Selections from "Morale and national character." by Gregory Bateson

Gregory Bateson

"Communication."

in The Natural History of an Interview by Bateson, Birdwhistell, Brosin, Hockett, McQuown1968 [1960].

CHAPTER ONE "COMMUNICATION"
by Gregory Bateson

This chapter is from The Natural History of an Interview by Bateson, Birdwhistell, Brosin, Hockett, McQuown was a multi-disciplinary approach to the analysis of the same piece of film data. The disciplines were anthropology, linguistics, and psychiatry. The book was never published because of release problems with the film. The existing chapters are on microfilm but this is a copy Bateson manuscript. The reference is:

"Introduction" in The Natural History of an Interview. University of Chicago Library Microfilm Collection of Manuscripts in Cultural Anthropology, series 15, Nos, 95-98. 1971.

Background

At the time of the outbreak of World War II, the most promising insights in the behavioral sciences were those derivative from Freudian analysis, Gestalt psychology, and cultural relativity. Linguistics had begun to take on new life under the leadership of Sapir (192i, 1925, 1933a, 1933b) and Bloomfield (1933, 1939). Psychiatry was evolving away from the exclusive study of the individual patient towards the study of human relationships, most dramatically under the influence of Sullivan (1940); and already there were moves towards a mathematics of human relationship under Kurt Lewin (1935) and L.F. Richardson (1939).

During World War II and immediately following that period of confusion,a series of exceedingly important new approaches were evolving more or less independently in a number of different places, but the possible relevance to behavioral science of the work of George Boole (1854), and of Whitehead and Russel" (1910-1913) was still unexplored. All of these scattered advances were precipitated by the war-time development of electronic engineering. A partial list of names and locations of the principal advances will give an idea of what is happening.

Rosenblueth at Cambridge and Mexico, and Wiener and Bigelow (1943) at Massachusetts Institute of Technology and Princeton respectively were laying foundations of what has come to be called cybernetics, extending what the engineers and mathematicians had learned about self-correcting mechanisms to the fields of biology and social organization.

Von Neumann and Morgenstern (1944) at Princeton were laying the basis of the theory of games.

Craik (1952), in Cambridge, England, before his premature death, wrote: "The Nature of Explanation" raising the whole question of how messages are coded in a reticulate central nervous system.

Attneave (1959), Stroud (1949), and others at Stanford, happened to see Craik"s little book and by it were inspired to a new approach to the problems of perception and adaptive action.

In Vienna, von Bertalanffy (1952) was building the beginnings of systems theory with a special emphasis upon those systems (e.g., organisms) which have a continuous source of energy derived from the environment.

Shannon (1949), and others working with the Bell Telephone Laboratories, were building.

Ashby (1952, 1956), in Gloucester, England, was devising new models for theories of learning.

Other names [McCulloch, Pitte (1943, 1946), Lorente de No (1922, 1933), Rashevsky (1948), Walter (1953), Tingergen (1953), Lorenz (1952)] might be mentioned as contributing to this general trend.

What has happened has been the introduction into the behavioral sciences of a number of very simple, elegant, end powerful ideas all of which have to do with the nature of communication in the widest sense of the word. The steps and sequences of logic have been coded into the causal sequences of computing madlines and, as a result, the Principia Mathematica has become a cornerstone of science.

The Natural History

The present book is an attempt at synthesis. It is written by five persons who are professionally concerned with communication problems in diverse fields and we attempt a synthesis of a wide and abstract kind, starting from the most concrete data.

We start from a particular interview on a particular day between two identified persons in the presence of a child, a camera and a cameraman. Our priaary data are the multitudinous details of vocal and bodily action recorded on this film. We call our treatment of such data a "natural history" because a minimum of theory guided the collection of the data. The cameraman inevitably made some selection in his shooting; and "Doris," the subject of the interview, was selected for study not only because she and her husband were w1lling to be studied in this way but also because this family suffered from inter-personal difficulties which had led them to seek special psychiatric aid,

These materials, then, while collected under circumstances unusual in human relationship, nonetheless provide the natural history of two human beings over a brief span of time, and the data themselves are sufficiently uncorrupted by theory so that the five authors, each with a particular theoretical bias and interest, could simultaneously approach this mass of detail. Moreover, we shared something less tangible than the common data: certain theories or preconceptions about what happens when two people interact.

My major task in this preliminary chapter is to outline those theoretical premises which were engendered in us by the recent advances in the study of human communication.

Freudian

From Freudian theory we accept the premise (1) that only limited aspects of a part of what happens in human communication are accessible to the consciousness of the participants, Our position however differs from that of many early Freudians in two respects which are minor so far as theory is concerned, but major in their implications for method. The important corrective which the Freudian applied to man"s thinking about human nature was an insistence upon the unconscious. The error to be corrected was the notion that in human beings the mental process is preponderantly or entirely conscious. This error has roots in eighteenth century culture and back into the Reformation and into earlier Judeo-Greek philosophies of free will. But this error today seems almost fantastic.

