Living Control Systems Publishing; dedicated to Perceptual Control Theory, a new conception of how all living things function

Search Living Control Systems


pic of Mindread

Mind Readings

Please have a look at the content of this book
as if you were flipping pages in a store.

    About Rick Marken

     
         
Part Paper #

Contents

Page  
         
    Foreword, William T. Powers vii  
    Introduction 1  
Part I Purposeful Behavior    
1 The Nature of Behavior:
Control as Fact and Theory
11  
Part II Mind Reading    
  2 Intentional and Accidental Behavior:
A Control Theory Analysis
35  
  3 Behavior in the First Degree 41  
Part III The Causal Circle    
  4 The Cause of Control Movements
in a Tracking Task
61  
  5 Closed-Loop Behavior:
Human Performance as Control of Input
45  
Part IV Control of Consequences    
  6 Selection of Consequences:
Adaptive Behavior from Random Reinforcement
79  
  7 Random-Walk Chemotaxis:
Trial and Error as a Control Process
87  
Part V Hierarchical Control    
  8 Levels of Intention in Behavior 109  
  9 Spreadsheet Analysis of a
Hierarchical Control System Model of Behavior
133  
Part VI Coordination    
  10 Perceptual Organization of Behavior:
A Hierarchical Control Model of Coordinated Action
159  
  11 Degrees of Freedom in Behavior 185  
Part VII Applications    
  12 Human Factors and Human Nature:
Is Psychological Theory Really Necessary?
207  
           
   

Foreword, William T. Powers

 
 

 

 
 

This is a book that can show a willing psychologist how to do a new kind of research. The theme that runs through all these papers is modeling, the ultimate way of finding out what a theory really means. Richard Marken is a skilled modeler, as will be seen. But he has a talent that goes beyond putting ideas into the form of working simulations, a talent that can be admired but is hard to imitate. He finds the essence of a problem and an elegantly simple way to cast it in the form of a demonstration or an experiment.

Sometimes the demonstration is so elegant and so simple that it slips past the reader without being noticed (more to the point, it slips past journal referees without being recognized as a fundamental contribution). Nearly every model in these papers, which did make it past the referees, is the sort that ought to convey to a reader a straightforward message: if the phenomenon you see here really works as this model shows it to work, then a whole segment of the scientific literature needs to be deposited in the wastebasket.

That is a message that ought at least to be discussed. To appreciate it, however, one has to understand what Marken's models are doing in those papers where models are compared against human behavior. If you pay attention, you will see that they are all dose imitations of a human being doing a task. In effect, the model, a computer simulation, is placed into the same experimental situation that a human participant experiences, and it behaves by producing the same kind of action the human being produces. The disturbances that are applied are just as much a surprise to the simulated person as they are to the real person.

The computer necessarily senses the situation differently than the way the person does, and produces its actions by different means. But the organization of the computer program, which represents a guess as to how the human being is internally organized, is sufficient to create behaviors that not only "resemble" the human behavior, but quantitatively reproduce it. That quantitative reproduction is the main point to which you should pay close attention.

The plotted behaviors of the simulations are not produced by curve-fitting. They are generated in real time in exactly the way the record of the human participant's behavior is generated, one point at a time. Every point on these plots is a test of the model. What is reproduced is not just a trend, or an average, or some abstract characteristic of the whole data set. Instead, the details of unfolding behavior are recreated independently, on the basis of the continuous functioning of assumed internal processes. The model can be run after the real behavior, to match its characteristics to those of the real system, or before, to predict new behavior under new conditions.

This, then, is truly the ultimate test of an idea about how any kind of behavior is created. The idea is cast in the form of a "generative model," one that will behave as it must behave according to the organization it has been given. Once the model is set up as a network of related processes, the theoretician is committed: whatever that model then does in the experimental situation is the meaning of one's idea about the organization. It then no longer matters what one thought an organization of that kind would do: it is evident what that kind of organization actually does.

The literature of psychology is full of organizational diagrams. In most cases, these diagrams represent ideas about how some behaviors are caused by internal processes or external interactions. But in most cases, you will not find any test of the model inherent in these diagrams. The diagrams are descriptions of how their authors think the real systems are organized; they are not working models. It wouldn't be possible, on the basis of the definitions given, to cast the models in the form of devices or simulations, turn them on, and see what they do. It's simply assumed that the depicted functions and relationships would do the same things that the real organism does, because that's how it seems to the author.

