A Test of "Collective Control" Theory

Yes, it’s right except that in PCT, if not in the “collective control model”, what is controlled is a perceptual variable not an environmental variable. Perceptual variables are functions of environmental variables. Controlled perceptual variables do not exist as entities in the environment; they exist only in systems (such as living organism) that can perceive them. This is an important point, as you will see below, since what I proposed as one of the variables that is being controlled by individuals in a collective is “civility”, which is a perception – a function of variables in the environment – and not something that exists as an entity in the environment.

This is a very good question (except for the CCEV part) and I will answer it after I answer your next one:

Then we are in complete agreement. I also agree that it is the “value” of civility that matters. When I say that civility is a perception I always mean perceptual variable; we control variables, not constants. So, as you say, civility is a perceptual variable that ranges in value from uncivil to polite and friendly.

So now I can answer your question about why I predicted that there should be a “lower police ratio in bigger collectives”. My reasoning went as follows: the collective control model (CCM) says that as the size of the collective increases, the variability of the virtual reference state of the virtual controlled variable – civility in this case – decreases. This happens because each member of the collective is trying to get the civility of the collective to be at their reference by behaving appropriately; people who want a civil society are behaving very civilly and those who want an uncivil society are behaving very uncivilly.

Assuming, as you do, that there is an absolute level of civility, in the form of criminal law, that defines the type of civil behavior that requires police action, then in collectives with highly variable virtual reference states, there will often be the need for those wanting an uncivil society to behave very uncivilly in order to bring the virtual state of civility back toward their reference; and these are the people for whom police are needed. In collectives with highly stable virtual reference states of civility, on the other hand, there will rarely be a need for uncivil people to go out of their way to bring the virtual state of civility back toward their reference since it is constantly the same distance from their reference so they will just keep acting as they are. So it’s the high variability of the virtual reference for civility in small collectives that leads uncivil people to behave even more uncivilly when the virtual reference for civility wanders up to much toward civility.

I hope this is relatively understandable. Of course, since the prediction based on this reasoning was not consistent with the data, either my reasoning is wrong or the model is simply not relevant to this phenomenon (the relationship of collective size to police ratio) and some other explanation will have to be found. But no matter what the explanation of the failure of the collective control model in this case turns out be, it appears to be a fact (in the US at least) that there is a relatively strong positive relationship between city size (the size of a collective) and police ratio (the number of police per capita in the collective); an interesting social phenomenon that should be explained.

Best regards, Rick

Depending on the loop gain in each individual. If individuals’ loop gain is quite low, the sum is not going to be very high. A smaller number, each controlling with high gain, could easily produce greater stability.

It does not become interesting, really, until we consider what it means to the participants for these stabilities (however initiated) to be present in the shared environment of the participants. The quote from p. 271 again: “Stabilized cultural forms can facilitate the coordination of people’s interactions, which helps to make their collective control activities more cooperative than conflictive.” The stabilities become useful within environmental feedback paths for controlling other variables that need not be related to those that were involved in initiating the stability (either physical or cultural). People come to depend upon them. Whatever conflict may have engendered them initially (if that was the origin) is superseded by cooperation in maintaining them for the sake of other purposes.

So you’re inquiring what phenomena Kent has had in mind from his field of sociology as he has been developing this part of PCT.

Oh. Well, then you’re just not interested enough. That’s OK too. But no, I won’t do the work for you. That would be unfair to Kent.

In your reply to Eetu:

How do you know that this is not in Kent’s model? Quoting again Kent’s very brief identification of three explanatory mechanisms, on p. 271:

Yes. It’s good as far as it goes. It does not recognize Bill Labov’s main finding in “The social motivation of a sound change” and in his subsequent work in New York. It’s not a matter of people talking like their interlocutors. Its a matter of adolescents deciding which of their interlocutors to talk like.

Your model does present a way of explaining a good range of other phenomena. The phenomenon of people talking like their interlocutors is seen for example in what is called social register, i.e. speaking to an interviewer in a relatively formal register, then being overheard speaking in a less formal way by the interviewer in when a telephone call from a friend interrupts, or shifting from one dialect to another. The two phenomena are of a different logical type, one depending upon the other.

There is a like logical type distinction needed in an earlier chapter. Using boids and the crowd simulation as examples of cooperation calls for a distinction between purposeful cooperation (such as your first example of jointly moving a piece of furniture) and the appearance of cooperation or stabilization as a side effect of controlling variables other than that result.

I think it’s implied that the increase in stability with the size of hte collective occurs with “other things being equal”. But it’s unlikely that, say, two people controlling the same variable could have the same collective effect on that variable as when the entire population of NYC is controlling the variable. Unless, of course, one of the two were Superman (I’m saving you the trouble of pointing that out;-)

What is interesting about that statement is that it is completely false. What the collective control model implies about what it means to participants in this kind of collective control is that nearly all are experiencing some degree of chronic error; those with references close to the virtual reference state experiencing little error and those with references quite far from the virtual reference state experiencing large error.

You betcha.

