Powers' Model of a PCT-Based Research Program

Phenomena
1

In 1979 Powers described his view of what a research program based on PCT – based, that is, on an understanding of organisms as living control systems (Powers, W. T. (1979) A cybernetic model for research in human development. In M. Ozer (Ed) A cybernetic approach to the assessment of children: Toward a more humane use of human beings , New York: Routledge, pp 11-66.). This paper is reprinted in Powers’ 1989 collection of papers entitled Living Control Systems: Selected Papers of William T. Powers, CSG publishing. Powers’ vision of a research program based on PCT is described on pp. 215- 218.

In that paper Powers gives a rather high level description of a PCT research program. Of course, it involves testing for controlled variables. But it also involves collecting many examples of controlled variables and then classifying them by type. Powers doesn’t say much about how to go about carrying out this “classification” phase of the research program. But he does say this:

This phase will call for new kinds of data gathering, aimed at sharpening methods for distinguishing levels of organization, or for otherwise mapping relationships among control activities. There will be plenty of room for ingenuity.

This classification phase of PCT research can’t even begin until there is a “…sufficient amount of information available in terms of variables proven to be controllable by human beings” (Powers wrote these words in this paper which was devoted to ideas about how to study human development). I have done some research that involves testing for controlled variables. So I have been involved in collecting information about the variables proven to be controllable by human beings. But I don’t think we yet have anything close to a sufficient amount of information about these. Nevertheless, I would like to start thinking about how to go about classifying controlled variables by type once we do have a sufficient amount of information about them.

So I’m posting this to see if I can get those of you who are interested in doing PCT-based research to think about this with me. To start this discussion I suggest you read Powers 1979 paper in Living Control Systems, with an emphasis on the last part of that paper describing the research program, and then post any ideas you might have about 1) what kind of information we need to collect about the variables that are controllable by humans 2) what would constitute a sufficient amount of information about those variables and/or 3) how to go about classifying these variables once we have a database containing a sufficient number of them.

No rush. It’s already been 40 years since Powers described his vision of a PCT research program. But I think we’ve had enough time to put off thinking about it and conventional psychologists have had enough time to ignore it. I think it’s time to just go ahead and try to build a PCT research program. And, perhaps, if we build it, they will come.

1 Like
2

Since not everyone has access to LCS or the Ozer book, I have taken the liberty of scanning that important 1979 paper and making the PDF image that you can download with this link

I trust the gods of copyright and the gods of domestic tranquility alike will forgive me.

1 Like
3

Hi everyone. I have two questions. First, should our method be inductive or deductive? I’m saying this because Bill did a lot of introspective work already to identify the levels of CVs that might be controlled. Should we be testing these out as hypotheses first?
Second, and relatedly, should we be trying to classify the CVs as they might be completely open-ended, dependent on the nature of the environment that the organisms is situated within, learned through reorganisation to be optimal at control. For example, I am sure most of us can control for producing legible text by typing with our thumbs on a smartphone, but would we ever have developed the ability to control the variables involved in that activity if the smartphone hadn’t been invented? Should we be testing the fundamental principles of PCT only and having an open mind that we can build individual models by inferring CVs for specific tasks in specific individuals under certain constraints, but that these models are multiple, heterogeneous and changing through reorganisation as the demands change?
Or maybe I could have misunderstood the question?
All the best
Warren

4

Hi Warren

WM: Hi everyone. I have two questions. First, should our method be inductive or deductive? I’m saying this because Bill did a lot of introspective work already to identify the levels of CVs that might be controlled. Should we be testing these out as hypotheses first?