This assumption is almost always wrong. In the first place, to repeat, such putative models are seldom defined in enough detail to turn them into an actual working simulation. But even more important, even the simplest diagrams with only a few boxes and arrows carry implications that their own inventors can't possibly forsee. The only way to discover those implications is to construct an example of the diagram with specific numbers and functional forms to make it work, turn it on, and watch it run by itself. Any experienced modeler will tell you that the result will be a surprise—usually an unpleasant surprise.

All working models do something. What is hoped, of course, is that their behavior will be very close to some specific behavior of a real person. But before that hope can even be tested, the model itself must be specified completely enough to behave. That behavior shows what the consequences of your assumptions are. Only when those consequences have been generated and examined can any comparison with real behavior be made. This is the step that is missing from ordinary behavioral research. Ordinarily, the creation of a conceptual model ends the process. In truth, however, it is only the beginning. Only when the conceptual model is turned into a working simulation can its behavior be compared against the real behavior. Only through that comparison can the differences between the modeled behavior and the real behavior be seen. And only when those differences—which can be alarmingly large—are seen can one go back and refine the model to make it behave more correctly.

Modeling is therefore an iterative process through which one's assumptions can be systematically changed until the behavior of the model—the actual behavior, not the imagined behavior—is as dose as possible to the real behavior. That is the method of modeling, and the secret of all the successful physical sciences. Where it has been tried in the life sciences, it has been equally successful. There is every reason to think that it should be the central method of any enterprise that lays claim to the name of science.

Most of the papers you will find here are lessons in the use of the method of modeling. If you look at them that way, you will learn more than just some facts about behavioral organization. You will learn how to raise psychological research to a new level of competence. Even though the method of modeling might take longer than turning a statistical crank, the result in the end will be a new experience for most experimental psychologists: models that work with great precision, and that teach you things you didn't already know. You will develop a new eye for nonsense and obfuscation: once you have created a model that actually works and matches real behavior, you will see just how little some theories in psychology have to do with reality.

This little collection of papers will someday be required reading in any course in psychology. You are fortunate to be able to begin reading them now.

William T. Powers
Durango, Colorado
April 1992

 
   

Introduction

 
 

 

 
 

The papers collected in this book are the result of a decade of research on the control-theory model of purposeful behavior developed by William T. Powers (1973). I decided to gather them together in a single volume for several reasons. First, I felt that a collection of papers describing experimental tests and demonstrations of control theory would be a useful supplement to existing theoretical (Powers, 1989) and textbook (Robertson and Powers, 1990) treatments of the subject. I also felt that my published research covered a broad enough range of topics to make a book like this feasible. Finally, and on a personal note, the publication of this collection marks the end of an era in which my research focused largely on what is wrong with current theories of behavior and the beginning of an era in which my research will focus almost exclusively on what is right with control theory.

Purposeful Behavior

I used to think that it is a scientist's job to show what is wrong with one theory before proposing a new one to replace it. But I have learned that things are not so simple with control theory. The problem is that control theory is not really a replacement for any existing theory of behavior. Rather, it is an explanation of a phenomenon that is not even recognized by current theories of behavior—the phenomenon of control. Control is the process of producing consistent results in the face of unpredictable disturbances. Control can be as simple as keeping your car in its lane on a windy day or as complex as


 
   

The Nature of Behavior:
Control as Fact and Theory

 
 

 

 
 

Failure to understand the nature of control has hindered efforts to apply control theory in the behavioral sciences. Control theory was developed to explain the phenomenon of control, which involves the production of consistent results in the face of environmental disturbance. Comparison of a quantitative description of the phenomenon of control with a physical analysis of behavior shows that the events referred to as behavior always involve control. The appearance of behavior as programmed output or response to stimulation is shown to be a consequence of ignoring two fundamental characteristics of control—reference states and disturbance resistance. The recognition of behavior as control requires a new approach to psychological research that emphasizes the discovery of controlled variables. This paper deals with living and non-living systems; the emphasis is on living systems at the individual (organism) level.