It is now a platitude to state that mental process depends upon hierarchic organization. Whether we think of mental levels or of a brain evolved by a process of successive telencephalization, we envisage a hierarchy of both anatomy and function. And our knowledge of hierarchic function -- in machines, embryology, physiology, and in human social organization -" indicates as a truism that under no circumstances can the upper echelons of any hierarchic system handle total information about the processes and events which occur at superordinate or peripheral levels. By the same token, the upper echelons can handle only limited reports -- can be only partially conscious -- of all that happens at their own upper level. To provide these upper echelons with total reporting would be to add to the system still higher echelons -- themselves, in turn, largely unconscious. To us, then, the fact that most mental process (including, especially, the process of perception itself) cannot be inspected by consciousness is a matter of course and what is surprising, and therefore needs explanation, is the fact of consciousness. Unconsciousness is a necessity of the economics of hierarchic organization (Sapir, 1927).

This does not mean, of course, that economy of effort or the economic use of the channels of communication to avoid jamming is tbe only factor determining want information shall be allowed to reach the upper echelons of consciousness. The analogy of human social organizations would indicate very clearly both that upper echelons are commonly "motivated" not to receive information about certain peripheral events and that there are many events which the subordinate echelons are:"motivated" not to transmit upwards. There are, therefore, many matters wich remain "in the unconscious" for reasons other than those of economy, and the unconscious becomes a repository for material which is repressed in the Freudian sense.

The second difference between our position and the classical Freudian results from our emphasis upon communication. We are interested in such questions as "what signals are emitted and what orders of awareness does the signaler show by emitting other signals about these signals? Can he plan them? Can he recall them?" And we are interested to know what signals reach the receiver and what signals he knows he has received. Our emphasis is thus upon perception and communication rather than on the internal hierarchies of mental process. From where we sit, the distinction between conscious and unconscious becomes significantly comparable to the distinction between foveal and peripheral vision.

A second premise related to Freudian theory holds (2) that everything which occurs is meaningful in the sense of being a part of the interchange as well as non-accidental. The Freudian emphasis was upon psychic determinism- that "no word uttered and no detail of a dream experienced can be accidental. A man cannot just dream." Our emphasis in this book will extend this psychological idea into the realm of inter-personal process. We shall attempt to see every detail of word, vocalization, and bodily movement as playing its part in determining the ongoing stream of words and bodily movements which is the interchange between the persons. We shall endeavor to think not only in terms of psychic determinism but in terms of a larger interpersonal determinism. Two people cannot "just agree" or "just quarrel".

Also, from Freudian theory, we accept the idea (3) that all messages, whether verbal or non-verbal, are mediated in their creation by primary process and therefore contain, either implicitly or explicitly, all the multiple reference characteristics of dream or fantasy. If it be possible for a man to seem to talk only about the overt subject of conversation, this is achieved only by vigorous ego-function which carefully excludes or conceals the multiple over-tones of implicit content. Further, we expect that the minute analysis of speech and movement will disclose that the messages in both these modalities contain a large proportion of unconscious material with primary process characteristics, that, for example, an unconscious fingering of the dress is likely to dcnote (or to be a resultant of) sexual interest and/or for its puritanical denial.

Also from Freudian theory, we accept (4) a generalized notion of transference: that any person emitting learned signals does so upon the (usually unconscious) assumption that the receiver of these signals will understand them "correctly" -- i.e., he assumes that his vis-a-vis at the given moment will resemble psychologically some former (or even fictitious) vis-a-vis from whom he originally acquired his conmunicational habits.

Closely related to the notion of transference is (5) the notion of projection. This explanatory principle differs, however, from transference in that it does not invoke some third historical or fictitious person. When A "projects" upon B, he is merely assuming that B's signals are to be interpreted as A would interpret these signals if he himself had emitted them. That is, A assumes that B operates according to systems of codification similar to his own. Both transference and projection may, of course, be carried into the future. A may expect that B will exhibit meaningful action of a sort which some historical figure in A"s life would have exhibited under similar circumstances (transference); or he may expect that B will behave as he himself would behave in similar circumstances (projection).

Identification must also be mentioned. This explanatory principle invokes (6) the idea "if you can't lick 'em, join 'em" -- or, at least imitate them as you see them. A is said to identify with B when he starts to mold his own meaningful action in terms of what he thinks are B"s principles of codification.

Notably, all of these principles -- transference, projection, and identification -- are likely to be unconscious in their operation and to be in some degree coercive. That is, any errors which A may make in his assumptions about B are likely to cause A to act in such a way that B is put under pressure to validate these errors by acting as if A"s assumptions were true. An especially interesting case arises when A acts in a way which will coerce B into identifying with A"s self-image -- which may be false.

Moreover, it must not be supposed that these explanatory or descriptive principles are mutually exclusive. In a given instance, A may consciously or unconsciously assume that B is parental (transference). But A's technique for dealing with his parent may have involved identification.1 He will then adopt towards B that role which he formerly adopted towards the parent.

l The word "identification" was perhaps an unfortunate choice for two reasons, in the phrase "A identifies B as father" is a statement of transference. And the phrase "A achieves ego-identity" suggests (as an ideal) A's escape from all the errors of transference, projection and identification.