I don’t know how you figure that it would be unfair to Kent? But I did read the paper and saw no description of a phenomenon that fit the model. I saw some social phenomena mentioned but no description of how the model actually applies to them. The paper is just filled with assertions, like the one above about what collective control is like for participants, with absolutely no evidence to back the assertion.

I can find nothing in that long quote that talks about the degree of conflict in a collective controlling a variable.

Boy, you just can’t find anything worthwhile in my work;-)

What I found to be interesting is that my simple imitation model not only accounts for the emergence of different pronunciations in different subpopulations but it also accounts for the fact that these differences stabilize at these new pronunciations; stability that emerges, by the way, sans conflict! Of course, this work also demonstrates how a control model of individuals can account for actual social data – the lovely data on regional differences in pronunciation collected by Labov.

Yes, it’s an extremely important distinction that I discuss in sections 7.1.4 and 7.1.5.

Best, Rick

My, how eager you are to claim that the collective control part of PCT is not part of PCT. Why?

“More cooperative than conflictive” tells you that it’s a gradient, not a binary choice. Some degree of chronic error is perfectly compatible with control activities being “more cooperative than conflictive” when “stabilized cultural forms facilitate the coordination of people’s interactions”.

No, and that is a common complaint of people living in an institution-ridden built environment. But they rely on the subway though they dislike many things about it, and some individuals and groups of individuals control a perception of “those responsible” removing or ameliorating chronic sources of disturbances to collectively controlled variables like running on time, being clean, and being safe,
with higher gain than others do, and control to perceive (find or create) effective environmental feedback paths for such control. Others employ similar means for more individual purposes. Of such is politics made.

Sorry all, this is too interesting topic, even though I have no time to focus on it properly.

RM: in PCT, if not in the “collective control model”, what is controlled is a perceptual variable not an environmental variable.

That is true, every individual concrete controller controls their own perceptual variables. But the virtual collective controller does not even have perceptions. Instead it stabilizes some variables in the physical environment. It is not even necessary that all individual controllers control the “same” perceptual variable. It is enough, for example, that some of the environmental argument variables of the perceptual functions are same. Collective control can even realize via side effects of individual controls.

Secondly, I don’t believe that most crimes are committed because the criminals control for low civility. It is possible, but far more usual may be that they commit crimes instead of their possible control for civility. They control for getting money, drugs, power or something else what they do control for. Crime is a means of control for something else.

Thirdly, the wealthiest people who – as well known(?) – control with high gain for their physical and economical safety, can use police forces as means for that. Because they have much more resources they can control for something with much more gain than poor people. So we are back in economical inequality, which I think cannot be overlooked even if we want to explain that phenomenon of big-cities-high-police-ratio using PCT.

Hi Eetu

RM: in PCT, if not in the “collective control model”, what is controlled is a perceptual variable not an environmental variable.

EP: That is true, every individual concrete controller controls their own perceptual variables.

RM: Yes, but in the collective control model, all controllers control the same (or nearly the same) perceptual variable. That’s why there is conflict, resulting in the perpetual variable being stabilized in a virtual reference state.

EP: But the virtual collective controller does not even have perceptions.

RM: . There is no “virtual collective controller”; the collective controller is a collection of real controllers. What is virtual is the reference state of the perceptual variable that those controllers are all controlling.

EP: No, Instead it stabilizes some variables in the physical environment.

RM: No, like the controlling done by any control system, a collective of controllers stabilizes a perceptual variable – a variable that is a function of variables in the physical environment.

EP: It is not even necessary that all individual controllers control the “same” perceptual variable.

RM: That’s true. But the perceptual variables controlled by the different controllers in the collective have got to be very close to being the same or there will be no conflict and, thus, no stabilization of a variable in a virtual reference state. For a collective of fixed size, the best stabilization of a perceptual variable that can be achieved by Kent’s model occurs when all controllers in the collective are controlling exactly the same perceptual variable.

EP: Secondly, I don’t believe that most crimes are committed because the criminals control for low civility. It is possible, but far more usual may be that they commit crimes instead of their possible control for civility. They control for getting money, drugs, power or something else what they do control for. Crime is a means of control for something else.

EP: Thirdly, the wealthiest people who – as well known(?) – control with high gain for their physical and economical safety, can use police forces as means for that. Because they have much more resources they can control for something with much more gain than poor people. So we are back in economical inequality, which I think cannot be overlooked even if we want to explain that phenomenon of big-cities-high-police-ratio using PCT.

RM: I completely agree with all this. But none of these explanations are anything like Kent’s collective control model of social phenomena. Maybe what you are saying is that the collective control model doesn’t apply to this aspect of society – the relationship between the size of a collective and the amount of policing per capita in that collective. If so, I agree. And I would just like someone to show me just one social phenomenon that the model does apply to, and how it explains it. Right now my only problem with the collective control model is that it seems to be what I would call a “Seinfeld model”: a model that is about nothing;-)

Best, Rick

I have no idea where you get this stuff. But, for the record, Kent’s collective control model is exactly a PCT model. Indeed, the virtual reference state that results when a perceptual variable is controlled by different systems relative to different reference levels was described (and named ) by Bill in B:CP.

