CYBERNETICS

CIRCULAR CAUSAL AND FEEDBACK MECHANISMS

IN BIOLOGICAL AND SOCIAL SYSTEMS

I
Transactions of the Eighth Çonference O Ñ C YÉEIR-YÍT-Í-ICS .
March 13-16, 1931, New York, N. Y.

Edited by
HEINZ VON FOERSTER
DEPARTMENT OF ELECTRICAL ENGINEERING
UNIVERSITY OF ILLINOIS
CHAMPAIGN, ILLINOIS

Assistant Editors

MARGARET MEAD
AMERICAN MUSEUM OF NATURAL HISTORY
NEW YORK, N. Y.

HANS LUKAS TEUBER

DEPARTMENT OF NEUROLOGY
NEW YORK UNIVERSITY COLLEGE OF MEDICINE
NEW YORK, N. Y.

Sponsored by the

JOSIAH MACY, JR. FOUNDATION

•/

565 PARK AVENUE, NEW YORK 21, N. Y.

 PARTICIPANTS
Eighth Conference on Cybernetics

members
WARREN S. McCULLOCH, Chairman
Department of Psychiatry, University of Illinois College of Medicine
Chicago, 111.
HEINZ VON FOERSTER, Secretary
Department of Electrical Engineering, University of Illinois
Champaign, 111.
GREGORY BATESON*
Veterans Administration Hospital
Palo Alto, Calif.
ALEX BAVELAS
Department of Economics and Social Science, Massachusetts Institute of Technology
Copyright 1952, by the Cambridge, Mass.
JULIAN H. BIGELOW
JOSIAH MACY, JR. FOUNDATION Electronic Computer Project, Institute for Advanced Study
Princeton, N. J.
Price: $4.00 HENRY W. BROSIN
Western Psychiatric Institute and Clinics
Pittsburgh, Pa.
e\
azuiu
n LAWRENCE K. FRANK ^
(pr 72 Perry Street
New York, N. Y.
RALPH W. GERARD
Department of Physiology, School of Medicine, University of Chicago
Chicago, 111.

A 1 GEORGE EVELYN HUTCHINSON

'n 1 Department of Zoology, Yale University
New Haven, Conn.

HEINRICH KLÜVER
Division of the Biological Sciences, University of Chicago
Chicago, 111.
LAWRENCE S. KUBIE
Department of Psychiatry and Mentar Hygiene, Yale University School of Medicine
l i s ) New Haven, Conn.
RAFAEL LORENTE DE NO*
c ' Rockefeller Institute for Medical Research
New York, N. Y.
DONALD G. MARQUIS*
Department of Psychology, University of Michigan
Ann Arbor, Mich.
MARGARET MEAD
Printed in the United States of America American Museum of Natural History
New York, N. Y.
by Progress Associates, Inc., Caldwell, N. J.
F. S. C. NORTHROP*
Department of Philosophy, Yale University
New Haven, Conn.

* Absent
 WALTER Pirrs
Department of Mathematics, Massachusetts Institute of Technology
TABLE OF CONTENTS
Cambridge, Mass.

arturo s. rosenblueth
Departments of Physiology and Pharmacology, Instituto Nacional de Cardiologia
Mexico City, D. F., Mexico

leonard j. savage Josiah Macy, Jr. Foundation Conference Program:

Committee on Statistics, University of Chicago
Chicago, 111. Frank Fremont-Smith vii
theodore c. schneirla*
American Museum of Natural History A Note by the Editors;
New York, N. Y.
Heinz Von Foerster, Margaret Mead, Hans Lukas Teuber xi
hans lukas teuber
Department of Neurology, New York University College of Medicine
New York, N. Y. Communication Patterns in Problem-Solving Groups:
gerhardt VON bonin Alex Bavelas :: 1
Department of Anatomy, University of Illinois College of Medicine Group Interchange
Chicago, 111.
john von neumann*
Department of Mathematics, Institute for Advanced Study Communication Between Men: Meaning of Language:
Princeton, N. J. Ivor A. Richards 45
norbert wiener* Group Interchange
Department of Mathematics, Massachusetts Institute of Technology
Cambridge, Mass.
Communication Between Sane and Insane: Hypnosis:
Lawrence S. Kubie 92
GUESTS
Group Interchange
herbert g. birch
Department of Psychology, City College of New York
New York, N. Y. Communication Between Animals:
john r. bowman Herbert G. Birch ^ 134
Department of Physical Chemistry, Mellon Institute of Industrial Research Group Interchange
University of Pittsburgh
Pittsburgh, Pa.
donald m. mACkay Presentation of a Maze-Solving Machine:
Department of Physics, King's College, University of London Claude A. Shannon - 173
London, England
Group Interchange
ivor a. richards
Graduate School of Education, Harvard University
Cambridge, Mass. In Search of Basic Symbols:
david MCKenzie rioch Donald M. MacKay 181
Division of Neuropsychiatry, Army Medical Service Graduate School
Army Medical Center Group Interchange
Washington, D. C.
claude shannon Appendix I
Murray Hill Laboratory, Bell Telephone Laboratories Inc.
Murray Hill, N.J.
The Nomenclature of Information Theory:
Donald M. MacKay 222
Absent

Appendix II
Josiah Macy, Jr. Foundation
References 236
frank fremont-smith, Medical Director
janet freed, Assistant for the Conference Program V
... e îj-"^

JOSIAH MACY, JR. FOUNDATION

CONFERENCE PROGRAM

As AN INTRODUCTION to these Transactions o£ the Eighth Conference

on Cybernetics I should like to outline what it is that the Foundation
hopes to accomplish in its Conference Program. We are interested,
first of all, in furthering knowledge about cybernetics, and to this end
the participants were brought together to exchange ideas, experiences,
data, and methods. In addition to this particular goal, however, there
is a further, and perhaps more fundamental, aim which is shared by
all our conference groups: the promotion of meaningful communica­
tion between scientific disciplines.
The problem of communication between disciplines we feel to be a
very real and very urgent one, the most effective advancement of the
whole of science being to a large extent dependent upon it. Because of
the accelerating rate at which new knowledge is accumulating and
because discoveries in one field so often result from information gained
in quite another, channels must be established for the most relevant
dissemination of this knowledge.
The increasing realization that nature itself recognizes no boundaries
makes it evident also that the continued isolation of the several branches
of science is a serious obstacle to scientific progress. Particularly is it
so in medicine that the limited view through the lens of one discipline
is no longer enough. For example, today medicine must be well versed
in nuclear physics because of the tracer techniques and the injury which
can result from radiation. At the other extreme, medicine is certainly
a social science and, through mental health, must be concerned with
economic and social questions. The answer, then, is not further frag­
mentation into increasingly isolated specialities, disciplines, and depart­
ments, but the integration of science and scientific knowledge for the
enrichment of all branches. This integration, we feel, can be encour­
aged by providing opportunities for a multiprofessional approach to
given topics.
Although the fertility of the multidiscipline approach is recognized,
adequate provision is not made for it by our universities, scientific soci­
eties, and journals. And -perhaps the presence of other hindering fac­
tors must be admitted. Partly semantic in nature, they may also to some
degree be psychological. Admittedly, it is oftentimes difficult to accept
data derived from methods with which one is unfamiliar. By making
vii
free and informal discussion the central core of our meetings, we hope Unfortunately, however, in the finished products which are presented
to achieve an atmosphere which minimizes as much as possible these to the world through research reports this integral part of scientific
emotional barriers. endeavor is shriveled by the cold, white light of logic. By preserving
Thus, our meetings are in contrast to the usual scientific gatherings. the informality of our conferences in the published transactions, we
They are not designed to present neat solutions to tidy problems but to hope to give a truer picture of what actually goes on in the minds of
elicit provocative discussion of the difficulties which are being encoun­ scientists and of the role which creativity plays.
tered in research and practice. For this reason, we ask that the presen­
FRANK FREMONT-SMITH, M.D.
tations be relatively brief and that emphasis be placed on discussion as
Medical Director
the heart of the meeting. Our hope is that the participants will come
prepared not to defend a single point of view but to take advantage of
the meeting as an opportunity to speak with representatives of other
disciplines in much the same way that they would talk with their own
colleagues in their own laboratories.
We have, now, thirteen groups functioning under the Conference
Program. The following topics are covered: adrenal cortex, aging,
blood clotting, cold injury, connective tissues, consciousness, cyber­
netics, infancy and childhood, liver injury, metabolic interrelations,
nerve impulse, renal function, and shock and circulatory homeostasis.

When a new conference is organized, the Chairman, in consultation

with the Foundation, selects fifteen scientists to be the nucleus of the
group, and every effort is made to include representatives from all
pertinent disciplines. From time to time new members are added by
the group to fill gaps in viewpoint or technique. A limited number of
guests are invited to attend each meeting, but, for the purpose of pro­
moting full participation by all members and guests, attendance at any
meeting is limited to twenty-five. It is inevitable that in no topic can
we possibly include more than a small fraction of the key investigators
in the field, and one of the difficulties in forming a group like this is
that it is necessary to leave out so many people whom we would like
to include.
The transactions of these meetings are recorded and published. This
is done because the Foundation wishes to make current thinking in a
field available to all those working in it, and because it believes that
conveying to those in other fields who are concerned with science, for
example, government officials, administrators, etc., the essential nature
of scientific research is also an important problem in communication.
Logic is a vital aspect of science, but equally essential is the intuitive
or creative aspect. Research is as creative as the painting of a portrait
or the composing of a symphony. Although logic is, of course, neces­
sary in order to rearrange, to test, and to validate, research thrives on
creativity which has its source in unconscious, nonrational processes.
viii ix
A NOTE BY THE EDITORS
To THE READER of this somewhat unusual document, a few words of
explanation, and caution. This is not a book in the usual sense, nor
the well-rounded transcript of a symposium. These pages should rather
be received as the partial account of conversations within a group, a
group whose interchange actually extends beyond the confines of the
two day meeting reported herein. This account attempts to capture a
fragment of the group interchange in all its evanescence, because it
represents to us one of the few concerted efforts at interdisciplinary
communication.
The members of this group share the belief that one can and must
attempt communication across the boundaries, and often chasms, which
separate the various sciences. The participants have come from many
fields; they are physicists, mathematicians, electrical engineers, physi­
ologists, neurologists, experimental psychologists, psychiatrists, sociolo­
gists, and cultural anthropologists. That such a gathering failed to
produce the Babylonian confusion that might have been expected is
probably the most remarkable result of this meeting and of those
which preceded it.
This ability to remain in touch with each other, to sustain the dia­
logue across departmental boundaries and, in particular, across the gulf
between natural and social sciences is due to the unifying effect of cer­
tain key problems with which all members are concerned: the problems
of communication and of self-integrating mechanisms. Revolving
around these concepts, the discussion was communication about com­
munication, necessarily obscure in places and for more than one reason.
Yet the actual outcome was far more intelligible than one might think,
so that the editors felt enjoined to reproduce the transcript as faithfully
as possible.
The social process, of which these transactions are an incomplete
residue, was not a sequence of formal "papers" read by individual par­
ticipants and punctuated by prepared discussion. With few exceptions
people spoke freely and without notes. Unavoidably some speakers
produced inaccurate memories of their own facts, or of those of others,
and trends of thought were often left incomplete. The printed record
preserves the essential nature of this interchange in which partial asso­
ciations were permitted on the assumption that closure would take
place, at some other time, producing new ideas or reinforcing those
that were thought of in passing.
Stimulation, for many scientists, comes from such partial, and some­
times even inaccurate, reproductions of material from widely separated
fields, fields which seem dissimilar except for the logical structure of
xi
their central problem. If the reader wishes a formal statement of the almost complete absence of an idiosyncratic vocabulary. In spite of
work and point of view of individual participants, he will have to con­ their six years of association, these twenty-five people have not devel­
sult other sources. This can be done with ease since most of the contrib­ oped any rigid, in-gróup language of their own. Our idioms are limited
utors have provided references to previously published material. to a handful of terms borrowed from each other: analogical and digital
The reader should be warned that the presentations and discussion devices, feedback and servomechanisms, and circular causal processes.
tend to be responsive to previous meetings of the group. Some state­ Even these terms are used only with diffidence by most of the members,
ments were designed to answer questions asked months or years before, and a philologist given to word-frequency counts might discover that
or designed to evoke some long-anticipated answer from a fellow mem­ the originators of "cybernetics" use less of its lingo than do their more
ber. Radical changes in the manuscript would have been necessary had recent followers. The scarcity of jargon may perhaps be a sign of
we attempted to rid the group of its history. Such changes would have genuine effort to learn the language of other disciplines, or it may be
been distortions, and would prevent the reader from noticing the unfin­ that the common point of view provided sufficient basis for group
ished state of the group's affairs. coherence.
Our editorial procedure, nevertheless, involved some revision of the This common ground covered more than the mere belief in the worth-
transcript prior to publication. A verbatim record based on the steno- whileness of interdisciplinary discussion. All of the members have an
typed protocol, even if perfect, would in fact have been an incomplete interest in certain conceptual models which they consider potentially
and misleading account. It would have given the verbal content, but applicable to problems in many sciences. The concepts suggest a similar
the tones of voice, the gestures, the attention directed by the turn of approach in widely diverse situations; by agreeing on the usefulness of
head toward one person or another would all be missing. For this these models, we get glimpses of a new lingua franca of science, frag­
reason, we adopted a more traditional procedure. Each participant was ments of a common tongue likely to counteract some of the confusion
supplied with a mimeographed copy of his transcribed remarks, and and complexity of our language.
was given the chance to revise his material for the sake of clarity and Chief among these conceptual models are those supplied by the
coherence. Not all of the participants availed themselves of this oppor­ theory of information.^' ^ This theory has arisen under the pressure
tunity so that our working copy represented a mixture of revised and of engineering needs; the efficient design of electronic communication
unrevised contributions. In the unrevised passages the editors corrected devices (telephone, radio, radar, and television) depended on achieving
only those statements which seemed to them obvious errors in record­ favorable ""signal-to-noise ratios." Application of mathematical tools
ing. For the rest, they confined their censorial activities to occasional to these problems had to wait for an adequate formulation of ""informa­
deletions of overlapping or repetitive passages and of a few all-too- tion" as contrasted to ""noise."
cryptic digressions. Most of the asides, such as jokes or acidities, were If noise is defined as random activity, then information can be con­
preserved as long as they seemed intelligible to people outside the sidered as order wrenched from disorder, as improbable structure in
group. contrast to the greater probability of randomness. With the concept of
The editors were eager to retain the participants' first names in the entropy, classical thermodynamics expressed the universal trend toward
printed record when they had been used during the discussion, but this more probable states: any physiochemical change tending to produce a
would have been an imnecessary handicap to the reader. However, the more nearly random distribution of particles. Information can thus be
reader should realize that most speakers addressed each other informally formulated as negative entropy, and a precise measure of certain classes
as a consequence of acquaintance outside the framework of these par­ of information can be found by referring to degrees of improbability of
ticular meetings. The occasional shifts to more formal modes of ad­ a state.
dressing each other was therefore indicative of distance and sometimes The improbable distribution of slots in a slotted card, or the improb­
of disagreement. The use of last names also underscored the special able arrangement of nucleic acids in the highly specific pattern of a
role of the invited guests and of the subtle differences in pace and tone ^SHANNON, C. E.: A mathematical theory of communication. Bell System. Tech. J. 27, 379-
which some of them introduced into the meeting. 423 and 623-6S6 (1948).
It is noteworthy that these invited guests cannot be identified by any ^SHANNON, C. E., and WEAVER, W.: The Mathematical Theory of Communication. Urbana:
obvious differences between their vocabulary and that of the regular University of Illinois Press, 1949 (p. 116).

members. One of the most surprising features of the group is the ^WIENER, N.: Cybernetics or Control and Communication in the Animal and the Machine.
New York: Wiley, 1948 (p. 194).

xii xiii
«
gene both can be considered as "coded" information, the one decoded of Gibbsian physicochemical systems^^ to the steady-states of human
in the course of a technical (cultural) process, the other in the course blood," and to integration in social groups, down to miniature social
of embryogeny. In both instances, that of the slotted card and that of systems." Quite independently, social scientists had been tending in
the gene, we are faced not only with carriers of information but with the same direction, as witnessed by the work of the functional anthro­
powerful mechanisms of control: the slotted card can control long pologists Radcliffe-Brown" and Bateson.^®- " In ecology, the concept
series of processes in a plant (without itself furnishing any of the requi­ of circular causal systems has been employed by Hutchinson,^® and
site energy) ; the gene, as an organic template, somehow provides for further applications in statistical biology and genetics can be expected.
its own reproduction and governs the building of a multicellular organ­ The remarkable constancy in the concentration of certain substances
ism from a single cell. In the latter case, mere rearrangement of sub- in the fluid matrix of the body led Claude Bernard originally to posit
microscopic particles can apparently lead to mutations, improving or the fixity of the "milieu intérieur" as one of the elementary conditions
corrupting the organism's plans as the case may be. Such rearrangement of life." Cannon^® designated as "homeostasis" those functions that
may indeed be similar to the difference brought about by the transposing restore a disturbed equilibrium in the internal environment—the com­
of digits in numbers, 724 to 472, or by transposing letters in words such plex self-regulatory processes which guarantee a relative constancy of
as art and rat.^ blood sugar level, of osmotic pressure, of hydronium ion concentration,
Extension of information theory to problems of language structure or of body température. Many of these processes are at least partially
has been furthered by psychologists and statisticians.®- ^ There are understood, but, as Klüver^^ has pointed out, we know next to nothing
unexploited opportunities for additional applications of the theory in of the physiological functions which underlie our perceptual "con­
comparative linguistics, and more particularly in studies of the pathol­ stancies."
ogy of language. Yet available work is sufficient to show how com­ Normal perception is reaction to relations, to universals such as
munication considered from this standpoint can be investigated in size, shape, and color. Perceived objects tend to remain invariant in
mechanical systems, in organisms, in social groups;^" and the logical their size while distance from the observer varies; perceived shapes and
and mathematical problems that go into the construction of modern colors retain subjective identity in varying positions and under varying
automata, in particular the large electronic computers,^^ have at least illumination. This crucial problem for the physiological psychology of
partial application to our theorizing about nervous systems and social perception is rarely faced^^ and the neural correlates for our reaction
interactions. to universals are still stib jtidice.
A second concept, now closely allied to information theory, is the Recent attempts at identifying a possible neural basis for our re-
notion of circular causal processes. A state reproducing itself, like an
organism, or a social system in equilibrium, or a physiochemical- I^GLBBS, J. W.: On the equilibrium of heterogeneous substances. The Collected Works. Vol. 1.
aggregate in a steady-state, defied analysis until the simple notion of Thermodynamics. New York: Longmans, 1928 (pp. 5 5-371).
I^HENDERSON, L. J.: Blood: A Study in General Physiology. New Haven: Yale University Press,
one-dimensional cause-and-effect chains was replaced by the bidimen-
1928 (p. 390).
sional notion of a circular process. The need for such reasoning was 14 ; physican and patient as a social system. New England J. Med. 212, S19 (1935).
clear to L. J. Henderson, the physiologist, when he applied the logic I5RADCLIFFE-BRO-WN, A. R.: The Andaman Islanders. A Study in Social Anthropology. Cam­
bridge, England; Cambridge University Press, 1922 (pp. XIV and 504).
i^Bateson G.: Naven: A Survey of the Problems Suggested by a Composite Picture o/ the
^GERARD, R. W.: Unresting Cells. New York: Harper, 1940. Culture 'of a New Guinea Tribe, Drawn from Three Points of View. Cambridge: Cambridge
^FRICK, F. C., and MILLER, G. A.: Statistical behavioristics and sequences of responses. Psychol. University Press, 193 6 (p. 286).
Rev. 56, }I1 (1949). 17 ; Bali: the value system of a steady state. In: Social Structure. Studies Presented
"MILLER, G. A.: Language and Communication. New York: McGraw-Hill, 1951. to A. R. Radcliffe-Brown. Oxford University Press, 1949 (p. 3 5).
^MILLER, G. A., and SELFRIDGE, J. A.: Verbal context and the recall of meaningful material. ^®Htn-CHINSON, G. E.: Circular causal systems in ecology. Ann. New York Acad. Sc. 50, 221
Am. J. Psychol. 63, 176 (1950). (1948).
^NEWMAN, E. B.; Computational methods useful in analyzing series of binary data. Am. J. I!>BERNARD, C.: Leçons sur les phénomènes de la vie communes aux animaux et aux végétaux.
Psychol. 64, 252 (1951). Two volumes. Paris: J. B. Baillière, 1878-79.
0 ; The pattern of vowels and consonants in various languages. Ibid., }69. ^oCannon, w. B.: The Wisdom of the Body. New York: Norton, 1932 (p. xv and 312).
"BAVELAS, A.: A mathematical model for group structures. Appi. Anthropol. 7, (part i), 16 2iKlOver, H.: Behavior Mechanisms in Monkeys. Chicago: University of Chicago Press, 1933
(1948). (p. xviii and 3 87).
'IVON NEUMANN, J.: The general and logical theory of automata. Cerebral Mechanisms in 22 ; Functional significance of the gcniculo-striate system. Biol. Symposia 7, 253
Behavior (The Hixon Symposium). Jeffress, L. A., editor. New York: Wiley, 1951 (pp. 1-41). (1942).

xiv XV
actions to universals'''' have adduced hypothetical sustained activity in spindles) which monitor a state of tension in the muscle and, by in­
neuronal circuits as one of the prerequisites for the central processes creasing their rate of centripetal firing, set off centrifugal volleys which
which guarantee perceptual constancies. Persistent circular activity in shorten the muscle in response to imposed stretch.
nervous nets had been postulated on theoretical grounds by Kubie^^ Numerous analogues for such recalibrating mechanisms can be found
over twenty years ago, thereby anticipating the subsequent empirical in those modern electronic devices in which output is regulated by
demonstration of such reverberating circuits by Lorente de Nó.^® The constant comparison with input. The automatic volume control circuit
importance of Lorente de No's disclosures for neurological theory lies of a radio receiver prevents ""blasting" by decreasing the volume as the
in the fact that, earlier in the century, many investigators considered signal is increased and counteracts "fading" by increasing the volume.
the central nervous system as a mere reflex-organ; the mode of action A speed control unit slows a motor down when its revolutions exceed
of this organ, despite all the evidence to the contrary, was viewed as a desired value and speeds it up when revolutions fall below this value.
limited to the relating of input to output, stimulus and response cor­ Such "feedback" or ""servomechanisms"®" are man-made models of
responding to cause and effect. The possibility of self-sustained central homeostatic processes. They are not exclusively found among electronic
activity in the nervous system was overlooked. Thence the denial of devices. In the days of the thermal engine, MaxwelP^ developed the
memory-images in early behaviorism, the emphasis on chain reflexes in theory of the mechanicàl "governor" of steam engines. Small versions
attempts at explaining coordinated action. of this governor are still found today in old-fashioned phonograph
To this day, many psychologists tend to see the prototype of all turntables. Two massive metal spheres are suspended by movable links
learning in elementary conditioned reflexes, a tendency which cannot from a vertical shaft which spins with the main shaft of the machine.
be understood unless one assumes, with Lashley,^® that these psycholo­ As speed increases, centrifugal force drives the metal spheres apart and
gists are still handicapped by "peripheralistic" notions, the unwarranted increases the drag on the shaft; the machine slows down. Again, with
idea that central nervous activity cannot endure in the absence of con­ the spheres sinking low, the drag on the shaft is decreased and the
tinuing specific input from the periphery. Undoubtedly, the action of machine speeds up. Such a governor insures approximate constancy of
reverberating circuits can be overgeneralized, and has been abused. speed in the engine, by the simplest mechanical means, and, in contrast
Long-range memory may need more permanent neural changes, but the to many more complicated devices, the mechanical governor shows little
notion of such circuits has suggested models of neural activity which likelihood of going into uncontrollable oscillations.
are potentially testable and therefore of value.^''' Recent complex electronic devices are not only '"error-controlled
Activity in closed central loops thus has to be carefully distinguished (like a mechanical governor), but can be so built as to ""seek" a certain
from the older and simpler notions of neural circuits, circuits which join state, like ""goal-seeking" missiles which predict the future position of
periphery and centers, as in the classical conception of postural reflexes. a moving target (at time of impact) by extrapolation from its earlier
Sir Charles BelP® spoke of a ""nervous circle which connects the volun­ positions during pursuit. Such devices embody electronic computing
tary muscles with the brain." Through such circuits, muscles in a limb circuits, and the appearance of "purpose" in their behavior (a feedback
maintain a given tension, as long as motor impulses flow into the over the target!) has intrigued the theorists®^-and prompted the
muscles according to the sensory signals which issue from these same construction of such likeable robots as Shannon s electronic rat de­
muscles. The idea antedated the discovery of the sense organs (muscle scribed in this volume.
The fascination of watching Shannon's innocent rat negotiate its
23pjtts, W., and McCulloch, W. S.: How we know universals; the perception of auditory and maze does not derive from any obvious similarity between the machine
visual forms. Bull. Math. Biophys. 9, 127 (1947).
^^Kubie, L. S.: a theoretical application to some neurological problems of the properties of and a real rat; they are, in fact, rather dissimilar. The mechanism, how­
excitation waves which move in closed circuits. Brain 53, 166 (1930). ever, is strikingly similar to the notions held by certain learning the­
^^Lorente de No, R.: Analysis of the activity of the chains of internuncial neuronç. /. Neu-
rophysiol. 1, 207 (1938). orists about rats and about organisms in general. Shannon's construc-
2®Lashley, K. S.: Discussion. In: Cerebral Mechanisms in Behavior (The Hixon Symposium).
Jeffress, L. A., editor. New York: Wiley, 1951 (p. 82). 3»MacColl, L. R.; fundamental Theory of Servo-Mechanisms. New York: VanNostrand, 1945.
2'^McCulloch, W. S.; A heterarchy of values determined by the topology of nerve nets. Bull. ^^Maxwell, C.: On governors. Vroc. Koy. Soc. 16, 270 (1868).
Math. Biophys. 7, S9 (1945).
32ROSENBLUETH, A., WiENER, N., and BiGELO\P, J.: Bchavior, purpose and teleology. Philos.
2®Hebb, D. O.: Organization of Behavior: A Neuropsychological Theory, New York: Wiley,
1949 (p. XIX and 335). of Sc. 10, IS (1943).
-^Bell, C.: On the nervous circlc which connects tlie voluntary muscles with the brain. Pror. 3-^Northrop, F. S. C.: The neurological and behavioristic psycliological basts of the ordering of
Koy. Soc. 2, 266 (1826). society by means of ideas. Science 107, 411 (1948).

