Every science, like every person, has a duty toward its neighbors, not perhaps to love them as itself, but still to lend them its tools, to borrow tools from them, and, generally, to keep the neighboring sciences straight. We may perhaps judge of the importance of an advance in any one science in terms of the changes which this advance compels the neigh-boring sciences to make in their methods and in their thinking. But always there is the rule of parsimony. The changes which we in the behavioral sciences may ask for in genetics, or in philosophy, or in information theory must always be minimal. The unity of science as a whole is achieved by this system of minimal demands imposed by each science upon its neighbors, and—not a little—by the lending of conceptual tools and patterns which occurs among the various sciences.
My purpose, therefore, in the present lecture is not so much to discuss the particular theory of schizophrenia which we have been developing at Palo Alto. Rather, I want to indicate to you that this theory and others like it have impact upon ideas about the very nature of explanation. I have used the title "Minimal Requirements for a Theory of Schizophrenia," and what I had in mind in choosing this title was a discussion of the implications of the double bind theory for the wider field of behavioral science and even, beyond that, its effect upon evolutionary theory and biological epistemology. What minimal changes does this theory demand in related sciences?
I want to deal with questions about the impact of an experiential theory of schizophrenia upon that triad of related sciences, learning theory, genetics, and evolution.
The hypothesis may first be briefly described. In its essentials, the idea appeals only to everyday experience, and elementary common sense. The first proposition from which the hypothesis is derived is that learning occurs always in some context which has formal characteristics. You may think, if you will, of the formal characteristics of an instrumental avoidance sequence, or of the formal characteristics of a Pavlovian experiment. To learn to lift a paw in a Pavlovian context is different from learning the same action in a context of instrumental reward.
Further, the hypothesis depends upon the idea that this structured context also occurs within a wider context—a metacontext if you will—and that this sequence of contexts is an open, and conceivably infinite, series.
The hypothesis also assumes that what occurs within the narrow context (e.g., instrumental avoidance) will be affected by the wider context within which this smaller one has its being. There may be incongruence or conflict between context and metacontext. A context of Pavlovian learning may, for example, be set within a
* Second Annual Albert D. Lasker Memorial Lecture, delivered at the Institute for Psychosomatic and Psychiatric Research and Training of the Michael Reese Hospital, Chicago, April 7, 1959. This lecture is here reprinted by permission of the A.M.A. Archives of General Psychiatry where it appeared in 1960, Vol. 2, pp. 477-491.
metacontext which would punish learning of this kind, perhaps by insisting upon insight. The organism is then faced with the dilemma either of being wrong in the primary context or of being right for the wrong reasons or in a wrong way. This is the so-called double bind. We are investigating the hypothesis that schizophrenic communication is learned and be-comes habitual as a result of continual traumata of this kind.
But even these "common-sense" assumptions break away from the classical rules of scientific epistemology. We have learned from the paradigm of the freely falling body —and from many similar paradigms in many other sciences —to approach scientific problems in a peculiar way: the problems are to be simplified by ignoring — or postponing consideration of—the possibility that the larger context may influence the smaller. Our hypothesis runs counter to this rule, and is focused precisely upon the determining relations between larger and smaller contexts.
Even more shocking is the fact that our hypothesis suggests —but does not stand or fall with the suggestion—that there may be an infinite regress of such relevant contexts.
In all of this, the hypothesis requires and reinforces that revision in scientific thought which has been occurring in many fields, from physics to biology. The observer must be included within the focus of observation, and what can be studied is always a relationship or an infinite regress of relationships. Never a "thing."
An example will make clear the relevance of the larger contexts. Let us consider the larger context within which a learning experiment might be conducted using a schizophrenic as a subject. The schizophrenic is what is called a patient, vis-a-vis a member of a superior and unloved organization, the hospital staff. If the patient were a good pragmatic Newtonian, he would be able to say to himself: "The cigarettes which I can get by doing what this fellow expects me to do are after all only cigarettes, and as an applied scientist I will go ahead and do what he wants me to do. I will solve the experimental problem and obtain the cigarettes." But human beings, and especially schizophrenics, do not always see the matter this way. They are affected by the circumstance that the experiment is being conducted by somebody whom they would rather not please. They may even feel that there would be a certain shamelessness about seeking to please some one whom they dislike. It thus comes about that the sign of the signal which the experimenter emits, giving or withholding cigarettes, is reversed. What the experimenter thought was a reward turns out to be a message of partial indignity, and what the experimenter thought was a punishment becomes in part a source of satisfaction.
Consider the acute pain of the mental patient in a large hospital who is momentarily treated as a human being by a member of the staff.
To explain the observed phenomena we always have to consider the wider context of the learning experiment, and every transaction between persons is a context of learning.
The double bind hypothesis, then, depends upon attributing certain characteristics to the learning process. If this hypothesis is even approximately true, room must be made for it within the theory of learning. In particular, learning theory must be made discontinuous so as to accommodate the discontinuities of the hierarchy of the contexts of learning to which I have referred.