The problem with Kent’s model isn’t that it’s not part of PCT; it’s that it is a model about nothing. No one has shown how the model applies to any particular collective control phenomenon. There is more to PCT than just theory; you also have to show how the theory applies to actual phenomena. And this has to involve a bit more than simply asserting that it does.

But that’s just an assertion. In order to be compared, “cooperative” and “conflictive” have to be measured. How were they measured? Where are these measures presented?

I see nothing that I would call “cooperation” occurring in Kent’s model. The virtual reference state of the controlled variable is a side effect of each individual’s controlling; the state of the controlled variable could be unwanted by all the participants so that everyone is experiencing considerable error. It may look like “cooperation” to an external observer but I don’t think it would feel much like cooperation to the participants in that kind of collective control. How cooperative would you feel in a tug of war against a team of equal strength to yours?

In the models of cooperative control that are described in The Study of Living Control Systems (pp. 104-113) none of the participants in the collective controlling experiences much error at all.

Best, Rick

Two themes here. First, you keep insisting that Kent’s model of collective control requires conflict. (Instant examples below.) Kent says that it does not. Please acknowledge his denial of your claim (on p. 271 of his chapter) and explain why you think his understanding of his model is wrong.

Secondly, yes data to be modeled are needed. But the data should not be oversimplified to fit preconceptions. Your demo of individuals’ numerical variables converging as a model of people coming to speak like their most frequent interlocutors is like the boids and the crowd demo. Bill’s data adds perceptual variables at a higher level: two kinds of people, Islanders and summer people, and a decision that an adolescent makes at the system concept and principle levels as to which of these kinds of people he or she is and will be. This accounts for exceptions in the two regions, where up-Island adolescents were more likely to identify as Islanders and down-Island adolescents were more likely to identify with summer people and be thinking about further education and a career that takes them away from the Island. Even though folk down Island in the larger towns and the harbors where tourists come and go more frequently have off-Island folk as interlocutors, as compared with folk in the smaller and more rural up-Island towns, kids in each place went against the expectation set by your model.

It also explains the sound change which is the principal topic of Bill’s paper. Yes, kids identifying as Islanders spoke the aw dipthong in shout loud about it with a centralized a vowel (as in “duh”). The sound change was brought about because they overshot the mark, producing this ‘higher centralization index’ in the ay diphthong of can’t abide a knife fight.

DIalect mapping in the 19th and 20th centuries had shown this feature of 18th century American English (along with other features like post-vocalic r, which came into the colonies from south of london) disappearing all over New England and New York, being superseded by high prestige ways of talking in the cities (New York and Boston leading the way, as imported from London). So the leading question that started Bill’s investigation was how come this steady erosion was being reversed on Martha’s Vineyard? And then the sharper question was, oops! How come the other diphthong is going along with it, toward a pronunciation of the ay diphthong that was never part of 18th century English.

Your numeric variable is proposed to correspond to the ‘centralization index’ measured in sound spectrograms of people’s speech, and I’m sure you get a nice time-varying convergence, though Bill Labov did not get corresponding time-varying data for these to support such a model. My own experience of my eldest daughter’s speech changing almost overnight to talking more like a Gloucester tough than anyone in the teen crowd she ran with tells me that system level control of the identity we present to others can control such variables with astonishingly high gain. Bill didn’t capture such individual change data on the Vineyard, but there’s lots of data out there for code switching, dialect alternation, social register switching, and the like. None of this can even be thought about in a single-level model that presumes that collective control can only arise out of interpersonal conflict.

I was all set to repent, thinking it entirely possible that I had accidentally said such a silly thing. But I looked at all your quotes and found that I said no such thing in any of them. Here are the quotes again:

In none of these quotes do I say anything like " Kent’s model of collective control requires conflict". What I did say (in the first and last quote) is: conflict is required for the commonly controlled variable to be stabilized in a virtual reference state!

When all members of the collective have the same reference for the commonly controlled variable, then that variable is being stabilized in an actual, not a virtual, reference state; it is the reference state in which all controllers in the collective actually want the variable to be.

Which leads me to suggest that you might want to get in touch with Kent and let him know that in future publications he should make it clear that a collective of controllers is a “giant virtual control system” stabilizing a virtual controlled variable only when all the controllers have different references for the controlled variable and are, thus, in conflict,

There should already be data available because the model should have been built to account for data. In fact the model came first and has never been tested against data. This is the theory first (or, really, theory only) approach to PCT in spades and it is of no interest to me at all.

I tried to introduce a possible way to test the collective control model and I got repaid with scorn; that got me all tangled up in blue so I do believe I’ve had enough and I’ll go seek some shelter from the storm;-).