xvi xvii
tion serves to bring these notions into bold relief. analogues requires experiment. These experiments would not have been
Recent emphasis on giant electronic computers as analogues of the considered if the analogue had not been proposed, and new observations
human brain should perhaps be considered in the same light. The on biological and social systems result from an empirical demonstration
logical and mathematical theories demanded by the construction of of the shortcomings of our models. It is characteristic that we tend to
these computers raise problems similar to those faced on considering think of the intricacies of living systems in terms of non-living models
certain aspects of the nervous system or of social structures.^^ It is no which are obviously less intricate. Still, the reader will admit that, in
accident that John Von Neumann, a mathematician who is currently some respects, these models are rather convincing facsimiles of organ-
concerned with the theory of computers, should be more generally ismic or social processes—^not of the organism or social group as a
known for his analysis of human interaction in games and economic whole, but of significant parts.
behavior,®^ and that Norbert Wiener, after working on computers and How this way of thinking emerged in the group is difficult to recon­
guided missiles, turned to the consideration of the social significance struct. From the outset, John Yon Neumann and Norbert Wiener furn­
of these mechanisms.®® Brief consideration of computers may therefore ished the mathematical and logical tools. Warren McCulloch, as the
be in order. group's "chronic chairman" infused it with enthusiasm and insisted on
Computers are constructed on either of two principles: they may be not respecting any of the boundaries between disciplines. The Josiah
digital or analogical. In an analogical device, numbers are represented Macy, Jr. Foundation, through Dr. Frank Fremont-Smith, provided the
by a continuous variation of some physical quantity, a voltage, say, or physical setting but actually much more than that: the social sanction
a distance on a disc. A digital device, however, represents numbers as for so unorthodox an undertaking. The confidence of the Foundation
discrete units which may or may not be present, e.g., a circuit that may and of Dr. Frank Fremont-Smith made it possible to obtain a type of
be open or closed, and the basic alphabet of the machine may be a cross-fertilization which has proved rewarding over a period of six
simple yes or no, zero or one. years.
Peripheral neurons act on an all-or-none principle, and synapses in The gradual growth of the principles can be recognized from dates
the central nervous system are frequently considered to act similarly. and titles of successive Conferences. A nucleus of the current Cyber­
Theories of central nervous activity have consequently often paralleled netics Conference seems to have been formed in May 1942 at the
those required for digital rather than analogue machines (cf., Pitts and Macy Foundation Conference on "Cerebral Inhibition." Among the
McCulloch^®). The applicability of digital notions to the actions of the participants were: Gregory Bateson, Lawrence K. Frank, Frank
central nervous system has been questioned,®® but the calculus worked Fremont-Smith, Lawrence Kubie, Warren McCulloch, Margaret Mead,
out for handling them is certainly applicable to electronic digital com­ and Arturo Rosenblueth. All of these later became members of the
puters,®^ and the very fact that testable theories of nerve action have continuing group devoted to the discussion of "cybernetics." The publi­
been proposed is due to the availability of the electronic models. cation of the article on "Behavior, Purpose and Teleology" by Rosen­
We all know that we ought to study the organism, and not the com­ blueth, Wiener, and Bigelow in 1943®^ focused attention on several of
puters, if we wish to understand the organism. Differences in levels of the problems which led to the organization of the first Macy Founda­
organization may be more than quantitative.®^ But the computing robot tion Conference in March 1946 devoted to "Feedback Mechanisms and
provides us with analogues that are helpful as far as they seem to hold, Circular Causal Systems in Biological and Social Systems."
and no less helpful whenever they break down. To find out in what The fall of 1946 foimd the group very active. Two meetings spon­
ways a nervous system (or a social group) differs from our man-made sored by the Macy Foundation followed rapidly upon each other: first,
in September, a special meeting on "Teleological Mechanisms in Soci­
34VON NEUMANN, J., and MORGENSTERN, O:. Theory of Games and Economic Behavior, ety"; then, in October, the second Conference on Teleological Mecha­
Princeton: Princeton University Press, 1944.
söWiENER, N.: The Human Use of Human Beings. Boston: Houghton MifiQin, 1950.
nisms and Circular Causal Systems. Next, the group formed the nucleus
2®GERARD, R. W.: Some of the problems concerning digital notions in the central nervous sys­ of a formal symposium on "Teleological Mechanisms" held under the
tem. Cybernetics, von Foerster, H., editor. Trans. Seventh Conf. New York: Josiah Macy, auspices of the New York Academy of Sciences.®®
Jr. Foundation, 1950 (p. 11).
In the following year, 1947, the third Macy Foundation conference
^^MCCULLOCH, W. S., and PITTS, W.: A logical calculus of the ideas immanent in nervous
activity. Bull. Math, Biophys. 5, i 15 (1943). was held, retaining the title of the second meeting; the fourth and
^^ScHNEiRLA, T. C.: Problems in the biopsychology of social organization. /. Abn. Soc. Psychol.
41, 385 (1946). s^Frank, L. K., et al.: Teleological mechanisms. Ann. New York Acad. Sc. 50, IS9 (1948).

xviii xix
fifth conferences, in 1948, were entitled: Conference on Circular Causal COMMUNICATION PATTERNS IN
and Feedback Mechanisms in Biological and Social Systems. The fifth
conference concerned itself particularly with considerations of the
PROBLEM-SOLVING GROUPS
structure of language.
With the publication of Norbert Wiener's book Cybernetics, a term
appeared which was unanimously chosen as title for the sixth confer­ ALEX BAVELAS
ence in the spring of 1949. The title Cybernetics was maintained for Department of Economics and Social Science
the seventh and the present eighth conference 1950, 1951 with the sub­ Massachusetts Institute of Technology
title Circular Causal and Feedback Mechanisms in Biological and Social
Systems.
Through the fifth conference, no transactions were published. With I AM GOING to describe two very simple experiments which are related
the sixth conference (1949) our program of publication began, so that and which I have chosen because I think they give a good picture of
the two preceding conferences, the sixth^" and seventh,^^ are available the way we are trying to get acquainted with our problem. I believe
in print. The reader might suspect that this urge to fix the group they give an idea of the motivation and the spirit of the work. They
process in printed form is the beginning of fossilization. He may be are not elegant experiments, and the work is in an exploratory stage.
right, but we prefer to think of it in terms more favorable to the group. We are not striving for niceness of design but, rather, for a maximum
For well over two thousand years, the "symposium" has been a set­ of interplay between the experimenter and his material.
ting for the matching and sharpening of ideas. Evolved from the Attic I should like to state the problem, describe the experiments, and then
stage, the literary form created by Plato has persisted through the tell you about some of the notions from which we are trying to build a
Middle Ages and the Renaissance. Until recently, it was unrestrained theoretical framework for understanding what is happening. The
by stenotypists and tape recorders. Few of the classical symposia were problem is this: If a task is of such a nature that it must be performed
anything but prose poems of one man's making. They brought a simple by a group rather than by a single individual, communication is usually
message, stated in contrapuntal fashion. Now we have our modern necessary; but does it make any difference who may communicate with
devices for the recording and storing of information. Communication whom.? In other words, if I draw circles to indicate people, and draw
transmitted has been infinitely multiplied in volume, but the thinking arrows between people (Figure 1) showing that a message may go
of simplifying ideas has not kept pace. from one to the other, does it make any difference, in a group situation,
Whether the meetings here recorded contain such simplifying ideas, where and in what direction these arrows go ì The experiments I shall
the editors would not presume to say. Some of us believe we can see describe deal entirely with a special set of such patterns which we call
such ideas here and there. For this reason, we preserved the record, and "connected groups." By "connected group" we mean such a group
exhibit it to others for their judgment. that, taking any pair of individuals, it is possible, over some route, for
Pressure of time made it impossible in the last two years to sum up the individuals that make that pair to exchange messages, i.e., the
the historical background and to formulate an editorial policy. The group shown in Figure 2 would not be a connected group, but the
published records of the sixth and seventh conferences, in 1949 and groups shown in Figures 3, 4, and 5 would be connected groups.
1950, therefore appeared without any introduction. We hope that this Of course, as long as there is an experimenter who imposes a task
note will serve as a preface to the earlier reports, as well as to the and who looks at the results later, I suppose we ought to indicate that
present publication. he is present as a sink and a source. We feel, however, that the experi­
heinz von foerster ments were so conducted that so far as the subjects were concerned,
margaret mead
hans lukas teuber they were operating without the experimenter in a sink-source role
New York, N. Y. for them.
January 29, 1952 If you were a subject in one of the experiments that I shall describe,
the first thing that would happen would be that I or some other mem­
^Cybernetics. Von Foerster, H., editor. Trans. Sixth Conf. New York: Josiah Macy, Jr.
Foundation, 1949. ber of the research group would appear in class about five minutes
^^Cybernetics, Von Foerster, H., editor. Trans. Seventh Conf. New York: Josiah Macy, Jr. after the class had begun. I would give you a souped-up story about a
Foundation, 19Î0.
XX
1
2 Cybernetics Communication Patterns 3

research experiment that was being carried on and I would explain

that it was supported by the Air Force, that it was quite important,
that it had to do with communication, and that we would like some
volunteers.
"When you arrived at the laboratory, you would be asked to sit down
and you would be presented with five cards. On each of those cards
would appear five symbols—perhaps a cross, an asterisk, a square, a
circle, and a triangle. Each set would be different. They would be so
arranged that each symbol would appear on four of the five cards, but
one symbol would appear on all five cards. You would be told to look
at those five cards and tell as quickly as you could which symbol was
common to the five cards. You would be asked to do this for several
sets of cards.
Then we would say: "What you have just done is the experimental
task. Instead of doing it yourself by being able to see all five cards,
the task will be done by five people. You will be separated from the
others, and each of you will have one of the five cards. By writing
notes to the others, you will determine as quickly as possible what the
common symbol is." Then you would be led to another room where
you and the other four subjects would sit around a table. The table
would have partitions built on it so that you couldn't see the others.
Each subject would have a little box full of cards of the same color as
the cubicle in which he was sitting. The cards would be numbered in
sequence. If you wanted to send a message you would write it on a
card and put it into the slot that had the color of its destination. You
would be instructed to keep all the messages you received; anything
you sent out would have to be on cards of the color assigned to you.
There would be no restriction as to what you might write on the
cards. You could write anything you pleased. As soon as you knew
the common symbol, you would turn to a little box with six switches
on it, each switch labeled with one of the six symbols, and you would
press the appropriate switch. You could send the answer out to the
others, too. These switches would operate a lightboard in another room
on which the responses Vere counted. A single trial would end as soon
as each man had a light on; in other words, it would end when each
man had pressed a switch.
Savage: Does each individual know by looking at his own setup
whom he can reach, and does he also know, presumably by having it
explained to him, whom the other participants can reach, or does he
have to learn through participation this kind of information ?
Bavelas: We began these experiments with the notion that we would
FIGURES l through 5. Reprinted by permission from BAVELAS, A. show each group what the network was. After they had sat at their
/. Acoustical So. Am. 22, 125 (1950). places, they would look at a little diagram which would show the net-
4 Cybernetics Communication Patterns 5
logical properties in each place are the same; that is, any differences
work; but we found that we couldn't discover a way of doing that
with respect to some of the things we were measuring and hoped to
without biasing the performance of the group. We decided finally to
locate in pattern relations should be zero here. We hoped that it would
tell the subjects nothing about the network, so that when each indi­
give us some notion of how big a difference was a real difference.
vidual sat down at the table all he knew was that there were two slots
Another pattern chosen is the one shown in Figure 4. The last one is
that were open, that one was colored green and the other blue, and
that when he sent a message in that slot it went to somebody designated shown in Figure 5.
In discussing the results, I am not going to present numbers; rather,
by blue; but he didn't know where that person was sitting, whether it
was next to him or not. with words, I shall give you a general notion of what seemed to hap­
Fremont-Smith: He has only two slots ?
pen, and I shall mention only those things for which the numbers give
us some reason to believe that the differences are significant.
Bavelas: He may have one, two, three, or four, depending on the
First, of all, there is the matter of speed. We measured speed by
net, but each man knows what he has. Gradually, they discover what
counting the number of seconds between the beginning of the trial and
the network is, in some patterns. In other patterns they never quite
know. the end of the trial. The pattern shown in Figure 3 turns out to be
quite slow; the pattern shown in Figure 5 is quite fast; and the one
Kubie: Some men will have one or two slots open, and others will
have three or four in the same trial ? shown in Figure 4 falls somewhere in between.
Fremont-Smith: Do you want to give us a rough idea of the duration
Bavelas: They might; the net is imposed in the beginning.
Kubie: But they don't all have the same number of open slots ? of a trial ?
Bavelas: Yes. Each group does fifteen trials. By the time they have
Bavelas: They might or might not.
done six, the picture with respect to time is fairly steady; that is, it
von Bonin: He doesn't know who is RedHe doesn't know with
whom Blue can communicate.? doesn't change much after that; in fact, it hardly changes at all. After
Bavelas: Not unless Blue tells him. Blue, of course, may tell him.
they have reached this steadiness, the pattern shown in Figure 3 is
This is a very interesting thing to watch. doing a trial at about 60 to 75 seconds.
Pitts: By the sixth trial they would be assumed to know the net­
Savage: How can Blue tell, if he doesn't know where he is himself Ì
work. Is that correct? Or does it change? The fifteen trials are all
Bavelas: What frequently happens is that a man's first message to
the next man is, "I have a square, and I can send messages only to you." run with the same connectivity ?
Bavelas: Yes, and the group is never used twice.
Bigelow: Does the curtain go up on everybody at the same moment?
Savage: So one group faces only one connectivity ?
Bavelas: You mean when we "uncover" the table?
Bigelow: Yes. Bavelas: Yes.
Savage: And then they are presumed to learn the connectivity in the
Bavelas: Yes. After they have taken their places, we uncover the
table and say, "Are you ready for trial No. 1 ?" They know what the early trials, so that in the later ones there are typically no messages
task is, and we have described the table with pictures, and so on, to saying, "I can speak to So-and-So" ?
them, so they know what to do. They go right to work. There are Bavelas: That's right.
rarely any mixups. Savage: And is it possible, at least legally possible, for them to form

Hutchinson: Is there any effect due to psychological associations with

a conspiracy, as it were; that is, to agree on a way of getting the thing
the color names ? over with quickly ?
Bavelas: Oh, yes, and of course, they do. They are working against
Bavelas: We don't know of any.
Savage: Are the individuals permitted, or even encouraged, to keep
time. They are told that other groups are doing this, that we are trying
notes and diagrams of their own? Does each have a blank sheet of to find the best group, and that they are to be compared with those
paper ? other groups.
Pitts: Do you find it takes longer to learn one connectivity than an­
Bavelas: They have blank paper and a number of pencils."
We chose three patterns of groups to experiment with; I shall report other, in the sense of leveling off ?
Bavelas: Some are never really well understood. In general, an indi­
only on these three, although we have done work with some others.
We picked one pattern, which is shown in Figure 3, because the topo- vidual in a net does not know what the net looks like, beyond two
6 Cybernetics Communication Patterns 1
transmission links; along the shortest path, of course. in the pattern shown in Figure 5, that after a few trials messages just
Mead: There is no illegality in giving people information ? go into the center and the answer comes back out ? Looking at it that
Bavelas: No. way, the pattern shown in Figure 3 never acquires anything you could
Mead: That is important to point out, I think. say was a stable pattern for sending the messages around.
Bavelas: There is no information that is barred. Bigelow: You mean stable within a single group or stable within
Savage: When you say some are not learned at all, do you mean that, group to group?
even among these three patterns, there are some that are not learned Bavelas: Within a single group. In other words, if you look at any
at all} one of the groups which operated in the "circle" pattern, you can't tell
Bavelas: Yes. In the pattern shown in Figure 4, the individual at from one trial to the next who is likely to send, in which of the two
one end doesn't usually know what is at the other end. directions he may send, or where the answer will occur first, and so on.
Bigelow: Whereas in the pattern shown in Figure 3 he would know Ì What the individuals in these groups tend to do is to send the informa­
Bavelas: Yes, and in the pattern shown in Figure 5 he would know. tion they have in both directions as fast as they can, and sooner or later
We have tried other nets, but by and large it looks as though more somebody gets all of it.
than two links away they really don't know; they guess. If you plot After the experiment is over, the subjects are interviewed, and one
the things they tell you, it is very hard to get a clear picture. question that is asked them is, "Did your group have a leader?" Now,
The pattern shown in Figure 5 averages 20 to 40 seconds per trial. perhaps one of the most interesting findings as a result of this question
The pattern shown in Figure 4 lies between the other two. is that nobody ever asks, "What do you mean by a leader?" Figure 6
Another thing we tried to measure was errors. If we look at the gives the percentage of responses which indicate a leader by position
errors made and average them for the nineteen groups and make a in the pattern.
graph, we find that they, too, become remarkably stable after the fifth Fremont-Smith: No special attachment to color?
or sixth trial. The relative differences are the same between the pat­ Bavelas: No.
terns. The pattern shown in Figure 3 makes very many errors. It stabi­ Savage: It is relative to ego. Is that right? In the first pattern it is
lizes at about 15 per cent. The pattern shown in Figure 5 stabilizes really perfectly symmetrical; and when you say there is such-and-such
below 1 per cent. It makes very few errors. And the pattern shown in relative per cent designated as leaders, they are so designated relative
Figure 4 lies between 4 and 5 per cent when it settles down. to ego, which is some place on the diagram ?
Savage: The group is never informed whether its last trials were Bavelas: Yes.
correct, so it can't learn by experience to avoid errors Ì Kubie: When you say 8, 12, 4, and 6, you are starting from some one
Bavelas: No. The group may infer that it has made an error, but individual viewpoint. You are starting from the viewpoint of 4, 12,
the experimenter never says, "You three made an error," or something 8, or 6?
of that sort.
Savage: Does the experimenter ever say, "You five made an error".?
Bavelas: No, he never does. But it is possible for them to know
that they have made an error from evidence inside the net.
Savage: How could that happen Ì It would mean talking in the next
trial about the last trial, which they presumably don't have time to do.
Bavelas: Well, it happens because an individual may throw his
switch but not be able to correct it in time. However, he gets conclu­
sive evidence just a moment later that it was wrong.
Another thing in which we were interested was what happens with
respect to the emergence of organization. We have looked at that in 12% 6% 0% 0% 0% 0%
two ways. First of all, if you plot the frequency with which messages
go from one place to another, do you get stability at all? In other FIGURE 6. Emergence of recognized leaders. Reprinted by permission
words, does an operational pattern emerge ? Do you find, for instance, from BAVELAS, A.: J. Acoustical So. Am. 22,125 (1950).
8 Cybernetics Communication Patterns 9

Bavelas: Well, each, person who said there was a leader was asked Savage: I suggest that it may be important, especially in so symmet­
who it was, and he said, "Red." We tabulate those. rical a situation.
Savage: Oh, you tabulate the color, not the position ? Bavelas: Yes, I agree with you.
Bavelas: The color, yes. Pitts: Then the percentages you have depend wholly upon color,
Fremont-Smith: What happens if he says his own color? upon whatever associations a man may have with colors?
Bavelas: We tabulate that color. Bavelas: Well, I don't know that color really enters into the choice.
Fremont-Smith: But you make no distinction ? Mead: Oh, no.
Bavelas: No. Bavelas: I regard the differences among the positions in the pattern
Savage: Should you not, in dealing with that pattern, recognize that in Figure 3 as pretty much zero differences—in the light of the dif­
it is really perfectly symmetrical (except for such connotations as may ferences one gets in the other patterns.
be attached to color names), and that therefore the position of the Pitts: You don't consider the differences of statistical significance
presumed leader relative to the individual who is speaking—that is, at all ?
the one I have been calling ego—is of central importance ? This general Bavelas: I wouldn't say that that is important.
idea seems to me of primary sociological and psychological interest here, Gerard: But those aren't fixed figures. You have just given a random
particularly in connection with highly symmetrical patterns. sample there.
Bavelas: You mean, we would like to know whether an individual Bavelas: They are pretty close to right, I think, for nineteen groups.
in a given position is more likely to indicate one of the adjoining men Rosenblueth: If they are right, they can only mean color selection.
as a leader ? I don't see what else they could mean.
Savage: Yes. Savage: They could be just at random.
Fremont-Smith: Or himself. Bavelas: There is a question as to whether this difference really
Savage: Yes, or himself. means anything.
Fremont-Smith: It seems to me it would have a particular signifi­ Gerard: Put it this way: Are the twelve always in the same color?
cance when he does indicate himself. That is the point.
Kubie: But he doesn't know who is adjoining. Bavelas: Of course not. I'm sorry. In tabulating the data, you have
Bavelas: Of course he doesn't know who is adjoining him. this difficulty, that the colors in position don't have any very good ref­
Fremont-Smith: But he does know himself. erence point unless you start with the ego point for reference. I am
Bavelas: Yes, he does know himself. sorry I don't have that analysis. Color and position were deliberately—
Fremont-Smith: Therefore, the man who chooses himself does Bigelow: Permuted?
something— Bavelas: Yes, permuted.
Bavelas: Oh, I can tell you that nobody chose himself. Savage: But, again, all that can be tabulated. The tabulations should
Fremont-Smith: Nobody did? Ah! be made with reference to position; thus you would be able to find out
Bavelas: Nobody did here, no. such things as that the man in the center frequently refuses to name
Fremont-Smith: That is very interesting. himself, that those on the periphery invariably name the center. Does
Savage: But he knows with whom he can communicate, that is, to the man on one end of the linelike configuration see himself as a leader,
whom he is, in a topological sense, adjacent. or does he see no leader? Is it only the man in the center of that con­
Kubie: This is a spaceless experiment from the point of view of this figuration who sees leaders, and so on ?
subject. Bavelas: I think the full answer to your question can be obtained. I
Bavelas: You have a wire but you don't know where it goes. am sorry I don't have it.
Savage: But there is adjacency, in a sense. Those with whom you can Hutchinson: Is there ever an individual present who is more inter­
immediately communicate may properly be called adjacent to you. ested in analyzing than in doing a job for you, who shows curiosity
Does he pick those as his leaders or the next two or— about this system and is a deviant from the standpoint of your experi­
Bavelas: I am sorry but I can't give you that analysis. I am sure I ment?
could dig it out very quickly. Bavelas: Well, after the experiment is over and the subjects have
10 Cyhernetks Communication Patterns 11

been interviewed, we let them sit around and talk if they wish. There a procedure. You are going to tell them indirectly when they are right,
is a great deal of curiosity and many ideas as to how the job could be so that they would be using you for a test device in the experiment.
done better. Savage: But the rules are such that if each participant has thrown
Hutchinson: But nobody ever holds up the experiment by getting too one lever, and one has thrown an incorrect lever, the group loses; that
interested in the formal properties 1 is, it is scored "incorrect" on that particular trial.
Bavelas: This was never true when speed was the main incentive. Bigelow: How about stopping the experiment? You said you stop­
When we ran other experiments and said, "We don't care about speed, ped the experiment.
but you must try to do the job in the fewest number of messages pos­ Bavelas: Yes, when the trial was over.
sible," the group spent a lot of time in the first two or three trials study­ Bigelow: Don't they know that the trial is over when they get a cor­
ing the thing out, and then performed much faster than the groups rect answer?
which were working against time. Bavelas: No, when five answers have been given, the trial is over.