Moreover, these discontinuities are of a peculiar nature. I have said that the larger context may change the sign of the reinforcement proposed by a given message, and evidently the larger context may also change the mode—may place the message in the category of humor, metaphor, etc. The setting may make the message inappropriate. The message may be out of tune with the larger context, and so on. But there are limits to these modifications. The context may tell the recipient anything about the message, but it cannot ever destroy or directly contradict the latter. "I was lying when I said ~The cat is on the mat' " tells the vis-a-vis nothing about the location of the cat. It tells him only something about the reliability of his previous information. There is a gulf between context and message (or between metamessage and message) which is of the same nature as the gulf between a thing and the word or sign which stands for it, or between the members of a class and the name of the class. The context (or metamessage) classifies the message, but can never meet it on equal terms.
In order to fit these discontinuities into learning theory, it is necessary to enlarge the scope of what is to be included within the concept of learning. What the experimenters have described as "learning" are in general changes in what an organism does in response to a given signal. The experimenter observes, for example, that at first the buzzer evokes no regular response, but that after repeated trials in which the buzzer has been followed by meat powder, the animal will begin to salivate whenever it hears the buzzer. We may say loosely that the animal has begun to attach significance or meaning to the buzzer.
A change has occurred. In order to construct a hierarchic series, we pick on the word "change." Series such as we are interested in are in general built in two ways. Within the field of pure communications theory, the steps of an hierarchic series may be constructed by successive use of the word "about," or "meta." Our hierarchic series will then consist of message, metamessage, meta-metamessage, and so on. Where we deal with phenomena marginal to communications theory, similar hierarchies may be constructed by the piling up of "change" upon "change." In classical physics, the sequence: position; velocity (i.e., change in position); acceleration (i.e., change in velocity or change in change of position); change of acceleration, etc., is an example of such a hierarchy.
Further complications are added—rarely in classical physics but commonly in human communication—by noting that messages may be about (or "meta" to) the relationship between messages of different levels. The smell of the experimental harness may tell the dog that the buzzer will mean meat powder. We will then say that the message of the harness is meta to the message of the buzzer. But in human relations another sort of complexity may be generated; e.g., messages may be emitted forbidding the subject to make the meta connection. An alcoholic parent may punish a child for showing that he knows that he should look out for storms whenever the parent gets the bottle out of the cupboard. The hierarchy of messages and contexts thus becomes a complex branching structure.
So we can construct a similar hierarchic classification within learning theory in substantially the same way as the physicists. What the experimenters have investigated is change in the receipt of a signal. But, clearly, to receive a signal already denotes change—a change of a simpler or lower order than that which the experimenters have investigated. This gives us the two first steps in a hierarchy of learning, and above these an infinite series can be imagined. This hierarchy93 can now be laid out as follows :
(1) The Receipt of a Signal I am working at my desk on which there is a paper bag, containing my lunch. I hear the hospital whistle, and from this I know that it is twelve o'clock. I reach out and take my lunch. The whistle may be regarded as an answer to a question laid down in my mind by previous learning of the second order; but the single event—the receiving of this piece of information—is a piece of learning, and is demonstrated to be so by the fact that having received it, I am now changed and respond in a special way to the paper bag.
(2) Those Learnings Which Are Changes in (1) These are exemplified by the classical learning experiments of various kinds: Pavlovian, instrumental reward, instrumental avoidance, rote, and so on.
(3) Those Learnings Which Constitute Changes in Second-Order Learning I have in the past, unfortunately, called these phenomena "deutero-learning," and have translated this as "learning to learn." It would have been more correct to coin the word trito-learning and to translate it as "learning to learn to receive signals." These are the phenomena in which the psychiatrist is preponderantly interested, namely, the changes whereby an individual comes to expect his world to be structured in one way rather than an-other. These are the phenomena which underlie "transference"— the expectation on a patient's part that the relationship with the therapist will contain the same sorts of contexts of learning that the patient has previously met with in dealing with his parents.
(4) Changes in Those Processes of Change Referred to in (3) Whether learning of this fourth order occurs in human beings is unknown. What the psychotherapist attempts to produce in his patient is usually a third-order learning, but it is possible, and certainly conceivable, that some of the slow and unconscious changes may be shifts in sign of some higher derivative in the learning process.
At this point it is necessary to compare three types of hierarchy with which we are faced: (a) the hierarchy of orders of learning; (b) the hierarchy of contexts of learning, and (c) hierarchies of circuit structure which we may—indeed, must — expect to find in a telencephalized brain.
It is my contention that (a) and (b) are synonymous in the sense that all statements made in terms of contexts of learning could be translated (without loss or gain) into statements in terms of orders of learning, and, further, that the classification or hierarchy of contexts must be isomorphic with the classification or hierarchy of orders of learning. Beyond this, I believe that we should look forward to a classification or hierarchy of neurophysiological structures which will be isomorphic with the other two classifications.
This synonymy between statements about context and statements about orders of learning seems to me to be self-evident, but experience shows that it must be spelled out. "The truth cannot be said so as to be understood, and not be believed," but, conversely, it cannot be believed until it is said so as to be understood.