With two instances of the ‘Little Man’ simulation, model a handshake. Include higher levels as follows:

For a deceptively simple example that illustrates how everyday interactions can include complex combinations of protocols and collective control, consider the custom of a handshake, where one person extends a hand for the other to shake, and vice versa. A handshake is structured by a highly ritualized protocol, the initial stage of which requires that the two partners control symmetric perceptions, with each partner controlling for the other to perceive his or her willingness to shake hands with the other. To continue the performance of the protocol, however, each partner must independently control his or her own low-level perceptions of extending a hand, grasping the hand of the other, and then making an up-and-down motion.

Simultaneously, at a higher level of perception, the physical actions of the other party (along with the haptic, kinesthetic, and visual feedback from the participant’s own actions) provide each party with atenfels for perceiving that a handshake ritual is in progress—a perception that they collectively control. At some higher level, the handshake ritual may function as one step in a culturally prescribed protocol for greeting old acquaintances or being introduced to new ones, in which the participants may play their parts by taking turns to choose from a variety of culturally stereotyped comments, thus controlling different perceptions in the service of collectively controlling a yet higher-level perception of sociability and goodwill between the two of them.

In the handshake many culturally relevant messages may pass from one person to the other, such as the assertion of dominance by a strong grip and firm control of the hand motion, an acceptance of submissiveness by letting one’s hand be squeezed, the giving of affection by different subtleties of grip, and so forth. Such messages may in some cases be collectively controlled and distinctive across cultures, while others may be as cross-culturally relevant as the smiles that Darwin associated with all primates, or may be as idiosyncratic as the “secret handshake” that members of a secret society use when they meet other members of the club.

This is on pp. 242-243 of Kent’s chapter, immediately following the first of the two passages that I have previously quoted, which is:

The important point to remember here is that collective control can result in stabilization of some features of shared environments whether or not conflict is involved.

Ordinarily in a handshake the two parties control so as to avoid or prevent conflict over the perceptions controlled in shaking hands. This purposeful avoidance of conflict is the main message of an ordinary, friendly handshake. One or both parties can introduce conflict at one or more levels of the protocol “such as the assertion of dominance by a strong grip and firm control of the hand motion”, etc., and this purposeful introduction conflict into a normally non-conflictual protocol constitutes a communication about a conflict in respect to control of other perceptions apart from the handshake itself. That could be part of your model.

Quantitative data could be gathered with sensors in gloves. Funding might be obtained for research relevant to human-like robotics.

More generally, the ‘same variable’ may be controlled by two controllers without the expected mechanically escalating conflict when ‘up a level’ one or both are controlling a purpose of avoiding or minimizing conflict.

Even if only one participant does this, with the appearance of ‘giving in’ to the other to avoid conflict, that one may still be controlling the seemingly abandoned variable, looking for alternative means or opportunities of control, or may be seeking alternative inputs for the higher perceptual input function to which that variable had been input. This can be modeled if the simulation includes appropriate connections to higher levels. As is generally the case with higher levels, getting quantitative data will be more difficult than with variables like hand pressure. That’s a very important methodological problem that I had hoped your book would help with. But maybe “not all that is important can be counted.”


You say that CCM is a model about nothing, that it does not explain anything. But think about one very common and in principle simple situation: Two parties or individual controllers are in conflict because they want to utilize the same resource, say a common car. Then they negotiate a pact: one party may use the car in mornings and the other in the evenings. Or they are in conflict about the position of the borderline between their fields, and they negotiate (perhaps in a court) and make an agreement about the position of the borderline. CCM explains these stabilizations. The interesting social situations are usually not so simple and straightforward, but it does not mean that CCM is useless in them. It only must be applied in suitably complicated ways and take many things into account – not just one perceptual variable and the size of the collective.

More in the next msg.

RM: In none of these quotes do I say anything like " Kent’s model of collective control requires conflict". What I did say (in the first and last quote) is: conflict is required for the commonly controlled variable to be stabilized in a virtual reference state!

Of course a model does not require anything, but in fact you insist that the use of it requires conflict, and there is not much difference. But even this is not true. What is important and interesting in CCM is that we can take into account and see the interdependence between perceptual and environmental variables: as you said earlier, a perceptual variable is a function of some environmental variables (EV). These EVs are arguments of the function, and the end result of the function, the perceptual variable (PV), depends on their values. The output of the control affects these EVs and so they depend respectively on the PV (or rather its reference value and of course the disturbance).

Now think about your What is size demo (Control of Size ). Then change the settings so that there are three different interfaces – display and an input device- in different rooms. In room 1 there is subject A who sees in the display a horizontal line a length of which (X) she is advised to control. Similarly in the second room is subject B controlling the length of a vertical line (Y). The computer program causes random disturbances for both A and Y. In the third room the display shows a rectangle where its width is X and height is Y. The subject C is a PCT literate who is advised to apply TCV and determine whether the size of the box is controlled by someone or not.

Now, what is the result of C’s TCV? According to CCM it should be that the size of the box (X*Y) IS controlled. But it is controlled neither by A nor B, nor anyone else but the “giant virtual controller”. If I have thought this experiment and the idea of CCM right, then this shows that in collective control there need not necessarily be neither conflict nor cooperation (= the PVs controlled by participants of the collective control need not be same) and that the “virtual controller” stabilizes EVs (because a nonextant controller cannot even have PVs). These stabilized EVs then affect the PVs of the same or still other concrete controllers. That is why collective control causes so surprising results.