To go on, we asked some questions in the interview with respect to You get the picture: those with more errors are happy, and those with
things like this: "How well did you like the job you were doing.? How fewer errors are quite unhappy. When you ask them how good a group
good a group do you think this is ? What kind of things in your opinion this is, the individual in the center feels the'group wasn't quite as good
prevented the group from doing better?" In general, based on the an­ as it should have been, but the others think it was all right.
swers to these questions, it is quite clear that the "circle" groups were Bigelow: Do you get a great deal more statistical variation in these

quite happy with the way things were going. They were slow and they centrally arranged groups than you do in the behavior of the ring?
made a lot of errors, but they were quite satisfied. They would volun­ Bavelas: That is a very interesting question. All of the error in the

teer to come back and do many more trials. They would often hang pattern shown in Figure 5 was contributed by two groups, which be­
around the laboratory to discuss the experiment and to tell us they could came confused, and apparently never got over it. This leads us to the
do better if they could try it again. In the pattern shown in Figure 5, second experiment—
except for the man in the center, the individuals were either quite ag­ Birch: Apparently there are two problems in this type of situation,
gressive or apathetic about the whole business. The evidence of this is when we begin to deal with the happiness-unhappiness dimension. On
very striking if you look at the messages they were writing. If you take superficial examination, it appears that the individuals are happiest
the messages and give them to a number of individuals and say to them, when they are (a) either unable to locate a leader in the group situa­
"Can you pick out of this group of messages those which seem to you tion or (b) where they have a certain degree of probability of assum­
aggressive or which say derogatory things about other people, or those ing that they are determining some kind of leader, whereas in your final
that are intended to stop the operation of the group?" they have no situation they have a leader imposed on them, and this leader is im­
trouble picking them out. One individual, for instance, sends out in­ posed on them by the structure of the situation, something which did
formation but sends it in Spanish; another tries to get his neighbor to not arise out of their interrelations but out of your manipulations. Now,
play tick-tack-toe instead of doing the job; another tears up the mes­ given that as the setup, a second point arises: which of these methods
sages he receives instead of saving them. would be objectively most efficient in solving a problem? Apparently,
Bigelow: Do you ever get any smart-aleck situations ? For example, from my point of view, I would think that this one central source as a
as you were speaking, it occurred to me that if the subjects decided that collator would be the most efficient objective method for problem solu­
the idea was to effect the quickest termination of the experiment, they tion, and that people might be happy with such a situation if they were
could easily set it up to produce a trivial solution along this line: Send to be permitted to design such a setup for themselves and to select a
one message to every member of the group, telling everybody to take central individual who would be doing their collating. In that sense
the first item on the card, then pull the lever for that; then take the they would have a leader who would be their representative and not an
second item on the card and pull the lever for that; take the third item individual upon whom they were dependent. I wonder if you began to
on the card and pull the lever for that, and when you get a coincidence, examine that feature of the relationship? It seems to me that this is
you will ring the bell. somewhat contrived, for that reason.
Bavelas: But you can have only one lever. Pitts: I should think that that could be controlled nicely in one very
Bigelow: What I mean is this: Suppose everybody does follow such simple way; namely, one would combine the advantages of symmetry
12 Cybernetics Communication Patterns 13
with the advantages with respect to the position of the participants and Gerard: Yes. '
those of having freedom of communication by considering the case Bavelas: We don't think that is a factor because the problem is so
where everybody can communicate with everybody else. In that case low-level. We know it becomes a very important factor if the problem
the group can select its own leader if it pleases. involves not only the collection of data with a very transparent opera­
Birch: That's right. tion to get the answer, but complexities as well.
Bavelas: We have run perhaps a half a dozen fully connected groups Gerard: I thought it would have made some difference in the speed
on this program. The result is complete chaos. and the success with which they discovered the actual network. There
Birch: Are they happy ? was no tendency for a particularly able person to become the acknowl­
Bavelas: They may take as much as a half-hour to do the problem. edged leader, aside from his position ?
No, they get frustrated, and there is always a battle going on as to Bavelas: Not for this kind of problem; there certainly was for more
which channels should or shouldn't be used. They want to cut down complicated problems.
on the number of channels and they don't have any good way of doing Richards: There is a new preliteracy test that Dr. P. J. Rulon at the
it. Harvard Graduate School of Education recently brought out that is al­
Teuber: Before you turn to the second experiment— most the same as your cards. You might find that if you got down to
Bavelas: Just one comment, first. If you look at the chain, there is testing on the preliteracy levels, you could find a relevance between the
no logical reason why all the messages should not go up to the top or abilities of these participants and their position. It might be worth
why the answer should not come back down. It doesn't take any longer looking at.
or any more messages than to have them come to the center and back Bavelas: I am very glad to know that.
out. As to the degree of imposition of leader, it is not quite as severe Rosenblueth: The emphasis seems to be on the way the group is or­
as it might appear to be. I know the board has been tilted by the im­ ganized, but then you mentioned that you had tried other problems.
position of the network. You are quite correct in that; but it isn't com­ Would those figures be very much altered? I would expect that they
pletely determined. would be altered, depending on the nature of the problem.
Teuber; Your speed score indicates how soon the experiment is over, Bavelas: Oh, yes.
doesn't it ? Rosenblueth: I mean, you are testing two things. One is the way you
Bavelas: The average of trial times. organize your group, but the other one is the problem you give them.
Birch: Does that mean successes or determination ? Bavelas: Yes. As a matter of fact, it seems to me that the only useful
Bavelas: Either way; you get the same relative standing. concept here is a kind of task-net concept, and not that of a net which
Teuber: But that score does not differentiate between the time it has properties without respect to the task. These things mean something
takes them to figure out their own group structure and the time it takes together but not separately.
them to arrive at a solution of the problem. It has been suggested to me that I should ask you to let me present
Bavelas: No, except that we know pretty well that the pattern is rec­ this companion experiment very quickly, very briefly, so I can then get
ognized and known by the second trial in the pattern shown in Figure on with what kind of notions we have as to an explanation of what
5, and by about the sixth or seventh trial in the pattern shown in happened.
Figure 4. You remember that the "circle" group was slow, made a great many
Birch: So these messages would be correlated so closely that it errors, but was quite happy. The "star" group shown in Figure 5 was
wouldn't matter? a very fast group, with very few errors, but many of the members were
Bavelas: No, because here they know the range. The stability in the unhappy. Now, two other types of things happened, which suggested
message-sending doesn't emerge, though. the next experiment. If you remember, each subject had in his cubicle a
Gerard: Did you have any data on the abilities of these people, and, box of switches so that he knew that the right answer must be one of
if so, did you get any correlation of abilities with what happens ? those six. On his card he had five symbols, so he really knew it must
Bavelas: Ability? be one of those five. Since he was looking for a common symbol, he
Gerard: Abilities—plural. could just as well send the symbol that was not on his card as send the
Bavelas: Of the individual ? five that he had. It was just as good a way of getting the answer. But
14 Cybernetics Conimunication Patterns .15
the task was so stated that it was more "natural" to send what one had opened the sixteenth box, the five marbles were there, but they were all
rather than what one had not. This insight occurred frequently in each mottled, a milky, mixed color. If one individual could have seen all
of these patterns. It was always acted upon in the "circle" group and five boxes, he could have easily picked the common marble. However,
never acted on in the "star" group. It never occurred to the individual to do this by means of descriptions was another matter. The disturbance
in the center of this highly centralized star group that this other method might be called "semantic" noise. What happens is interesting.
could be used, I think, really, because he was too busy. You see, he was First of all, the errors per trial for both the circle and the star group
getting all the messages coming in, and he had to get the answer out as increase sharply, and they reach about the same level. The star group
quickly as possible. At any rate, he did not get the idea. But the idea goes on making about the same average number of errors throughout
occurred as frequently in the other people of this group as it did in the the remaining fifteen trials. The circle group comes down very sharply,
people of the circle group. The messages sent in the star group give and by the twentieth trial is doing as well under the new condition as
some rather amusing interchanges. An individual would say, "We it did under the old conditions.
ought to send what we have not." The answer would come back, Klüver: The leaderless group comes down first?
"We're doing all right." The suggestion would come again, and the an­ Bavelas: Yes. In the case of the circle group a common language
swer would be, "Forget it," and so on. developed; that is, a word is found which fits a marble, and it is used.
This interested us very much. Another thing that interested us was In the star group no common language seems to develop. What hap­
that in the star group there was a considerable volume of messages sug­ pens in the star group, from what we can tell, is that the messages come
gesting changes or new ways of doing things, new ways of organizing in, and the man in the center assumes that the individuals he is working
the group. These came from the peripheral members who, you remem­ with are unreliable because they are describing the same marble, he
ber, were not very happy with the way things were being done. These feels, with different names. You see, the same marble may be called
suggestions were not accepted by the center man as anything to be acted "ginger ale," or "amber," or "light brown," and so on. He just guesses
upon. Also, in one star group where the idea of sending what you have at what must be the truth. He goes on, very often unaware of the errors
not occurred to two members simultaneously, they began sending this that are being committed.
information in spite of the central man's resistance, and consequently The feeling of the individuals with respect to the problem or the job
they so fouled up his calculations in the center that the group made a remain the same as in the previous experiment. The percentage of
great many errors. leader emergence stands the same. What changes mainly is this picture
It occurred to us, therefore, that perhaps we ought to study groups of errors, if you make a study of the speed with which correct answers
which had been permitted to "stabilize" and whose operation was then are turned out after the "shock," the circle group is, in fact, faster than
in some way disturbed, and to ask; "Now, what happens.? Is there a the star group.
difference in the ease with which the different groups will adjust to the Now, what I would like to do is to tell you a bit of how we are think­
new circumstance?" ing about this problem theoretically, and here I expect to get some
The following experiment was done. All the conditions were like help from you. We started originally by looking at these patterns in
those in the one I described, but instead of showing the subjects a card terms of their topology. We asked the question, "How many "steps' are
with five symbols, a little mechanism fed them a box in which there required to get from any place to any other place in the same network
were five marbles. They were colored red, green, white, blue, and so along the shortest path there is?" For the end man on a chain of five
forth. The problem in this case was to find the common marble, since people (Figure 4), the number of steps along the shortest path to all
only one color was present in every box—every other color occurring the others is4-f-3-l-2-(-l. We add it up and say it is ten steps from
in only four out of the five boxes. In this case, however, when the com­ the end to everyone else. We would do this for each man separately,
mon color is known, the individual picks up the appropriate marble and add it all up, and say: ""This chain group has a dispersion of 40, and
drops it into a tube which delivers it to a counting mechanism. You the circle group has a dispersion of 30." We felt that that number
will notice that in this case the error is irretrievable. might bear some relation to the measures we were making of the group's
Each group in this experiment did thirty trials. The first fifteen trials operation.
compared roughly with those of the previous experiment. When the We went a little bit further. In the chain group (Figure 4) the
group reached the sixteenth trial they had a "shock." When they distance from the next-to-the-end man to all the others was 7, whereas
 Cyhernetks Communication Patterns 17
16
in the star group the distance from the center man to all others was 4. else to do, he merely waits for the answer to come back.
What we did was to make a fraction—in the first case 40/7, and in MacKay: I was wondering which was the primary variable, and
the second case 32/4—and arrive at a number for the position which which might be merely dependent and more or less predictable from it.
we called its relative centrality. We calculated one more number. If Bavelas: I don't know. We have begun to think of the description
you take the most central position in the network and then take the dif­ of these nets in different terms. What we are doing now is something
ference between his centrality and the centrality of X (any other posi­ like this: We plot the topology of a net as a matrix. We put I's where
tion) , that difference we called the peripherality of X. there is a link, and O's where there is no direct link. Actually, the im­
In other words, we asked this: First of all, how spread out is the portant numbers here are the O's, not the I's, because the I's tell you
group? who is closest to everybody? and how far away is anyone from that there may be a communication between those people, but the O's
the one who is closest to everybody? We have worked with a limited tell you that there will not be. The 0 is a firm figure, but the 1 merely
number of patterns, and it appears that if you take the peripherality in­ indicates that there is a probability that a message will pass along that
dex—how far a given individual is from the person who is closest to link.
all—and plot against this index, the morale measurements we were able We are attempting now to describe the operation of a net in terms
to make have a very good correspondence. of the probability that there will be a message from' A to B m a. given
McCulloch What sort of scatter do you get on your experimental time period, and in terms of the probability that an item of information
at any point in this net will appear at another point in the net within a
results ?
Bavelas: It is quite tight.
certain time. If we accept this way of describing a net, then we can
Pitts: There are two possible curves you could plot: morale against play an interesting game. We play it on paper. This is a calculation,
distance from the most central person; and morale against the man s not an experiment.
own centrality. We make these rules: first, that every person in the net will send a
Bavelas: Yes. Now, the correlation is much better with distance message at every moment of time; second, that he will send all he
from the most central person than with one's own centrality. knows every time; third, that if he has more than one channel on which
Pitts: You mean it scatters much more if you plot it against his own to send, he chooses a channel at random. The probabilities of choosing
any one channel are equal. The last rule is that a person cannot send
centrality ?
the same message to the same person twice. If we play out this game
Bavelas: Yes.
Pitts: The curve is the same in general shape, though? on paper, we get almost exactly the operation one gets with an actual
Bavelas: The curve is the same in general shape, yes. group. These rules for operating apparently give the same statistical
Savage: You say morale goes down if he is close to the central picture that a group of subjects will.
Savage: Well, there is an ambiguity. If a person is excluded because
person ? r i i
Bavelas: No. The further away he is in distance from the central of this "no repetition" rule, then the probabilities to be assigned to the
person, the lower the morale tends to be. remainder are the so-called conditional probabilities.
Savage: And yet in the star-shaped configuration, where everybody Bavelas: They are equally partitioned for the remainder.
is proximal to a central person, morale was low in the periphery? Savage: Well, suppose his probabilities were already a third, a
Bavelas: Yes, but the centrality of the central man in the star is quarter, or whatever the figure may be.
32/4. The centrality of the rest is 32/7. The difference between Bavelas: No, they are always equally partitioned in this game we
these two is big. It is the biggest difference we have in these three are playing.
groups. In other words, the more you centralize your net, the bigger Savage: Oh, it isn't the real probabilities that occur in the matrix?
Bavelas: No, this is a game, a game being played on a sheet of paper
will be the peripherality difference.
MacKay: Have you plotted any index of morale against the number with a pencil and a random number table.
of messages sent by each individual ? Savage: I, too, refer to such a game, but a more complicated one.
Bavelas: Yes, and that correlates too. And the number of messages Arbitrary probabilities might be assigned in advance, perhaps in the
sent, of course, correlates highly with where you are in the group. You light of empirical work, but you are referring to the special case in
see, in the chain the end man sends a message and then he has nothing which the probabilities are divided equally among the opportunities.
 Cybernetics
Communication Patterns 19
18
Pitts: He might also send a random sample of the bits of informa­
Bavelas: That's right.
tion he has in front of him. He might have only one item of infor­
Savage: I see.
Bigelow: Why do you rule out the duplication ?
mation exchanged at every stage.
Bavelas: Yes. If we take a group in which there are no restrictions
Bavelas: If you don't apply the duplication rule, you get some very
nice curves, but they are not the curves which the subjects give you. If placed upon linkage and which sends at each moment the very best
you put this "against redundancy" sort of stipulation in, then you get message possible to exactly the right place, then the average amount of
a curve which is very, very close to what the subjects really give you. information across the entire net from moment to moment is given by
Bigelow: Isn't there an objection to this on philosophical grounds, a curve which is described by loggn, "n" being the number qf nodes.
that you interfere with the randomness and hence put a bias on the Using the rules I first gave you and playing the game, you can, for
any given pattern, find who gets the five items first, on the average.
results ?
Bavelas: Well, perhaps it is not a good way to do it.
And when this person gets those five items, you can discover how much
Bigelow: I am not saying that. I am only saying that if one tries to information is present at the other positions. This profile correlates
think of this in terms of a network in which, at certain points, the very well with the emergence of the recognized leader and with morale.
process is completely random, then perhaps you should allow the dupli­ So this leads us to ask whether, for a group that is working in this
laboratory situation, and who have a very specific task to perform, say­
cation.
Savage: It is not that random. Nobody is so foolish as to say the
ing that So-and-So is a leader really means that he is a person from
same thing to the same person twice. Only human beings do that. whom information which can lead you to successful completion of the
task regularly comes. In other words, the man recognized as leader is
Mead: This refers to human beings.
McCulloch: How about children? Might we not get a better fit for
the man at the position where information accumulates most rapidly.
very young children if you leave out that requirement? This line of thinking leads us to make the following general theory.
An individual at any time has certain hypotheses that are very important
Bavelas: I don't know.
Pitts: When you say that the message must contain all the man
to him, and some that are less important. He is interested in optimizing
knows, does that include the fact that he has sent messages to a given the probabilities that these hypotheses are correct—for instance that he
person in the past, of a given sort, and received messages from other is a successful person or that he is a loved person or that he is a good
person. He does this primarily through information from other people.
people ? .
Bavelas: No. You see, if he kept every message he received, in­ The sources of such information are very important to him.
cluding the stuff he started out with, in front of him, every message Imagine a research group that has been working on a program, and
imagine that they are very discouraged and that they feel they are fail­
would have all of this on it.
ing. Now we observe the following phenomenon: a man walks in; a
Pitts: I see.
Kubie: It seems to me there is one assumption here which you may smile and a word from him and the picture changes. They don't feel
have to make for mathematical purposes, but which is so contrary to the so discouraged. They feel that they may succeed after all. What I
way in which human beings operate that it troubles me. That is the as­ would say has happened is that an information process has occurred
sumption that everything a human being knows is equally available and by which the probabilities that certain hypotheses are correct have
accessible to him at any particular moment. You are resting on an as­ changed, and that—
sumption which is contrary to fact, as the basis of operation of your Kubie: Why information probabilities?
Bavelas: Because I would like to define a change in probability as
experiment.
Bavelas: Well, by "everything he knows," all I meant was what he being an indication of an information process occurring.
Kubie: That is a circular assumption.
knows about these five symbols.
Bavelas: Yes, of course.
Kubie: Yes, that is what I mean.
Kubie: But because this is a circular conference doesn't mean that
Bavelas: These are available to him.
all circular assumptions are valid.
Kubie: Not equally.
Bavelas: All I meant is that they are on a piece of paper in front of Bavelas: Let me try to explain what I am about in this way: If I set
up a situation in which a person must try to pick the one of eight boxes
him and that he can copy them.
 Communication Patterns 21
20 Cybernetics

which has a $10 bill in it, and the distribution o£ choices tends to be colored marbles, these marbles now may appear as very, very difi^erent.
rectilinear, then I would say that he isn't getting any information with He is looking for really minute differences, which always exist, such as
respect to this problem. But if I have a person look through a wmdow bubbles in the marbles or slight differences in size. He may, in effect,
at him and make faces, and this event changes the distribution, I say find he is alerted to search for differences.
Gerard: You are answering a somewhat different question from what
that, by definition, information has been received, and that I should
be able to calculate how much was received. Now, I am beginning to I had in mind. Was the performance of the different types of networks
think that what we call social needs are hypotheses, culture-given with the mottled marbles used from the start the same as that when the
hypotheses. The individual who accepts the culture tries to make the mottled ones followed the solid-colored ones 1 i
Bavelas: The difference is not as great, but they are in the same
probabilities of correctness of these hypotheses as good as he can. Many
of the things that happen in a group process depend on how this infor­ relative position.
Gerard: In other words, even without a predisposition, the central
mation is being transmitted and received, whether an individual is be­
low a tolerable level of uncertainty concerning these hypotheses. I can man does badly.?
Bavelas: Yes.
tolerate considerable uncertainty with respect to whether or not my
Gerard: Is the variation between groups greater with the mottled-
dessert will be chocolate ice cream or vanilla because this hypothesis
is not as important to me as some others; but I am not willing to accept marble problem as a function of the central man Ì In other words, I am
that level of uncertainty for other things. The uncertainty concerning again asking about this factor of ability.
Bavelas: Personal ability in the level of problems we have used does
hypotheses of social relations between myself and others can be changed
mainly—if I am a normal person—by information from others. In not seem to enter as a factor.
Gerard: Did some of your star groups do perfectly with the mottled
other words, an individual tries somehow to make the probability that
certain hypotheses are correct as high as possible. Some of these cannot marbles, for example?
be made higher than a certain amount, but if they are high enough he Bavelas: None did.

behaves as though they were really so. I cross the street as though it Gerard: Never?
were certain that I would reach the other side. I know that it is not Bavelas: That's right.
Savage: I should like to call attention to a fact brought out by your
certain; but when the uncertainty is great enough then I begin to be­
later discussion; namely, that it is essential to your experiment that the
have "queerly."
I am suggesting, therefore, that a good deal of the behavior one ob­ individuals communicate serially. When one writes a note, he can't
serves in organizations is not apart from the communication network write it with carbon paper to everybody he is in a position to write to,
that obtains in the organization. A person may well be in a situation but must take quite a bit of time writing to one after another. Though
where he gets no information about important hypotheses, although the some real-life situations may be subject to such a limitation, the situa­
leaders of the organization may feel they are giving him a great deal of tion is often quite different. Consider, for example, the number of peo­
information. The pamphlets written by the boss or by a personnel ex­ ple to whom I can simultaneously transmit information simply by talk­
pert are distributed to the workers with their pay checks. These pam­ ing out loud. I suppose that an experiment parallel to yours, but per­
phlets, found in the gutter outside the plant, contain information— mitting such simultaneous communication, might be rather dull, for in
probably from the boss to himself—that is not of much value to the such a situation it would be sensible for a person simply to keep every­
one with whom he could directly communicate up to date on all in­
workers.
Now, I think I have talked enough to give you some feeling as to the formation which reached him, and the actual experiment might very
direction in which we are trying to go theoretically, and I would be in­ well simulate such ideal behavior. Though this situation could not be
terested to hear what you think about all this. expected to lead to so interesting a set of experiments as those on which
Gerard: I would like to go back to your second experiment, if I may, you have been reporting, it may be a model for as many, or possibly
and ask you two questions about it. Did you try the mottled marbles even more, real-life situations.
with the group without the preliminary runs on the clean-cut situation ? While I have the floor, I should like to say that there is a chapter in
Bavelas: That is being done now. When a group starts with the modern statistics which I think is very intimately associated with com­
mottled marbles, and then goes from the mottled marbles to the solid- munication within groups. I don't want to tell it now, but I do want to
22 Cybernetics
Conimunication Patterns 23

ask that if ten or fifteen minutes can be found at this conference, I be cards and say, "Now, gentlemen, you are going to line up." Let's as­
sume we have at least 52 people. "Walk by and take a card off. the
permitted to discuss this topic.
McCulloch: Can we hold that, then, until we come to a place where deck. You take the one that is on top when you get there. Now, will
you indicate in a secret ballot which position you want in this line?"
it might fit ?
Savage: Any time you like. I am just asking now, while I think of it, You get a distribution. Of course, nobody wants to be first or last.
Most people want to be between 15th and 20th in line. Then you spend
for time to say it.
McCulloch: Right. I should like to hear from Shannon and from
as much time as is necessary in proving to them that it makes no dif­
MacKay, if they will, as to the extent to which the notion of informa­ ference where they stand in line. After they are "convinced," you ask
tion, as you have used it, Bavelas, is the same as it is in their scheme of for another secret ballot, telling them that this ballot will determine
where each one will be in the line. Well, how much do you think you
things. Will you take it, Shannon ?
Shannon: Well, I don't see too close a connection between the no­
will have shifted the distribution ?
Savage: I don't think that is a fair example. If you had taken an
tion of information as we use it in communication engineering and
what you are doing here. I have a feeling that the problem here is not example in which the information imparted had practical consequences,
so much finding the best encoding of the symbols, because you do not I would have been impressed. But here, you see, it has none. Suppose
have a limited amount of paper to write on, which corresponds to the you argue yourself blue in the face that position in the line doesn't mat­
engineering problem of encoding messages to use the smallest amount ter, then I say to myself; "Yes, perhaps it doesn't matter, but, after all,
of channel, but, rather, the determination of the semantic question of the guy might be wrong. If he's right, it follows that there is no harm
what to send and to whom to send it. I don't see quite how you measure in my playing my hunch; if he is wrong, I might much better play it."
any of these things in terms of channel capacity, bits, and so on. I think If you take a situation where the information has a practical conse­
you are in somewhat higher levels semantically than we who are deal­ quence—for example, you could arrange one where the first place is
ing with the straight communication problem. actually best, but looks worst—then you will arOuse a real conflict in
Bavelas: May I just give an example of how one of my students is the subject after instruction. He will say, "Gee, the guy says I should
trying to measure this? What he says he wants to do is this; He says go first. It looks as though I should go last. Maybe he's got an argu­
that the group starts out with a probability of one-fifth, that any one ment." In such a situation you might shift somebody, but when you tell
of the five symbols each individual has is the right one. Now, at the me that my prejudices are simply irrelevant, that they are superstition,
end of "X" minutes the experiment is halted, and he says to a partici­ and so on, then I might as well follow them, because it doesn't cost a
pant, "Now, of these five symbols, which can you, with certainty, cross thing to have the best both of your world and of mine.
Klüver: I have a question of some practical importance. Some years
off the list?" He hopes in this way to get some indication of how much
ago I came across a French book in the field of personnel research.
information has really been received.
Shannon: There was one remark you made which intrigued me a bit. After several hundred pages the author concluded that a man running
You relate information to the change in probabilities. I gather by that an army or an industrial empire should never communicate or deal with
you mean subjective probabilities, the probabilities a person estimates more than five or six people. Any attempt to have personal contacts
for such-an-such an event—^whether or not there is any valid reason for with more than six people would turn out to be a disaster either for
his estimate. From the communication point of view, the subjective army or for industry. I wonder whether your work is ultimately going
probabilities do not enter at all. I wonder if we shouldn t somehow to throw light on such practical problems.
Rioch: Along the same line, Dr. Bavelas stated at a recent seminar
distinguish between valid information and information that the man
thinks he has and acts on, perhaps, but which isn't based on any logical that the size of the groups under study was limited to five, since with
six or more the number of possible configurations became unwieldy.
reasoning.
Bavelas: Of course, I am biased in this direction, because I think the It occurred to me later that in team sports we subdivide teams of more
change in the subjective probabilities is the thing that is worth talking than five members into three, four, or five subgroups; also, in other
about, if you are thinking about people at all. You know the demon­ group activities we similarly deal with a small number of subgroups.
stration that is frequently done in classrooms. You come to class and Gerard: What happens when you lecture to a class ?
put a deck of cards on the table. You put a $10 bill within the deck of Rioch: You deal with just one, then. You are dealing with a class.
 Cybernetics Communication Patterns 25
24
Gerard: You call that an individual, not a subgroup? MacKay: But I thought our question was whether, when Bavelas
Rioch: When lecturing I find I usually pick out one, two, or three talks about "information," he means what you and I do. I think if we
people in the class and talk to them, treating the rest of the class as use the word "selective" to distinguish it, then we are all three talking
about selective information.
a unit.
Bavelas: I must admit that the concept I tried to describe is very at­
Gerard: I am not convinced.
MacKay: I wonder if I could come back to Bavelas's concept of in­ tractive to me, but it always leads to a very puzzling conclusions. If I
formation, because I don't agree with you. Shannon, that it is essentially may do so, I will cite one.
diflferent from yours. I think what is different is the type of ensemble "The following experiment was done by some students. Five individ­
with respect to which you measure it. Suppose we define "selective" in­ uals were placed in separate cubicles without linkage one to another.
formation, which is the concept used by Shannon in communication They had no communication among themselves in terms of direct link­
theory, as that of which the function is to specify a selective operation: age. All the subjects could see was an indicator. They were told that
that is to say, you think of yourself, the receiver, as having prepared a the experimenter would cause a number to appear here, and that it
filing cabinet of possibilities, and the receipt of information enables would be some number from 0 to 25. All could see the number. Each
you to make a selection from it or, at any rate, to narrow down the of them was to write on a slip of paper a number from 0 to 5. The ex­
probability-distribution over the possible commands that you have to perimenter would collect the contributions and add them up. Then the
give to your own selective mechanism. Then, I think Bavelas's results sum of their contributions would appear, and they would hear him say;
are a consistent illustration of this process, particularly when he in­ "Your target was 17. Your summation was 15. Try again." They also
troduces symbols differing from one another by nonsignificant character­ wrote down why they picked the number.
istics, and then goes back to the original symbols. When the nonsignifi­ An interesting result here is that when the subjects are thus given the
cant characteristics appear, their effect is to enlarge the space of possi­ size and the sign of the error, they don't do as well as a group which
bilities out of which the receipt of information enables you, the re­ is merely told, "You were wrong,'" try again." Certainly, it looks as
ceiver, to select. Originally, you started with a space with just five pos­ though you are giving them more information when you tell them the
sibilities in it, or six possibilities; later, you had to widen your space of size and the direction of the error; but if you look at the "real" array
possibilities to include such things as mottling, and so on. Then, when from which the choices are made, the array expands when you give
you come back to the original symbols, the selective operation you are them this additional "information." Without the size and sign of the
performing at first is still one on the larger space, until you learn that error, the subjective array from which they select is smaller. Now, in
certain dimensions of it are "no longer being used," so to speak. I agree the first variation, are they really being given more information ?
with Bavelas that it is the subjective probabilities that determine the en­ Bigelow: No matter how much time they have, that is true ?

