LCS_I .INF 

Book information:

LIVING CONTROL SYSTEMS vol I

Selected Papers of William T. Powers

Contents, Foreword and Preface:

CONTENTS

Foreword, Richard S. Marken vii

Preface, William T. Powers xiii

A Note on the Text, Gregory Williams xix

A General Feedback Theory of Human Behavior Part I 1

A General Feedback Theory of Human Behavior Part II 25

A Feedback Model for Behavior

Application to a Rat Experiment 7

Feedback: Beyond Behaviorism 61

Behaviorism and Feedback Control 79

Applied Epistemology 89

The Cybernetic Revolution in Psychology 103

Quantitative Analysis of Purposive Systems:

Some Spadework at the Foundations

of Scientific Psychology 129

A Cybernetic Model for Research in Human Development 167

Degrees of Freedom in Social Interactions 221

On Purpose 237

Control Theory and Cybernetics 245

The Asymmetry of Control 251

An Outline of Control Theory 253

Published Works by William T. Powers

on Living Control Systems, 1957-1988 295

FOREWORD

Some of the best science is done by people who refuse to take the obvious for granted. Copernicus didn't take the sun's daily trek across the sky for granted, Einstein didn't take the regular tick of time for granted, and William T. Powers didn't take the appearance of behavior for granted. The results of not taking things for granted can be powerful new ways of looking at the obvious, but the value of the new point of view is rarely appreciated immediately. Gregor Mendel who didn't take the blending of hereditary traits for granted, approached the study of heredity in a completely new way, using a combination of botany and mathematics. For his creativity and hard work, having single-handedly invented the field of genetics, he was rewarded with complete and utter neglect. His work was finally recognized 30 years after his death (small comfort to Gregor) by three scientists who independently rediscovered his laws.

Mendel's story illustrates a rule of scientific discovery that is too often followed: "... look at a problem from a totally new angle and people won't so much disagree with you as completely misunderstand you. They won't grasp what you are talking about and will ignore you." (Maitland A. Edey and Donald C. Johanson, _Blueprints_, Little, Brown and Co., Boston, Toronto, and London, 1989, p. 105) Powers has looked at the phenomenon of behavior from a totally new angle and, sure enough, people have misunderstood him and ignored him, but they have rarely disagreed with him The lack of disagreement is rather surprising, since Powers' ideas about behavior contradict the fundamental assumptions of scientific psychology. Conventional psychology views behavior as evoked motor output; Powers argues that behavior is controlled perceptual input. These approaches to behavior could hardly be more different.

Powers developed his ideas about control after taking a close look at a phenomenon that most psychologists have taken for granted --behavior. The conventional wisdom in psychology is that behavior is what organisms do. This seems obvious. Behavior appears to flow out of organisms like water from a spigot or printout from a computer. It is difficult to believe that this view of behavior could be wrong, but it is. Powers saw that behavior is not produced solely by the organism, but by the organism in concert with unpredictable and usually undetectable environmental disturbances. These disturbances are pervasive but difficult to notice because behavior is ordinarily quite consistent. Organisms weave webs, migrate to specific destinations, build dams--and they do these things over and over again. Powers, looking at behavior through the eyes of a trained physicist and engineer, saw that such consistency was quite surprising. He realized that organisms can produce consistent results (a web, a landing in San Juan Capistrano, a dam) only by continually adjusting their actions to compensate for disturbances. It is as though they intend to produce these results and vary their actions appropriately in order to do so. Organisms seem to behave on purpose.

Psychologists before Powers had noticed the purposiveness of behavior. They saw, for example, that organisms produce consistent results using highly variable actions. But most psychologists ended up attributing this variability to "statistical noise"; Powers, on the other hand, saw it as essential. If actions did not vary, behavioral results would repeat only by chance, fluctuating as a result of the random effects of environmental disturbances. Instead, actions vary to compensate for the effects of disturbance, producing consistent results in an inconsistent world--a process called "control." Powers realized that the events we call behavior, from lever presses to religious ceremonies, are controlled results of action; to behave is to control. This was a momentous observation, and it needed an explanation. How were organisms able to control? Fortunately, the basic answer had already been discovered--it was control theory.

Powers is not the inventor of control theory. Nor is he the first to have applied it to behavior. He is, however, the first to have used control theory to explain behavior as an example of the phenomenon that control theory was designed to explain --control. Previous attempts to apply control theory to behavior put the cart before the horse, so to speak. People were more familiar with the theory than with the phenomenon of control. Thus, control theory was applied to behavior before anyone had any idea that behavior involved control. This was a bit like trying to develop a theory of evolution before there was any evidence that evolution had occurred. It was bound to lead to confusion and disappointment. Before Powers came along, control theory was already on the wane as a model for behavior. And for good reason. Control theory is the wrong model of behavior if behavior is evoked motor output. But it is the right model of behavior if behavior is control.