What is behavior? The answer seems so simple that the question is rarely asked. Behavior is what organisms do—things like walking, talking, and playing chess. Scientific psychology is built on the assumption that behavior is output—the last step in a causal chain that begins in the environment or the brain. This is an axiom of psychological research; a fact beyond question. Nevertheless, there is an alternative. Powers (1973) has argued that behavior is not output but a controlled consequence of output: behavior is control.

Control is a real, objective phenomenon that involves the production of consistent results under varying environmental conditions. Control theory was developed to explain how control occurs (Black, 1934; Buckley, 1968; Jones, 1973, Maxwell, 1868; Wiener, 1948). While behavioral scientists have

Reprinted with permission of publisher from: Marken, R. S. The Nature of Behavior: Control as Fact and Theory. Behavioral Science, 1988, 33,196-206.


 
     
   

Intentional and Accidental Behavior:
A Control Theory Analysis

 
 

 

 
 

Summary.—A procedure based on control theory was used to determine which of two simultaneous activities was the one performed intentionally by a subject. The procedure provides an objective means of discriminating intended from accidental behaviors.

Psychology, which bills itself as the study of behavior, has yet to provide a universally accepted definition of its subject matter. The term "behavior" typically refers to some observable result of an organism's actions, such as a "lever press." But actions produce many observable results, any one of which could be considered the organism's behavior (Powers, 1973, p. 50). The actions which produce a lever press also move a limb, close an electrical circuit, move air molecules near the lever, and produce a food pellet. Which results should count as behaviors of the organism? Some have argued that only intentionally produced results should count as behavior, other results being accidental side effects of actions (Powers, 1973; Searle, 1981). This approach to defining behavior is rejected by many psychologists who consider intentions both unnecessary and unobservable (Schwartz, 1978). This report shows how intentions can be observed and why the concept of intention is necessary in order to know what an organism is doing.

The problem of defining behavior without reference to intention is illustrated in a simple demonstration. A subject is seated in front of a video monitor showing two vertical lines,


Reprinted with permission of publisher from: Marken, R. "Intentional and accidental behavior: a control theory analysis." Psychological Reports, 1982, 50, 647-650.


 
     
   

Behavior in the First Degree

 
 

 

 
 

Control theory has been around for some time now, the basic equations having been worked out in the 1920s by H. S. Black (Waldhauer, 1982). Yet the application of control theory in the behavioral sciences is still considered novel and, to some extent, revolutionary (Powers, 1978). What is revolutionary about control theory is not so much its content as its subject matter. The subject matter of conventional psychological theories is behavior; the subject matter of control theory is control (Marken, 1988).

The difference between behavior and control is obscured by the fact that both phenomena are typically referred to by the same name—"behavior." But these phenomena are not the same and the difference is significant. Behavior, as the term is typically used in scientific psychology, refers to any observable result of an organism's muscle actions—lever presses, rating responses, and verbal reports are familiar examples. Control, on the other hand, refers only to intended results of an organism's actions. Control is behavior in the first degree.

In order to properly apply the theory of control we must be able to tell the difference between control and other kinds of behavior. This amounts to distinguishing intentional from unintentional (accidental) behavior. Many psychologists have tacitly assumed that the difference is obvious. Freud, for example, wrote about the significance of unintentional behavior, taking for granted that he knew an accident when he saw one


Reprinted with permission of publisher from: Marken, R S. "Behavior in the first degree." Volitional Action, W. A. Hershberger (Ed.), Elsevier Science Publishers B. V., 1989, 299-314.

 
     
   

The Cause of Control Movements in a Tracking Task

 
 

 

 
 

Summary.—The classical cause-effect or input-output model of behavior breaks down when there is feedback from response to stimulus. Using a compensatory tracking task it is shown that response variations on different occasions can be nearly identical while stimulus variations on these occasions are completely unrelated. This result seems to rule out stimulus variations as the cause of responses which control (stabilize) the stimulus. When feedback exists, the cause of control must be viewed as an internal reference rather than an external stimulus.