semble from which we estimate amount of selective information here, Bavelas: There is no time limit at all.
but, with that proviso, I think it is the same concept as Shannon's. In Bigelow: Another point of the same sort; In the previous case,
communication you ask; "In the ensemble of expected messages, how where you were discussing the precision of the process, it isn't clear to
rare is this one which has been selected?" In this problem, we ask: In me under what circumstances the people were allowed to take as much
the ensemble of expected responses, or expected commands to respond, time as they needed. For example, in the case where you had the ring,
how rare is the response that I have made?" were they allowed to take as much time as they needed with mottled
Shannon: I would certainly agree that information is being trans­ marbles? Also, were the people with the central sitaution ever given
mitted here. What I was trying to say was that I don't feel that this conditions where they could take as much time as they needed in order
group of people will tend necessarily to the most efficient way of doing to build up the nomenclature ?
it. This makes it difficult to apply any of the results of information Bavelas: No. In those experiments the instructions were; "Get it

theory. I would expect that they might settle—in fact, some of your right. Get the right one as quickly as possible."
experiments indicate that they do—on a means of communication other Bigelow: But has anything been done to explore what happens when

than the most efficient. For instance, they do very poorly on the con­ you eliminated the necessity to do it quickly?
Bavelas: Only with the five symbols; in that case, I told you what the
nected graph, which surely should be better from the straight communi­
cation point of view, if it weren't for psychological factors. facts were. But you see in this last example what this may mean. The
 Communication Patterns 27
26 Cybernetics
Even though the target is 16, and the sum comes out to be 15 five times
linkage has something to do with how much information the group
in a row, he won't change. He will say that his job is to stay constant
can use. When the group was told only that they were wrong, they
because somebody else will adjust. Some subjects take the opposite at­
were selecting from a limited array—let us suppose the target was 17.
titude. They keep trying to compensate for what the others are doing.
A subject would reason, "If I put in 3 and the others do, too, it will
MacKay: May I return to the question as to whether or not the error-
make 15." Some subjects—most of them—put in 3 or 4. An occasional
feedback procedure is giving information? The whole thing is com­
2 was contributed, but 0, 1, and 5 were never selected. When they are
pletely consistent with the information picture. The general point is
told the size of the error and the direction of the error, however, the
that the specification of more possibilities enlarges the space. That
reasoning may go this way: "We're too low. We should go up. But
increases the number of bits necessary to perform a given selective
everybody will think of that; therefore I should put in less than I did
operation, and so you would expect that your variance would go up if
last time. But everybody will think of that, so maybe I should put in 5
what you are dealing with is some kind of random process which is be­
to overcorrect." So you can see that the range from which choices are
ing modulated, as it were, in its probability peaks by the information
made has been increased.
that is coming in. I don't see any paradox. What is happening, if you
Bigelow: No matter how long you give them? It is amazing that
like, is that the information that is coming in is "structural" in its effect.
they don't realize that they should take the desired result "n" and divide
It is "structural information;" that is to say, its effect is to increase the
it by 5, and come closer to it? There is no tendency to do that? If
number of dimensions of the information space. Then, of course, you
a person is above the "nth" portion, and the result is still too low, he
have a much larger number of possibilities, so you would expect the
certainly ought to stay where he is and let the other converge. Did
normal flux of information to have a proportionately smaller selective
they never learn that point ?
Balevas: But this is what they do. If the target is 17, they know effect.
Savage: I don't see it at all. You say if I tell people their error, that
there are five of them.
should naturally make them do worse than before, and you are not a
Teuber: Do you finally get a corrfect result?
Bavelas: Oh, yes.
bit surprised?
MacKay: Yes.
Teuber: After how many trials ?
Bavelas: There is no linkage here, you see.
Bavelas: Oh, the number of trials tends to get shorter; in other
MacKay: Let's put it this way: By naming something, you raise in
words, they do learn something. But it might be three or four trials.
the man's mind a complex array of possibilities corresponding to what
Teuber: Only three or four?
you have named. If you name the range of errors, you modify the
von Bonin: Suppose they all put in 5. All right, it will be 25. The
structure of the filing cabinet from which the almost random mental
experimenter says, "You have 25 instead of 17 ? '
process is making its selections, and it is not in the least surprising that
Bavelas: Yes.
the result is to increase the spread of the total. It is evidence, if you
von Bonin: Do they imdercorrect then ? Do they all go down ?
like, of the often ignored distinction between human thinking and
Bavelas: The difficulty is, you never get 25 if the target is 17.
von Bonin: Just by way of example— human logic.
Gerard: Isn't it much like what you said about the difficulty of
Bavelas: What happens, very quickly, under either condition, is that
some individuals, although they have information as to what the others selecting the clear-colored marbles after they have worked with the
are doing, begin to structure the group. They say, "My job is to do so- mottled ones, that they are looking for more difficulties ?
Bavelas: Yes, the space has been enlarged.
and-so, and somebody else is doing that." They get very complicated
Kubie: Because we so often fool ourselves that we are basing our
hypotheses about how the group operates. They assume a role and
choices on purely external criteria when in reality we are choosing in
give other people roles, almost as though they couldn't stand the situa­
terms of our own unconscious processes, the probability of error in­
tion if it didn't have some social structure.
creases as we go from random choice to unclear ad hoc hypotheses.
Bigelow: Does anybody take the position that they should stay con­
This was proved in the selection of pilots for naval aviation before
stant to allow others to adjust the small difference ?
the last war. Attempts were made first to select pilots on the basis of
Bavelas: That is what some of them do. A subject may put in 3 ten
a variety of theoretical assumptions, but it was found that choosing
times in a row, regardless of what the summation is. He won't change.
28 Cyhernetks Communication Patterns 29

men at random worked at least as well as did these initial attempts to Savage; This is one of those points of elegance you mentioned
guide the choices in accordance with some pet theories. earlier. In so far as you have statistical sophistication about the experi­
Next I want to record my disagreement with Dr. Bavelas's first state­ ment, you will remove the significance of color and of the emblems by
ment, that this is not an "elegant experiment." On the contrary, I think suitable randomness and balancing. You have already indicated that
it is both beautiful and elegant. I derived from it the same pleasure some such precautions were taken, particularly with respect to color.
that I get out of any precise and adroit and clear intellectual or physical With ingenuity and with circumstances at all favorable, you should be
activity. able to abolish utterly any of these effects. Strictly speaking, they can­
Bavelas: Well, I can easily demonstrate that it is not precise. not be really abolished, because they continue to be facts of life, but
Kubk: That may be. However, apart from either its beauty or its you can see to it that they cancel out, so that no apparently systematic
precision there are elements in it that bother me. In the first place, in results of the experiment, depend, for example, on the fact that some
addition to the point which I have just raised about the shift from ran­ particular emblem is phallic, or anything like that.
dom choice to ad hoc hypotheses, we must include among possible Mead; That is fine for the experiment, but I think the important
sources of error one which is so disturbing that I hate even to consider thing here is that we are going to take these experiments and use them
it. Even if it should turn out to be true, I frankly do not want to be­ in planning for organization in real life. In a sense, once we have been
lieve it. Nevertheless, it must be kept in mind. This is the issue of elegant and taken them all out of the experiment, the next elegant
parapsychology, and whether any extrasensory perceptions exercise an thing will be to put them in in such a way that we can learn something
influence from one individual to another in this experimental situation. about them.
Actually, it seemed to me that the experiment is so designed that it Savage; But you put in the ones in which you are interested. You
might even throw light on this problem. will never be interested in real life in whatever phallic significance an
Bavelas: Would this effect cease if you give them information on the asterisk may have.
size and direction of the errors ? Mead: I disagree. Have you ever had proofreaders on your hands ?
Kubie; Yes, because then each individual becomes a restrictive indi­ Savage; If I say "never," I mean "hardly ever." Particular symbols
vidual when he tries to fit something to his individual hypothesis, for emblems which have been chosen for the experiment under discus­
whereas if he is working at random choice, if there is such a thing as sion are largely accidental so far as any wish to explore the deep psy­
parapsychological influence—^which I hope there isn't, because life is chological significance of shapes is concerned, though that subject is
complicated enough as it is—then we give him an opportunity, when also worthy of experimentation. I therefore reiterate that in this experi­
all five are working at random, which he wouldn't have when he is ment you will want to remove the psychological effects of shapes by
obeying his little secret contrived hypothesis. I think an issue comes up randomness and balance in design.
in relation to the other one of these experiments. Kubie; May I put in just one point, to make it cohesive ?
Bavelas; In other words, we should measure the departure of the McCulloch; Go ahead.
results of this group—^without error information—from those derived Kubie: Keep in mind the meaningfulness of the whole experimental
from the tossing of five six-sided dies. situation, that is, of the cubicle. To some people the cubicle would be
Kubie: Correct. My next point, of course, concerns the meaning of of claustrophobic import; to others it would not be.
"meaningless." This is most uncertain. It has come up repeatedly in Klüver: In your experiments. Dr. Bavelas, the items with respect to
connection with color. We are all aware that color exerts emotional which communication occurs are such items as circles, triangles, aster­
influences and has other subtler significance. But these emblems (for isks, marbles, and so forth; that is, items easily named and specified in
example, the asterisk and the quadrilateral figures) are also not devoid English. But suppose, you deal with such an item as "an irregularly-
of meaning. In fact; this is demonstrated by the data presented here. shaped, wet, smooth object changing rhythmically in brightness." Per­
I do not know to what extent these unconscious meanings influence the ceptually, this is something very simple, since such an object is easily
material; but the possibility must be kept in mind, in terms of your identified and recognized; to designate it unambiguously in English,
conclusions. Certainly, they will have different meanings to different however, requires telling a rather long story. Items with respect to
people, to different sexes, and at different ages. which communication may occur often differ in "word-nearness" or
Bavelas: The thing is whether the effect we are trying to look at is "word-remoteness," that is, in linguistic distance. I wonder, therefore,
overwhelmingly stronger. whether you have any data on how the mechanism of communication
30 Cybernetics Conimunication Patterns 31
is influenced by systematically changing the linguistic distance of the after a time, improved somewhat. How much was that? That is, in
various items entering the communication network. the case where you have both the error and the target specified.
Bavelas: We have done nothing on that point. Bavelas: I can't give you exact numbers. There are two ways of look­
Rosenblueth: But there should be one thing you are trying to meas­ ing at it. One is, how many tries—
ure, to learn or to study with this experiment. I think that it is an Pitts: That's what I mean.
extremely elegant experiment, that you correlate certain things, and Bavelas: —and the other is, what is the larger error ?
that you get very nice correlations. But you don t know what you are Pitts: I mean how many tries.
after. It seems to me that you have given a little too much importance Bavelas: It tends to go down. How much, I don't know, but there is
to the notion of information in your judgment of the experiments. a general decrease of the number of tries on the average.
Solving the problem is not strictly a matter of conveying information. Pitts: Does it ever fall below the number of tries taken by the group
Here, the nature of the problem is obviously very important for the re­ with only the target and no additional information about their own
sults that you get, from the data you gave us. You get quite opposite errors supplied to them?
results with the different groups when you give them different prob­ Bavelas: Not consistently.
lems, because you have introduced different things to solve. I can t Pitts: Never ?
quite make out what this is giving you a measure of. Bavelas: No.
Bavelas: The question we have been trying to answer is simply this; Savage: That group also learns. There is learning in both cases; is
Does it make a difference if you have groups that must communicate that right?
within the constraints of given patterns when all of the patterns under Bavelas: Yes.
consideration are equally logical with respect to their adequacy for Teuber: I was trying to ask Dr. Rosenblueth's question; let me try it
solving a task? Does the pattern itself have any psychological effects once more; In all these situations you have two things. Your group
which cannot be explained on the basis of anything else but the fact has to communicate; that is, they have to talk to each other, and they
that this pattern was imposed to start with ? And does this imposition have to solve a problem. Now, to rephrase our question; What dif­
generate other psychological effects ? The answer to these questions is ference does it make if their problem is merely to talk to each other
very likely yes. and to understand what they are saying, without using this communi­
Rosenblueth: Given a specific task, yes, but what about different cation process to solve a problem imposed upon them by the experi­
tasks ? How can you generalize, and how can you evaluate what those menter? I suppose you know the experiment that George Miller did
different tasks have given you when they lead to quite opposite results ? with the Harvard articulation series (1). He has a triad, a group of
Bavelas: That, we don't know. We know this, however, that if we three people, hooked up in different ways by telephone. With one
change the nature of the task in such a way that certain individuals hookup any person can talk to every other person, and they can reply
possess more information regarding that task, then the effects we get to one another; with another hookup, they have to talk around the tri­
here can be explained only on the basis of where the information is at angle, but cannot reverse the flow of communication, and things of that
a given time. You see, what we have done here is to guarantee, because sort. He comes out, I believe, with results that are somewhat different
of the nature of the experimental task and the topology of the net, that from the situation in which the problem would be not just to under­
information will accumulate at certain places more rapidly than at stand what the message is, but to solve a problem on the basis of
others. When you change the task, one thing that changes is the origi­ messages.
nal disposition of relevant information. But again you can say, "What Bavelas: Yes, but, you see, it isn't so different, really, is it?
matters here is where is the greatest amount of relevant information Teuber: That is what I am asking.
and what is the path over which it must travel to get to the others." Bavelas: In an articulation test I think you have a list of words.
McCulloch: Before we go on, there are now four people whose Something comes over the wire, and that is supposed to help you make
names I have on the fire for questions. I think Pitts's question should a proper or correct choice. In those terms I don't think the problem is
come first. very different. This group also has a task which consists of selecting
Pitts: It is purely a numerical one. In the case where information of an item out of an array, and the means are signals from others.
error is provided, in the second experiment, you said that the groups. Teuber: But isn't the outcome different?
32 Cybernetics Communication Patterns 33

Bavelas: Not very. certain circumstances are worse than the most incredibly bad computing
Teuber: The same connectivities, the same effects.? machine you could possibly imagine. They cannot add 2 and 2, so to
Bavelas: In general. He finds in that situation that the three or four speak, under certain circumstances of duress. This is a concrete fact
patterns that he used cannot be distinguished until noise is introduced, and, as far as I am concerned, it is a new one.
and then they are distinguishable in their results. I think that the people in the social sciences, the people in the more
Rioch: I was wondering, in this last experiment you described, if difficult sciences, ought to realize the tremendously valuable lesson un­
that had any relation at all to the class demonstration attributed to Ray­ covering facts of this sort. It is from these that people in the cleaner
mond Pearl. He asked his class in statistics to guess the length of a and easier sciences have constructed fairly substantial theoretical build­
line drawn on the blackboard, and showed that the average of the ings. To worry whether this is a matter of the real world is fine and
guesses was always closer to the "measured" length of the line than has a purpose; but this experiment has shown some facts, and they are
any one person's guess. the really sound part of it. The facts are that people under certain
Savage: I don't think that has any real relation at all, because it does circumstances apparently cannot do simple computation. The circum­
not really seem to refer to psychology. If the line had been measured stances enormously affect their ability. This is, furthermore, a very
by an instrument such that human error could be considered negligible good illustration of the difference between what one thinks of as a
compared with instrumental error, the results would have been the computing mechanism and what one thinks of as a neurological process
same. We are dealing here with the phenomenon of statistical stability, that goes on. The neurological processes are very well known to be
the tendency of averages to be close to their expected values. much more elegant in some respects than a computing machine. On
Bigelow: I should like to rub in a few points very hard that occa­ the other hand, they have these shortcomings of breakdown during
sionally seem to have been missed in this group. We are a group of duress. It is a very concrete and very definite point.
very mixed backgrounds, as everybody knows; and I should like to say Pitts: I have a suggestion which I think opens up all sorts of inter­
that to me this experiment which has been described seems particularly esting possibilities. So far, in these experiments, you have imposed a
elegant and has in it a drastic lesson for all of us, outside of the par­ common purpose on the group in the sense that the members of it are
ticular form in which it appears. The lesson, I think, is that in ap­ all trying to discover which is the common symbol on their list, and
proaching a very complex field, what one does is to take a very simple then transmit that information to all the other members. Now, suppose
model, which is almost always picked purely on an intuitive basis. The you change it in this way: offer a prize to the person who first discovers
intuition may be good or bad, depending upon elements that nobody the correct answer in rank or order. Then you have all sorts of possi­
can judge, so far as I know. Then you set the model up, look at it a bilities of coalitions forming against given people, and so forth. This
while, and sometimes you come to some concrete, although narrow and situation has the great advantage that there exists a mathematical in­
specific, facts. Now, these may constitute the most useful part of the strument for analyzing it.
experiment, rather than what you originally intended the experiment to Savage: It doesn't analyze it at all.
suggest. That is sometimes useful in a different way, but the most Bavelas: May I give an example of something in our own venture
definite and concrete part of the experiment may be the fact that it does close to the game theory. We set up what we thought was, and what I
or does not show information on one specific point. still think is, an example of the classical duel situation. This is set up
Now, one of the questions which was directed to the speaker, I think, for two people; in order to get the "bugs" out of the machine, we in­
concerned the maimer in which this experiment is related to the real tend it to be a game between two groups. Two individuals sit on oppo­
world. Suppose there are affect values and relationships to the symbols, site sides of a table. They are separated by a panel on which there is a
and so on. I think that it ought to be understood clearly that such big green light, which flashes on certain occasions which I shall explain
questions are really irrelevant; that in making this measurement, the in a moment. In addition, there are sixteen small bulbs across the
person who is making it is presumably not trying to say, "This is a panel. Under each bulb is a button which can be pushed. An electrode
model of the real world," when nobody can make a model of the real is attached to each' individual's arm. The setting is the same for both
world. If you try to make a model which contains enough to describe subjects. At the start of the experiment, the subject is told that one of
what all of us feel when we see these cards, it is obviously impossible. these sixteen buttons is the correct one, and if he pushes the correct
But lo and behold, this, for example, tells us that human beings under button he will shock the other person and also trip a relay so he him-
34 Cybernetics Communication Patterns 35
self can't be shocked. At the beginning of the experiment all the lights the present time no consistent operational definition of the term "moti­
are on. Every five seconds, one of the lights goes off; therefore, if he vation."
waits long enough, he will know which button is the correct one; but Bigelow: Can you tell us what the increase of adequacy is ?
the other fellow's chances are getting better, too. If one subject presses Rioch; If a person is faced with a difficult decision, for example,
the wrong button, the big green light flashes, and the other subject differentiating shades of color with the expectation of punishment, for
knows he has missed. example, electric shock, in case of error he may respond with being
We ran a series of trials under the condition that each man would "alerted" or he may respond with "anxiety"-—Kurt Goldstein's catas­
have only two "shots"; we ran others with four "shots." We let each trophic reaction. The physical sensation of tension may be quite simi­
pair play long enough under one condition so that they really came lar. Our concepts of motivation are clouded because they are not
very, very close to doing what the mathematical analysis gives as the adequately differentiated from the conscious sensations or feelings
best time to press the button. This interesting thing happens, however; which accompany the act.
if they have four shots to fire, the first shot—and we placed particular Pitts; I expect that when an experimental psychologist uses "motiva­
importance on that one for our measurement—^is fired at the seventh or tion," he is in essence defining it by the experimental arrangement he
the eighth light, that is, when there are seven or eight lights remaining. sets up to produce it in a given way, such as the $10 bill or electric
If, however, you start this game, the same game, not with sixteen lights shock. He means the external situation which he has imposed, not the
but with ten lights, they don't fire at seven or eight; they fire later, individual's reaction to it.
although, of course, the probabilities are the same when there are seven Rioch: But he doesn't measure it. That is just the point. He doesn't
or eight lights left! use methods which are operationally directed towards investigating this
Pitts: Oh, yes. I would say the game theory in relation to'a situation problem.
of that kind is roughly analogous to information theory in relation to Bigelow: He tends to eliminate the abstraction "motivation" by util­
your other case; namely, where, in both cases, the theory provides izing the end result, by measuring the end result.
maxima or theoretical optima, without, of course, providing any mech­ Pitts; Certainly, the experimental psychologist in general simply pro­
anism for analyzing cases where people don't behave in that way; but vides what he has reason to suppose is a desirable prospect or the
if you do apply a rank order for the people who arrive at the correct chance of avoiding an undesirable one, and then simply defines motiva­
answer first, you have a very interesting combination or conflict of tion as a state produced in his subject by that, without considering that
motivation with information, since, of course, information from the it may vary considerably, depending upon his other conditions.
other people is necessary for anyone to have the answer. I think that Rioch; Yes, having defined his desirable and undesirable by what he
might possibly be important, while still retaining the great advantages thinks about it, not by what the rat thinks about it.
of a controlled system where the communication can be analyzed very Bigelow; Aren't there plenty of examples of learning as a result of
carefully. hunger ?
Bavelas: It would be interesting, as a matter of fact, to set up groups Birch; Oh, yes.
in which, if everyone in the group gets the answer, there is a certain Bigelow: Aren't these perfectly good continuous variable measure­
payoff to everyone, but with provision for an individual to get a private ments in the sense that he is—
reward even if the group fails. Birch: No, no.
Pitts: Yes, make the individual find the correct answer himself first, Bigelow; To some extent?
which he must do, of course, by means of information from the other Rioch: No.
people. Bigelow: To a considerable extent?
Rioch: I want to say just one thing about this concept of motivation. Rioch: No.
I don't think we have any basis for using the term, because there is no Birch: Well, the problem is a rather difficult one. Let us take some
differentiation made between "motivation" which is increased alertness of the systematic attempts to study the quantitative variation in moti­
and "motivation" which is deterioration of adequacy. We speak about vational states in animal learning or in animal performance — the
"motivation" when we have a physical feeling, but such somatic feel­ Columbia studies in the early thirties, for example, that were done
ings may represent any one of several personal operations. There is at under the aegis of Warden, in Warden's laboratory (2) (3). Under
36 Cybernetics Communication Patterns 37
those conditions you tried to compare such things as the relative groups than you do for the twelve-hour deprived groups. This was a
strengths of maternal drive, sex drive, himger need, and so forth, in function not of the time of deprivation but of the time of day that it
the frequency and in the rate of the animal crossing of a grid, let us was possible to introduce such a period of deprivation and still have
say. Warden found that he could get certain curves that would be re­ the animals work. The twelve-hour period could be introduced only
lated to the presumed intensity of drive; actually, however, he was not overnight, because the animals would not work at certain other times,
discussing drive but deprivation, which is something else again. He in accordance with their own rhythms. They had a life history in which
was discussing what he was doing to the animal rather than the effect they had not been fed during the night, anyhow. Therefore, vmder
of what he was doing upon the animal, so that a number of problems those conditions, you had a noncontinuous relationship between the
arise. periods of deprivation and the effects of these removals upon the
For example, if you deprive an animal of water, what happens to its animals.
sensitivity to shock, independently of how many times it crosses the Gerard: I am sorry, but all this seems to me irrelevant to Bigelow's
grid.? In other words, the problem of sensory threshold 'and sensory question, as the whole discussion is irrelevant to the paper. What you
threshold changes under conditions of deprivation as derived or as an gentlemen are saying is that poor experiments have been done, not that
ancillary effect of the condition of deprivation have never been studied; the experiments can't be done properly, which is what he means.
what has been assumed is that once you have deprived the animal of McCulloch: I think they can't be done properly, and I am going to
something, you have automatically a direct relation between this depri­ say it very simply. Under absolutely standardized conditions, you can
vation and the amount of motivation which it itself has. It does not have organisms that prefer A to B and ß to C and C to A. The con­
follow. Therefore, you get a quantitative relationship between two sequence is that the strength of motivation with respect to the three
abstractions that you have developed, rather than a quantitative rela­ items is multidimensional and cannot be measured by any one quantity.
tionship between two processes that may be occurring for the animal. I think that some time we had better sit down with the problem of
Rioch: As a matter of fact, what we need is a definition of motiva­ motivation, because it is a tremendously complicated one.
tion in terms of information theory, because the whole concept of Gerard: I will agree with that, but that is still not the point Bigelow
motivation is in terms of energy mechanics. is making — whether you cannot get graded, instead of Yes and No
Bigelow: You mean, there are no ways to eliminate secondary effects answers. I think these experiments are potentially designed to give
upon the animal itself, of any of the usual incentives, when varied graded answers.
quantitatively ? Bigelow: It is not a question of whether they are easily separable
McCulloch: That is the awkward thing— but rather whether they are separable—
Bigelow: Is that true, sir.? Rioch: They are separable on the basis of the frame of reference in
McCulloch: —that motivations to and from turn out not to be added which you are looking at them.
or subtracted but entirely haywire in relation. Bigelow: This is my question—
Bigelow: But you might suspect that, for example, one might take McCulloch: Will you hold it one second. There is a question of
the height a rat will jump over a fence one or two or four hours after Shannon's still on the floor.
food to get to a given target— Shannon: Well, the remark was about the theory-of-games possi­
McCulloch: Well, in three hours, the rat is practically dead. The bility. It seems to me that all of the experiments you did could be
rat must eat oftener than that. interpreted as games. The chief trouble is that they are not zero-sum
Bigelow: But there are no ways to eliminate these secondary effects two-person games, which is the simple case. Most of them are non­
in experiments? zero sum, because co-operative effort gains for everybody; also, they
Birch: You get peculiar reversals, too, which apparently indicate that require more than two persons. This type of game is much more com­
what is happening is that a different qualitative process is becoming plicated in the sense that there can be several strategies, several systems
involved rather than that a nice homogeneous continuum is being ex­ of imputations among the various players. I think that perhaps what
panded. For example, in one study of my own on the effect of differ­ you are seeing emerge in some of these experiments are those patterns
ential food deprivation on the problems of behavior in chimpanzees, of imputations, particularly in the experiment where people were trying
you find that you get a greater predicted effect for six-hour deprived to match a given sum, and where one person would take on a certain
38 Cybernetics Communication Patterns 39