Gestalt

From Gestalt psychology, we have accepted a premise of very great importance: that experience is punctuated. We do not experience a continuum: on the contrary, or experience is broken up into what seem to us to be events and objects. In Gestalt psychology this idea is basic to the figure-ground hypothesis. And for us it is related to the premise that nothing never happens -- i.e., that both sender and receiver of sienals are so organized that they can and must use, for their understanding of what is going on, the fact that certain possible signals are not present The first step in building the figure-ground hypothesis is a postulate of this kind. In order to recognize that there are stars in the night sky, we must use the fact that certain retinal end-organs are not stimulated by the darkness. In human relations, no silence is insignificant and the absence of tears may speak volumes.

More must be said concerning the punctuation of interpersonal events. Our whole procedure and, indeed, any analysis of communicational data is shaped by premises which define the units into which the stream of data is to be divided. First, in a macroscopic examination of the interview we assumed that the 400 feet of film on which the interview is recorded2 can be punctuated into incidents for microscopic study. Our macroscopic study serves to direct our more particular attention. And, while we narrow our focus from the interview and then downward to the finer and finer detail of these incidents, we work throughout with similar assumptions about punctuation of the stream of signals.

The historical basis of this assumption will make clearer what is meant. Historically, scientific linguistics has progressed most rapidly since the time when certain popular and preponderantly occidental notions about language were adopted, made rigorous, and extrapolated into the study of minute detail.

In their popular form, these notions are, for example, that speech is subdivisible into sentences which in turn are subdivisible into words, which in turn can be subdivided into letters. -Profound modifications have been introduced into this hierarchy by the linguists who needed to describe speech rather than written language, but the essential idea that a stream of communicative material must, of necesaity, be susceptible of such multiple re-subdivision is fundamental in linguistics and in that part of communications theory which deals with coded communication -- a much wider field than the conventionally linguistic. A major contribution of the linguists is the demonstration that the stream of communication contains positive signals by which its units are delimited. Moreover, Gestalt theory presupposes a hierarchy of subdivisions characteristic of the process of perception. We do not perceive the firing of unit endorgans but, from the showers of neural impulse started by that firing, we build images of identifiables and larger meaningful complexes of identifiables. We can argue from perception to communication: If an organism"s perception is characterized by Gestalten and this organism is capable of emitting complex streams of communication, then these streams must be dissectible into a hierarchy of successive subdivisions. @@uv@ such analyses will be possible, and there will be one which will represent correctly the natural history of the organism. The film also has irrelevant gaps while the camera was being reloaded after each 100 feet. The longest consecutive recording is about 3 1/2 minutes.

We deal, after all, not merely with the fact that a commtdiicational stream can be dissected but also with the question: in which of the many ways possible could this particular stream be dissected? What we know about language and communication is general indicates that there will always be one or more hierarchies of Gestalten which will be correct in the sense of describing how the message stream is created and/or how it is received and interpreted by the hearer. The Freudian findings also indicate that in any given instance several different interpretations may be correct. A particular message maybe simultaneously interpreted in different ways by different levels of them@nd: we face problems of multiple coding.

The linguists are ahead of the other natural historians in their study of the hierarchy of Gestalten in terms of which a particular kind of behavior should be dissected. Their studies are being fortified by cross-cultural comparison, by dialectal (subcultural) comparison, and by statistics of individual variation. By way of contrast, kinesics -- the study of body motion, position, and action as a modality of communication -- is a relatively recent development which, like linguistics, is achieving a firm scientific base by the rigorous dissection of the kinesic stream into a hierarchy of Gestalten and subdivisions of Gestalten.

In a later chapter, Birdwhistell will outline the hierarchy of units which he is devising for kinesic description. He is proceeding in a manner comparable to but not identical with that methodology of description which has proved valuable in linguistics. The ultimate validation of this approach in kinesics will, of course, depend upon the results obtained, but there is very strong a priori argument in favor of the correctness of the approach from all that we know about communications theory in general and about human communication and perception in particular.

Returning for a moment to linguistics, other types of description which the linguists have achieved must here be mentioned. The very complex question of "meaning" is too large for discussion in this chapter but this must may be said -- that a tape recording of human speech contains a great deal more than the signals correlated with the lexical meaning of what was said.

If a tape recording is transcribed into ordinary script, although some of this more-than-lexical content will be lost, some will still survive in the transcription. Indeed, to reduce a speech to its merely lexical content would require very drastic procedures (in the course of which other and probably inappropriate non-lexical overtones will inevitably be added). It would first be necessary to strip the speech of all indications of the context in which it was uttered and by whom and to whom it was uttered. But there would still remain cadences and overtones of e non-lexical nature. To getrid of these it would be necessary to translate the epeech into some other language and to use as a translator some hypothetical person (or machine) totally insensitive to non-lexical content both of the language from whichhe is translating and of the language into which he is translating.