In compensatory tracking tasks a subject is asked to control a cursor, keeping it aligned with a stationary target. To accomplish this the subject must make responses (for example, vary the position of a handle) to compensate for disturbances of the cursor's position. Much of the research on this task concerns the effects of temporal characteristics of disturbances on the accuracy of control of movements (1). This paper addresses a different question, namely, "How is this control effected?" The conventional answer is that some aspect of the stimulus (such as the position or rate of change in position of the cursor) is transformed into responses (handle positions) which control the cursor, keeping it stabilized near the target (3, 4, 8). Powers (5, 6) has taken pains to explain that when there is feedback from response to stimulus, such that there is a closed loop of cause and effect, conventional explanations which treat stimulus as cause and response as effect are no longer appropriate.


Reprinted with permission of publisher from: Marken, R. "The cause of control movements in a tracking task" Perceptual and Motor Skills, 1980, 51, 755-758.

 
     
   

Closed-Loop Behavior:
Human Performance as Control of Input

 
 

 

 
 

For the past couple of years I have been developing demonstration-experiments to show that input-output models of behavior are inappropriate when there is feedback from output to input. My aim is to encourage a new approach to the study of living organisms: an approach based on viewing organisms as feedback control systems.

One experimental paradigm I have been using to demonstrate feedback phenomena is the compensatory tracking task. In this task, a subject is asked to hold a cursor aligned with a stationary target by varying the position of a handle to compensate for disturbances of the cursor's position. The task is similar to that of driving a car down a straight road on a windy day. The driver must turn the wheel to compensate for lateral wind forces, the disturbances which would otherwise push the car out of its lane. In these experiments, the disturbance is low pass filtered noise which is added to the position of the cursor. If the subject does nothing, the disturbance causes the cursor to move slowly and irregularly back and forth. By moving the handle appropriately, the subject can compensate for the disturbance and keep the cursor on target.

Conventional theories of behavior hold that tracking occurs because disturbance-created deviations of the cursor from the target are transformed, via processes within the subject, into handle movements which bring the cursor into alignment with the target. This is an input-output model of performance: input (deviation from the target) is identified as the cause of


Paper presented at Western Psychological Association Meeting, Los Angeles, 1981.


 
     
   

Selection of Consequences:
Adaptive Behavior from Random Reinforcement

 
 

 

 
 

Summary.—The behavior of subjects in a human operant conditioning experiment was "shaped" using a random reinforcement contingency. Bar-press responses kept a moving cursor near a target although the consequence of each response was a random change in the direction of the cursor. The apparent effect of reinforcement on behavior is shown to be an illusion created by ignoring the consistency of behavioral results.

Reinforcement theory holds that adaptive behavior is selected by its consequences. Shaping behavior by reinforcement is viewed as analogous to the evolution of species by natural selection (7). According to this view, the environment must provide the correct reinforcement contingencies if a particular behavior pattern is to occur. The behavior of a duckling following its mother, for example, should only occur if the duckling is reinforced for making movements which achieve this result. If the environment provides reinforcements randomly (such that all movements are reinforced with about equal probability), achievement of the result would be unlikely; the duckling would wander off into oblivion.

An experiment to test the effect of a random reinforcement contingency on behavior was suggested by studies of the behavioral systems of E. coli bacteria (1). These organisms navigate through chemical gradients although the consequence of steering movements is a completely random change in direction. The experiment described in this report is based on a


Reprinted with permission of publisher from: Marken, R. "Selection of consequences: adaptive behavior from random reinforcement." Psychological Reports, 1985, 56, 379-383.

 
     
   

[with William T. Powers]
Random-Walk Chemotaxis:
Trial and Error as a Control Process

 
 

 

 
 

The biased random-walk chemotaxis of the bacterium Escherichia coli is a remarkably effective method of navigation based on random trial-and-error responding rather than steering. Humans restricted to the same mode of responding are able to navigate to target locations, just like the bacterium. This mode of navigation can be modeled as an input control process that selectively retains favorable and rejects unfavorable consequences of the random responses. The selection process is determined by the internal organization of the system rather than the external influence of the environment (as in natural selection or reinforcement).