Powers built a model of behavior based on control theory. The basic tenet of the model is that organisms control perceptual input, not motor output. This is a fact of control system operation. Control systems act to keep their perceptions matching reference images of what those perceptions should be. They do this by acting on the environment, producing effects which, when combined with prevailing environmental disturbances, produce the desired perceptions. Living control systems are no different than any other control system in this respect. When we watch the behavior of organisms, we are watching living control systems "from the outside"--systems that are controlling their own perceptual experience. Behavior is, as Powers put it in the title of his classic book on the subject, "the control of perception."

To understand the behavior of a living control system, the observer must learn what perceptions the system is controlling: what reference images the system is trying to match. Living control systems produce many results, some of which may be controlled and others not. The observer must learn which results correspond to the perceptual variables that the system is actually controlling. These results are called controlled variables. Powers has described an objective method, called "the test for the controlled variable" (or simply "the test"), for finding out what variables a control system is controlling. When applied to living control systems, "the test" constitutes a new research methodology for psychology, with a new goal--the discovery of controlled variables.

Powers understood that the variables controlled by living control systems can be quite complex, as evidenced by the complex behaviors produced by organisms--behaviors like building skyscrapers and writing piano concertos. As a control engineer, Powers knew that a control system could be designed to do anything that it could perceive. Thus, a control system could produce complex behaviors if it could perceive complex variables. Powers showed how a hierarchy of systems controlling different classes of perceptual variables could produce the kind of complex behavioral results produced by living organisms. The model has a satisfying consistency with what we know of the functional and structural organization of the nervous system. Whether or not it proves to be completely correct (it has survived numerous tests, but there are a great many more to be done), it has served its purpose by showing that all behavior, from tensing muscles to writing poems, can be modeled by control theory.

A number of scientists, impressed by the power and beauty of control theory as applied to behavior, have devoted their research efforts to testing and expanding Powers' ideas on living control systems. Obviously, I am one of them. I knew after reading Behavior: The Control of Perception that Powers had something very important to say; I just wasn't sure what it was. It isn't easy to understand control theory at first reading, especially if you are thoroughly imbued with the concepts of conventional behavioral science, as I was. It takes a while to understand that control systems compensate for disturbances rather than respond to stimuli; that stimuli are controlled and not in control; that living control systems control and cannot be controlled.

In order to get a grasp of the control model, I sought out other works by Powers. I was able to find them, but it wasn't always easy. Now it is. You no longer have to be a fanatic to obtain Powers' finest publications: they are gathered together in this book. You can learn a great deal about living control systems by reading this book. But don't expect to find the answers to all your questions about life. Control theory provides a new foundation for the study of living systems, but it is just the foundation--it is not the edifice. A great deal of work must be done to build on this foundation, but construction can progress with confidence because the foundation is solid. Behavior is the control of perception. Understand that, and you understand the basic organizing principle of living systems. The rest, as Einstein said, are details.

Richard S. Marken

Los Angeles

July 1989

PREFACE

For uncomfortably close to 30 years I have been writing an article called "Control Theory for the Life Sciences." I have published this article in books and journals, newsletters and proceedings. It has been aimed at behaviorists, cyberneticians, linguists, biologists, social scientists, and anyone else I thought had a glimmer of interest in the subject. I have spoken this article to seminars of graduate students in several disciplines, to medical students, and to faculty members, in classrooms, lecture halls, and brown-bag lunchrooms. You will see the article here, in most of its incarnations. This is getting extremely tiresome, not only for those who have heard the message too many times, but for the one who has heard it the most often of all: me.

This persistence was not in the original plan, which was to communicate the basic theory worked out by R.K. Clark, R.L. McFarland, and myself in the 1950s and published in 1960, then to find a place to work and develop the basic ideas into a full-blown discipline. The theory was so elegant, so close to being self-evident, so clearly useful and explanatory, that neither I nor my collaborators anticipated any problems in gaining support for it. We knew that others were moving in the same general direction and thought they would welcome real signs of progress. I look back now and wonder how we could have predicted the future so poorly.

It is now clear that a new theory is quite welcome in the sciences of life, but only if _it does not call for revision of important beliefs_. It's all right, for example, to propose a theory saying that an organism's susceptibility to reinforcement by food might be modified by the use of water deprivation. It is not all right to propose a theory that says, in effect, that there is no such thing as reinforcement. There are simply too many scientists who rely on the concept of reinforcement as their main explanation of behavior, the foundation of their theorizing; take away that tool and they have nothing left. The same thing would happen to control theorists if the principles of control were shown to be spurious. Every science needs explanatory principles on which it can rely; if the principles of a science were reorganized yearly, no organized concepts could ever develop. I was naive to think that control theory could become influential in the life sciences over a period of a year or two, or even a decade or two. No idea that can change the course of a science that easily could be anything but a fad. A science cannot change its system concepts overnight, for precisely the same reason that an individual can't do the same thing.

A system concept is an attitude, an understanding, a world view. It's a sense of orderliness and coherence that we see in a body of principles and generalizations. It lives in an individual. It not only forms out of coalescing principles, but it determines which principles belong in the system and which do not. The process is one of assimilation and accommodation, simultaneous mutual adjustment between levels.