As we climb the hierarchic ladder of Gestalten from the most microscopic particles of vocalization towards the most macroscopic units of speech, each step on this ladder is surmounted by placing the units of the lower level in context.

"Meaning, as this word is vulgarly used, emerges at a very high level in this hierarchy. We discriminate the initial phoneme of the word "peter" from the initial phoneme of the word "butter" but these phonemes are in themselves meaningless apart from their setting in a stream of phonemes. Even the syllables, "pete" and "buta" are in themselves meaningless or multivalent (except insofar as their possible meaning is restricted when we know how they are placed in a stream of syllables). With each step towards a larger unit -- the larger unit being always the small unit plus its immediate getting -- there is a more and more drastic limitation of possible referents. "meaning: therefore, is a function of this restriction of possible meanings. Even the words:"peter" and"butter" are still multivalent. When the word "blue" is added, the hearer may be pretty sure that the referent of "peter"is a flag. But still, there is room for doubt.

The "Blue Peter" may be referred to as an actual object of action or observation in the larger context of a ship about to leave a particular port. Or the reference may only be metaphoric if the term is used on land, Or the usage of the term may be neither metaphoric nor direct but may be part of a lesson in maritime communication. Or -- as here upon this page -- the words "Blue Peter" may be mentioned only as an example of communicational phenomena.

Meaning approaches univalence or non-ambiguity only when very large units of the communicational stream are admitted to examination. And even then, the approach to non-ambiguity will be asymptotic. As larger and larger bodies of data are admitted, the probability of a given interpretation will be increased but proof will never be achieved. The situation is essentially the same as that which obtains in science where no theory is ever proved.

This book is concerned with trying to put together those parts of the communicational stream which the professional linguist studies phonemes, morphemes, phrases, vocal modifiers, junctures, etc.) with those parts of the stream which are studied in kinesics (kines, kinemorphs, etc.). A central question, therefore, which we shall have to face when we analyze the data is the extent to which there is a mutual relationship of "context" between kinesics and linguistic elements.

We face phenomena so structured that there is perhaps no definable upper limit to the size -- either spatial or temporal -- of the Gestalten. In practice, this would mean that no finite collection of data would confer complete non-ambiguity upon any item within the collection; that however widely "context" be defined, there may always be wider contexts a knowledge of which would reverse or modify our understanding of particular items.

Context

These considerations force us to a method of inquiry which will postpone the question of "meaning" When faced with a given sequence of signals, we shall delay the question "what do these signals mean" for as long as possible. We shall ask, rather, the collateral question "would the meaning be changed by a given change in the sequence or by a given change in the context?" This is a question which can be asked and answered without too much difficulty. We shall, for example, not ask whether the word "Peter" refers to an apostle or a flag, but rather whether its meaning when the word "Peter" follows the word "Blue," is peculiarly appropriate to the new context.

In kinesic analysis, we shall similarly delay the question of @bat ismca@t by the rapid closing and opening of an eye which is visible to the visa-vis and shall ask rather, for example, whether the meaning of this signal would be altered (a) if the other eye were blinked simultaneously and (b) if the blinked eye were one which is invisible to the vis-a-vis. Parenthetically, we may also ask whether the meaning of the word "Peter" is altered by winking one eye.

It is, after all, only an historic accident -- a past pathway in the evolution of science -- that has lead to the circumstance that linguists study data which can be heard, while the kineticist studies data which can be seen. That the scientists have become specialized in this particular way does not indicate a fundamental separateness between these modalities in the stream of communication. It is for this reason that the work of this book starts from concrete natural history -- from the recorded interaction between Doris" speech and movement and the speech and movement of Gregory. This placing of every signal in the context of all other signals is an essential discipline of our work. A great part of the work which Birdwhistell, Hockett and McQuown have had to do has involved a grueling process of synchronization. The audible stream for which Hockett and McQuown are specialists was recorded on tape and on film with an unsatisfactory sound track. The analysts had to work frame by frame through the film to establish the point in the audible sequence at which, for example, Doris turned her head or let her shoe fall away from her heel. I described our data loosely above as the aggregate of signals recorded on the film. More accurately, I should have said that our data are the individual signals or messages, each in its immediate and extended context.

But the context of a signal emitted by Doris is not merely those other signals which she has recently emitted plus those which she emits soon after; it is also the room in which she is speaking, the sofa on which she is sitting, and the signals emitted by Gregory with whom she is talking, and by the little boy, Billy, and the interrelationships among all of these.

Interaction

At this point, our concept of communication becomes interactional and our intellectual debt is to G.H. Mead (1934) and to Sullivan (1940) rather than to Freud and the Gestalt psychologists. The system which we now study is no longer merely a descriptive synthesis of Doris" body motion and speech, but the larger aggregate of what goes on between Doris and Gregory.

This larger frame determines meaning for what each person does and says. Rilke"s "Unicorn" is present in every conversation between persons and this fictitious beast evolves and changes, dissolves and is recrystalized in new shapes with every move and message. Denial of the Unicorn will not prevent its existence -- but only cause it to become monstrous.