RM: To the extent that I understand what inductive and deductive research is, I guess I would say it should be both. I think the main thrust of Bill’s pitch in the Cybernetic Model paper is that the research program be inductive. He suggests starting by creating a database of controlled variables. This database would contain entries something like those in Table 1, as shown here:

image
image830×855 10.5 KB

RM: The column labeled “Variable” are verbal descriptions of the controlled variables. I think BIll’s idea was once you have a large database of these variables you would then start classifying them by type and testing to see if there are hierarchical relationships between the different types. Unfortunately he didn’t say how to go about the process of classification, but that process, whatever it is, would be inductive, as I understand that term: going from specific instances (the controlled variables) to a general conclusion (many specific instances being of the same general type).

RM: I think the process of identifying controlled variables – using some version of the TCV – could be considered deductive: going from general principles (the principles of control) to a specific conclusion (that a specific variable is being controlled).

WM: Second, and relatedly, should we be trying to classify the CVs as they might be completely open-ended, dependent on the nature of the environment that the organisms is situated within, learned through reorganisation to be optimal at control.

RM: I think the answer is yes. CVs are just CVs. In the “Cybernetic Model” paper Bill’s idea, I believe, is to somehow classify the CVs themselves by type, regardless of the nature of the environment in which they are controlled or how the ability to control them developed.

WM: For example, I am sure most of us can control for producing legible text by typing with our thumbs on a smartphone, but would we ever have developed the ability to control the variables involved in that activity if the smartphone hadn’t been invented?

RM: I think we’ll find out if this is the case by putting as many CVs as we can come up with into the database. If we find CVs that couldn’t be “typed” because they are specific to things like typing with the thumbs on a smartphone then that would be interesting (but, I think, highly unlikely) and would certainly require a major revision of the HPCT model.

WM: Should we be testing the fundamental principles of PCT only and having an open mind that we can build individual models by inferring CVs for specific tasks in specific individuals under certain constraints, but that these models are multiple, heterogeneous and changing through reorganisation as the demands change? Or maybe I could have misunderstood the question?

RM: As I understand Bill’s proposal, I would say the research is aimed at testing fundamental assumptions of the model (like the assumption of a hierarchy of different types of controlled variables) by doing research (and developing models) on all kinds of different behavior. Actually, part of Bill’s proposal about how to identify types of controlled variables is to look at when in development the ability to control these variables comes “on-line”, which is one way reorganization might come into this research program since new control skills presumably develop after an identifiable period of reorganization.

Best

Rick

5

7 posts were merged into an existing topic: Observed reference value, inferred reference signal

6

Hi Warren

RM: Since you are the only one so far who has commented (substantively) on this thread I would appreciate if you could give me your comments on Bill’s proposal for a program of PCT-based research that he described in his Cybernetic model for research in human development paper (posted here by Bruce Nevin) when you get a chance. This might get the ball rolling. If you do, I promise to give you a more complete answer to your comment here:

WM: For example, I am sure most of us can control for producing legible text by typing with our thumbs on a smartphone, but would we ever have developed the ability to control the variables involved in that activity if the smartphone hadn’t been invented? Should we be testing the fundamental principles of PCT only and having an open mind that we can build individual models by inferring CVs for specific tasks in specific individuals under certain constraints, but that these models are multiple, heterogeneous and changing through reorganisation as the demands change?

RM: This is related to something I’ve always wondered about which is why humans developed the ability to do all the complicated stuff they do – like building a Large Hadron Collider and Jumbo jets, etc – when their brains stopped evolving at the point when it provided them the ability to do things like build fires, live in groups and talk to each other. I think the PCT hierarchy provides the answer.

7

I think I remember discussion of creating a cloud spreadsheet (Google doc, probably) to accumulate these examples. I couldn’t find the CSGnet thread.

In the paper I presented in Manchester I identified quite a few controlled variables, hierarchically ordered. I want some more clarity about how we’re organizing this before I spell them out this way. There’s no clear way to show those hierarchical relationships except by repetition of the same word label first in the Means column (for the higher level) and then in the Behavior column (for the next level down) with its own means, variable, and reference state. We could try to manage it with linked content or with pivot tables. Maybe you have some ideas, since you’ve done some pretty sophisticated things with spreadsheets.