RM: I tried to introduce a possible way to test the collective control model and I got repaid with scorn; that got me all tangled up in blue so I do believe I’ve had enough and I’ll go seek some shelter from the storm;-).

I appreciate that you did this – even thought it was not a very successful way. I am sorry for your feeling. Perhaps the reaction followed because you stated your idea in a way which made it possible to interpret that you had falsified CCM with it. Anyway you should be glad that it caused discussion because it is the only way how science can proceed.


Building a working model is a nice way to illustrate how the model works. So if you think a model of shaking hands would be a good illustration of how the collective control model works then I think you or some other purveyor of that model should build such a model. I am interested in how models fit actual data so what would be more helpful to me is a description of the kind of data that would test the collective control model along with an explanation of why it would.

Could you please give me an example of a feature of the shared environment that is stabilized by collective control?

Best, Rick

This can also happen when the gain of one of the two opposing controllers is much lower than that of the other. I discovered this when I was doing some research on conflict. It turns out that when there is a large gain difference between two agents controlling the same variable in a conflict the agent with the higher gain controls better than he would have without the opposition.

This result was a huge surprise to me and I thought it was inconsistent with the PCT model of the agents, which would have required some revision of the model. But I ran simulations with two PCT agents of very unequal gain controlling the same variable and the results were the same as with a real person as the high gain controller; it works that way when there is a transport lag in the high gain agent is longer than that in the low gain agent.

This finding was dubbed (by Bill) the “Marken effect” and was discussed at some length in the earliest days of CSGNet. Here’s one discussion of it by Bill from 1994:

[From Bill Powers (940602.2040 MDT)]
RE: the Marken Effect.
Discovering that the model had to reproduce the real subject’s transport lag in order to get this effect did, as you suggest, reveal the conditions under which this effect is seen. But it also explained_ the effect.
The explanation is this. The auxiliary control system, with a modest loop gain, simply tried to keep the controlled variable constant, operating in mild conflict with the main control system, either Rick or the model of Rick. What made control a little better with the auxiliary system in operation was the fact that it did NOT have a transport lag in it. Thus, on the average, a disturbance that caused a change in the controlled variable was counteracted, to some extent, by the auxiliary control system during Rick’s transport lag. This reduced the effective disturbance that Rick or the model of Rick experienced, resulting in slightly but reliably better control. A test of the Marken Effect by simulation failed at first because the model used for Rick’s behavior did not include a transport lag. The model of Rick could therefore act just as fast as the auxiliary control system could, so there was a simple conflict and no improvement in control. When the model of Rick was changed to match Rick’s actual transport lag, the improvement reappeared. This explanation goes considerably beyond merely noting the conditions under which the Marken Effect appears.

It’s amazing what you can learn when you actually compare the behavior of the model to that of an actual person.

My book didn’t include this this “very important methodological point” because I never encountered the problem of having to “include appropriate connections to higher levels” in my models (such as the model of the “Marken effect” described above). Maybe you could give me an example of where you have encountered the problem in your own research and I’ll include a discussion of that “very important methodological point” in the second edition of the book.

Best, Rick

Not quite. I say it is about nothing because, to my knowledge, it has never been shown to fit actual data.

So you say. But I have never seen the CCM model fit to data on what you say are the stabilizations that result from negotiation. Indeed, I didn’t even know that negotiation was part of CCM. An important part of CCM is that it shows that stabilizations occur even when there is conflict between the parties controlling the variable being stabilized.

My criticism of CCM is that it might not be true, not that it might not be useful.

Best, Rick

RM: What I said above is not an insistence that the use of the model requires conflict. What I said is that conflict is required for the collectively controlled variable to be stabilized in a virtual reference state; that is, a reference state that doesn’t correspond to the actual reference specifications of some or all members of the collective. As I said, if the reference specifications of all members of the collective were exactly the same (rather unlikely in a real collective) the reference state of the collectively controlled variable would be “actual”, not “virtual”.

RM: There are a few problems here. First of all, this demo doesn’t let us “take into account and see the interdependence between perceptual and environmental variables” any better than does any plain vanilla PCT model. Second, this is not an application of CCM. In CCM, all members of the collective are controlling the same perceptual variable, relative to the same reference (no conflict) or different references (conflict); in your demo, the members of the collective (A and B) are each controlling two completely different perceptions (X and Y, respectively) and the area (XY) that is seen by observer C as being controlled is a side-effect of the controlling done by agents A and B. Third, a proper application of the TCV would readily reveal that XY is a side effect of the controlling done by A and B and not the result of the controlling done by a “giant virtual controller”. This is because the last step in the TCV, after determining that a variable appears to be under control, is to trace the source of the actions that keep the variable under control. The result of this trace would be to find that the actions are those of A and B, each independently controlling X and Y, respectively. Thus C would see that neither A nor B is controlling XY and he would correctly conclude that the constancy of XY is a side effect of the independent controlling done by A and B.