This poetic fancy must be made scientifically real to the reader if he is to understand what this book is about.

For every human being there is an edge of uncertainty about what sort of messages he is emitting, and we all need, in the final analysis, to see how our messages are received in order to discover what they were. For the schizophrenic this is often dramatically and conspicuously true.

Let us illustrate by an example. A schizophrenic patient tells me that he built the China Wall, rowed across the Pacific and landed in Seattle. He then walked to California where he was "affriended by those people." This narrative he offers as if it were a statement of fact. But whether it is a statement of fact for him depends upon my response. If say,"Nonsense. You were born in California," I have thereby verified for him the notion that his narrative is to be taken literally. I have denied it as if it were a literal statement and now it exists for him as a literal statement which must be defended. From there on, we shall get into an argument not about the question "is this narrative a statement of facts?" but about the red-herring question "is this a true statement of the facts?"

The response which we get tells us about the state of the hearer after he has received the signals which we emitted. It may be evident that he misunderstood the message either grossly or subtly. The status quo, however, which obtained when we emitted the message, no longer exists and merely to repeat the message will not do. We are now communicating with a person whose relationship to us is different from what it was a moment ago. And within the framework of this new relationship we must now speak.

All of the elements and vicissitudes of formation and re-formation of relationships, perhaps the most interesting is that process whereby people establish common rules for the creation and understanding of messages. Whatever reply I may make to the patient"s delusional narrative proposes a pact to govern us both in our understanding of the message. If I deny the factual truth of the narrative, I implicitly propose that we agree to treat it as literal. If, on the other hand, I ask him whether he thinks his parents had a part in building the "China Wall: which separates him from them, I have proposed that we agree upon a different set of rules for the creation and understanding of such messages.

The possible systems of rules which two persons may share are many and complex. Among them must be mentioned a system which has been characterized as symbiotic. Such a label refers to, as I understand it, a system of nonverbalized and usually unconscious pacts in which, for example, A and B "agree" to accept each other"s messages in some spirit other than that in which they are coined. By ignoring overtones and implications, or by reading in overtones which were not intended, the persons maintain a strange semblance of understanding.

Code Distortion

In this book we shall pay but little attention to those failures of communication which are due to the randomization of signals occasioned either by background noise or by imperfect resolving processes in the receiving sense organ. We are concerned with a more subtle phenomenon -- the distortion of messages which occurs when the persons involved differ from each other in their rules or assumptions governing the making and understanding of messages -- their explicit and implicit rules of coding.

Imagine a machine which has the function of transmitting a half-tone block, (a picture formed entirely of rows of dots), over a wire to another machine.

The transmitting machine will transmit over the wire a sequence of electrical impulses such that each impulse or absence of impulse is "yes" or "no" answer to the question "is there a dot in the given space?" When the transmitting machine comes to the end of a line of spaces, it will either transmit a special signal which will cause the receiving machine to go on to the next line, or the machines will have to have been so adjusted to each other, th&t they operate in terms of a common pact governing how many dots there shall be in line. A discrepancy regarding the terms of this fact will introduce code distortion. In such a case, the receiving machine will create a picture which may be an absolutely correct record of the sequence of signals emitted by the sensor but which, considered as a picture, will be a distortion of the original, Figure 1 is a picture to be transmitted, B is the distorted version which is created when the receiving machine acts upon the premise that there are only 16 squares in each line instead of 17.

In Figure 1, the effect of code distortion is shown, and it is worthwhile to stress the basic difference between such distortion and the loss of information due to entropic noise. In the case of entropic noise, the information which is lost is irretrievable, but in the case of code noise what has occurred is a systematic distortion which could conceivably be rectified.

All that is needed for this correction is that there be some means whereby the transmitter and receiver can communicate about the rules of communication. This presents special difficulties, but it is a fundamental thesis of this book that at the human level such communication about these rules of communication occurs constantly. This, in fact, is the process whereby the "Unicorn" is continually created and recreated. When my patient tells his story of the "China Wall." Whatever reply I make is a communication to him about how I received his message and therefore indicates to him (ideally) how he should restate it in order to have me receive that message which he wants me to receive. It tells him how to code his messages so as to elicit an appropriate response from me.

It is necessary again to insist upon the unconscious character of most communication. We are almost totally unaware of the processes by which we make our messages and the processes by which we understand and respond to the messages of others. We are commonly unaware also of many characteristics and components of the messages themselves. We do not notice at which moments in a conversation we cross and uncross our legs or at which moments we puff on our cigarettes or blink our eyes or raise our brows. But the fact that we do not notice these things does not imply that all these details of personal interaction are irrelevant to the ongoing relationship. Just as we are in the main unconscious of the fleeting pacts which we enter into as to how messages are to be understood, so also we are unconscious of the continual dialogue about these pacts.