Control theory is commonly applied to human goal-seeking behavior in situations where behavior has moderately predictable influences on environmental processes, and those processes simultaneously have regular influences on behavior (Powers, 1973). A few years ago, however, we became interested in an apparent goal-seeking phenomenon that takes place through a highly irregular, in fact random, link (Marken, 1985). This phenomenon was described by Koshland (1980); it is the method that the bacterium Escherichia coli uses to make its way up concentration gradients of attractants and down gradients of repellents. We have simulated this behavior of E. coli using a control-system model and have extended the principle to experiments with human beings.

The control-theoretic analysis may improve understanding of the trial-and-error phase of learning, the phenomenon that Campbell (1960) called "blind variation and selective survival"


Copyright 1989 by the American Psychological Association. Reprinted by permission of the publisher. Marken, R. S., and Powers, W. T. Random-walk chemotaxis: Trial and error as a control process. Behavioral Neuroscience, 103, 1348-1355.

 
     
   

[with William T. Powers]
Levels of Intention in Behavior

 
 

 

 
 

The nervous system can be modeled as a hierarchy of control systems (Albus, 1981; Arbib, 1972; Powers, Clark, & McFarland, 1960a, 1960b). Hierarchical models are motivated, in part, by the hierarchical appearance of the behavior produced by the nervous system (Lashley, 1951; Miller, Galanter, & Pribram, 1960). Complex behavior can be seen as the result of simpler behaviors that are themselves the result of even simpler behaviors, this reduction stopping only at the level of muscle tensions. Thus, going to the store is done by driving a car which is done by turning a wheel, which is done, ultimately, by tensing muscles.

While hierarchical models can be devised that reproduce a given behavior (Pew, 1966; Rosenbaum, Kenny, & Derr, 1983; Marken, 1986), it is difficult to show that the same behavior could not equally well be produced by a single-level model that is no more complex (Klein, 1983). Nevertheless, the concept of hierarchical organization in behavior is accepted by most psychologists. This acceptance is based largely on persuasive descriptions of brain structure and plausible—but untested—diagrams of neural organization (Davis, 1976). Laying out an organizational scheme that could work is only the beginning. The model thus proposed must then be used to predict behavior quantitatively to show what the model would, in fad, do, instead of only asserting that it would behave in the proper way. This kind of modeling amounts to simulation of


Reprinted with permission of publisher from: Marken, R.S. & Powers, W.T. "Levels of Intention in Behavior." Volitional Action, W. A. Hershberger (Ed.), Elsevier Science Publishers B. V., 1989, 409-430.

 
     
   

Spreadsheet Analysis of a Hierarchical
Control System Model of Behavior

 
 

 

 
 

The behavior of a hierarchy of control systems can be simulated with an electronic spreadsheet. Each control system is a column of three cells representing the reference signal, perceptual signal, and output variable of the system. All of the control systems are closed-loop, the input to each system being a function of its output. The circular references in the spreadsheet are resolved through iterative recalculation. When the parameters of each control system (amplification and slowing factors) are set to appropriate values, all control systems in the hierarchy continue to match perceptual signals with reference signals. A three-level hierarchy with four systems at each level is simulated in the spreadsheet. The spreadsheet model makes it possible to observe the dynamic behavior of the control systems as they correct for the effects of environmental disturbance and changes in higher-order reference signals. It is possible to "reorganize" the system by changing the perceptions controlled by systems at different levels of the hierarchy. The user can also test to determine the variables being controlled by the system.

Powers (1973, 1989) has proposed a hierarchical control system model of the purposive behavior of organisms. In this paper a method of implementing the model on an electronic spreadsheet will be described. The spreadsheet system makes it relatively easy to explore the model's behavior with a personal computer. A spreadsheet implementation was selected because many (perhaps most) personal computer users have access to spreadsheet software and are familiar with its use. Moreover, the matrix design of the spreadsheet provides an excellent format for representing a control system hierarchy. The spreadsheet model is designed to be a self-instructional


Reprinted with permission of publisher from: Marken, R. S. "Spreadsheet analysis of a hierarchical control system model of behavior." Behavior Research Methods, Instruments & Computers, 22, 349-359.