role and another person a different role. The various ways that these The really interesting aspect of the experiment, though, is its depar­
roles could be assumed correspond to different good strategies for the ture from the Von Neumann concept. Thus, the team is not permitted
game. to get together before the whole experiment, armed with foreknowl­
I also think that the poorer results you obtained when they were edge about its general nature. In contrast, suppose five of us, having
given the amount of error were really an indication of the irrationality heard Bavelas's lecture, were to go up to the Massachusetts Institute
of people, because in a game situation the additional information cer­ of Technology on the train to participate in this experiment. We would
tainly could not hurt you if you were perfectly rational; that is, if you sit in the parlor car and discuss how we should work, what codes and
choose the best strategy, additional information can only make it better shorthands we should use, and in general make agreements as to how
from the point of view of reducing errors, if you play the game by the to behave in the face of the various contingencies which might arise.
most rational means. If you had five Von Neumanns sitting in your We might then go up there and beat the pants off Bavelas. But the .
cubicles, the answer should be better or at least as good with the addi­ experiment as it is actually done carefully precludes all such pre-
tional information. arrangements.
Bavelas: You mean if there were five Von Neumanns playing the Therefore, interesting and invigorating though Von Neumann's the­
game. No. 1 would know what No. 2 would do ? ory is in many contexts, it seems here to have been dragged across our
Bigelow: Not by identity. path in two different directions as a red herring.
Shannon: Each one would choose the best strategy for this game, Bigelow: If time urgency were eliminated, though, wouldn't this be
and if that involved making use of the additional information, that possible? If they were allowed all the time they wanted to devise a
could only help them. If it didn't, they would completely ignore it. strategy, then this communication with pieces of paper would do that.
McCulloch: There is one question of Savage's, I believe, still on Savage: Then you might say that early parts of the game should not
games. Do you want to bring it in now.? be considered as a game; they should be considered in the light of the
Savage: Yes, I do. When Pitts alluded to the possibility of doing a prearrangements made on the hypothetical trip up to M.I.T.
modified Bavelas experiment, which would be a many-person zero-sum Bigelow: There, I agree with you.
game, as a check on Von Neumann's theory of such games (4), I tried Savage: Where there is a long breaking-in period, the participants
to speak then and there. I wanted to say that so far as I have been could begin playing the Von Neumann game, which, as I have said,
able to understand. Von Neumann's theory simply doesn't make test­ ought to be considered as a one-person one. But, of course, a one-
able predictions about many-person games. Though a lot of mathe­ person Von Neumann game is rather dull.
matical machinery is constructed in this connection. Von Neumann, Pitts: That is exactly the reason why I suggested introducing the con­
neither in writing nor in conversation, seems to me to make at all clear flict of motivations, because, in effect, it does become a real game on
what empirical consequences this machinery may suggest. The situation the assumption that everybody knows the structure, which you could
is totally different from that for two-person games, about which Von easily make, but has a motive different from the others.
Neumann's writing suggests quite definite consequences. Thus, though Savage: But that's a poor game.
such an experimenter as Bavelas can and does test the consequences of Pitts: I know; it's terrible. In the first place, you can say that the
the two-person zero-sum theory, I think he would not even know what theory of games naturally cannot make any empirical predictions as to
to look for in the case of many-person games. what any group of people playing a game will actually do. Being
While I am on the subject, I should also like to say a word about the mathematical theory, you can't have any empirical consequences di­
experiment Bavelas did do, as a game. Shannon has said, and I think rectly; but it does provide certain extreme possibilities which you can
at least partly correctly, that it could be considered as an example of a list; then you can inquire whether the actual behavior of the group
non-zero-sum game; but as such it would not fall under Von Neu­ shows a tendency to approximate those, and, so to speak, provide you
mann's definition, because there are artificial constraints on the com­ with reference points and a mode of characterizing particular sorts of
munication of players. Von Neumann would presumably prefer to con­ situations.
sider it as a one-person game, in much the same way as he considers Gerard: I want to go back to a comment that Bigelow made. He
ordinary bridge as a two-person game, neither player of which knows stated that, to him, the most important consequence of this research was
the whole of his own mind—is schizophrenic, so to speak. the demonstration that human beings under certain conditions calcu-
 40 Cybernetics. Communication Patterns 41

lated less well than the most primitive calculating machine. It worried deductive ability. If the mechanism operating here is some kind of
me at the time and is still worrying me, and I am going to ask Bavelas random search process in the mind or the mechanism of the individuals
if he agrees with that statement. The reason, it seems to me, why it is concerned, then we should make comparisons not with the ideal logical
not a correct interpretation is that when one starts with a rational series mechanism to deduce logical sequences, but rather with the ideal ran­
or set of operating conditions, as in the paper-solving of these prob­ dom mechanism to induce structure.
lems, one finds that the subjects do precisely the same thing that hap­ Bigelow: Well, it certainly starts off with a random input in the
pens when the problem is being solved in reality. In other words, I sense that you never know to whom you are going to talk. But beyond
wonder whether it isn't a matter of the people doing the best that is that, the question would be whether these people, on the basis of in­
possible with the information available at any time? Would you accept formation that they have at any given time, make choices which are less
. the conclusion that Bigelow drew, that human beings are worse than than optimum. Each individual may or may not make optimum choices
the calculating machine? with the information he has. Now, I would guess that the evidence
Bavelas: In these circumstances I would, because in the experiment here is strongly that they do, and that an individual computing ma­
where you have the individuals in a circle, it can be demonstrated that chine, given any information—they may start out with random search
the problem can be done in just three transmissions; but we've never and then increase the information in a certain flow pattern which is not
had a group do it even under specific requests to attempt it. exactly the same in each case — would make inferences from these
Savage: But six machines can't do it in three transmission move­ which are far more reliable than those of human beings.
ments; one machine could. It doesn't seem to me the individuals here MacKay: Yes, I think the importance of Shannon's point is that we
are guilty of miscalculation. They are guilty of international warfare should ask ourselves with what type of mechanism it is fair to compare
or something like that; that is, they are guilty of not getting together a human performer. In attempting to make sense of Bavelas's results,
into the best combines and subteams. where the subject has to decide whether to fire on the fifth or the sev­
Rosenblueth: Given the same instructions that you gave your five enth light, our analysis will be quite different according to whether we
people, I suspect five machines would not do it any better. assume the performer to be designed as a logical calculator or as a
Bigelow: I certainly do not. I mean, if you give them more informa­ mechanism adapted to the kinds of situation that confront a human
tion than they need, the machines will certainly not do worse as a being. If we ask whether a human being is as good as a machine, or
result; but the people did. whether an action is "unreasonable," the answer will depend on which
Rosenblueth: I mean, give them the same information. data and assumptions we want our ideal mechanism to use. Psycho­
Bigelow: This is a case where excessive information would produce logically, it appears that the subject treats the array of lights as a uni­
a negative result. form assembly and picks something near the middle as a typical point
Rosenblueth: Excessive information of the wrong sort. in it. He is presumably using a random mechanism whose chief law is,
MacKay: It seems to me that this is a two-level problem, and that is "Thou shalt show no bias."
why we are arguing. There is, first, the problem of formulation. You Bigelow: Surely, but that is the point, that here is a striking example
see, these people are locked in, and the problem is formulated only in where the human being is not behaving like a computing mechanism.
vague terms, and they've got to be able to develop for themselves a MacKay: I am sorry, but I think it is a case in which the human being
model of their situation in terms of which they may draw their deduc­ is behaving, certainly not like a normal computing mechanism, but like
tions. Second, there is the question of drawing the deduction. If you a random-search-computing artefact, if you like.
put a man in the situation in which the problem is so formulated that Bigelow: Fine, but then it is not a computing mechanism in the
the question is as simple as, "What is 2 and 2?"—if that is simple— ordinary sense.
then he will make his deductions as reliably as the computing machine. MacKay: Not like the conventional type of computer, no; but not
But the first problem here is essentially one of abstracting a structure necessarily inferior (or superior) to it. It has to be ready for a wide
from a pattern of experience, isn't it ? and any child is better equipped range of quite different tasks.
than a normal computer for that job. It is possible that a suitable Pitts: The computing machine can certainly be designed to carry out
machine could be made to do better than a human being at both tasks, processes of random search in a way which minimizes the number of
but it must meet a criterion of performance different from that of expected steps to acquire the desired information, and certainly such a
42 Cybernetics Communication Patterns 43