This dialogue is not only between persons and about tbe pacts which they form, it is also and more strangely a dialogue which governs what each person is. When A makes overtures which B brushes aside, this experience is to A more than a hint about how eo code messages when dealing with B. In everyday language we say that a person"s self-esteem is enhanced or reduced by the responses of others. Or we say that "be sees himself differently." In communicational terms, we can translate this into a statement that the very rules of self-perception, the rules governing the formation of a self-image, are modified by the way in which others receive our messages.

Learning and Pathogenesis

In part, this book is a study of how communication works between two persons, but it is also a study of how communication fails to work -- that is, of certain pathologies of communication. Our collaborating team includes not only the two linguists and the founder of kinesics, but also two psychiatrists and the writer, whose initial training in anthropology has finally led him to study schizophrenic communication. It is therefore appropriate to examine a little more closely the relation between psychiatric pathology in the individual patient and the pathologies of communication which may develop between persons. In order to keep the subject matter simple, I will exclude from consideration those psychiatric abnormalities which have an established base in organic lesion.

To build a bridge between the study of psychiatric functional pathology and the pathologies of communication, it is necessary to insist upon the existence of the facts of learning and conditioning. Two considerations become especially relevant. First, every failure of communication is painful. Second, the learning organism always generalizes from experience. Further, the business of communication is a continuous learning to communicate. Codes and languages are not static systems which can be learned once and for all. They are, rather, shifting systems of pacts and premises which govern how messages are to be made and interpreted. Every signal which establishes a new premise or pact bringing the persons closer together or giving them 8reater freedom may be a source of joy. But every signal which falls by the wayside is in some degree a source of pain to both. The ongoing stream of communication is thus, for each individual, a continuous chain of contexts of learning and, specifically, learning about premises of communication.

At this point, it is necessary to consider certain aspects of the learning process and to expand conventional learning theory to make it relevant to an analysis of the interchange of signals between persons. A typical learning experiment involves two entities: an experimenter and a subject, and the theoretical conclusions derived from such experiments are commonly stated as psychological regularities descriptive of the subject. In contrast, I shall here view the experimental situation as an interaction involving two entities in whose relationship I am interested. I shall regard their relationship as formally characterized by an interchange which is repeated in successive"trials" and shall assume that not only the subject, but also the experimenter is undergoing a learning process determined, at ieast in part, by reinforcements which the subject provides.

As a preliminary to this, it is necessary to define a hierarchy of orders of learning. This may be done as follows:

1. It appears that the simplest learning phenomenon is the receipt of information or command. The event of perceiving a whistle may con

investigate and has not been an immediate object of experimental study. It has, however, been a major focus of theory. What seems to have happened is that in order to arrive at a theory to describe what I call the second order of learning, the psychologists have had to provide some description of this first order process -- some verbalization of what message the dog received. The "effect" theory proposes that this message is a promise of reward or a threat of punishment, whereas the associational theory proposes a more automatic and less purposive description of the dog's response.

2. The second order is the learning or conditioning upon which the vast mass of experimental work has been conducted. Here the word "learning" refers to a change in the dog"s ability to act upon percepts or signals received. What the experimenters study is changes in the dog's behavior resulting from a sequence of trials. The phenomena studied are of a higher order than those discussed in 1 above. The question asked is not "what change occurs in a dog when he hears a whistle?", but "what changes have occurred in the change which a dog undergoes when he hears a whistle?" This subtle difference in the question asked by the experimenter makes it formally impossible for the theorists to deduce answers to the first question from data collected to answer the second. The behaviorists had logic on their side when they insisted that we never ask about the subjective experience of the dog. To try to deduce the dog's experience from data which could only throw light upon change in his experience is to attempt the logically impossible. From the characteristics of a class, I can make no deductions about what a member of that class might be.

3. The third order of learning is a familiar laboratory phenomenon but has received only slight attention from the experimentalists. If we describe the second order of learning as "learning to receive signals," then the third may be described as learning to learn to receive signals. What happens in the laboratory is that the animal having been subjected to experiments of the second order becomes "wise," That is, when faced , with an entirely new experiment of this sort, the animal requires now a smaller number of trials to achieve that learning of the second order which the experimental situation demands. The animal has acquired a knack or skill for second order learning. This phenomenon has been measured by Hull (1940) studying rote learning of nonsense syllables, and by Harlow (1949) studying problem solving in rhesus monkeys.

4. There is no theoretical reason to deny the possibility of fourth and higher orders of learning, though none of these have been demonstrated. The nature of the hierarchy which we are discussing is such that there is no upper limit to the series other than that set by the limitations of brain structure. The number of neurons being finite, it is certain that for any organism there is a practical upper limit to the number of orders of learning of which it is capable.

Unfortunately, in an earlier theoretical paper (Bateson, 1942) I have used for this third order of learning the term "deutero-learning." This was due to my failure to recognize the receipt of a meaningful signal or the receipt of a bit, of information as an example of the simplest order of learning. To achieve apy! analogy between the mechanical computers and the brain it is necessary to insist that any receipt of information is, in a broad sense, learning.