8

RM: Yes, good memory. I think the spreadsheet is available as a shared link at:

RM: I think we should start by getting a better idea of what information about the controlled variables we should put into the data sheet (which is what the “Behavior is Control” spreadsheet is). I think the basic idea of Bill’s research program is 1) to develop a database of controlled variables and 2) based on the data in this database figure out a way to see if they group into types, whether these types correspond to those Bill proposed and determine what the relationship between these variables is; and do this without preconceptions about the types of variables in the database or the nature of the relationship between them (hierarchical, heterarchical).

RM: I’m looking for ideas about how to do both 1) and 2). I honestly have only a vague idea of how to do either one, but I am particularly perplexed about how to do number 2.

9

Thanks for the link.

Perceptions are categorized first of all by their hierarchical relations. Whether these fall always and cleanly into orders or levels of perception should be an open question to be investigated without preconception–“without fear or favor” might be an appropriate way to put it when it comes to communicating findings to the faithful. It is important but seldom mentioned that the actual hierarchical relations that have been found to be a practical necessity in simulations like the arm and the little man do not always align with the received notion of levels. The ‘Levels’ column imposes a pre-judgement and should be removed. The date and contributor ID should go to the right of Comments or those two columns should be deleted and the content relegated to a comment attached to the first cell of the row.

The procedure should be to work out what is involved in the given behavior in a separate spreadsheet and then copy from there to the shared document. To segregate the categories that emerge by their hierarchical relations, link across multiple sheets in the spreadsheet file. The data in the Means column should always point to data in the Controlled Variable(s) column of another sheet. For example,

Behavior
Controlled Variable(s)
Reference State
Means
Disturbances
Type
Context
SippingTea
Position of Cup
At lips
Lift, tip cup
Amount of tea in cup
Relationship

‘Lift, tip cup’ is complex. What CVs, references, means, and disturbances are involved in implementing these means? ‘Sipping’ indicates a reference for ingesting a small amount at a time (different for quenching desperate thirst, clearing a bad taste, getting a caffeine fix before a meeting you’re late to, taking a medicinal brew, etc.).

10

Hi Rick, it’s great reading this article again. I particularly like how he explains the program level. I’m not sure what to say regarding comments except that it’s what many of us are doing already but in a piecemeal fashion. You especially, but to varying degrees, Frans Plooij & Henry Yin too. More needs to be done!

Really it’s about covering the entire territory of psychology - but doing it the TCV analysis way, systematically- specifying not just the CV at the top of the hierarchy, but all the way down, for the various skills & activities that we see humans do.

In itself it’s a venture to be done within a PCT community. But to the extent that it counters existing interpretations, or provides new, impactful, applications, it would have something to say for the wider community.

Besides my other questions - around individuality - would be where to start? Do we start from age zero? Do we start from where we have most expertise? Or from the simplest activity to model? Or the one that if we understand it, would have the most impact?

Human (bidepal) walking seems to be a good place to start? Rupert would be interested. However, we’d need some decent technology to create the various disturbances & monitor the effects. I could try to bring in my contacts in engineering & physiotherapy. How about through an interdisciplinary grant on ‘balance’?

Talk to you soon!
Warren

11

Hi Bruce

BN: Thanks for the link.

Perceptions are categorized first of all by their hierarchical relations.

RM: Right now they are (probably incorrectly) categorized by type: intensity, sensation, etc. The categorization is probably incorrect because it was just done “off the top of my head”. I think what we need are two things to improve the spreadsheet database: 1) better – more precise – definitions of the controlled variables and 2) an objective basis for grouping the variables into types “without favor or fear” as you say. The “better” definition of a controlled variable is one that makes it possible to bring this variable “into the lab” for testing. The testing I would involve measuring the time to react to disturbances of that variable. Then for each controlled variable we would have a measure of reaction time that could be used as a basis for grouping controlled variables into types.