 
     
   

Perceptual Organization of Behavior:
A Hierarchical Control Model of Coordinated Action

 
 

 

 
 

The behavior of individual subjects is compared with a hierarchical control system model of behavioral organization. Subjects varied the position of two control handles simultaneously to keep the distance constant between two pairs of lines. Three variations on this basic experiment that illustrate some fundamental properties of coordinated action are shown: first, how independent actions, compensating for unpredictable and undetectable disturbances, can produce a single behavioral result; second, how the ability to produce a particular result is maintained when the connection between action and result is changed; and third, how two independent outputs can appear to be related as coordinative structures when one output disturbs a result being controlled by the other. The correlation between the behavior of subjects and model in all experiments is typically on the order of .99. A detailed examination of the operation of the model shows that actions are structured by perception, not by central commands or equations of constraint.

Behavior of any significance (such as driving, talking, and playing piano sonatas) requires coordination of several motor outputs in an unpredictably changing environment. Current models of behavior place the burden for coordination of multiple outputs on motor programs that purportedly transform a central command, like "make a left turn," into the outputs that produce the intended result (Keele, 1968; Kelso, 1977; Pew, 1974; Schmidt, 1976, 1980). A great deal of effort has also been devoted to the problem of how a system might be designed to regulate, simultaneously, the many outputs (degrees of freedom) required to perform even the simplest behavior (Bern-


Copyright 1986 by the American Psychological Association. Reprinted by permission of the publisher. Marken, R. S. Perceptual organization of behavior A hierarchical control model of coordinated action. Journal of Experimental Psychology: Human Perception and Performance, 12, 267-276.

 
     
   

Degrees of Freedom in Behavior

 
 

 

 
 

Abstract—Coordinated behavior is typically explained in terms of motor programs or coordinative structures. The experiments described in this report suggest that a better explanation may be provided by control theory. In all experiments, subjects controlled the two-dimensional position of a cursor. By varying the disturbances to the two degrees of freedom of cursor movement, it was possible to elicit task-specific, unidimensional movement patterns resembling coordinative structures. Perturbation of one dimension of the movement pattern failed to produce a concomitant response in the other dimension unless there was a link between the degrees of freedom of action and perception. The subjects' behavior was accurately simulated by the behavior of independent feedback control systems. The results show that the problem of coordinating many degrees of freedom of action can be solved by viewing behavior as control of the perceptual consequences of these actions.

Consistent behaviors are produced by adjusting many dimensions of action simultaneously. Even a simple behavior, such as lifting a glass of water, is accomplished by adjusting the tension in many muscles of the hand, arm, and back, each muscle representing an independently adjustable degree of freedom. Researchers have wondered how the nervous system computes the appropriate values for so many degrees of freedom simultaneously to produce consistent behavioral results (Turvey, Fitch, & Tuner, 1982; Kelso and Tuner, 1980. This has been called the degrees of freedom problem. The problem is par-


Copyright 1991 by Cambridge University Press. Reprinted with permission front Marken, R. S. Degrees of Freedom in Behavior. Psychological Science, 2, 92-100.

 
     
   

Human Factors and Human Nature:
Is Psychological Theory Really Necessary?

 
 

 

 
 

Psychologists try to understand how people work. Human factors engineers try to design systems that help people work better. As a psychologist working as a human factors engineer I sometimes wonder whether these two enterprises are actually related. What do we really need to know about human nature in order to design the systems that humans use? It seems that much of the best work in human factors is done without any reliance on psychological theory at all. In fact, there is good reason to believe that the basic model of human nature accepted in human factors engineering is wrong (Powers, 1978). Like celestial navigation prior to Copernicus, human factors engineering works in spite of its theories. I will argue that it could do even better with the right theory.

The Systems Approach

There are really two different approaches to human factors engineering—one focuses on systems, the other on people. The great successes in human factors have come from the systems approach, which requires little in the way of psychological theory. The systems approach succeeds by focusing on the person as a component of a human-machine system rather than as a system in his or her own right. The goal is to maximize system output in terms of productivity, safety, and the like. Thus, the typist-keyboard system is designed to produce the the most pages of correctly typed copy; the operator-control


Reprinted with permission from Human Factors Society Bulletin. Copyright 1986 by The Human Factors Society, Inc. All rights reserved.

 
     
           


© 2004–2017 Living Control Systems Publishing
Dag Forssell, Webmaster
dag at livingcontrolsystems dot com