Inspection of this hierarchy of learning reveals that the difference between any order of learning and the next higher order is essentially a difference in size of Gestalt. The higher order is always documented by demonstrating change which results from a larger Gestalt, this larger Gestalt being in general built up of a multiplicity of the Gestalten characteristic of the lower order. But while this generally seems to be the case, there is no theoretical premise by which we might estimate the multiplication factor, and it is necessary to consider as at least theoretically possible the case in which this factor would be unity.

A single increment in what appears to be a context of lower order learning might conceivably precipitate major changes of some higher order, whereby all experience of the lower order would be reframed and reorganized. We face herean unpredictability of a sort which I noted earlier when discussing the indeterminacy of meaning. Larger and larger bodies of data will provide greater and greater certainty of interpretation but it is never possible to be sure that the next increment of data will not compel us to a totally new interpretation. There -s thus an anology -- perhaps amounting to identity -- between those hierarchies of Gestalten which determine meaning and that hierarchy of Gestalten which we here call contexts of learning.

These abstract matters become clearer when we state that learning of the third or higher order is, in popular parlance, called "change in character."Let us suppose that an organism becomes "wise" in dealing with contexts of Pavlovian learning. The change which we here refer to may be describes both has a change in the organism"s expectations and as a change in its learning habits. If we speak in terms of expectations, we will say that the organism now preponderantly expects the universe of experience to be punctuated into sequences resembling the Pavlovian context; i.e., sequences in which certain percepts can be used as a basis for predicting later events. Or, if we do speak in terms of learning habits, we will say that this organism will respond to the predicted certainty of that which is to come, (e.g., by salivating), but will not endeavor to change the course of events. In a word, the organism has become "fatalistic" and examination of the formal characteristics of the learning context has provided us with a formal definition of one particular sort of "fatalism."

The psychiatrist is interested largely in learning of the third or higher order. If a patient tells him that she can use a typewriter, the psychiatrist will pay but little attention. She has reported only a result of second order learning. But when the patient goes on to describe the context in which she learned to typewrite and tells him that her teacher punished every error she made but never praised her progress, the psychiatrist will prick up his ears. He will see in this narrative a statement of what effect the context of learning to type may have had upon the patient"s habits and expectations -- i.e., upon the patient"s character, This enlargement of learning theory to discriminate orders of learning makes this body of experimental knowledge especially relevant to the psychiatrist. Actually the old barrier between experimentalists and clinicians seems to have grown out of this: that the experimentalists have mainly studied learning of the second order, while the psychiatrist is interested chiefly in effects of the third order, These effects he tries to evaluate in his diagnosis or to achieve in his therapy.

If this description of learning is substantially correct, that is, if there really is a hierarchy of orders of this phenomenon and the description of these orders is something more than an artifact of description, then it becomes theoretically probable that there exist complex sequences of experience and action such that learning of one order will in some degree contradict the learning of some other order. We can imagine, for example, that a human subject might experience a long sequence of Pavlovian learnings but might be penalized (Bateson and Jackson, 1956) for exhibiting "fatalism." Or he might be trained towards obedience but be continually penalized for the finer detail of every obedient act. As between adults, this is familiar enough and may make for bad "personal relations." As between parents and small children, I believe that it is -- under some circumstances -- pathogenic.

Pathogenic Contexts

It is now clear, however, in an abstract and formal way, what patterns of interchange we should look for in our data, The discussion which preceded this reexamination of learning theory concerned the establishment of pacts and premises of communication. But evidently a premise of communication, a rule governing how messages are to be constructed or interpreted, bears the same relationship to the given message as occurs between a higher and lower order of learning.

The acceptance of what I have canceled a premise of communication is the same phenomenon as the acceptance of a role -- a momentary of enduring shift in habit and expectation. And "role" is only a word for some phase of character change, be it brief or enduring. It is a description of the pattern exhibited by one person in that two-person system which constitutes a context of learning.

It follows that what we have to look for in the data is sequences and, at the meta-level, sequences of sequences. The relevant units will be those segments of the stream which constitute contexts of learning. Problems of pathology within the stream will become recognizable when we see instances so constructed that learning in a given small sequence would be contradicted by learning in some larger sequence of which the smaller is a component. Theoretically, we may expect instances in which part and whole will be identical -- where the multiplication factor relating the part to the whole is unity. A single context (seen in two different ways) may propose contradictory learning at different levels.

One other peculiar phenomenon must now be mentioned -- namely, that the premises of communication are commonly self-validating. By their operation they may create that consensus which will seem to validate them. He who believes

that all the world is his friend -- or enemy -- will emit messages and act meaningfully in terms of his premise, He will meet the world in a way which puts pressure upon this very world to validate his belief, which belief he acquired in the first place by the cumulative impact of those contexts of learning which were his communication with some earlier person.

An inquiry into the functional psychopathologies this becomes an investigation of the dynamics of past communication. But curiously enough, because of this fact that communicational premises are self-validating, it is often not necessary to delve into the past in order to investigate their etiology. The premises are self-validating in the Present and therefore the disturbed -- like the normal -- is continually creating around himself that environment which provides the typical etiology for his communicational habits -- his symptoms.