Unfortunately you didn’t get CCM right but no harm done because no one said that conflict (or cooperation) is needed for collective control to occur in CCM. When a collective controls the same variable that variable will be kept in a reference state; if there is conflict among the members of the collective then that reference state will be virtual; if there is no conflict among the members of the collective then the reference state will be actual – the reference state desired by all members.

RM: Thanks Eetu. But don’t worry. I’ve been doing this for over forty years so, like Inigo Montoya of Princess Bride fame, I’ve had to get used to disappointment.

Best, Rick

Four dollars.
A handshake.
Each of the thirteen words above and here.
The translations into Dutch of these answers to your request.
The concept ‘controlled variable’.
That these are answers to your request.

What phenomena does this model?

I can think of a very large class of phenomena where not one but many autonomous control systems are each controlling at low gain some perceived aspect or part of an environmental situation in which a number of aspects or parts are interdependent. It is often the case that there are also one or more ‘maintainers’ who control perceptions of that situation with high gain, but with delays between control outputs. And in many cases there are long delays between control outputs by the low-gain controllers, but they are many, and the aggregate effect is a continuous low-gain control. This is analogous to Bill’s useful fiction of the ‘neural current’, useful because it can be represented in the model and in simulations by a single numerical variable.

RM: Are these names of the variables that are being collectively controlled, or the reference states of those variables? If the former, could you please pick one or two (or all, if you would be so kind) and tell me the possible states of those variables; if the latter, could you tell be whether these are virtual or actual reference states of the variables. Thanks

RM: The improved ability to control when placed in conflict with a low-gain, 0-transport lag control system.

Best, Rick

Actually, you did have that problem for one of the examples in your book. The data presented in “The social motivation of a sound change” (Labov 1963) crucially depend upon control of perceptions as high as the system concept of self-image. You limited your model to the lowest-level data that he presented. You modeled a convergence of numerical values, analogous to the ‘flocking’ of boids or the rings and arcs of the crowd demo. Labov handed you numerical data related to control at the relationship level. The data about higher levels were not handed to you as quantitative values, so you ignored them.

You may not have noticed that in doing this you exemplified one way of quantifying higher-level perceptions, even though in this case it was applied to perceptions at the relationship level. But the ‘centralization index’ that you took as data is not a measure of an output quantity Qo. It is not even a measure of individual behavior.

The over-all degree of centralization for each speaker is expressed by the mean of the numerical values of the grades of each instance listed on the chart. Thus on Figure 4, the centralization index for /ai/ (CI /ai/), is 0.75, and the index for /au/ (CI /au/), is 0.39. We can then find the mean CI for any group of persons by averaging the CI for the members of the group. ([Labov1963:291](http://languagelog.ldc.upenn.edu/myl/Labov1963.pdf))

So by accepting Labov’s centralization data you were using statistical methods much as Brian D’Agostino used statistical methods to identify correlated reference values at the principle level controlled as inputs to a self-perception system concept.

(See the first addendum to this post, below, for some of the issues with the ‘centralization index’ measure.)

One is above, another is below.

I’ve given other examples which you’ve rejected outright because you require me to give data first, but the methodological problem we’re talking about here is how to get quantified data about these higher levels. One way is statistical measures like those used by D’Agostino and (above) by Labov (and therefore by you). Another way is a form of Turing test; more on that below.

Here is a discussion of where we were 28 years ago in February of 1994:

“As the model grows,” he said, and the model has grown since then “to encompass more of what is observed and experienced” at higher levels.

I am asking you to participate, rather than resisting. Your model of the lowest level of Labov’s data is a good start. Keep going. For anyone who understands what Labov did, it’s kind of an embarrassment, until it’s clear that you reported only the first steps toward a model.

OK, on to a different example.

The methodological problem we’re talking about here is how to get quantified data for higher levels of control so that a model of this type can be built.

You’re the experimental psychologist. You’ve written as an authority on methodology of psychology. You have skills and experience obtaining quantitative measures of the observable aspects of behavior. I should think you would relish the challenge of extending the ‘Little Man’ program to run two instances at once, and instead of one hand tracking a cursor have each hand control the location of the other’s hand. Leave out for now the problems of programming a hand that can sense and grasp configurations, such as the configuration of another hand; simply bringing the two hands into (some representation of) contact is enough for now.

What higher-level perceptions are controlled by two autonomous agents bringing their hands in contact and producing the appearance of a hand-shake?

How do you compare model performance to living performance? You don’t need quantitative data about how people move their hands in a handshake to validate the performance of a simulation, a kind of Turing test by the observer is sufficient: you know it when you see it. Trial and error until it looks right.

Then inquire, and submit to public inquiry here, whether or not the posited higher-level CVs and their references are plausible. Inquire whether control of other higher-level variables might have the same observed appearance of a handshake. We all have memories of diverse experiences of shaking hands with another person.