93 ' 1971. In my final version of this hierarchy of orders of learning, published in this volume as "The Logical Categories of Learning and Communication," (see p. 283) I have used a different system of numbering. The receipt of a signal is there called "Zero Learning"; changes in Zero Learning are called Learning I; "deuterolearning" is called Learning II, etc
It is necessary first to insist that in the world of communication the only relevant entities or "realities" are messages, including in this term parts of messages, relations between messages, significant gaps in messages, and so on. The perception of an event or object or relation is real. It is a neurophysiological message. But the 'event itself or the object itself cannot enter this world and is, therefore, irrelevant and, to that extent, unreal. Conversely, a message has no reality or relevance qua message, in the Newtonian world: it there is reduced to sound waves or printer's ink.
By the same token, the "contexts" and "contexts of con-texts" upon which I am insisting are only real or relevant insofar as they are communicationally effective, i.e., function as messages or modifiers of messages.
The difference between the Newtonian world and the world of communication is simply this: that the Newtonian world ascribes reality to objects and achieves its simplicity by excluding the context of the context—excluding indeed all metarelationships — a fortiori excluding an infinite regress of such relations. In contrast, the theorist of communication insists upon examining the metarelationships while achieving its simplicity by excluding all objects.
This world, of communication, is a Berkeleyan world, but the good bishop was guilty of understatement. Relevance or reality must be denied not only to the sound of the tree which falls unheard in the forest but also to this chair which I can see and on which I am sitting. My perception of the chair is communicationally real, and that on which I sit is, for me, only an idea, a message in which I put my trust.
"In my thought, one thing is as good as another in this world, and the shoe of a horse will do," because in thought and in experience there are no things, but only messages and the like.
In this world, indeed, I, as a material object, have no relevance and, in this sense, no reality. "I," however, exist in the communicational world as an essential element in the syntax of my experience and in the experience of others, and the communications of others may damage my identity, even to the point of breaking up the organization of my experience.
Perhaps one day, an ultimate synthesis will be achieved to combine the Newtonian and the communicational worlds. But that is not the purpose of the present discussion. Here I am concerned to make clear the relation between the contexts and the orders of learning, and to do this it was first necessary to bring into focus the difference between Newtonian and communicational discourse.
With this introductory statement, however, it becomes clear that the separation between contexts and orders of learning is only an artifact of the contrast between these two sorts of discourse. The separation is only maintained by saying that the contexts have location outside the physical individual, while the orders of learning are located inside. But in the communicational world, this dichotomy is irrelevant and meaningless. The contexts have communicational reality only insofar as they are effective as messages, i.e., insofar as they are represented or reflected (correctly or with distortion) in multiple parts of the communicational system which we are studying; and this system is not the physical individual but a wide network of pathways of messages. Some of these pathways happen to be located outside the physical individual, others inside; but the characteristics of the system are in no way dependent upon any boundary lines which we may superpose upon the communicational map. It is not communicationally meaningful to ask whether the blind man's stick or the scientist's microscope are "parts" of the man who uses them.
Both stick and microscope are important pathways of communication and, as such, are parts of the network in which we are interested; but no boundary line —e.g., halfway up the stick—can be relevant in a description of the topology of this net.
However, this discarding of the boundary of the physical individual does not imply (as some might fear) that communicational discourse is necessarily chaotic. On the contrary, the proposed hierarchic classification of learning and/or context is an ordering of what to the Newtonian looks like chaos, and it is this ordering that is demanded by the double-bind hypothesis.
Man must be the sort of animal whose learning is characterized by hierarchic discontinuities of this sort, else he could not become schizophrenic under the frustrations of the double bind.
On the evidential side, there is beginning to be a body of experiment demonstrating the reality of third-order learning94; but on the precise point of discontinuity between these orders of learning there is, so far as I know, very little evidence. The experiments of John Stroud are worth quoting. These were tracking experiments. The subject is faced with a screen on which a spot moves to represent a moving target. A second spot, representing the aim of a gun, can be controlled by the subject, who operates a pair of knobs. The subject is challenged to maintain coincidence between the target spot and the spot over which he has control. In such an experiment the target can be given various sorts of motion, characterized by second-, third-, or higher-order derivatives. Stroud showed that, as there is a discontinuity in the orders of the equations which a mathematician might use to describe the movements of the target spot, so also there is a discontinuity in the learning of the experimental subject. It is as if a new learning process were involved with each step to a higher order of complexity in the movement of the target.
It is to me fascinating to find that what one had supposed was a pure artifact of mathematical description is also apparently an inbuilt characteristic of the human brain, in spite of the fact that this brain certainly does not operate by means of mathematical equations in such a task.
There is also evidence of a more general nature which would support the notion of discontinuity between the orders of learning. There is, for example, the curious fact that psychologists have not habitually regarded what I call learning of the first order, the receipt of a meaningful signal, as learning at all; and the other curious fact, that psychologists have until recently shown very little appreciation of that third order of learning, in which the psychiatrist is predominantly interested. There is a formidable gulf between the thinking of the experimental psychologist and the thinking of the psychiatrist or anthropologist. This gulf I believe to be due to the discontinuity in the hierarchic structure.
Was this article helpful?