BN: The ‘Levels’ column imposes a pre-judgement and should be removed.

RM: I think you must mean the “Types” column. I would like to keep it just to see how the groupings that emerge from classification in terms of reaction time relate to our intuitive classifications based on out understanding of Powers’ proposed classification.

BN: The date and contributor ID should go to the right of Comments or those two columns should be deleted and the content relegated to a comment attached to the first cell of the row.

RM: I am planning to “nicen up” the spreadsheet and I’ll take those recommendations into consideration.

BN: The procedure should be to work out what is involved in the given behavior in a separate spreadsheet and then copy from there to the shared document. To segregate the categories that emerge by their hierarchical relations, link across multiple sheets in the spreadsheet file. The data in the Means column should always point to data in the Controlled Variable(s) column of another sheet.

RM: This is a great idea but, I think, one that should come after we have done the typing. If that works, then looking for hierarchical relationships (or, as you note, the absence thereof) should be based on some other kind of objective determination. But putting the results of that into something like a relational database is a super idea.

RM: These comments have been very helpful to me, Bruce. Thanks. I’ll work on creating a new and, hopefully, improved version of the controlled variable database and, when it’s done, look forward to your comments, criticisms and suggestions about how to continue with this process.

Best

Rick

12

Hi Warren

WM: Hi Rick, it’s great reading this article again. I particularly like how he explains the program level. I’m not sure what to say regarding comments except that it’s what many of us are doing already but in a piecemeal fashion. You especially, but to varying degrees, Frans Plooij & Henry Yin too. More needs to be done!

RM: Yes, indeed. I think a good project for students who might be interested in starting a PCT-based research program would be populating the on-line “Behavior is Control” database:

Behavior Is Control - Google Sheets)

with examples of controlled variables from the existing scientific literature. I think you can find some pretty good examples of controlled variables in the conventional behavioral science literature if you know what to look for. But my work, that of the Plooijs and that done in Henry’s lab would provide a good source of examples of controlled variables because those variables have passed either formal or informal examples of the TCV.

WM: Really it’s about covering the entire territory of psychology - but doing it the TCV analysis way, systematically- specifying not just the CV at the top of the hierarchy, but all the way down, for the various skills & activities that we see humans do.

RM: Yes, indeed. I’d say that these examples of controlled variables should be culled from the entire territory of the behavior of organisms but probably with emphasis on the behavior of humans. It would be nice to specify all the CV’s involved in any particular behavior but I don’t think that’s necessary (or even feasible) right now. If you can get that kind of data it would be great but I think what we should focus on now is just identifying all kinds of controlled variables – involved in all kinds of behavior – and describe them in a way that we makes them testable in a laboratory situation.

WM: Besides my other questions - around individuality - would be where to start? Do we start from age zero? Do we start from where we have most expertise? Or from the simplest activity to model? Or the one that if we understand it, would have the most impact?

RM: I would suggest starting to look for controlled variables in the behavioral science literature – without prejudice regarding the “importance” of the variables. Some of the best examples of controlled variables would be the ones that are known to be controlled because they were the object of research conducted by people who were wearing “control theory glasses” and, therefore, knew how to see control when it was happening. But there are also good examples of controlled variables in research done by people who were not looking at behavior through control theory glasses.

RM: What your post makes me realize is that there are a lot of possible archival sources that can be culled for examples of controlled variables. These examples of controlled variables – or possible controlled variables – can be used to populate the “Behavior is Control” database of controlled controlled variables. This would make a great project for a upper level undergrad or graduate student with an interest in PCT.

WM: Human (bidepal) walking seems to be a good place to start? Rupert would be interested. However, we’d need some decent technology to create the various disturbances & monitor the effects. I could try to bring in my contacts in engineering & physiotherapy. How about through an interdisciplinary grant on ‘balance’?