Quantitative I/O and d measures are straightforward for motor control but the higher we go in the hierarchy quantities are notoriously more and more difficult to obtain (or pretend to obtain, statistically or otherwise) for perceptual input, output affecting that input, and disturbances affecting that input. I am proposing that something like a Turing test is an appropriate methodological solution to this problem. Until you as the author of textbooks on methodology can propose how we can get quantified data at higher levels of control, this is the only recourse that I see. Until you or someone else extends our methodology so we can produce such quantified data, and satisfy the criteria that you are demanding before you recognize observations as legitimate phenomena, then our PCT methodology at higher levels is to create a model that we think will pass a ‘Turing test’ and then see if it does.

Would you consider the achievement of a handshake cooperative? It is controlled by two participants (and by observers, who may be controlling perceptions of its higher-level significance). So it is a collectively controlled perception. There is no conflict; it is achieved by mutual avoidance of conflict. As the grip is achieved a handshake may be made conflictual, e.g. to control a perception of a dominance/subservience relationship or a perception that a demand for such a relationship has been communicated. Mutual avoidance of conflict communicates an intention to avoid conflict when controlling in a common environment (cf. Bateson “A theory of play and fantasy”).

It sometimes happens that you end up modeling something different from the phenomenon that you intended to model. Your ‘Marken effect’ is an example.

Your methodological requirement that you lay on me is that one should start out with a phenomenon and the model comes later. In this case, the phenomenon emerged unexpectedly from performance of a model of controlling with a slight disturbance. You wanted it to model an ‘inanimate’ disturbance, such as wind blowing, but you used a low-gain controller to generate the disturbance.

Since the disturbance was in fact generated by a ‘computer generated control system’ or “CGCS”, which is not in fact an ‘inanimate’ source of disturbance, to make sure the results were valid you recorded the outputs of the CGCS in a table and ran the experiment again with the canned data as the successive values of the disturbance variable. Unexpectedly, the effect went away.

When the CGCS was controlling in very slight conflict with the main controller (either you or a model control system replicating your tracking behavior), all was well, but when the very same numerical data was stored in a table and then input from the table in a subsequent run, it did not work.

Bill suggested that you include a value for a transport lag in your model of your performance.

Without the lag the stabilization of control is a consequence of two control systems in conflict, one with low gain. Bill described this case as follows:

In general, the “dynamics of the output” of a control system mirror the ‘dynamics’ of a disturbance. When the source of disturbance to the subject’s control is another control system, then because the outputs of the subject are reciprocally a disturbance to the disturbing control system the “dynamics of the output” of the disturbing control system mirror the “dynamics of the output” of the subject. Bill’s phrase “as much as” is not numerically true (because of the difference in gain), but it is true as a kind of synonym of “in the same way as”.

This is true of any conflict that does not go into runaway escalation. In this case, the low gain of the disturbing controller prevents runaway escalation of the conflict. The low-gain controller is unable to control completely, but does have some effect.

It follows that varying the disturbance at values stored in a table would result in different outputs from the subject or the model of the subject, because this reciprocal interaction is not present.

It appears that this is the (or a) mechanism by which collective control increases stability of control.

I wonder if it is essential that the disturbing control system “tried to keep the controlled variable constant”. I bet it would not matter if it had a varying reference value.

The same stabilizing effect should result if the higher-gain system controls from time to time or on a schedule and the disturbing system controls more frequently or (mostly) at other times. Likewise if there is a population of low-gain controllers which control in the aggregate with greater frequency than the higher-gain system, even if they control at different values, or if they control different aspects of a complex higher-level variable.

The stabilization should result if the low-gain controllers do not all control the same aspect of a complex perceptual variable which the high-gain controller is controlling as a whole.

The relevance to collective control is evident.

Earlier, I quoted three things that you had written and said that they were examples of your claim that conflict is a prerequisite to collective control. You denied that. You were correct about one of the three:

In haste and late at night I didn’t cut and paste this in a different location as I had intended. My intention was to agree with you about the complexity of what is perceived, and to point out that different participants in collective control may stabilize some but not all the physical variables of which the observer’s perception is a function. Indeed, what one agent controls may only intersect the set of physical correlates that another agent’s controlling affects, or the set of lower-level perceptual inputs of one agent’s perception of a collectively controlled variable may only intersect another agent’s corresponding lower-level perceptions. They may be controlling different higher-level perceptions which happen to have perceptual inputs in common.

Dave crosses at the crosswalk to get to the bookstore; Alice crosses the other way on her way to the college campus; Jane’s mother driving down Main street gets stopped by a policeman for failure to stop at the crosswalk; Jane’s father, irate, books a separate visit to Northampton to contest the ticket in court, presents photos showing that the crosswalk lines have been worn to near invisibility; the judge agrees and dismisses the ticket; Ben in the DPW has the repainting of crosswalks and bike lanes on his schedule during the college break; he gets a call from the judge’s secretary telling him he’d better get the crosswalks done sooner. The painted lines on the street are collectively controlled.

Quantify that variable.

Addenda follow.