computing machine would do better than a group of human beings. It human being, and that the building of a machine is copying the human
could not do any worse. being over a narrow range of the human capacities but multiplying the
MacKay: Yes. The question is, first, whether, in the present prob­ human capacities in that particular narrow range. He described the tre­
lem, there is anything paradoxical, and, second, whether we are justified mendous resistance to making machines previous to the time of the
in comparing the performance of the human operator in this situation Renaissance. This anxiety is represented in all sorts of way, both in
with that of an optimum computing machine, which might not have terms of the danger of copying the human body and in terms of the
nearly as good an all-around performance. danger of dissecting or experimenting with the human body. The
Gerard: We are all talking about simple computing machines, not Koran, for example, forbids the copying of the human body, and a good
rather complicated ones. Mohammedan has no portraits.
Fremont-Smith: It is evident how intrigued we are with the problem In another very interesting scientific revolution, which occurred at
that comes up again and again: whether the computing machine would the end of the nineteenth century, we have the curious situation of the
do better or worse than the human being. I think that the degree to machine coming between the observer and the phenomenon. The
which that intrigues us is one of the most interesting things. manometer comes between the physiologist and the blood pressure, and
Savage: That used to be our title subject. We used to be a seminar the camera comes between the astronomer and the star; along with that,
or meeting on the subject of computing machines and we naturally do we get a great deal of anxiety about the machine. The exact verbalized
revert to the theme of the analogy between computing machines and form that it takes is different from that which it took in the Renais­
human and social behavior, because at one time, at any rate, it was one sance, but the general problem is still there. However, I think it would
of our most important theses, that we might find something fruitful in be a horrible comment upon the people who build computing machines
that analogy — not that the machines would be like people, because if they couldn't build one which, over a narrow range, was better than
they were made by people — after all, we don't expect a neon lamp the human being over the same limited range.
to be like a person because it is made by a person — rather, we had Pitts: That is perfectly true in principle. I think we have made a
at that time some hope, or some of us had some hope, and it was widely wrong comparison, really, one which is unfair to the human being. We
advertised, that computing machines would be like people because they have compared the human being to a machine which is designed to per­
did so many human things. It seemed to us worth while to examine form a particular purpose, and naturally the machine will perform that
these analogies, especially to discover, if ever we could, limitations to purpose as well as it is possible in principle to do so. We really should
the analogy. make an analogy between a machine designed to perform a particular
We have had, throughout the existence of the group, the important purpose and one designed for a purpose different from the one we are
problem of seeing if there is anything that people do, that can be pre­ considering.
cisely stated, that may not be done by a machine. We have always Teuber: You could rephrase that and say that the machine might do
known that that hope was in some sense chimerical because there is better in any individual situation, but an organism of the complexity
the famous theorem of Turing, which we used to hear about at every of man might still answer with relative adequacy to a much wider range
meeting, to the effect that if you could only tell precisely what you of situations than any presently constructed machine. I think Bavelas's
want, you could make a machine that would do it. And yet we have concern was not so much that, but the question whether we can use
continued to hope in spite of this theorem, which in a sense should end these particular models, at this point, as a measure of rationality, or as
all our hopes that we might discover some kind of behavior which does a measure of optimum performance against which we can measure the
not deserve to be called mechanical. This has not prevented us from actual performance of the group. Or, if the group tends towards opti­
simultaneously hoping the opposite, so we have repeatedly sought me­ mum performance, how can we use these mechanisms and mathematical
chanical analogues to the various aspects of human behavior. But it is theories in predicting how a group will act under certain simplified
by no subconscious accident that we revert to a theme which, as I say, conditions? I think that is a somewhat different approach.
used to be incorporated in the very title of our group. Gerard: Along the same lines, it seems to me that, although I may
Rioch: It is interesting to note the historical concepts of this. Hanns not recognize a good deal of unconscious anxiety as to whether we are
Sachs wrote a paper, "Why the Delay in Civilization," making the or are not better than our machines (which may lead to a preoccupation
point that the machine operationally represents a magnified part of the with it), the conscious basis for preoccupation is with the extent to
44 Cybernetics COMMUNICATION BETWEEN MEN: THE
which we can get useful clues from studying the machines, as to how MEANING OF LANGUAGE
our brains or our social groups work; and—
Teuber: Exactly.
IVOR A. RICHARDS
Gerard: —that is not a matter of saying, "It is," or, "It isn't," but of
Graduate School of Education
saying, "In what respects is it?" Harvard University
Mead: May I make one point more? Dr. Bigelow made the point
about the experiment that is limited and simple but gives us some clues
ALL SORTS OF doubts have waylaid the preparation of these remarks,
on the relation between the natural and the social sciences. I think we
had a very nice illustration of both uses here. When Bavelas produced doubts little and great. As an example of the smallest-scale doubt,
the patterns, Lawrence Kubie showed points that might be relevant. there was the queer classification in the advance notice for this meeting.
Now, it is perfectly true that you can take those out; you can set up an If you remember, we were originally going to study language in bees,
experiment and randomize them so that they are not relevant. Part of vertebrates, and men, which set me to feeling my backbone, but I know
what you said, Dr. Savage, was fine, that you should have different enough about this group and its history and its interests not to be too
experiments to study these deep psychological relevances as a step to much distressed by that. I knew there would be no special creation of
doing something else. You also said, "Who cares about an asterisk?" I Adam lurking in the background of this picture.
believe. Weren't you the one who said that? A larger doubt, of course, is the appalling scope of my assignment—
Savage: I did, and I said I was misinterpreted.
"Communication Between Men, and the Meaning of Language." Obvi­
Mead: That is the point which I think is dangerous. We have to dis­
ously, I could only sketch some aspects, and I tried to pick out aspects
tinguish here between the fact that you can do experiments, you can which might engage with, or fit into, the special interests of the varied
simplify them, you can make nice tight experiments, and you can use groups here present. I should like to stress for a second the variety
information about the unconscious to be sure that you have randomized that must be present. You can fill it in better than I can.
the right elements; but if you carry that over into real life, with a state­ It has been some comfort to me in listening to this morning's dis­
ment that a thing like an asterisk is of very little importance, and then cussion to see how many of the topics that I had taken up have actually
make an estimate about the amount of error that you will get in proof­ engaged discussion this morning already. From time to time, I will
readers in a year in the world today, with the amount of printing that point them out, if I can remember to do so; but you won't, some of you,
there is, you would make an enormous error. We keep vacillating be­ have any difficulty at all in seeing preoccupations which have already
tween the two. showed their heads this morning. One of them, of course, is the thing
Bigelow: The danger point is where you make use of such a specific
I just referred to—the heterogeneity of intellectual interests in a group
conclusion from an experiment. like this. The problem is whether this is one group or whether it is a
Mead: But that is one of the difficulties that comes between the
number of groups, and if so, how you would separate them into groups
groups, that when people introduce "real life" they are only introduc­ within the group. I can indicate some of them, for I think it will help
ing "real life" to make the experiment tighter; say, you had better the strategy of what I am going to say.
randomize this, or you had better take this into account, but don't Language I take to be pre-eminently the learned activity of man, and
generalize from the experiment in such a way as to say, "Who cares learning itself to be the major point of procedure of evolution in a con­
about an asterisk?" jectural world process. I am struck, accordingly, by the vigorous efforts
Savage: Let me defend myself about the asterisk, if I may. I said
which language-theory specialists—and here I had better go into in­
what you said I did, and you criticized it rightly. What I should like verted commas—or "linguistic scientists" (as they are proud to call
to have said is this: Who will pretend that the six symbols which themselves, with a great deal of punch behind the word "scientist" as
happen to enter into this experiment are the best six to study with re­ opposed to any other description) have made to achieve something like
spect to their psychological import? The six are accidental and, in that autarchy, or independence, of the whole intellectual confraternity, if
sense, no one cares about them, although, of course, one or more of we may give it such a pleasant name. There has been an attempt at
them might turn out to be important in some psychological context. isolationism among linguistic scientists—it keeps recurring in different
45
46 Cybernetics
Communication Between Men 47
modes—and among theorists of the study of behavior. There is a revolt worst, shoulder all of them. It has to represent all the tensions and
against, and a withdravi^al from, other studies that amounts at times to troubles of language, and these in turn represent all the troubles and
a breaking off of diplomatic relations. tensions of the most complex modes of life.
Now, I take philosophy to be the over-all name nowadays of the I have used the word ""represent" here, and begged with it, inevitably
diplomatic agency which endeavors to keep studies in some touch with and typically, typically and inevitably, most of the questions that any­
one another. I think this attempt to break away, to secede, on the part body could raise about what language is and how it works. Does it
of new sciences—and linguistics and behavior are two new sciences we work through representation? What in the world is representation?
might think of in this connection—is dangerous for them and for others. Perhaps if we considered those uses closely enough and with an atten­
Both dream, I think, of intellectual world conquest. It seems a strange tion of enough resolving power, we should see that they had to be dif­
sort of design to see behind a proposal for autarchy or isolationism, ferent, in meaning and, even more important, in type of meaning.
doesn't it? And to the observer of these things, the diplomat, the lin­ My point is simply this: The very instruments we use, if we try to
guistic scientists, and the students of behavior alike show a certain say something which is not trivial about any aspect of language, em­
young ruthlessness and disregard. body in themselves the problems we hope to use them to explore. The
Now, all this is just to explain what' follows and why. I am con­ doubt comes up, therefore—and it is a very familiar doubt, if we may
cerned with the diplomatic-philosophic strategy for the further study linger with it for a second—as to how far we can hope to be understood
of language rather than with any attempt at an over-all presentation or even to understand ourselves when we use such words (as we all
or any report on any investigation of my own or anyone else's. As I inevitably do, all the time) ; and in the lucidity of this doubt, the
proceed, I drop the suggestion that among new sciences that may literature of the subject, I think, can take on a queer appearance. If I
emerge there might be a theory of analogies, systematically developed. may quote one sentence only from the last time I tried to write on this
There is this great analogy between the state, the political world, and topic, in the Philosophic Review of March, 1948 (2)—quite a time ago,
man as the epitome of the political universe—the kind of thing done by I realize—I had this to say: '"It is odd indeed how the artificialities of
Plato, which I suppose has ben the most fertile analogy in the Western these ancient rituals"—the rituals of philosophic and linguistic discus­
tradition. As always with such analogies, it works both ways. There is sion—"are maintained, how writer after writer will lay on at his op­
traffic in both directions. The political concepts developed on the one ponent with words like "know' and "true' and "say' and "be' and "mean'
side shaped beliefs about personality structure on the other, and these and 'believe' and 'understand,' as though such strokes of tongue or pen
in turn are reflected. It is hardly possible to discuss one set apart from could hit anything, and as though finding out how those words may in
the other. fact work were not after all and underneath all the forever neglected,
The only other analogy that strikes me as having an equal past, an though ostentatiously paraded, aim of the procession."
equal present, and perhaps an equal future is the analogy with which Well, now, this situation which I have tried to highlight with that
you as a group have been especially concerned and about which we picturesque antic is not, of course, peculiar to the study of language, as
talked this morning—the analogy between an organism and a machine. we all realize very fully. All studies suffer from, and thrive through,
That, again, is one of these prodigious analogies which should be under this : that the properties of the instruments or apparatus employed enter
special study in relation to other analogies which may be used for into, contribute to, belong with, and confine the scope of the investiga­
similar purpose. tion.
To come back to my doubts: a penetrating doubt, truly a bosom That, again, is a topic that I am happy to notice arose this morning,
doubt, concerns the sort of language with which one may profitably try fleetingly. I am going to put a great deal of stress upon this. The prop­
to talk about language. I think the most shocking understatement that erties of any apparatus used—and the apparatus will include pre­
I have ever met in my reading came to my attention quite recently. It eminently the language of discussion and whatever else you like to put
appears in C. L. Stevenson's engaging but somewhat tangled book, behind the interpretation of the language—^enter into the investigation,
Ethics and Language (l). There he says, ""Language about language
and not only enter into it but belong essentially to it and contribute to
must share some of the complexities of all language." I do think that it and form and shape it, and, I suspect, confine it.
is a prize winner of an understatement. Of course, it shares; but not so A way to put a little familiar backing behind this, and to show the
much share them as it must shoulder them, and shoulder them at their general relevance of this strategic problem of how we are to talk about
48 Cybernetics Communication Between Men 49
language, will be to quote to you from Oppenheimer's popular survey then whatever you put above poetics, getting very near the highest of
of science for fifty years in the Scientific American, in its jubilee num­ heights, which I am going to list as dialectic. That would be the con­
ber of September, 1950. Here is Oppenheimer talking about Bohr's ventional increase of complexity scale for the history of the universe.
Principle of Complementarity: Furthermore, it might be held that the higher you go in this scale of
"The basic finding was that in the atomic world it is not possible to complexity, the more of the mind we bring in as apparatus or instru­
describe the atomic system under investigation . . . [<?] in abstraction ment of the inquiry. Here we have a word which we've got to put in
from the apparatus used for the investigation; [¿] by a single, unique quotes, and you will hear the quotes in my voice, I hope. Later I am
objective model" (3). going to do some vocal antics to show what kind of quotes I am using,
I have intruded upon Oppenheimer's sentence with a and b. Oppen- and "mind" is a very suitable word for this sort of experiment. I am
heimer does not separate those two things; for his purposes he does not going to suggest that we leave the problem of what we might mean
need to. But there is a Principle of Complementarity which applies to by it, and say that the more complex the whole investigation is, and
the limitations of the model, and another principle that it would be well the more complex the subject matter (if you can separate the subject
to have a name for. I haven't thought of the right name. It might be matter from the modes of inquiry), the more necessary it is to bring
the "Principle of Instrumentality": you cannot describe any system in more of the mind.
under investigation in abstraction from the instrument used for the in­ I am suggesting that the mathematician, as a mathematician, uses one
vestigation. Oppenheimer goes on to a familiar point: branch only. It may be a prodigious branch of human activity, but, in
"Rather, a variety of models, each corresponding to a possible experi­ comparison, the anthropologist has to reveal many more sides. As for
mental arrangement and all required for a complete description of pos­ the student of poetics—^I venture to rely on Coleridge here—he is the
sible physical experience, stand in a complementary relation to one only student who has to bring in the whole of the mind, with the ex­
another, in that the actual realization of any one model excludes the ception of the dialectician, who manages to bring in a little more.
realization of others, yet each is a necessary part of the complete de­ You will remember that Coleridge, at the end of the fourteenth
scription of experience in the atomic world." Well, all that is in terms chapter of Biographia Literaria, said, "The poet described in ideal
of the atomic world, so Oppenheimer very properly goes on to say: "It perfection brings the whole soul of man into activity, with the subordi­
is not yet fully clear how characteristically or how frequently we shall nation of its faculties to each other, according to their relative worth
meet instances in other fields, above all in the study of biological, psy­ and dignity." Well, that is what the great investigation is. What are
chological, and cultural problems" (3). the relative worths and dignities -of all the instruments, the faculties,
What I am going to do essentially is to point out some places in the that together make up that which may and must, perhaps, enter into any
field of the over-all strategy of linguistics where you will find principles investigation and be its instrument?
of complementarity and instrumentality intruding. Here, I am very Now, corresponding to all these studies are characteristic uses of
much at your mercy, but it is amusing to speculate for a minute or two language; and poetics, I suggest, is faced by the most complex of sub­
on the sequence of fields in which recognition of some such principle ject matters and is required to undertake the most complex activity itself
as the Principle of Instrumentality would arise. in the process of the investigation. I have put dialectic above it. Dia­
Where would you most expect to find it recognized ? I conjecture— lectic is above it because it is concerned with the relations of all the
and I speak very humbly here—that mathematics may have been the subordinate studies in a way in which poetics is not. So dialectic is the
earliest study forced to ask itself about its own intellectual viewpoint, supreme study, as it were—^with philosophy serving as a sort of Harry
and the influence of its symbolism on its scope. This may suggest that Hopkins.
the more abstract the properties of the instruments, the easier it may be I want to linger for a moment, if I may, with anthropology, because
to take account of their presence and not overlook them. If so, we of the close ties and great influence that anthropology has lately had
might get the familiar complexity sequence of events and their corre­ with linguistics. There has been a bit of a "ganging" around there. The
sponding subjects: mathematics, physics, chemistry, biology, sociology. chief methodological problem, I take it, of anthropology is very closely
Should psychology or anthropology come next.? You may like to paralleled in linguistics, if it is not identical; though perhaps lin­
change them around. After that, we are sure to have a quarrel. What guistics, by its humbleness of scope, by severe limitations put upon itself
is more complex and concrete still than anthropology? Poetics, and as to what it shall and shall not take into the investigation, by confining
50 Cybernetics Communication Between Men 51
itself largely to morphology, and by something that sometimes looks No, the linguistic scientists are working out an over-all apparatus
like a kind of panic flight from semantics, has been making the better which can, they hope, be used for an examination and description of all
progress. language. It may very well be that no one such apparatus will prove
This chief methodological problem I think I can state by another to be feasible, in which case you will be back in what will then be a very
quote taken from the same handy source, the Scientific American, the familiar Principle of Complementarity situation, with a number of dif­
same number, in a later article. Kroeber says, ""Anthropologists now ferent systems of description, each valid within its limits but not com­
agree that each culture must be examined in terms of its own structure patible with one another, or as yet reducible to an over-all united
and values, instead of being rated by the standards of some other civili­ system.
zation exalted as absolute" (4). He puts in a historical comment on So far the linguistic scientist's apparatus keeps within very narrow
that which I think would not stand up very long. He says, ""—which in limits as to which of the features and functions of languages it is yet
practice, of course, is always our own civilization." I think study of the ready to give an account of or even examine. Here is a neat statement
curious ways in which modern culture will judge, not according to their of this:
own standards in the least, but by Judaic, early Greek standards, and ""The linguist focuses his attention upon those selected aspects of
so on, would make trouble for Kroeber. But still he says, ""This prin­ language which he believes his methodological equipment gives him
ciple leads, it is true, to a relativistic or pluralistic philosophy, to a be­ the authority to investigate." Limitation by the equipment. He doesn't
lief in many values rather than a simple value system. But why not, attempt to study the totality of , language phenomena. Again, "The
if the facts so demand?" sharp separation which exists between linguistics and other disciplines
I venture to suggest that the basic principle of the relativistic ap­ is, at best, an arbitrary one based on the development of an isolated
proach which he is preaching here is plainly a halfway house, not per­ methodology and not on any empirical division of subject matter" (5).
manently tenable. It is a methodological self-denying ordinance. And This brings me back to what might be called the marked ""Ishmaelite
better not say a halfway house, for, good heavens ! it is an early staging complex" that haunts linguistic scientists and students of behavior.
place, not nearly halfway, perhaps not even a millionth part of the way, Their hand is against all men. I don't think all men's hands are against
if one can imagine such a progress being metaphorically measured. It them, though they seem sometimes to think so, which is enough. But,
is a negative, defensive step, this relativistic approach, an anti-imperi­ above all, their hand is against others who like to think about meaning
alist move, necessary and desirable, of course, but not at all a suflicient with methods other than their own. That may be explained by the dif­
principle for an over-all comparative study. It is a parallel to the lin­ ficult labor they are having in working out an over-all descriptive tech­
guistic principle that the structure of a language is not safely described nique for all the languages which are constantly, of course, showing
in terms of the structure which has been devised for describing some more and more variety as more and more people know more and more
other and different language. English language must not be described languages. In any case there is—and this is, again, something which
in terms of Latin, or Hopi in terms of English grammar. popped its head up in our discussion this morning in another context—a
"What the linguistic scientists have been doing is to fashion a grow­ hostility and even contempt frequently shown by them for those who
ing system of instruments for comparison, an apparatus for overall talk about meaning, and there are all sorts of rather virulent blasts that
survey among languages, able, they hope, to put diverse languages into they emit against ""mentalism." They share that with some students of
a common frame. They are not, of course, doing what Kroeber says behavior very markedly. I think there may have been in the background
they should do if you transfer it to the linguistic field. They are not at­ some academic or, at least, intellectual persecution of both these young
tempting to examine and describe Hopi in Hopi, or Kwakiutle in studies, but that should surely be over now. The boot, in fact, is rather
Kwakiutle. They don't know the languages well enough to do it in on the other leg. That is just an aside.
most cases, I believe. Anyway, it would be very difficult, because in I am coming to some rather deep water in a moment. I am no doubt
Kwakiutle, we are told, you cannot say, ""The farmer killed the duck­ influenced toward this frivolity by the trivial fact that the book I once
ling," without saying something in Kwakiutle about the space rela­ wrote with Ogden has been about equally attacked for being mental-
tions of the farmer, the duckling, its owner, and so forth, to the speaker istic—and a type specimen would be J. R. Kantor's Objective Psychology
and the listener. You have to put all those specifications in or you can't of Grammar (6)—and for being behavioristic, as recently as two years
say it at all. ago, by Max Black whom I mention because I wish so much that he
were here.
 Communication Between Men 53
52 Cyhernetks
That recognition is something that we should look into more seriously,
I have come back, you see, to my initial bosom doubt: With what
for as language study advances, the probability that it will upset funda­
language should we talk about language ? And typically, should it be
mentally the procedures of its own inquiry is high, and, if so, the
mentalistic or behavoristic ? I had better calm your worst fears at once
chances of Principle of Complementarity situations arising are good.
by saying that I do not propose to solve this problem this afternoon.
Meanwhile, pending such ambitious developments, there would be
I hope instead to refrain either from dismissing it or from assuming
gain if we merely appended to an important word indices showing, for
any answer to it, unless some analogue to Bohr's Principle of Com­
plementarity would be an answer or, rather, a staging place on the way example, how literally, and with reference to what- defining connec­
tions, we are using it. "Mind" would be an excellent subject for such
to further inquiry.
I have not forgotten that this is a conference on feedbacks and cir­ work. We tag it, when we can, to show what sort of use we are in­
cular mechanisms; but before I show you that I have been trying to viting our hearers and ourselves to make of it. Indices showing the
prepare for discussion of these things as they appear in language itself status—empirical, hypothetical, or systematic—we are giving a word
and pre-eminently in the learning of language, let me say a word or two in an exposition would be useful, or indices showing the mode of be­
about the possibilities that open up of inventing new languages or new lieving we consider appropriate to the utterance in which it occurs. We
are not ready to admit that there are a great number of varieties of be-
language devices, perhaps for use in linguistic theory.
There are the quotes, of course, we use to prevent some routine pat­ lievings, that the belief attitudes are almost as diverse as the belief ob­
terns of behavior from being exhibited by readers. It is a very important jects, or that they can be so treated with advantage, if we are ready to
device, but we are unregulated in our uses of it. Some time ago I sug­ depart from the framework of Standard Average European patterns.
gested some specialized uses of such quotes, replacing inverted commas In the Yana language, we are informed, you have to show by the
by sw—sw^ etcetera (7). In point of fact, this words you use whether a statement is known to be true and vouched
is an excellent teaching device. In a few moments I shall try out some for by you, or whether it is made on someone else's authority. Now, if
that were made mandatory for all remarks about language and lan­
vocal equivalents.
Formal linguistic description does, of course, use many more or less guages, I can tell you something of the silence that would ensue.
routine types of descriptive terms, many names such as phonemes, mor­ To come now to circular and feedback mechanisms: in what I am go­
phemes, tagmemes, lexemes, etcetera—nearly all nouns, please be it ing to say now, I suspect that if we could allow for the influence of the
noted. I don't think much experimentation has been done with other language I shall use and its vast difference from the languages that al­
categories of technical language, technical verbs. But if we take to most any of you, I think, would use—different though your languages
heart— and I don't know whether we should take it to heart; I defer to be from one another and great though the strain we experience in talk­
anyone who can check it—what Benjamin Lee Whorf (8), for example, ing at all in the same words, except from the speller's point of view or
had to say about Hopi handling of numeration, time-space, substance, the phonetician's—in spite of all that, we may have a good degree of
and matter, and notice his remark that in Hopi "generality of state­ agreement. I don't know; recall the embarrassing situation that came
ment is conveyed through the verb or predicator, not the noun," you up in our discussion this morning. AmT chiefly informing myself.? Is
will see what an enormous field the construction of speculative instru­ it a case of information given by myself to myself ? Or am I really talk­
ments other than our Standard Average European patterns might ing to people who have to take what I say, put it into very different lan­
give us. guage through diflFerent transformations in many sciences, fundamental­
It is the hardest thing for any study—I think this may be universal— ly recode it, and then come back, but through another receding process,
to notice and then abide by the new conditions set up by its own ad­ to try to put me right ?
vances. Scientists very often cut their own throats and then go on, pro­ You see, now we are coming to a nest of things that we talked about
ceeding as though their throats had not been cut. It is usually not a this morning. If the properties of the apparatus employed do enter in,
success. The problem is how to find out what we have done, take ac­ and belong with and confine the scope of an investigation, the problems
count of it, and proceed in the light of what we have done. I attempt to explore will be in certain respects essentially different from
Here is an advance—the recognition that the linguistic structure de­ the problems that you would explore in your intellectually native and
veloped in a group of languages, in the Standard Average European, acquired languages. How different they must be and will be, and how
say, may have structured even the fundamental concepts of their users. to investigate that are the methodological problems.
54 Cybernetics Communication Between Men 55
All this while, of course, I am exemplifying, and I would fain say il­ main senses of "end"—to the "acteevity" of searching. Something is
lustrating, circular and feedback processes in everything I say, as none fed forward by coincidence with which that "acteevity" reaches its
of us can avoid illustrating them amply and continuously in everything terminus and goal.
that we say. There is no escaping that. I need not insist here, of This word I have been using (^®know^®), can't be treated too
course—in other places I would—that what we say next will be in part warily. It is a great traitor. I don't suppose one should mean anything
controlled by feedback from what we have said up to that point, and more here than I have already said about it in terms of the behavior
the feedback from the reverberations of what we have said up to that involved in searching—the feedforward kind of thing. Nothing need
point. And at this point I must say that throughout I have been pre­ be implied as to how it is done, whether it is done through an image or
paring myself and you, I hope, to be receptive to, to expect, to attend through a substitute for an image, a schema, or anything else you like
to a certain peculiar stress on something which does not, I am surprised to think up, anything more generic and more symbolic. I know, gen-
to note, appear nearly so often as it used to. There has been a waning, erically, in these sorts of terms, now, the grave doubts that many people
commented upon this morning, in some of our attention to the original here must be feeling as to the sort of thing I am saying. My knowing
curiosities that brought members of this group into contact with one them in some sort of schematic way is my main guide, the feedforward
another. I am going back, perhaps, nearer to those original interests. through which I try to find further things to say, which may in some
Perhaps this thing on which I want to put the spotlight will be con­ measure meet them. So focus your doubts and bring them, apply them,
sidered to be included in some ingenious way under the word "feed­ if you will, to—^well, any sort of example.
back." But what I am going to stress stands in an obvious and super­ I am now going to feel in my pocket to see if I have, among the
ficial opposition to "feedback," and it will, in certain frames of thought, pennies which are there, a dime—a typical acteevity of searching. I
be given nearly, if not quite so much, importance, and sometimes more am going to know that by touch, possibly aided by hearing—with a bit
importance than feedback itself in certain connections. It is certainly of difficulty. I have fed forward certain things and—yes, here is the
as circular. You have no doubt fed forward enough to see that what I dime. I am too far off and the room's acoustics aren't so good that you
am going to talk about from now on is feedforward. I am going to try can hear what you have fed forward equally with me as the identifying
to suggest its importance in describing how language works and, above mark—the scrape of my thumbnail on its milled edge—by which I have
all, in determining how languages may best be learned. recognized my dime. That, I think, is an instance of characteristic feed­
Before going into that a bit, I should point out a few instances of forward behavior. I am trying to draw attention to what is the distinc­
feedforward at prelinguistic levels. We are familiar with them at all tive mark of what I am calling feedforward.
times. There is nothing novel in the least in what I am going to say, Very probably some of you will be thinking that there is nothing
but let's attend to it for a moment. here but "taping"—if I may use that as a technical term here, and I
Let me put a tag on a key word here—"activity"—which is capable think I can—a special taping for a special run of the manipulative-
of extremely diverse interpretation, as we all know. I put a vocal tag perceptual mechanism by which I conduct that particular bit of behavior
on that and pronounce it "acteevity." The Scots have a reputation— of finding the dime and not the penny. It is taping plus resort to a
which may be excessive—for knowing what they mean. Also, I am memory store, containing what happens when you scratch the edge of
indebted for my theory of "acteevity" very largely to Stout (9), and I a penny, and so on.
fancy that Stout came from the Shetland Islands. You may say that there is nothing distinctive about feedforward, and
Take the acteevity of looking for something, hunting, searching. We that this sort of general account in terms of a memory store and an
don't find anything, if you put the right mental-vocal tag on ^®find,^® ad hoc taping covers all that need worry us about all these very great
unless in some sense we are looking for it. You may happen on some­ words—about "purpose," "intention," "foresight," and the rest, and
thing, but in my technical tagged sense of the word ^®find,^® you do "feedforward" among them. If that is so—and I am not disposed to
not find it unless you are looking for it and—now, mind you; don't be dispute it—that turns the question simply into one about taping, about
too precise about that—in some sense, at least, know it when you find ad hoc tapings, their sources and their dependence upon more generic
it. You can't give an account of ^®find,^® with my tag on it which tapings. One of the things I want to offer for discussion is the de­
doesn't bring in the problem about ^^know'^® with my tag on it, and pendence of ad hoc taping upon more generic frames of tapings within
vice versa. They are correlative terms. Finding is the end phase, in both which it has a place. Tapings, I am suggesting, are hierarchic. All
 Communication Between Men 57
56 Cyhernetks
criticism and in the pedagogy of teaching the beginnings of a subject,
this is extraordinarily obvious. The only thing that may be puzzling
a flexible fitting in of means to ends—in more solemn language, de­
about it, perhaps, is the language I am trying to put it into. Tapings
seem to be hierarchic, or to form an enclosure series; the widest, most sign—is all-important.
For criticism, which is radically an evaluative ""acteevity," the dif­
inclusive, or over-all tapings being least determinate as to what will
ference between better and worse utterances is in design. Poor speech
end them, what will be their goal, their terminus. The higher, more in­
and writing are poor either because they are not attempting anything
clusive tapings, I suggest, issue for their own maintenance and execu­
worth trying or because they are inefficient. You can reduce the whole
tion, to preserve themselves and to achieve their goals, subordinate,
thing to efficiency, I would maintain. And the design that is there in
narrower, more specific instrumental tapings; and these do so again
the background, I should insist, may be, but need not be, witting.
and again and again and again, in the hierarchical pattern. For ex­
This principle of efficiency is, I think, no more than a recognition of
ample, consider an animal hunting for food. Let him be a sizable ani­
the enclosure series I have been mentioning—the hierarchical service,
mal that we can watch with ease, say, a grizzly. There is a very general,
the ad hoc tapings, serving wider aims, which serve wider and wider
all-inclusive taping—the search for food. As he scents this or that
aims up the hierarchy. And, of course, since language is, I think, ines­
possible source of food, subordinate tapings are issued. Suppose he
capably a social activity which only comes into existence and owes its
scents ants. He subtaps himself to follow up the scent. Suppose the
whole character to mutualities between men and within communities,
ants are under a boulder. Then enter subsubordinate sets of tapings,
study of language is inevitably dependent upon ethics. The study of
ad hoc tapings, based upon the size and shape of the boulder. When
language is concerned endlessly with the better and worse of utterance.
at last he gets to the ants, consummatory behavior occurs, and at that
It is normative through and through, as characteristically normative and
point there is coincidence with the goal-end point of his subordinate
inevitably normative as, for example, the study of medicine is, if you
tapings, and he stops; or else the smell of some berries sets him off on
will give a sufficiently inclusive, over-all taping to what you put be­
a new cycle of the same sort of pattern.
All this is a very familiar line of speculation, oh, so familiar! What hind medicine, that ""acteevity."
How many other studies must also be normative.? I am not certain
I am hoping it will say for me is this: with feedforward—and with
that they must not all be. How about biology.? Isn't it normative, radi­
that I am trying to name the peculiar character of tapings which arise in
cally normative throughout.? I mean, in the sense that about each or­
the service of less definite, more inclusive tapings—the adequacy of any
ganism studied the student must ask, '"How far is this a typical, normal
description or evaluation of any "acteevity" depends upon recognition
specimen? How far is it representative? How far is it usable,in the
of the sources of its feedforward.
Now, that has all been teleology and old stuff. I have some more old purposes of a general investigation?"
This is the situation. Surely, it is very familiar. I am merely going
stuff to add to it, and then I shall be making my way toward an end. I
am a little alarmed lest I go on too long and too continuously, but I to illustrate it with one glance again to Coleridge. Here is another
think that in a very few minutes from now I shall have sketched the declaration of the relativistic approach which ties in, I think, as a pro­
over-all picture; and, of course, on my own theories, it would only be visional staging place with' the Principle of Complementarity and es­
within the framework, the over-all taping of my general presentation tablishes esthetics and unites it with other evaluations. This is a scrap
that any of the points in it could properly receive attention. That is of Coleridge's which can be found in his Miscellaneous Criticism, his
an uncomfortable consequence of this sort of doctrine of composition. Shakespearian criticism. It is in the first volume, on page 196 (10).
This is the moment for me to trot out my two pet professional prej­ He says:
udices. Everybody's pet professional prejudice is ego-idealistic in this ""We call, for we see and we feel, the swan and the dove both beauti­
ful. As absurd as it would be to make a comparison between theii
connection; as an investigator they govern him. Well, mine are like
this: I am by profession a critic, concerned with the value of uses of separate claims to beauty from any abstract rule common to both, with­
language, and a pedagogue, principally concerned with teaching the out reference to the life and being of the animals themselves—say
rather if, having first seen the dove, we abstracted its outlines, gave
very beginnings of reading, A-B-C stuff, and the very early stages of
them a false generalization, called them principle or ideal of bird-beauty
teaching a second language. Criticism and pedagogy, thus, for me con­
and then proceeded to criticise the swan or the eagle—not less absurd
stitute two fairly high-level feedforward systems which tend and guide
is it to pass judgment on the works of a poet on the mere ground that
two extensive worlds of relatively ad hoc activity. And in both, in
38 Cybernetics Communication Between Men 59
they have been called by the same class-name with the works of other teaching or inquiring or searching in an area he thoroughly understands
poets of other times and circumstances, or on any ground indeed save has no patience for a second with the idea that because it is said widely
that of their inappropriateness to their own end and being, their want enough, it is right and that is the way you should talk. That is not the
of significance"—and then he upsets the whole thing with a very strong point of view of the over-all study of language. That is, I have sug­
word—"as symbol and physiognomy." gested, inescapably normative. It is concerned (as every speaker and
I think esthetics is there shriveled down to the judgment you com­ listener is always concerned) with improvement in the use of language.
monly pass upon somebody as to whether or not he or she is good- So this "scientific objectivity," of which so many linguistic scientists
looking. That doesn't mean that he or she has standard patterns of are so charmingly vain (like a boy with his first bicycle), is out of place
feature, physiognomy. You judge the part played by any element in the when it interferes, as it does, with education or criticism.
whole with reference to the whole, typically, in the case of apprecia­ This usage doctrine rules at very humble levels. I am just in the
tions of physiognomies. Anyway, here we have the anthropologist's middle of a weird and wonderful course at Harvard, where I have
relativistic approach back again in the service of esthetics, and I am freshmen, two hundred of them. I asked them to write on subjects
venturing to suggest that all studies whatsoever—and this is a very which you would think they couldn't possibly fail to be interested in
long flight—are normative in this sense, by the very fact that they use and speculative about. But the freshman lacks curiosity. I don t think
definitions and that their statements work only through agreements he knows a curiosity at the linguistic or educational or intellectual
among users; to use each word in such and such a way and not other­ levels. The curiosity he most shows is career curiosity. That is very
wise. Insofar as anyone does not use them, then the over-all taping, active. But on an intellectual level, if you ask him why he believes twice
the purpose of the discussion or meeting, is not served. And that gives, 2 are 4, two-thirds of them haven't the beginning of an idea of how
unless I am in need of a very great change of view, a sense in which all to speculate about it. All they can say is, "Teacher told me, and I
studies whatsoever are, I believe, both relativistic and normative, inso­ would have got dreadful grades if I had said "5.' " They are ruled, yovi
far as they depend at every step upon ad hoc tapings for the definitions see, by a purely usage doctrine on the multiplication technique. I think
of all the terms in the language used in the exposition; and those that is deplorable; and I notice, I am sorry to say, that this pattern {ad
choices of meanings are controlled by the hierarchy of tapings, ad hoc hoc examination taping) rules their intellectual behavior very widely.
tapings, wider and wider and wider in the hierarchy—all good tapings What they want to know is what is thought. They don't want to know
that we jointly share, I should say, with those which we have as human any "whys" or what it means.
beings, and under those common tapings we have as employers, let us Now, here is the point, you see, where philosophy, the diplomatic
say, of the Standard European language patterns. agency of dialectic, must intervene. It has to protect studies from the
Now I can progress quickly toward an end. You see, in terms of that interferences of other studies; but it has to do more than that. It has
sort of picture, why I started with a sort of naïve scientism which set up to go into studies and protect them and help them out of self-frustra-
autarchic policies for languages studies and for behavior studies. There tions owing to their neglect and ignorance of what other studies are up
are vast areas of purely descriptive linguistics which are a danger at to. And it is purposive, the whole activity of thought is as purposive
present to all the over-all purposes for which we use language. Here I throughout as you can make it.
get on to rather less speculative ground, ground on which I feel myself To come to my final topic: feedforward in teaching—especially in
much more secure, my professional ground. There are, I say, tech­ the teaching of the first steps in any subject, for example, beginning
niques—and standards derived from those techniques—^which are a reading. I am up against the discussion we had this morning about
threat to education, to the conduct of language generally. There is the motivation. I am going to dodge it, possibly, by saying "reinforcement,"
appeal to usage as sanctioning a mode of language. This very frequent where, if there hadn't been the discussion this morning, I might have
appeal I think is vicious. It illustrates all the dangers. Every useful wanted to say "motivation." My point is this: all language use, and
feature of language was not in use once upon a time. It had to come pre-eminently all language learning, depends on feedforward confirmed
in. Every degradation of language, too, has a starting point and a and regenerated, reinforced, by—well, the nearest word I can find for
spread; and behind usage at all times is the question of efficiency. In­ it is "success"—by enhanced power, a very general thing indeed—abil­
efficient features of language are not sanctioned by prevalence. We do ity, if you like, increase of ability. Success is the great general motivator,
not consent to allow them in our own special fields. Anybody who is certainly for beginning studies. In the beginning stages of a new activ-
6ü Cybernetics Communication Between Men 61
ity, it is all-important to avoid drops in what we might call "morale." changed as people, if you can imagine such a thing. And in most cases,
How to teach reading comes down to this : Keep avoidable mistakes we have found so far, unless there is something very wrong, there is
to a minimum. There is an elaborate technique by which they can be no reason why they shouldn't learn. What stops them is the harm that
kept to a minimum, and almost all current practice keeps them absurdly has been done to them by bad practice.
high. There are many sides to that. I have just stressed the motivation Well, now what is good practice ? Suppose we take seven letters—
side, or, if you like, the reinforcement side—that success is the great a, h, i, m, n, s, and t are seven good letters to take for various reasons—
generalized reinfo rcer. Compared with that, such local things as the all of which illustrate important theoretical considerations. With these
supposed interest of the child in the narrated doings of some children seven letters you can offer the beginning reader all the short sentences
with their pets don't weigh in the balance in the very least. he needs. All these statements should have to do with a concrete situa­
But there is another side to this keeping away from mistakes in the tion; they should be sentences whose meaning he can see, sentences
early stages of the subject, and it may go up high in the subject. A whose meaning he can act. All the materials you need for this very
mistake is a permanent source of weakness. When enough fatigue or novel activity of learning to read—the optical control, the first discrimi­
enough strain through new problems comes along, the mistake is very nations, all this tremendous accustoming himself to a new type of visual
likely to occur again; you can see that in any systematic observation of attention—can be supplied through seven letters which present minimal
early teaching. opportunities for mistakes. That is the formula. Now, make up some
We prevent mistakes by simplification, and here comes in the tech­ sentences with those letters. You will find there are many that you can
nique of simplification, say, for beginning reading. Twenty-six letters make. We add another letter fairly soon, and then another not too long
may not seem a lot, but try seven instead. We are all aware, at least after that; but we keep for a long time to only half the alphabet. It is
I am acutely aware, of the problem that arises when you meet twenty- far better to begin teaching with a film strip than with a book. With a
six persons for the first time, all together. How long is it going to take film strip you have a public-meeting atmosphere, a screen and a focal
me to stand, as it were, in a corner of the room and instantly see who point, and an enormous release of tension.
is talking with whom in groups of three, four, five, or six, in various I must not linger on this, but the first fifty frames of the first film
parts of the room? If, on the other hand, I had met seven people and strip use only twelve letters, and an elaborately graded presentation
had lived with seven people for several days, and then a stranger had can develop in words and sentences which use only half the alphabet.
been added to a group, there would have been no generation of un­ Which letters should be included? Obviously, you don't want sym­
certain choice points, hazardous speculation, as it were. Mere routine metrical. Any child is taped—I don't know how far back it goes—to
could take care of it. When you throw any letters of the alphabet along see that this and this [turning a pipe in four directions] are the same
with any other letters of the alphabet before a beginning reader, you thing. It is still the same thing, but outline it on a board and it is four
are putting a strain on him that he is going to suffer from for the re­ distinct letters—p, b, q, d. It is just a matter of how you turn it. The
mainder of his life, unless he is a person of extraordinary happy condi­ child couldn't live life unless he saw a knife, say, as a knife, no matter
tioning up to the moment, or in an unusual state at the moment. There which way up it was. It is bad technique to make a sudden transforma­
are some people who can do it, but the great majority of the ordinary tion to script, in which it is all-important whether the u is upside
population cannot. down—or is it the n that is upside down? We penalize the bright
The early history of reading for most minds is a history of frustra­ child by setting a whole set of bogus traps for him in the script we
tion, of great personality strain, all of which is irrelevant to a task begin to teach him. They don't belong to the subject. They just betray
which can be made simple. I speak with some feeling on this because him through his biological smartness.
we do teach people to read when nothing else can teach them. When The other set to watch, besides the complete oblique symmetrical,
you take boys of fifteen or sixteen, who have been hanging about, be­ are the familiar enclosure things. Here are what look to the child like
coming delinquency cases because of the frustrations due to their fail­ three attempts to do the same thing [writing o, c, and e on the black­
ure—the Ishmaelite complex of being the only persons in the society board}. Whether or not you finish them, what does it matter? It is
who hadn't learned to read—put them into another room and, a few easy to spot the probable sources of confusion between letters, and
weeks later, after they have taken the decisive step and have been to design a sequence which will avoid all this unnecessary difficulty. In­
taught to read, you find you have a roomful of new people, of people troduce only one of the mistakable pairs until that is so well established
 Communication Between Men 63
62 Cybernetics
chological phenomena to brain functions (12). Admittedly, science is
that the other position comes with a shock, and then you go on.
not possible without language. Language itself, however, is the final
What we do also—it is all part of the same avoidance of distrac­
product of an unbelievably complex development of the brain. There
tion—is to stabilize the syntax. We use only one verb (is, are, and so
will be nothing but confusion, these investigators believe, if we use
forth) all the way through learning to read. We cut down syntax vari­
everyday language to talk about the very brain which has created this
ations to a minimum and reduce meaning to something the learner can
language. What von Monakow said in effect was that we should not
actually see: the meaning of the sentence, as it were, through the
use ordinary English, German, or Russian in talking about brain func­
sentence, and so on, right through.
tions, but should invent a special language.
Above all, the interest of the exercise in these early stages is in learn­
Savage: But not with the brain.
ing to read and not in any adventures of Jack and Jill and their pets.
Klüver: He argued that all words dealing with psychological phe­
That is really quite important.
nomena have gradually acquired too many meanings and connotations,
Well, that is a practical outcome of these speculations. I am not sure
and that real progress in neuropsychology depends, therefore, on the
that they belong together. Again, I am not sure that the air of belong­
use of neologisms. It is for this reason that he introduced such terms
ing together which I tried to give them is more than a product of lite­
as klisis, ekklisis, protodiakrisis, kakon, and so forth.
rary composition. It is efficient in that sense, bless it. If it is inefficient,
Richards: But how does he define them? If you introduce enough
let's pull it to pieces. Thank you very much.
new technical terms, isn't there a sort of law of economy about the in­
Klüver: You used the example of an animal looking for food to il­
troduction of technical terms? One technical term is excellent business;
lustrate how sets of subordinate-subordinate tapings are involved in a
two technicalities in the same sentence are all my eye can take. There
general and all-inclusive taping, such as the search for food. Perhaps
is the problem of anchoring your new technical substitutes to observa­
one of the most beautiful illustrations of this kind of taping is to be
tions, isn't there? You have to go into cases.
found in the hunting behavior of the peregrine falcon as described by
Klüver: Perhaps von Monakow was of the opinion that desperate
Tinbergen (il). After leaving its perch, the falcon may start looking
situations demand desperate remedies. Historically, he has not been
for potential prey in a territory as large as ten miles in diameter. The
the only one to insist that the use of everyday words in referring to
sight of prey does not immediately elicit the consummatory act; there is
psychological phenomena will forever impede progress in the field of
instead, as "Tinbergen points out, a sequence of more and more specific
psychology and prevent a sharpening of its conceptual tools.
and restricted types of appetitive behavior, down to the final swoop and
Richards: I am very much inclined to experiment with not using con­
the catching of the individual prey. Unfortunately, the recognition of
ventional terms here. I think the introduction of images at this point
the fact that certain tapings are more specific or more general in the
would be absolutely disastrous, although many of the observations we
hierarchy of tapings is of no great help to the psychologist. He still
rely on to some extent are phrased in images. There is a typical anec­
has to cope with the difficult experimental job of determining the con­
dote which turns up—I should be able to give you the reference, but it
stellation of properties or the particular categories in terms of which
doesn't matter—of the hunter who is inexperienced, going out looking
particular tapings are made. It is of interest that work on interspecific
for deer. He never sees a deer. He does everything he can, until one
recognition in birds, especially on the recognition of predators by their
day he goes out with an old hunter who puts his hand on his shoulder
victims, has led to the conclusion that certain predators are recognized
and says, "Now, look, don't you see that deer?" The other says, "No."
on the basis of only two visual characters: namely, outline and move­
The old hunter begins to describe it, and as he describes it, it appears.
ment. When dealing with a swiftly moving enemy, it is apparently bet­
And then the advice is given, very judicious hunting advice: "Don't
ter to make use of very few and very general schemata than to be dead.
go out after anything unless you imagine the animal you are looking
Richards: Well, the ad hoc taping can be quite generic in that case.
for. Imagine it as concretely and as fully as possible."
Klüver: It can be generic ?
I begin to feel doubtful about that. In my own case, I can see that a
Richards: Yes.
visual image might act as an obstacle if I relied on it; but that illu­
Klüver: What you said about the complexities of language which
strates feedforward suggestion.
arise when using language about language reminds me of the com­
Mead: Dr. Richards, in connection with the point you were making
plexities that bothered such outstanding brain researchers, as von Mona­
about the dangers of autarchy, I think that this danger is very striking
kow and Mourgue when they tried to talk about the relation of psy­
64 Cybernetics Communication Between Men 65
at present for anthropology. By the isolation of linguistics from the fore, biology is normative." That doesn't seem to me to be at all fair.
study of the rest of culture—and linguistics is more isolated from" an­ "Wherever we must make a decision, we have, I would say, a normative
thropology than it has been for a very long time—we have lost in a question. "We must decide what it is right to do. And, in particular,
sense our capacity to look at other systematic aspects of human behavior, if the advance of anatomy is our problem, then what are the right
so that the argument in anthropology at present, which you summed up moves to advance anatomy? But that does not in itself imply that the
by picking up Kroeber's statement (4), is this; Yes, of course we know study of anatomy is per se—
that language has grammar; language is systematic and linguistics can Richards: "Well, may I answer that point? So far as the anatomist is
study it. It is something that is in a nice little box. The rest of human looking for a general study and wants to generalize from his observa­
behavior, however, is regarded as subject to some completely nonsys- tions, I am saying very little more, I think, than that part of his tech­
tematic set of principles which make cross-comparisons impossible, and nique in deciding that comes down to: Is it safe? The question is: Is
one comes out with Kroeber's sort of statement as to the uniqueness of it safe to generalize a description of this bone ?
each culture; whereas if one says that language as we know it, and Kubie: There would seem to be a subtle transition and overlapping
other systems of communication between people, other methods, parts of meanings here, beginning with the statement about the "ethical"
of behavior, all of which involve the whole body, are all systematic, be­ element that enters into the acquisition of language. This is different
cause they can be referred to the human organism; then it is possible from the ethical element of a language or of a science itself. The
to make the sort of cross-cultural comparisons that you are asking for process of acquiring the language, or the process by which a student ac­
and to use the uniqueness of each culture only as a point of reference quires a knowledge of chemistry or anatomy or anything else, carries
for particular observations within the culture, so that false equations the ethical values and systems with which he has learned to keep him­
are not made from one to another. self clean, to brush his teeth, to obey his parents, and so on. There is a
Richards: That is the danger, and it is very hard to avoid it, because right and a wrong, in the sense of a right and a wrong answer, of being
until you get an over-all descriptive system which has been sufficiently a good little boy or a bad little boy. This invades the whole learning
criticized, you are almost bound to make false identifications, aren't process inevitably. "Was there not some confusion here?
you? Richards: I think you are right. There is a hierarchy of the uses of
Mead: "Well, typical points would be that you might be able to get a the word "ethical," and I think probably Savage understood me to use
cross-cultural or cross-language language which is sufficient to think one that was fairly high up, whereas I am really talking about what we
about languages that are quite different and to think about them to­ usually call good investigatory practice, the sort of precautions that you
gether; but if you talk about whether you murder your grandmother or do observe, that your evidence is not biased by something you have not
not in one setting or another, without reference to the setting, you end considered. Take the anatomist and his bone: if somebody else can
up with cultural relativity. come along and say, "That is not a typical bone; and you are assuming
Richards: "Would you agree that all subjects are inevitably evaluative? a very great deal and I will show you a typical bone," then the problem
Mead: In the sense that you are saying it, yes. is, I think, about value, but very low down in the statement.
Richards: It does solve a great number of artificial problems—the By the way, I should like to take up a point about the way of teaching
classic warfare between strict neutral science which is not concerned a language. Of course, you are being given the culture; you are being
with value and poetry which is concerned with nothing else, as I see given the ethic. But this is the thing that makes me boil when I see it
it today. happening on the planetary scale, with billions of people. There is the
Savage: I am very much interested, too, in these remarks about nor­ question: How do you spell this wretched word? ["Writing on black­
mative aspects of science. You said several things about it; some of board, idear.^ My Harvard freshmen come from 3 per cent, you know,
them I thought were good, and some of them I don't yet understand. of the most favorably placed people in the country. Many of them
But one of them, at least, did seem to me to be a technical mistake; at write idear. They have made a mistake. They have not assimilated one
least I would so interpret it. You say: "Suppose that the biologist of the curious things about the culture. Of course, they spell it phoneti­
worries about whether this bone in the hand is typical of such bones. cally. It is what they hear—idear.
He must then face the question whether this is a good bone to describe. Kubie: It is how he feels.
"Well, that is a normative question. It is a biological question. There- Richards: Actually, it is not a crime; but bad spelling is put into a
66 Cybernetics Communication Between Men 61