One has only to examine the present family relations of a patient to find working today the constellation which is etiologic for his symptoms. Indeed, we may profitably examine ehe workings of any typical mental hospital for clues as to why its patients are mentally ill.

This broad description of the interchange between persons aa a sequence of contexts of learning contains the possibilities for two kinds of psychopathological result: the learning of particular error and the disruption or distortion of the learning process itself. historically, the first of these received most attention in the early days of psychoanalysis when emphasis was placed upon the fact that certain neuroses result from single and extremely painful experiences in childhood. In terms of what has been said above, we might rephrase this theory as a learning of error -- the error being an inappropriate generalization from some terrifying, painful, or over-rewarding experience. Today, less theoretical importance is attached to this kind of pathogenesis, but its occurrence is still undoubted.

In contrast, modern psychiatric theory insists more upon those psychopathological results which derive from continual and repeated experience rather than from isolated trauma. Here the probability that simple error will be generated in the learning individual is much less, since, after all, his opinions, stemming from a multitude of instances, are to that extent validated by the repetition of instances. What is rather to be expected from such an etiology is the distortion of the learning process itself -- a type of pathological result more abstract and intangible -- and more difficult to correct by any therapeutic experience, since whatever the patient learns from this experience will probably be learned by means of that process which is already distorted.

It is, however, necessary to give some substance to the phrase "distortion of learning" I have to indicate what sorts of interpersonal sequences might have this effect on one or on both of the participants.

A context of learning is a definitely structured segment of the stream of interchange between persons. We know from experimental data that while the structuring of contexts of learning is extremely variable, some structuring is always present. The events of which the context is composed -- conditioned stimulus, response and reinforcement -- may be variously related to each other and still constitute a structured whole. That is, we are here dealing with Gestalten (units of the interchange) and are therefore again face to face with the peculiar nature of all such units. Although they are in large part the creation of the individuals concerned, and are necessarily a product of the ways these individuals perceive and punctuate what is happening, their perception is inevitably guided by culture and convention. Such perception may be rigid or it may be flexible. But the essential fact is that the rules for this punctuation are a part of that system of pacts and premises upon which communication is based, and for their learning communication must be viewed as a sequence of contexts.

What I am describing is a strangely retroflexed procedure; a process which is in a way folded back upon itself. This may be said in many ways and perhaps most simply by stating that the communicational stream is a sequence of contexts both of learning and of learning to learn.

At this point, the phrase, "distortion of the learning processes" takes on meaning. It would refer to all those cases in which an individual punctuates the stream of communication in a way different from his vis-a-vis but which are reinforced nonetheless by the pain resulting from his idiosyncratic view. From the point of view of the speaker, it will seem to him that he has incurred punishment for what he thought he was communicating, whereas he is in face being punished for what his messages seemed to be, as perceived by the other.

It is clear that his line of thought, if substantially correct, will lead to a formal theory of stability and instability in human relations. We might, therefore, inquire into what the engineers call criteria of stability. Is it possible to classify the degree and orders of misunderstanding in such a way as to separate those conditions which will be corrected by the participants, so that the system continues in a steady state, from those others which lead to a progressive deterioration? At the present time such a question can only be posed in the most general terms and no meaningful answers can be imagined. One relevant matter must, however, be mentioned -- that we deal with entities whose behavior is by no means describable in terms of linear equations or monotone logic. What actually seems to happen in many instances is that when what seems to be progressive change sets in, the situation becomes more or less intolerable for one or both persons and some sort of climactic outburst occurs. Following this, the system either returns to a state which existed before the change began or entirely new patterns of communication may be evolved. There are, after all, larger and longer sequences of interchange than any which we meet within the brief spans of data upon which this book is based.

From what little we know of the relationship between the fine details cf human interaction and the longer cycles of the career line, there is reason to expect that the longer cycles will always be enlarged repetitions or repeated reflections of pattern contained in the fine detail. Indeed, this assumption that the microscopic will reflect the macroscopic is a major justification of most of our test procedures. A major function of the techniques of microanalysis is, therefore, to obtain from small quantities of data, accurately and completely recorded, insights into human relationship which could otherwise only be obtained either by long-time observation or from the notoriously unreliable data of anamnestic reconstruction.

In sum, we are concerned in this book to present the techniques for the microscopic examination of personal interaction. While, of course, the words that people say to each other have importance, the question with which we are concerned, the problem of describing the relationship between persons, is not a question which can be answered by any summary of the dictionary meaning of their messages. There is a vast difference between the mechanical description "A gave B such and such information" and the description of the interchange "A answered B"s question immediately." The ultimate goal of the procedures outlined in this book is a statement of the mechanism of relationships. No statement of mechanism without larger context can be of long-term interest; no statement of relationship, unsubstantiated by a statement of mechanism, can warrant confidence. In order to trace the path from mechanism to validated relationship, it is first necessary to lay out for the reader some description of how the flow of linguistic and kinesic material can be systematically described.