RM: I think it’s best not to limit the gathering of examples of controlled variables to any specific behavior. As Bill said in the 1979 paper “It may well be best to take examples at random” which is pretty much the way we started populating the “Behavior is Control” database. And this data gathering should definitely be done “without regard to any proposed hierarchy or any other preconceived notions”.

RM: I think the first step in this research program should be to populate the “Behavior is Control” database with a large array of controlled variables, defined in a way that makes them suitable for study in a lab situation (if they haven’t already been studied in a lab situation) and collect some data on these controlled variables that will allow them to be classified by type. If you can get a grant to support such a project I think that would be a huge plus for PCT science.

WM: Talk to you soon!

RM: That would be great. As I said to Bruce, I’ll update the current “Behavior is Control” spreadsheet and once that’s done I’ll look forward to hearing what you think about it.

Best

Rick

13

Hi Rick, it all makes sense. The spreadsheet is very informal and inferential though. We probably need to start a new one that is organised in a way that documents the robustness of evidence. A student with the calibre and motivation to do this doesn’t come along too often but I will definitely be on the look out. Max and Mak would be ideal but are overcommitted at the moment. Maybe someone reading this thread will volunteer!
Hope the book is going well,
Talk to you soon
Warren

14

Hi Warren

WM: Hi Rick, it all makes sense. The spreadsheet is very informal and inferential though. We probably need to start a new one that is organised in a way that documents the robustness of evidence.

RM: I agree! I’ll give it a try and see what you think.

WM: A student with the calibre and motivation to do this doesn’t come along too often but I will definitely be on the look out. Max and Mak would be ideal but are overcommitted at the moment. Maybe someone reading this thread will volunteer!

RM: That’s what I’m hoping for. Though I am used to working solo. I would love to develop a paper based on this. Even the little conversation we have had so far has helped me out.

WM: Hope the book is going well,

RM: It’s in production, I think. But it probably won’t be out until next year. But it will be out!

15

That’s great then!

16

Hi Warren

RM: I just remembered that I had promised to answer an interesting question of yours if you gave me some ideas about how to proceed with Powers’ PCT-based research program. And you did give me some ideas so here’s my answer to this question of yours:

WM:… should we be trying to classify the CVs as they might be completely open-ended, dependent on the nature of the environment that the organisms is situated within, learned through reorganisation to be optimal at control. For example, I am sure most of us can control for producing legible text by typing with our thumbs on a smartphone, but would we ever have developed the ability to control the variables involved in that activity if the smartphone hadn’t been invented?

RM: This, as I said, is related to a question I always had about the evolution of human capabilities. My question is sort of the inverse of yours. It is: How are people able to produce something as complex as a smartphone using brains that evolved in apes with brains that allowed them to produce things that were far less complex.

RM: My answer to both questions is the same: I propose that it is because homo sapiens evolved the ability to perceive the world in ways that their ancestors could not. Specifically, they evolved the ability to perceive the world in terms of new types of perceptual variables. Specifically, in PCT terms, their brains became capable of perceiving the world in terms of programs, principles and system concepts.

RM: The ability to perceive, and thus control, these new types of perceptual variables probably gave homo sapiens an adaptive advantage over earlier hominid species, particularly in the ability to wage war, I’m afraid. This is probably why there aren’t any earlier versions of the genus homo or, for that matter, any subspecies of homo sapiens (such as homo neanderthalis) around any more. The ability to formulate and carry out battle plans (programs) according to principles (such as those in Sun Tsu’s The Art of War) in order to preserve a system concept (like “my people”) would give homo sapiens a huge advantage in conflict over other hominids (and, incidentally, may be the reason why its so hard for current homo sapiens to put an end to war).