Centralization index as a datum:

In the case of ‘The social motivation of a sound change’, you took the ‘centralization index’ quantity to be a measure of the central position of the tongue within the oral cavity. Bill Labov did not measure the height at which speakers held their tongues. He measured the height of the second cluster of undamped harmonics in sound spectrograms of audio recordings sampled at points where he (as a native user of English) recognized that the speaker was producing a word with a diphthong in it that was under investigation. Such a cluster of undamped and resonance-reinforced harmonics in speech sounds is called a formant. The perceptual input function for the second vowel of the diphthong recognizes relationships between formants across the audible harmonic spectrum, or perhaps a configuration, if that is a signal that the auditory system generates—I am not certain which—and their relationships to the formants for the preceding a vowel. As I am sure you recall, the formants themselves are bands of harmonics with greater amplitude separated by damped regions of the audio spectrum; the frequency of a harmonic at the visually estimated center of the formant is taken to represent the frequency of the formant. These ‘center frequencies’ vary from speaker to speaker, and from one utterance to another with a given speaker, the ranges of the formant values for any one vowel intersect those for adjacent vowels, and those for the centralized vowel sounds of English especially intersect with one and another and with their neighbors. (Examples of centralized vowels: cup, butter, cigarette, random, etc.). What is constant is the configuration or relationships, whatever the absolute pitches. Various kinds of investigations have ascertained that the height of the first formant (graphed on a log or Mel scale; the Mel scale correlates physically measured frequency to perceived frequency perceptions, see the addendum at the end of this post) correlates to how open or closed the narrowest aperture in the oral cavity is (high=open), and the height of the second formant correlates with how far front or back the narrowest opening is (high=front). These variables define the auditory space and the correlated articulatory space within which vowels can be produced and perceived. The ‘apical’ or ‘quantal’ vowels at the extremes of the auditory space and of the correlated articulatory space so defined are

  • i (seed) F1 at its lowest, F2 at its highest, tongue any higher produces zh.
  • u (food) F1 at its lowest, F2 at its lowest, tongue any higher produces gh.
  • a (baah!) F1 at its highest, F2 at a midpoint, opening the oral cavity wider, if you can, makes no difference; variation in F2 produces a front-back range for ‘ah’-like vowels, making the picture of the acoustic space mapped onto the articulatory space a vowel quadrilateral rather than a vowel triangle.

A ‘centralized’ vowel is somewhere in a rather ill-defined central region of the acoustic space and articulatory space relative to these three apices, or two apices and the “ah” open vowel boundary.

Values from some speech synthesis work:
ə F1 399 Hz F2 1438 Hz (a centralized vowel)
a F1 708 Hz F1 1517 Hz
ɑ F1 703 Hz F2 1074 Hz

Bill Labov measured formant frequencies at the peak of the first formant during the transition from a to u o and from a to i for two productions, as shown in Bill’s Figure 2, below


The first (on the left) has a more open first vowel in the diphthong of ride, and the second (on the right) has a more centralized vowel. These are not from two speakers representative of the two populations; “a North Tisbury fisherman” produced both within a single sentence (p. 290, an example of stress as a phonetic condition for centralization). Nevertheless, generalization is possible:

“Despite the differences in· vowel placement, these seven speakers utilize the same dimension to produce the effect of centralized or open vowels: widely separated formants for centralized vowels, adjacent formants for open vowels.”
Labov (1963:288)

This “dimension” involving the relationship or configuration of two formants is an example of collectively controlled perceptual variables that constitute language. The difference between the configuration (or relationship) on the left and that on the right is the collectively controlled variable that is to be modeled at this lowest level of a model of the phenomenon of “a socially motivated sound change”.

Mel scale:

The responses of human listeners to even “ simple” nonspeech stimuli like sinusoidal signals is not simple. Psychoacoustic “scaling” experiments show that judgements of the relative pitch of two sinusoids are not equivalent to their arithmetic frequency ratio (Beranek, 1949; Fant, 1973; Nearey, 1976,1978). In other words, if you let a human listener hear a sinusoid whose frequency is 1000 Hz and then let him adjust the control of a frequency generator until he hears a sound that has twice the pitch of the 1000 Hz signal, he will not set the control to 2000 Hz. He will instead select a sinusoid whose frequency is about 3100 Hz. Judgement of relative perceived pitch can be related to the physical measure of frequency by the use of a “Mel” conversion scale. […] [T]he perceptual ratio between two frequencies depends on the absolute magnitude of the frequencies. […] A sinusoid whose frequency is 1000 Hz thus has a Mel value of 1000 Mel and is twice the pitch of a sinusoid having a pitch of 500 Mel. The frequency of a sinusoid having a pitch of 500 Mel is 400 Hz. A sound whose pitch is 3000 Mel will have twice the perceived pitch of 150C Mel but the frequency ratio of these two sounds is 9000/2000. The Mel scale is of particular value in regard to some of the acoustic relations that structure the phonetic theory of vowels…
(Lieberman & Blumstein 1988:154; a discussion of categorical perception begins on the next page).