worse category of cultural crimes than inefficient use of words. But I that if you looked at "irnportant" carefully enough, you would find a
am really punishing that boy with disapproval, with social scorn, which good ethical tag on it.
is the whip. What I do is to run through their own compositions and Mead: Well, my understanding, when I answered you that I agreed
throw them on the screen, and the whole two hundred howl. Of course, with your use of the word "normative," was that it not only applied to
they are all like wolves, to tear a bad speller to pieces. But if a boy good practice in the specific sense, but that it also applied to the fact
comes along with one of the most important tools of all Western that you had to consider all of life in the end as a whole and all the
morality, all world morality, and writes "disinterested" where he meant sciences in relation to one another, and that these mere, trivial, though
"uninterested," does anyone wince.? When the culture takes one of worth doing, considerations are part of the total pattern. One gets,
the most important instruments of discrimination we possess—differ­ therefore, the distinction that the difference between medicine and biol­
entiation between disinterested behavior and uninterested behavior— ogy is a difference in level; but they would both be normative in that
and throws it into the trash can, that class won't bat an eyelid. Has sense.
anything that mattered happened.? You can't persuade them. If you Savage: It seems to me tremendously important not to lose sight of
could find out how to teach the beginnings of the great subjects, the something that I think our ancestors gained in formulating the concept
problems that we are vexed by in the upper hierarchies would be of pure science. Though we have heard about it all our lives, it may be
vanishing. difficult to express what is meant by this notion, but the minimum is
Savage: It seems to me that precisely because words like "normative" this: One should not confine one's attention in science to what is super­
and "ethical" can be, and are, used at different levels, there is some ficially useful. The judgment of the scientist, the taste, and even the in­
sense in referring to some disciplines as normative and others as non- stinct of the scientist should be allowed a good deal of play and a good
normative. But it is commonly said that medicine is ethical in a sense deal of time in which to express themselves.
in which biology is not, and I think that is correct. Such a statement The advocacy of pure science means, for example, that a bacteriolo­
should not be regarded as nonsense and contemptuously thrown away gist should be free to study those organisms which seem to him bio­
just because we find some normative activity among biologists. logically interesting, without special reference to those which sour milk
Hutchinson: I think one needs at least three distinctions. There is or those which cause mumps.
the normativeness of medicine and the normativeness of good practice Richards: I am going to agree with that, of course, but I am also
in any investigation; and then there is a third one that is, it seems to going to try to appeal to this over-all grand analogy here. There is
me, escaping us a little bit, and that is the normativeness of whether something that corresponds to sovereignty in a separate study. It alone
it is worth while to make statements at all, however well the investiga­ can settle certain things about what it should do. It is its own authority;
tion has been carried out. If that is lacking, then the whole subject is but, on the other hand, much that it does—and your bacteriology is a
condemned. In mathematics, it is called trivial. I don't know what you good case—is the concern also of other sovereign studies. And here
call it in biology, but I think you generally say that it is the kind of comes in your diplomatic service, not to dictate to people inside bacteri­
thing that a man in a museum would do, or that the work is "mere." ology what studies they should pursue, but to represent public health.
Richards: In learning, "botanizing" is usually the description. Klüver: You referred to the fact that the same object in four dif­
Mead: "Blind alleys," too. ferent positions is still the same object and that a child, therefore, may
Hutchinson: Yes. And then there is the very interesting phenome­ have difficulties in learning that d, b, q, and p represent four distinct
non that most of the "mere" aspects of biology of twenty years ago now letters. Psychological investigations have shown that reversals, rota­
are exceedingly important. tions, and spatial displacements are characteristic of the perceptual
Bigelow: To some extent that is true in mathematics also. world of certain children or of certain age groups (13) (14). In such
Savage: What—the trivial parts of mathematics have become im­ children the process of learning to read may involve special difficulties.
portant ? Richards: I have looked into' it a little bit, and I don't think that
Bigelow: Well, the basis of axiomatics, for example. has any connection with the symmetricality of certain letters. But, on
Richards: This is uncovering what is behind this word "important." the other hand, I think this queer reversal, for instance, this tendency to
We throw the word "important" around with great ease because we reverse reading and writing, might possibly have some connection with
haven't defined it as ethical. Now, my little talk was a matter of saying the general stage of intelligibility of the form that is being studied in
68 Cybernetics Communication Between Men 69
the early stages of reading. What one wants to achieve, if one can, is Richards: I was struck this morning by a remark about the way in
a sequence of steps into reading, or anything else you like, which is which function and structure can change places as the point, temporari­
maximally intelligible. ly, of maximum interest and significance. I think there is something
Klüver: Reversals of letters, if they should occur, are undoubtedly very similar going on with process and structure in this language.
not isolated phenomena but are correlated with other perceptual de­ There are aspects of language which are, for people like that, purpose-
velopments. serving, rather than perhaps promoting, aspects which have been neg­
Richards: I agree. lected and should come in. I think there is a very long way to go.
von Bonin: I should like to bring up something which troubles me. But as to your remark on the general breakdown of communication
You talked about Kroeber and his anthropological ideas about studying inside this culture, I agree we've got to take it terribly seriously. I am
cultures, and the idea or the conception of studying each culture from raw with irritations caused by my Harvard representation of the in­
its own point of view as being a halfway house. "When you talked tellectual ""cream"—and I couldn't give you examples that would fright­
about literary criticism afterwards, I think you mentioned that the two en you as much as I have been frightened. I am terrified. I think it may
things that mattered were the thing the author wanted to express and be that they have been selected by an improper selection process, and, if
how efficiently he did it. Isn't that exactly the sort of thing that so, that is very serious.
Kroeber (4) tries to do in culture ? I am not quite clear about that. Bigelow: Is there any evidence that it is worse now than it has ever
Richards: Yes, you are right. I think it is quite probable that litera­ been ?
ry theory is at a very early staging place, and it probably does bring up Richards: It is all gossip and impressions. I don't know whether it
these characteristic complementarity situations. I don't feel very strong­ is sound or not. I should like to know. Of course, we have come to a
ly that there is one there at present. I think you nearly always have two short way of educating, as everybody knows, with so many more stu­
grounds for criticism, two prime questions which you ask in judging dents and, on the whole, education in larger classrooms, and we are
anything; What do you want to doand. Can you do itI don't think trying to teach them more about more things. All this background
that is quite the same as the Kroeber situation. I want to go back to mji philosophy of mine about the over-all united world view and about
remark that that was a defensive utterance of Kroeber's. He wanted to getting a headache every morning concerning every grief and trouble
damn the people who were invading a given culture with presupposi­ there are on the planet—we do that intellectually, too. I don't think
tions taken from another culture and describing it with their tools. it really helps the young initiate in the culture to learn what will grow
Frank: I hope this question is relevant. Can we say that we are be­ best in him.
coming increasingly concerned with problems of communication in re­ Mead: But don't you think also that what is happening in a period
cent years largely because the traditional usages, the long-accepted where we change as rapidly as this is that as teachers, instead of our
meanings of words have begun to break down.^ As we try to develop getting better, in a sense we get worse, because we can't learn by ex­
new ways of thinking and new ways of communicating what those new perience from teaching generation after generation of approximately
ways of thinking are, we are faced with some of these difficulties. the same pupils because our pupils are changing all the time. Now, I
Would it give us any better perspective on the very problem you have go away and I come back to teaching, sometimes with intervals of four
been putting before us if we thought of ourselves as engaged, here in or five years. The thing I am progressively struck by is that students
these meetings, in trying to create a new climate of opinion, a new way can understand now that which they couldn't understand ten years ago.
of thinking where we are moving from some,of the old static, analytic, Richards: A hopeful thought on which to end.
linear ideas to a way of thinking in terms of context and dynamic Hutchinson: Well, there was just one parting thought: we are ac­
processes? We are trying to establish the word "process," for example, tually, in all this discussion of language, throwing away a great deal
with the assumption that the same process may produce different prod­ of what we know. Lawrence Frank was talking about the substitution
ucts, depending upon where and how it works, as an attempt to get of ideas of process for static subjects. Every time we draw a graph
away from the static nouns and to put our generalities and abstrac­ with a time axis, that is what we are doing. We have been doing it
tions in terms of verbs. It seems to me perhaps that might be one way since the beginning of the eighteenth century. The only thing that we
in which we are trying to move out of an older way, and do what the don't do is to recognize in this group that that is what we have been
Hopis have done for a long time; namely, to use only verbs for ab­ doing half the morning. It is all in front of us, but it is compartmen-
stractions.
70 Cybernetics Communication Between Men 71
talized; and too many people in the literary disciplines regard it as a Pitts: I think that what is important in this is that what can easily
crude and scientific and inhuman thing to do. But it is perfectly stand­ be a matter of convenience can become a matter of principle; that is,
ard and well-integrated practice, and merely has to be spread out a it is perfectly possible for us in English to express every shade of in­
little bit over a few more Harvard and Yale freshmen and other people. decision or truth in a statement by prefacing it with words like, "It is
Mead: And you have to recognize that moving your eye is something hardly certain that . . ." We commonly dô that. I think the extreme
that is relevant. case of that occurs in mathematics. As Russell and Whitehead have
von Bonin: I think Heraclitus first said, "Everything flows." shown, it is possible for every mathematical computation to be ex­
Rosenblueth: One of the statements you made, Professor Richards, pressed in ordinary logic, in ordinary words containing no special
which impressed me very much and which I want to ask about, is this; mathematical notion such as number, and so forth. But if you do so,
You said that the first trouble you found, or the important one, was the simplest arithmetical computation becomes a very complex com­
what language to use to talk about language. That is an extremely im­ putation spread out over not less than thirty pages of text, and re­
portant question if we wish to talk about anything. On the other hand, sembling legal arguments about the Constitution which nobody can
when you told us why you had that trouble, you said, "Would you use possibly hope to follow. If mathematics had not been developed sepa­
mentalistic or behavioristic terms?" Is that really very important, as rately, we can be quite sure that although everybody, in principle, who
long as somebody makes the statement, "I am using these terms with has a training or intuition for syllogistic logic could make that infer­
this general connotation, with this general meaning?" ence, it would actually never occur to them under any circumstances
Richards: Yes. I think you can get out of it. I think, that twenty or to do that. If some tribe of American Indians has a language which
thirty years ago that was a barrier to a lot of people's honest thinking. causes them to express universal appearances by inflecting the verb in­
I think now, somehow, they have got accustomed to the situation and stead of by qualifications attached to the noun, it is perfectly true that
they see they can put their inverted commas, their special meanings, on we can say in English everything they could say. We can translate their
it, and continue. I don't think that would be an important point. There descriptions into English. But if their natural and brief descriptions
are others for which that would only serve as a model, or as an ex­ tend to become intolerable prolixity in English, the chances are we will
ample. I am not sure I should try to elaborate very much, but in places not see that aspect of the situation which to them is the simplest be­
where you do see a rather crucial choice as to conception, it turns out cause it is the most simply expressed.
to be a choice of language that you might employ. Richards: I entirely agree. It is almost a doctrine for the people I
Rosenblueth: The language which you employ would, of course, de­ am thinking of, the linguistic scientists, that anything can be covered
pend on your desires, on what you think is the right way to criticize in any language. It may not be so.
something, the language you think is the most useful one or the one Pitts: Perhaps it can, but not in any natural way.
that fits the purposes which you are going to follow. It will always be Richards: They may be simplifying their notion of covering.
nothing but a tool, an important tool in general. The best language, Bigelow: Do we believe it can ?
however, is nothing but a poor tool. But once one realizes that, I don't Richards: If you are willing to go to enough lengths to say it.
think there is any special trouble that arises. We just have to try to re­ Bigelow: Perhaps it can be done by the people who translate poetry.
fine it. Rosenblueth: I can give you an example in Spanish; cursi. I would
Richards: But there is a point where its defection may sometimes be have to write a book to tell you what that means in English.
its merit. That is constantly the line of progress in this kind of specula­ Savage: What does it mean?
tion; the breakdown of the tool. That really is the important part of Rosenblueth: It would take a book to tell you.
the observation. We are not trained to note that. We blame the lan­ Pitts: But in any book in which you vise that word, at the expense of
guage, and we should, instead, focus on the language. enough prolixity, you could translate it.
If I could have a last word, it would be this; if we could only take Rosenblueth: But not the feeling. You would not know what I am
account of our constant, habitual skill with language and translate that talking about. And if I gave you a good example, it wouldn't even
into a general theoretical understanding, we would be very near where tell you.
we wanted to be. We constantly know things in practice about lan­ Mead: I think you have a different point, though, the point that
guage which we are blind to, completely, intellectually. Bateson made in his paper on pidgin English. You can say anything in
72 Cybernetics Communication Between Men 73