RM: The reason this could be an answer to your and my questions is because once your brain is capable of experiencing the world in terms of a particular type of perceptual variable then you are capable of controlling the world in terms of all instances of that type of variable. For example, once humans were able to control the world in terms of programs, they could control any kind of program. While the first humans were probably controlling programs no more complex than hunting and foraging plans, their brains now had the ability to control programs as complex as building a smartphone. And to the extent that the operation of a smartphone requires controlling a program of thumb presses, the fact that humans have the ability to control programs means that they can learn to control that aspect of smart phone use.

RM: I believe that this answer to our questions is at least generally consistent with the PCT model of behavior. It is based on the assumption that the human brain has evolved to be capable of perceiving different types of perceptual variables. This doesn’t means that we don’t have to learn to perceive specific examples of controlled variables (such as the thumb pressing used to produce a message on a smart phone). But the specific examples of controlled variables are of particular types of perceptual variable that the brain is capable of constructing. According to PCT, a brain that is not capable of constructing, say, program perceptions, is not capable of learning to control specific examples of those that type of perception.

RM: I think this answer is testable by ethologists or comparative psychologists who could use the TCV to see what types of variables different species are capable of controlling. If PCT is right I think you would expect to find that as you ascend the phylogenetic tree you would find species able to control higher and higher level types of perceptual variables. Might be a fun area for research by a biologically savvy control theorist.

Best

Rick

17

That’s great.
What about also the idea that humans might be able to perceive (and therefore control) elements of their environment at the lower levels of the hierarchy through the reorganisation of input functions that other animals cannot - because their input functions are hardwired?
I’m thinking for example of writing - we can identify and reproduce letters - which are simple configurations, but no other animals can do this. They can produce elegant shapes - spiders webs, weaver bird nests - but these are formed as an emergent consequence of controlling for very simple, hardwired aspects of theIr perceived environment. I think there is evidence that the great apes can execute programs, for example when stripping a branch of leaves. I’m not sure though if this is learned or not - it’s certainly not learned by symbolic transfer of information.
Talk to you soon!
Warren

18

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.

This is surely not new, or limited to human learning. I don’t know where Relationship control structures are located in the brain, but I suspect that they also are structures of the Configuration ilk used for more ‘abstract’ purposes. In humans, a subhierarchy of perceptions up to the Relationship and Sequence levels are specialized for language. (I stop there because communication strategies for persuasion, elicitation of cooperation, etc. are not limited to or unique to language. Their means of control in language are evidenced in systematic word repetition across clauses, which is what makes discourse coherent.)

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. Trial and error eventually reaches fit conclusions where prescription, being an abstraction from experience, may become inapt dogma. For this reason, as we say about the current administration, watch what they do (their actual behavior–what they’re controlling), not what they say. Naturalistic observation precedes the TCV, to narrow the population of possible variables before setting up an artificially controlled environment where you know what the subject is perceiving and what the disturbances are to that perception. Still more challenging to measure q.o, d, and q.i.

Consider cooperative hunting strategies. Here’s a proposal for defining and classifying how groups of animals hunt together:

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

Here’s a public-facing overview (August 22, 2017):

To pick one example, orcas, which are a kind of overgrown dolphins, have become the apex predator in the Pacific by evolving their hunting strategies with changes in arctic ice cover. Reseachers captured video which was included in a 2014 Public Broadcasting System (PBS) program. The PBS website says they have technical problems making this video available ‘in my area’ but you may be able to view it; here’s the description.

A shift of power is taking place at the top of the world. The Arctic is undergoing a dramatic change, and with this change, one iconic Arctic hunter may soon have to give way to another as solid ice turns to open sea. The polar bear, once king of the North, needs ice to stalk its prey. Killer whales, or orca, on the other hand, are unable to hunt in an ocean locked in ice. As the ice increasingly disappears, the tables have turned. Polar bears are struggling to survive while the now open ocean provides bountiful new hunting grounds for the whales.

Larger whales can hide from them under the remaining ice sheet, but narwhals can’t do that – IIRC because they have to surface more often to breathe. Orcas have complex cooperative strategies for feasting on gray whales, which are much bigger than orcas.

