that he did not offer any references to his own research suggests that he has not yet worked in this area. I hope he will.

Meantime, all we have from neuroscientists is reports of research by people who designed their investigations and interpreted their results on the basis of the perverse assumptions of conventional Cognitive™ psychology.

I asked Henry about my interpretations of this research, and he replied that my interpretations are reasonable and even uncontroversial. I’m sure he would also say that the research itself is reasonable. You can equate the observations of people born without a cerebellum to those early 1960s and 1970s ablation studies, if you ignore the majority of the research which I cited, which the researchers making those clinical observations explicitly bring in as context for their observations.

That’s why I didn’t post it in this topic. Fortunately, the category/topic organization of Discourse enables us to limit such digressions.

Backtracking a bit:

This is precisely why my discussion derived from Frans’s comments and from some of the published literature does not belong in this topic.

Discussion of the relationship between data, intuition, and theoretical speculation belongs in the Science:Methodology subcategory. Suffice to say here that this reciprocal relationship has always been a necessary and essential part of doing science. In particular, Bill worked back and forth between neurological and behavioral data, phenomenological intuition, and theoretical speculation to arrive at his proposals about the orders of perception in the perceptual hierarchy, his “reasonable guesses about our actual construction” (Powers 1979:190). The same methodology is appropriate for PCT now. Just not in this present topic which has its own methodological strictures.

Hi Bruce

BN: I asked Henry about my interpretations of this [cerebellar] research, and he replied that my interpretations are reasonable and even uncontroversial.

RM: I have no idea what Henry might have meant by that so I’ll just tell you why I am not enthusiastic about your interpretations of that research.

RM: The main general finding of the cerebellar research is that the anterior cerebellum (AC) seems to deal with motor functions while the posterior cerebellum (PC) seems to deal with cognitive functions. Your PCT interpretation of these results was that that the AC deals with motor control – particularly control of relationship perceptions – while the PC is involved with cognitive control – which you took to mean performing mental tasks such as trial and error testing and planning: examples of control in imagination.

RM: I think the main problem with your interpretation is your idea that there is a part of the brain – the PC – that is dedicated to controlling in imagination. In PCT control in imagination can happen at all levels; composers can imagine (and plan) how things will sound (sensations and others), painters can imagine (and plan) how things will look (configuration perceptions and others), lovers can imagine (and plan) what it will be like to be with their loved one (relationship perceptions, and others), etc. So this kind of cognition – imagining a future perception, as in making the plans that gang oft aglay – can apparently happen for all kinds of perceptions and, to the extent that these different kinds of perceptions exist at different levels of the nervous system, it’s unlikely that the the ability to plan is located in only one structure in the brain.

RM: So why does it look like some cognitive functions are located in the PC? I believe it’s because the PC is involved in the control of perceptions that are often considered cognitions: the perceptual variables described in the Higher Levels chapter of B:CP. These “cognitive” perceptions are perceptions that seem to exist in our head rather than in the world outside. Powers starts these at relationships (is “on-ness” out there of in here) and goes up from these to programs, principles and system concepts. I think the PCT view of the difference in the roles of the AC and PC in behavior would be that it is likely a difference in the types of perceptions that are controlled by these brain structures.

RM: In the cerebellar research they measure cognitive loss due to stroke in the PC using various tests. I think if you took a careful look at these tests you might be able to come up with hypotheses about the perceptual variable(s) that a person is unable to control with a damaged PC. I think that would be a nice contribution to the PCT-based research program that Bill envisioned.

RM: As to the fact that the PC grows quickly once a child is able to control perceptions at all levels, there are various different possible interpretations of this fact from a PCT perspective but the least likely interpretation is that this happens because the PC is involved in making control at all levels more effective by allowing trial and error testing and planning in imagination. A more plausible explanation, it seems to me, is based on the assumption that the PC is involved in controlling higher level perceptions than those controlled by the AC. The idea is that once the entire hierarchy is formed more of the type of perceptions controlled by the PC are needed by the higher level (cortical) systems controlling perceptions of programs, principles and system concepts.

RM: I think the best way to proceed, though, would be to make a list of the tasks that can’t be done by a PC stroked person (tasks that are considered a measure of cognitive ability) that can be done by a non-PC stroked person. That should give us some hypotheses about the variables controlled at the PC level. And also make a list of the tasks that can’t be done by an AC stroked person that can’t be done by a non AC stroked person. That would give us some hypotheses about the variables that are controlled at the AC level. You seem to be quite an expert on the cerebellar research, both AC and PC, so it should be easier for you to get this information than it would be for me.

Best

Rick

I do not say that all imagination occurs there. I said that the abstract imagery subjectively experienced during mental manipulation of concepts is analogous to the control of relationships and transitions of configurations by motor control because the same kinds of cerebellar structures are involved. There are certainly many forms of imagination other than the mental manipulation of concepts. Among them, of course, is imagined control by functions in the A[nterior] C[erebellum] of configurations perceived to be present as objects in the environment or remembered or imagined so.

A second problem with your oversimplification to a straw man is that control of transitions (including events) and relationships of configurations in imagination can extend down the hierarchy to imagined control of sensations and intensities in vivid imagination.

Yes. That is what I’m talking about. And those perceptions exist “in the head” rather than in the world because they are grounded in imagined control at lower levels rather than being grounded in input from the environment.

Yes indeed. That’s what commands the trial-and-error testing by sending reference signals for relationships and transitions among configurations, and that’s what receives perceptual input resulting from those imagined control processes below them in the hierarchy.

Did you intend that to be sarcastic? I’m asking because it seems so to me, and I don’t want to make assumptions.

Piqued by what Frans has told me, and its relevance to my field of linguistics, I put a few hours into an unprofessional search of the literature about the cerebellum. My field is linguistics. I don’t have the resources to find out what tasks can be performed by people with different kinds of cerebellar strokes. Nor if I did do I have the time free from grant and other obligations, which is a reason for the long delay responding. But it’s a good suggestion.

(Relocated from Delay in control loops.)

RM: It’s hard for me to tell that that’s the case. Figure 4 says it shows the behavior of a participant tracking two different target types and Fig. 6 says it shows “model simulation accuracy” but it’s not clear whether that is accuracy of the model tracking the target or the accuracy of the model accounting for participant behavior. But it doesn’t really matter. This just isn’t the kind of research I would like to see done to test the PCT model. The kind of research I would like to see done is described in Powers (1979).

RM: I’m not that interested in research on how organisms overcome transport lags in control loops because we know they do. We know that because we see them controlling all kinds of different variables. What we don’t know is what all the different variables are, whether they are of a fixed number of types, whether they are controlled at different levels of a hierarchy of control systems, etc. These are the kind of question that distinguish PCT from other applications of control theory to understanding behavior.

RM: Once we know what variables are controlled and how they are related to each other, our models, when successful, will implicitly show us how transport lags (which will be different for different variables with the duration of some of these lags – as well as the duration of the variables themselves – being quite substantial; think about the control loop that controls for a particular program perception, for example, as in my demo of program control).

RM: Apparently there aren’t many (any?) people interested in doing such research but, if anyone is, I again invite them to discuss it in the
PCT-research Discourse topic that I created for that purpose.

Best

Rick