Acceptance of control theory requires a change in the beliefs of life scientists at the level of system concepts. System concepts bring order into principles; principles bring order into methods; methods bring order into symbolic representations; symbolic representations bring order into all lower levels of observation. Reorganizing a system concept therefore requires reorganizing everything else. The very way the world looks to us changes when a system concept changes. In fact, the system concept cannot change first. The whole system must reorganize at once. Newcomers to control theory do not all learn it the same way. One part of it is immediately clear to some, other parts to others. What we understand in one area of knowledge causes problems with what we thought we understood in other areas. Even in a willing individual, this reorganization can't take place overnight. It requires years. It requires changes at levels where we all find voluntary change difficult, mysterious, or even impossible.

In the life sciences, there is a widely-accepted system concept of what an organism is. When a scientist speaks from the viewpoint of this system concept, we can recognize it easily, although it's not easy to put into words. The words and descriptions we can find are only signposts pointing to the system concept. There's a dispassionate aspect to it, a distancing. There's an avoidance of empathy. There's a kind of sternness, an overcoming of natural sympathies, a pride in being immune to the weakness and sentimentality of the layman's view. There's a picture of an organism as a natural object, a bag of chemicals, a preparation of irritable tissue. The word most often used to symbolize this complex structure of attitudes is "objectivity." But objectivity is only evidence of the system concept: the system concept itself is a point of view from which all the rest, from principles on down, hang together and make sense. The system concept is the understanding of living systems that makes objectivity seem appropriate.

The control theory that you will find in this book is a collection of principles, methods, symbol systems, relationships, and observations of more detailed kinds. If I had it to do over again (and if I were a different and smarter person with a better education and a different way of growing up, and understood what I understand now), I would not persist so long in arguing at these levels. I would spend much more time trying to understand and express the difference in system concepts that separates control theory from all conventional theories. Control theory was never the only ingredient in this alternate view of organisms; I only made it seem that way. Even calling it "control theory" is an example of synecdoche, in the sense of referring to something by naming only one of its attributes.

Another important ingredient of this system concept is a view of what constitutes understanding of a system. I do not accept that statistical studies of behavior ever give us understanding of human behavior or human functioning. All they do is overwhelm us with random and unreliable facts. To understand a system, we must be able to see that it _must_, because of its inner nature, behave as we see it behaving. Its properties must grow out of its inner organization; its behavior must arise from its properties.

This principle is connected to a principle of explanation (or so, from my system concept, it seems). The only kind of explanation I can believe is one that demonstrates the principles that are proposed. It's not enough to say that a block diagram represents a system. One has to show that, in fact, a system organized in that way must behave in a particular way. If you can't deduce how a system would work from the explanation, then you don't have an explanation. The best way to prove that the explanation actually explains something is to cast it as a working simulation, turn it on, and let it operate by the rules you have put in it. If you can't do that, then you don't have a model or an explanation. All you have is more or less persuasive rhetoric.

I did not come down from a mountain with these principles tucked under my arm. They grew out of my training and my occupations, out of my successes and failures, out of my listening to others who were trying to solve the same problems. As the principles changed shape, the system concepts changed shape; as the system concepts changed shape, the principles became clearer. I still don't know how to express very clearly the system concept that contains control theory; now that I know this is needed, I will begin trying to do so. But I have learned that I don't have to provide understanding for others at this level. At best I can make it a little easier. But I don't really have to do it at all.

The reason I don't have to teach system concepts (aside from the fact that I can't) is that people can learn them on their own. The principles and methods of control theory can be taught; once a person grasps them beyond a certain point, they teach themselves. But those principles, everyone discovers, have implications that clash with the rest of one's knowledge of human behavior. Once one has understood an explanation of any one behavior from the standpoint of control theory, other explanations suddenly look different--more evasive and rhetorical, more conjectural, even wrong.

Understanding control theory just as a collection of logical and mathematical manipulations or as a diagram of relationships is relatively easy; those things can be taught to 30 people at once. You can learn those things from this book, reading it from either end. But grasping all the implications of control theory at the higher levels of understanding is hard, and can be done only within one individual at a time. Every control theorist I know has put a great deal at risk in the process of learning this subject. Most of them have suffered the embarrassment of seeing a former belief as foolish or naive, of realizing that they had been uncritical or even gullible. I know of none, however, who would go back to what they believed before, although there is nothing to prevent their doing so--and every encouragement from their colleagues to do just that.

Members of The Control Systems Group conceived the publication of this collection, organized it, and labored to make it real. I am profoundly grateful to them. Seeing all these papers brought together has, unexpectedly, shown me that a phase of my life is over. This book and the fact that I had so little to do with creating it have convinced me that I can stop writing that article over and over. The basic ideas are in good hands; I can let go of them now. Now, perhaps, I can try to remember what was supposed to come next.

William T. Powers

Northbrook, Illinois

July 1989