This description of aspects of the video identifies specific strategies that Orcas use:
The Killer Whale’s Killer Weapon — Its Brain

Around New Zealand they seem to be using sharks as bottles of nutritional supplements. The likely reason is that the shark carcass sinks, so they grab the best part, just as they do if they can’t get a gray whale carcass into shallow water.

19

Hi Warren

WM: What about also the idea that humans might be able to perceive (and therefore control) elements of their environment at the lower levels of the hierarchy through the reorganisation of input functions that other animals cannot - because their input functions are hardwired?

RM: My concept (and, I think, the PCT concept) of the input functions in all species is that they provide the neural architecture for computing different types of perceptual variables. It’s this architecture that is hardwired. I imagine this architecture to be functionally equivalent to a mathematical equation that describes the function. For example, in B:CP, Powers proposes that the architecture of input functions that produce sensation type perceptions is equivalent to a linear equation:

p = a +b.1 x s.1+b.2 x s.2…+b.n x s.n

where the s.i are the intensity perception inputs to the perceptual function and the b.i are the weights for these inputs. I view the linear equation as the “hard wired” aspect of the perceptual function and the input weights, b.i, and possibly the number of inputs as the potentially reorganizable aspect of the function.

RM: But I know of no evidence that perceptual functions – such as the functions that produce sensations such as color and timbre – are reorganizable, in humans or in animals. For example, the sensation of color is presumably a function of a linear combination of the inputs from three different color receptor (cone) cells (ignoring context effects). I know of no evidence that the weights of the linear function that produces the color sensation can change. Such a change would result in a change in the apparent color of an object. This can happen as a consequence of certain diseases, such as diabetes. But the change in color sensation that results from disease is caused by a loss of receptor input – loss of cone cells of a particular type – not from learning a new way to weight these inputs. But if you have evidence that humans can reorganize perceptual input functions while other animals cannot I would like to see it.

WM: I’m thinking for example of writing - we can identify and reproduce letters - which are simple configurations, but no other animals can do this…

RM: I think animals can identify different configurations Chimps, for example, can certainly identify different shapes. One of my friends in grad school was the lady who talked with the ape Sarah in David Premack’s ape language studies using cards where different shapes stood for different words.

RM: Of course, reproducing configurations is a whole different ball game. But if an animal can identify configurations it should be able to reproduce them if it has or can be given the means to do so. It should be possible to find some existing research that is relevant to this point.

RM: But this is a thread about how we might conduct a PCT-based research program. So I think the first order of business should be collecting examples of the controlled variables involved in various behaviors and developing a basis for classifying these variables into types. You can’t really study the involvement of reorganization in behavior until you know how organized behavior works, And that means knowing the variables around which any particular example of behavior is organized.

Best

Rick

20

Warren, for example primates don’t have the physiological means to control the phonemic contrasts of human languages, but it is well established that they can learn sign language, even from other primates. For speech, you control contrasts between vocal sound configurations. In primates, the vocal folds cannot close completely, they are said to lack sufficiently fine motor control of the jaw and tongue, and as Phil Lieberman found, the root of the tongue is too far forward in the oral-pharyngeal cavety. However, they do have the necessary dexterity to control contrasts between configurations of the hands and fingers for signing.

The learning of signing I would guess requires not only developing perceptual input functions for the specific configurations but also repurposing and reorganizing some of those ‘mirror neuron’ systems that we were reading about 15 years ago. Rick suggested at the time that the neural firing that was called ‘mirroring’ is the firing of reference signals. The idea is that on observing another doing X, the reference signals for doing X fire, but presumably controlling in imagination since the observing animal does not actually do X. But the ‘mirror neurons’ were observed not in the motor area but in Broca’s area, which in humans is involved with language. Bill thought this might localize the proposed Category level. The thread initiated in 2005 by Dick Robertson is Mirror Neurons ?