pidgin English (15). I have translated part of Alice in Wonderland tem of abstractions in the right order, as we are educating children, and
into it. It can be done. But the point is that you don't want to. There take Dr. Richards's point that we must not let them make a mistake,
is a way to say it formally, yes. You can translate Sapir's old state­ because the mistake will dog their footsteps forever. The best illustra­
ment that Kant could be written in Eskimo in such a way that it is tion I know is still that old study of bilingualism in Welsh and English
formally true, probably, but it leaves out of account the fact that you published in 1923 (18), in which bilingualism showed up as a defect
won't ever want to. Conversely, of course, after we have become ac­ worse at the college level than at the kindergarten and elementary level.
quainted with these concepts, such as the one you are mentioning or We could begin to use all these concepts of selectivity and types of fil­
those I find in each language I have learned, I have added a few things ing and hierarchies of generalizations. What you were giving us, as I
that I need. We can use our knowledge of other languages to shake understand it, was a model at the educational level.
us loose. That really is the point: not to take over the Hopi ways of Richards: Yes. It is work that has been going on very laboriously
saying things, but to shake our own system loose, so that we can build for a long time, and I am quite ready to say that we have evidence now,
new ones. Won't you agree to that? in teaching reading and English—I am not sure yet about French
Richards: Yes, I think that is true. and Spanish, but English—that the order of the operations, the order
Teuber: There was a German humanist, for whom I have the great­ of presentation of sentence-situation units, is decisive.
est respect, Stenzel, who wrote page after page on single Greek nouns, Mead: But I should like to add to that, now that we have a little bit
explaining that they could not be translated, and paraphrasing them in of evidence—it is only case-study evidence—that infants who have been
a great many ways, trying to show that logos, after all, expresses what reared on what was erroneously called ""self-demand" get a generalized
logos expresses, or that arete, meant, well, arete, and that nothing you capacity to reduce confusion.
could say in German would explain it (16, 17). He went all around Richards: That is quite right.
the bush in trying to give you a general evocative meaning for arete or Mead: So you get this sort of analogy, for instance, that if you try
logos, as he felt it should be understood, in the context of the Athenian to teach the child who has learned, who has been fed in relation to its
democracy of such and such a year, even part of a year, when it was own rhythms, too many transportation systems at once—for instance,
used by such and such an author. When you looked at some of the three bus routes—he will say: ""Stop. If I learn any more bus routes
translations that he gave in illustrating his points, the majority of the now, I will be confused."
words were Greek words, interspersed with a few German prepositions, Richards: They are right.
and then he went on to show that arete differed from dikaiosyne, and Mead: That is at six and seven years of age; so it is possible that we
there was more Greek to it than German. can begin this question of order virtually at birth, and then follow it
Mead: Before we end this discussion, I should like to go back to one along.
other aspect of what Dr. Richards said. It seems to me that he was Brosin: In your functions as critic, and more so as an educator, do
also dealing with this question of teaching and learning in the very you have systems whereby a person, an interrogator, and his vis-à-vis
young, and that what we know about all sorts of things can be trans­ can establish communication over barriers with more economy? This
lated into pedagogy in such a way that we will be able to educate in­ room can be used as an example. Can I, or other clinicians, when talk­
dividuals whose communication potential will be so much higher. It ing to a patient, use these devices which set a theater for the beginning
goes back to a question I asked Dr. Bigelow about two years ago: If of our exchange and serve as a basis for further exploration ? It is use­
he could re-educate himself with all he knows now, would he do it dif­ ful to remember Sullivan's phrase that the patient is a stranger (19).
ferently ? As I remember, he said he would settle for twice as large a Richards: Yes, exactly. I can give you two instances which will fit
brain. That was a conversation over coffee, and it may not be fair to into what you have asked. We are almost certain now that it is a more
quote it. But the description that Von Neumann gave of memory at the efficient, satisfactory procedure to give a beginning class of children,
time he talked about it, and the possible cross-referencing betweén filing learning to read or learning a second language, every sentence-in-a-
systems and senescence, for instance, and senescent memory, raises the situation on a screen in the artificial screen space, not in a book. And
problem in connection with what Dr. Richards is saying here. (I think it is certainly more efficient to give them the sentence, not as spoken by
you still all hold to the hypothesis, don't you, that we use only about the teacher, with all attendant complications, but from a recording.
a tenth of our brain in some way or other?) We could build our sys- They will do things to a recording in the way of parodying it, guying
 Communication Between Men 75
74 Cyhernetícs
it, that they would never dream of doing to a teacher's face, and they Brosin: You must jSroject this universe of expectations. I would re­
get a beautiful pronunciation in that way. That is a side product. The quest Dr. Richards to review his statement that criticism is almost lost,
actual point is that you have a release of exploratory action in the child and with it the writing at the level of, or with, the certainties of a
or the learner which is constantly not present even with a very good Goethe or the seventeenth- or eighteenth-century critics, or the Coler­
teacher. It is very interesting. idge which you quoted. I can understand your position in comparison
Brosin: May I follow that.? How do you establish communication with the mid-twentieth-century author, but, actually, each of us is a
with, say, an English-speaking person on a train, about subjects upon critic every day in every piece of communication.
which neither of you has knowledge of each other's theaters of action ? Richards: Oh, surely.
What are the formal steps ? Brosin: I am asking the impossible, of course, but that is the purpose
Richards: You mean, how do I do it, if I meet an English-speaking of this meeting. In your experience, what are the formal barriers and
stranger ? how do you operate, either from past experience or by very careful
Brosin: Yes. I do not mean in any trivial conventional sense now; pedagogic logic, to establish a theater, a frame of reference for more
but rather, what are the serious operations that you employ to get past meaningful communication in terms not only of the superficial values
these barriers and become acquainted, in the sense of exchanging maxi­ but of the nuances, the deeper appeals to past experience ?
mum information in a short time.? Richards: I can deal with just a little of that in a moment. We had,
Richards: As I understand it, I have no problem in this country, but I though, when I said criticism was almost lost, exactly the same situation
have a horrible problem in England. It is perfectly easy to talk to most that arose with Savage and me over "normative" and the bone and
people here, but it is very hard to get talking to anyone in England. "ethics." You see, you can send criticism to astronomic heights, and it
This is a technical problem of communication. becomes a sort of peculiar privilege of the literary critic to be concerned
von Bonin: The same problem, it seems to me, arises in literary criti­ with criticism. Most people say, "Oh, no, not us!" But you are right.
cism, on a somewhat different level. Somehow or other, you have to Everybody is being critical, in the most fundamental sense, with every
know or you have to feel or establish what the goal of the author was. utterance that he makes himself or takes from anyone else. Now, what
Of course, there are his printed words, but one has to read between the can break down the barriers ? The kind of thing I hope I have illus­
lines. The same thing arises for a historian when he wants to evaluate trated by comparing the ethics situation with this criticism situation.
whether Caesar did his job well or not, and things of that sort. The All the main troubles of language are endlessly recurring. They have
same thing arises in trivial and less trivial situations, but that is the certain common patterns. We continue with ad hoc analyses and im­
problem. provisations, to deal with misunderstandings as they come up. But
Richards: I think that is the heart of it. That was easy, relatively, where is there the generalized theory of misunderstanding, or the gen­
when you had a stable culture in which, when you encountered a new eralized technique for developing skill in knowing what are the prob­
text, you could assume that the people who were meeting it would have able meaning variations around a given utterance, taking the utterance
had a common literary experience leading them to it, even if it were just as it comes.? The hope of doing it would be through a structuring
only the Bible and Homer, and so on. Now, when you can't have any of the field of misunderstanding. If you could do that, you could begin
confident expectation that anything you mention will have been met to teach people to understand one another better. That is my hope.
before, criticism is almost lost. Rioch: I would be interested in another aspect of language, taking
Mead: Don't you mean that before you can talk to another person, a language as a central nervous system function. It seems to me to have
chance-met and totally unknown persoñ who, however, does speak your an entirely personal significance, quite apart from any social signifi­
language—those are your conditions, aren't they— cance, though the personal significance is almost certainly developed by
Brosin: Yes. the social function. But probably there are equivalent activities in forms
Mead: —that you have to work out his probable image and under­ or species without language. It is pretty clear that a rat confronted by
standing of who you are.? two doors sets up a hypothesis, orienting himself; he goes to one door
Brosin: Yes, indeed. or the other. The difference between that and the personal function of
Mead: And his willingness to talk to that image. language is that in some way we can remember something about the
Bigelow: As you did beforehand.? situation which calls up a word. That word then organizes the whole;
76 Cybernetics Communication Between Men 77
that is, the function of the word, or the activity of the central nervous content of the words you use is timeless.
system which is this word, then organizes the whole of the central nerv­ Brosin: The media for the expression of my meanings is a process
ous system to be oriented to a situation according to that word. It is an occurring both in time and in space.
entirely personal thing, and in that sense not— Mead: To that extent it is analogical, isn't it? The communication
Richards: I have a doubt about ""entirely personal." It has been that goes on between people that involves their whole persons, the
learned in interpersonal relations. stresses in their shoulders and the lifting of their eyebrows, and all the
Rioch: Yes, but now you have to consider learning instead of taking rest of it, doesn't have this digital character.
the activity as it occurs. Rioch: No. When you talk about communication between people,
Richards: I don't think they are really separable, are they ? That is then you have to divide things up much more than that. Let us consider
my point. language and vocalization. If we take a series of categories of either
Rioch: I think we have to separate them if we are going to think complexities, obscurities, or elaborateness of vocalization, it doesn't
clearly about them, because we have to know what point we have ar­ matter which category we use, we will at one end of the series have that
rived at in learning. situation which is probably the simplest, that in which there is a double
Richards: I am worried about the central nervous system. I want the integration of the organism with the environment, probably mediated
rest, any nervous systems you can provide, everything. I think it is by brain-stem mechanisms. In this case the vocalization is a part of the
nearly everything that comes into language. total activity of the organism as a unit.
Rioch: If we make it that broad, then it is very difficult to differenti­ Now, when one proceeds through a series of categories of complexity
ate anything from it, whereas I think we can select factors out of it, of vocalization, one arrives at the other extreme, in which, regardless of
especially when we see the effects in cases in which there are certain the tone of voice, the style of handwriting, any gesture, or anything
deficiencies. else that goes with it, the entire communication is contained in the con­
Richards: I want very much to hear how animals do communicate, tent of the words as defined by Webster. You have everything in be­
because that would throw a great deal of light, I think, upon how tween. But whenever a word is used, then you can separate the sig­
we do. nificance of that word in terms of its content, as something that stops
Rioch: We can tell a lot more about human beings and how they time.
communicate than about animals. A very curious thing about language McCulloch: Once for all.?
is that there is a continuous communication which is going on through Rioch: Once for all; and that is its main function, which is some­
time. When a word is used with respect to that communication, then thing about language which one notices when one deals with people
time is stopped and we no longer deal with time, because the word is who have difficulties with language. I think you notice it with children.
now either so or not so; and if it is so, it is always so, and if it is not so, You certainly notice it with patients. You can see a patient come from
it is never so. By the use of words we have introduced a digital system a situation in which the content of the word is essentially of no sig­
instead of — I don't know whether you can call it an analogizing nificance—the tone of voice is of tremendous significance—to a situa­
system — a system which is a continuous communication through time, tion in which the content of the word has entirely personal significance
with continuity of interaction. We put in a word which now destroys in a very limited situation, indicating what he is going to do in terms
time, and that is the only way in which we can deal with past or future of some very limited thing; to what we speak of as normal where the
time. That, I think, is one of the major functions of language in com­ content of the word is very important, but in the sense in which the
munication—to destroy time—and it is the thing that at the present language is socially learned. One of the most interesting things we ran
time has made language such a dangerous thing, because it has de­ into recently was a lobotomized patient who apparently did not have a
stroyed time in certain directions instead of in other directions, and concept of what was going to happen in continuous time; that is, there
people no longer trust what other people say. was no feeling that if he hit somebody the other person wouldn't like it.
Brosin: Has it destroyed time ? Meaning is independent of time and Partly as an experiment, we decided to give him a language formula as
space. While I am talking, that is a process in time. The meanings that a time-binding tool. We tried saying to him, ""If you hit somebody,
I have reference to have nothing to do with either time or space. they won't like it." Within two weeks, the hitting of people stopped.
Rioch: Your tone of voice is something proceeding in time, but the I don't know if the formula really had anything to do with the change
78 Cybernetics Communication Between Men 79
in the patient, or whether other implications or factors were responsible. Savage: He may say, "You're not sorry at all."
Be that as it may, the time-binding quality, either getting rid of time or Rioch: It still limits the thing that you can say. Now, this curious
making time permanent — it is the same thing — is very curious in function which is carried to, we might say, a pathological extreme in
language. the obsessional neurotic state is, I think, one of the chief feedback
Richards: May I put a note in here.? It is just a footnote to this defense mechanisms to prevent complete breakdown and anxiety. But
asterisk business. We are hearing one another make a series of what it is a function of language in which language is not being used in
appear to be assertions in the full indicative, yet I think if you asked terms of its content or in terms of its intonation. I think in your teach­
people you would find they were really talking in another mood alto­ ing experiments you are getting a very important use of this function.
gether. It is customary to talk in the indicative, but might it not be that The child gets the adequate response from the other person, which is
what you were saying was only the equivalent of an optative.? How predicted by the child when he reacts to the symbols properly. This may
would it be to conceive that ... ? There is all this going behind both have nothing to do with the child and the symbols, nothing to do with
of your remarks. Dr. Rioch, when you said "destroys time," I thought the child's orientation to the content of the symbols. You point that
I detected around the table a reaction, the sort of reaction that occurs out when you say that Jack and Jill are not important. It is very much
when people use a phrase which is characteristic of poetry in a situation concerned with being able to do something that will get the personal
which causes them to expect prose. I know I had to adjust myself for response that is predicted.
a moment. Richards: Oh, yes, that is all-important.
The question came up the other day whether there could be a Com­ Rioch: And that may be in a different direction from the direction of
plementarity Situation, "mentalism" and "behaviorism." It is much the content.
more than that; it is that we have modes of expression which we don't Richards: Oh, quite. I quite agree with that.
know how to replace but which do work, as it were, takeh from the Rosenblueth: This states that there are languages other than the
integral action of the mind. spoken or written word, and that language can be used for purposes
Rioch: But sometimes they don't work. other than the transmission of information. I would certainly be in
Richards: This one did work—in time—but it took several people in complete accord with what you say.
the room, I think, some time to tolerate this "destroying time." Hutchinson: If one takes that point of view too seriously, any repeti­
Bigelow: I still can't tolerate it! tive learned movement becomes a language. One learns a repetitive
Rioch: There is another aspect to vocal language which is quite enter­ movement of the legs because it is the same form as that used on a pre­
taining in terms of the information conveyed. Harry Stack Sullivan vious occasion. If there is going to be a personal element that does not
illustrated it as follows: "Good night, Mrs. Thomas; it has been a per­ involve social communication, then all repetitive actions throughout the
fectly foul party this evening, and you have been a horrible hostess." whole universe would have to be regarded as linguistic.
If you put the right intonation on it, the words don't matter and you Savage: That is a tour de force, isn't it ?
can actually get away with it. McCulloch: I think we had better get on. There are two more people
Savage: But can you? who have asked for the floor. One is MacKay; but just before he speaks
Rioch: There is one other aspect to this problem of language I should I should like to say that he was pointing out to me the other evening
like to mention; that is, the magic of communication that has nothing the importance of—
to do with the content. Going through the right formula does two Bigelow: Who is "he"?
things: It keeps the situation structuralized; but it also does another McCulloch: Donald MacKay—the importance of maintaining disci­
very important thing in that it limits the range over which communica­ pline or order, or whatever you want to call it, in the British Parlia­
tion may occur. You bump into somebody and say, "Oh, I'm so sorry." ment; that no matter how foul the remark or how personal the man
You're not sorry. It is not something you can be sorry about. You may may want to be in his attack on another, he may only address the chair.
feel chagrined, you may regret; there are several things you may feel, The very indirection makes the speech tolerable. Will you come in now,
but you are not sorry. But you say, "Oh, I'm so sorry." It keeps the on the main subject?
situation structuralized and limits the man as to what he can answer MacKay: I really wanted to get back to the question of "language
you. about language," and the way it ties up with what has just been said.
80 Cybernetics Communication Between Men 81
because I think there is an important distinction between a new lan­
guage and a shorthand. We can devise a shorthand for talking about
language, in which we make abbreviated noises that are really equiva­
lent to strings of statements in the ordinary language. On the other
hand, a new language must, I suggest, have a different set of referents
or must include at least some referents different from the other.
Now, if one thinks of the function of language, of communication,
as the instruction of the hearer to replicate in his brain or mind—put
it as you like—a representation of something in the mind of the com­
municator, if you think of any linguistic body of information as an
instruction, the intended response is essentially a selective operation
from the junk box of component parts ready to hand in the mind of
the hearer. Therefore, if you are dealing with someone from a different
background, you may require a complete book in order to represent a
word, as in Rosenblueth's example, simply because there does not exist
in the ensemble of component parts from which you are instructing him
to make a selection any one elementary thing, or a compresence to use
the whole spectrum of a Fourier series to describe an impulse if you are
confined linguistically to the logical space in which only frequency can with the language itself representing something else, caused the para­
be defined. I shall be talking more about this sort of thing tomorrow, dox. But I should like for a moment to put a picture on the board
so perhaps I ought to say no more now. But the question of formulat­ which I find helps me a little. Suppose you tried to represent the
ing a language about language does bring up this distinction, and in language functions, perhaps, like this (Figure 7). I have a suspicion
fact it seems that the most we can hope to do in this case is to formulate that you couldn't do with less than six, for the purpose of this after­
a shorthand. noon's discussion. For other purposes, you could do with a very much
Now, that doesn't get around the other difficulty that I wanted to ask simpler scheme. I think there has been a great confusion in the field
Dr. Richards about; namely, whether he has considered what happens through trying to bring in a ""reference function of language" and
when a logical sequence itself becomes the subject of the logical se­ ""emotive language" and ""promotional," and so on. But I want to
quence. In a deductive machine, for example, that would correspond, suggest that [l] you have what is essentially a pointing function. You
or could correspond, to an unstable situation leading to oscillation or pick out, you select, something you are going to talk about. This func­
blocking or something silly like that. In some logical disciplines it tion is normally not distinguished. [2] You have characterizing. You
emerges in things like Russell's paradox; and Popper (20) has raised characterize what you pick out in some vague or precise fashion. [3] A
a related question, namely, whether in principle a thinking machine function, which can be dropped, if you wish, is realizing. Your mode
could predict its own state. He shows there, I think in a watertight of selection or your mode of characterization plus something else invites
way, that it is in principle impossible for a calculating machine, given you to realize what is presented to you by the other person more or less
all the classical data, to predict its own future state; and hence he de­ vividly. youhiVQ appraising. [5] Here you have
duces that determinism, of the sort which asserts that a computing I suggest they all normally go together. [6] No. 6, which is left over,
machine of sufficient size could completely predict the future, is in is an organizing activity. This is the supreme praesidium, as it were;
principle untenable. this is the ultimate authority; this is the United Nations. It is still very
This seems to me to tie up very closely with what Dr. Richards was unsuccessful. But it endeavors to keep these various simultaneous com­
saying; but this is a case where you have íeeáback in language instead ponents of the whole language purpose from interfering with one an­
of ieedfortvard. I wondered if he had anything to say about it. other. Interference comes to a height in a battle between the subjects,
Richards: It is the crux, of course. I tried to do a little with that, but it is actually, I should say, stirring, as a potential self-frustration,
pointing out that the attempt to use language to represent language. inside anyone who says something to himself at any time.
 Communication Between Men 83
82 Cybernetics
MacKay: A simple way of saying it would be that within the rules
Your point was; What arises if you put in a proposition about this
function of organizing itself? You get opportunities for all the head­ of a game you can never find the justification for playing it.
Pitts: Whenever you have a language that is rich enough to talk
aches that Russell grappled with. I am not sufficiently interested, really,
about language, you always have an incomplete language.
in struggling with the precise processes that people have used symbolic­
Savage: I hope Walter's speech can be kept out of The New Yorker.
ally to get out of that fix. I only know you must be able to get out of it
Pitts: Well, it is always a one-way sequence. In the sense that if
somehow. Somebody ingenious enough will find a way in which you
language B contains all the logical apparatus necessary for describing
don't get anomalies. But if one could remember the complexity of every
the structure of a language A, then you don't need language A. You
utterance—though some of these functions sometimes drop to very low
have all the mechanics at your disposal in language B, so the language
values—it would help.
is necessarily more inclusive. It is a never-eriding series.
Vitts: The formal deductive systems which are capable of discussing
Bigelow: I should like to ask some questions essentially on data from
themselves in the sense of describing their own structure, and the deduc­
Professor Richards. It struck me as very interesting — the process of
tions that can be made from them, have, of course, been extensively
simplification that you have been going through in teaching language—
studied by such people as Godei. The conclusion they come to is not
and the following question occurs to me; Suppose that the students to
that there is in any sense an irretrievable paradox in having a logical
whom you teach the language develop an ability to read rudimentary
language which speaks of itself, but that whenever a logical language
sentences in which certain letters have been completely excluded. The
is rich enough to do so, there is a fundamental incompleteness in that
question would be whether or not these students had a residual tend­
system; that is, the possibility of describing a language in the language
ency to delete or to eliminate certain characters permanently in their
itself always means that the language itself is incomplete.
habits of reading and writing—
Richards: Would that have anything to do with any hierarchy of
Richards: No.
tapings ? Supreme tapings, you see, would be too much.
Bigelow: I am not quite through. One of the reasons one might ask
Pitts: It is hard to say. It always means that there is a more inclusive
this question is that it is very well known that language is highly re­
one outside.
dundant, especially the English language, and one might very well
Von Foerster: We can't escape Godei's Theorem.
suspect that a person who is taught systematically this way, and perhaps
Pitts: His theorem depends upon a method which enables you to
advanced to a considerable level by such instruction or such techniques,
construct a logical method within that system, using certain things, such
would actually learn to read most of those letters on a page by a process
as numbers, to describe the symbols in that system; it enables you to
of scanning in which he would never know whether or not the thing
formulate the logic of that system in arithmetic relation to numbers.
was complete.
Then you can do this; You can construct a sentence and have it assert
Richards: Well, he does, and we have to do something about it.
that it is not provable, or, rather, that is what it appears to do. What it
Bigelow: He does learn to read in that fashion?
really does, formally, is to assert that the number of that particular
Richards: He does learn without any knowledge of what letters the
sentence, when you calculate it, is the number of a nonprovable sen­
words are composed of.
tence. As you can easily see, that must be a sentence which is not
Savage: No, you are not in mesh.
decidable by the means of logic at the disposal of that system, because
Richards: Oh, I'm sorry. I missed it.
if it were provable it would not be true; so the assertion is not true. If
Mead: As I understand it, your question is whether learning this
you could prove the contradictory of it, well, then, it would be true, and
smaller set first out of the total alphabet will not establish a predispo­
then you would have arrived at a proof of it. But, of course, it asserts
sition to see only those letters and read, ignoring letters that are ac­
that it is not provable, so what follows is that that particular sentence
quired later.
which asserts its own unprovability is indeed undecidable either way,
Bigelow: Or a preference of some sort.
by the means at the disposal of the system. You can enlarge a system by
Richards: We have not noticed a trace of it, and the answer as to
simply adding that as an axiom; but then, in the new system, you could
why it shouldn't appear is that no particular attention is paid to the
construct the sentence which asserted that it was not provable in the
letters in early stages. They are only components, as it were, of phrases.
uses, even with the additional axiom, and that would become an addi­
Bigelow: But you are selecting them in a particular way.
tional undecidable sentence. You could go on enlarging it, but you
Richards: You have to work out what letters you are going to use.
would never reach an end.
84 Cyhernetícs Communication Between Men 85
You don't worry the children about that. As far as they know, they are one of a species of which you have been taught to examine the members
just using ordinary sentences. more closely, but the question is whether one might conjecture a redis­
Bigelow: But what happens in the case of a language in which indi­ tribution in the process of interpreting a printed page which would re­
vidual letters completely murder the meaning of the sentence.? sult from this ordering of the information. For example, what Pro­
Richards: Well, they don't know that. Beginning readers are only fessor Richards obviously does is to choose to eliminate right, left, up,
too ready to take it. down in the order of a given category. He does not choose to eliminate,
Bigelow: But suppose you are teaching Italian, where the last letter , for example, the ordering of A before B, because, obviously, the child
means singular or plural; don't you find some difficulty, if you start out must learn "no" before "on." "No" is different from "on." Then there
by choosing letters, later on in focusing attention on the individual let­ is a definite preference to the components of language which are being
ters you add, and find there is a hierarchy of preference and that certain fed in here. This would, presumably, have a certain significance (a) if
ones tend to be excluded ? it were true that this is the optimum component ordering; it would have
Richards: Well, except this, that each new letter that comes into the some significance in determining what are the essential characters of a
system comes in as a welcome entertainer, as a new problem, a new language that one must learn in order to understand it fastest and best;
diversion. All through this, we put the letters, whatever they are, in a but (b) one could explore later on and see whether vestigial traces do
sort of cellar box to the film strip frames, so at first there will be a, hi, exist in later performance. If the ordering does affect the preference
mn, St, with intervals left for the other letters. Then suddenly in comes for any appreciable length of time, then one ought to examine very
d, and, soon after, e, in the box next to it. After a bit the children are carefully how it is done, in order not to change the structure and tech­
fascinated by the great question, "When are we going to get another nique of reading.
letter?" Well, we say: "You don't get another letter until you can read Gerard: Concretely, you would expect, having started with these
what we are giving you. When you can read what we are giving you, original letters and added the g much later, that the child would have
we will give you another letter in the picture." Where does it come? some difficulty in distinguishing "sin" from "sign."
It comes in one of these spaces. Bigelow: I don't know, but in reading a page rapidly, this might be
Teuber: I wonder if this would help Bigelow: If you are not very a vestigial process that would crop out.
familiar with Chinese faces, you think they all look alike. Actually, you Mead: As I understand what Dr. Richards said, what the child learns
know they don't, but it is very hard for you to distinguish them unless is that letters make sense and that you can discriminate between them.
you have a lot of contact with Chinese. Now, you may have one or two The major learning is not the particular letter, but the fact that reading
Chinese students, perhaps only one or two over a year, but they have can be done; it works, and letters are discriminable. What you are try­
been associated with you over quite a time. From then on, you will no ing to teach first is the possibility of discrimination and then, within
longer have the same amount of trouble—and it is not a matter of that, adding discriminable objects. If that is the emphasis, then the
specific knowledge of those two faces. You now go out and see a group effect you are talking about would come.
of Chinese; they all look very different, and it is now much easier for Bigelow: I am talking about an indirect effect that would come from
you to recognize them, reidentify them on the subway individually or in doing it in that particular, selective way rather than by throwing all the
groups. It is quite a difficult thing to explain, but I think it is a very letters at the child and letting him pick at random to form words from
common phenomenon. them.
Mead: But what you have learned there is a cultural point, isn't it— Richards: You could get somebody to explore them. That would be
that the Chinese are identifiable and have "human" faces instead of the only way, I think. We couldn't do it in our heads.
looking all alike ? Rioch: There is—in a sense it isn't quite comparable—that observa­
Teuber: I don't know whether that is the main thing. Don Hebb tion of Nielsen's (22) that children who have been taught to read by
(21) likes to tell about a similar experience he had in Lashley's labora­ words, without learning the alphabet, never develop alitralia when they
tory. At first, all chimpanzees looked alike to him, but after working get aphasia.
with a few chimpanzees for a while, in a very concentrated fashion, he Richards: I didn't know that. That is interesting.
could no longer conceive of anyone confusing any of his chimpanzees. Savage: What does that mean in English?
Bigelow: My question is not exactly a matter of failing to recognize Rioch: That means they never show the symptom of not being able
 Communication Between Men 87
86 Cybernetics
before, either. Some languages are just harder than otliers.
to read single letters. Patients who have as children been taught the
Richards: Sure, they are.
alphabet and then to read will be unable to read words but will be able
Teuber: By the same token, I am inclined to believe that when lan­
to read letters, whereas those who have not been taught the alphabet
guage breaks down, in dysphasia or aphasia, it does not break apart
but taught to read words, according to Nielsen, will either be able to
into such neat categories as some of the clinical reports indicate. I
read words or not be able to read anything.
haven't seen too many forms of aphasia, but I have never come across
Bigelow: The elementary bits, then, are words, somehow ?
a single case of ""pure" dyslexia or alexia in which the recognition of
Mead: Yes.
letters was affected, and not the recognition of words, or vice versa. I
Savage: An aphasie knows only what he has never learned.
think these are matters of degree. We need much more empirical evi­
Klüver: I have a question which, I am sure. Dr. Richards, with his
dence before we can be convinced that the early reading habits or the
wide experience in teaching languages, can readily answer. It is gener­
ways in which reading was taught, really show up later in the way in
ally maintained that a young child put into a different linguistic envi­
which speech and reading might break down.
ronment will quickly acquire a new language and soon speak Finnish,
Bigelow: Isn't there considerable psychological evidence that prac­
Chinese, or any other language with great fluency. Are there any excep­
tically everything else you are taught early shows up later?
tions to this? Or are there at least some languages which are more
Teuber: No.
difficult than others for young children ?
Bigelow: There is not?
Richards: I don't know any, but the anthropologists would be more
Teuber: I wouldn't say so.
likely to have such evidence. I am not aware of any, but I may be quite
Bigelow: Maybe I am reading too much Freud.
wrong.
Hutchinson: If you could introduce, hypothetically, into a language
Mead: We have no evidence at present. Now, I have to add how bad
an element that would make it almost impossible for children to learn
our material is on ages of children. Most anthropologists are men who
the language, children would not learn the language. Presumably an­
don't pay a great deal of attention to small differences in age of chil­
other language modification would grow up that they could learn. It
dren, and we don't have birth records, so you have to add that. But if
seems to me it is exactly what you would expect, within a rather narrow
you add both those difliculties, we have no evidence that suggests that
limit—that a sort of natural selection keeps the language learnable.
it is harder for a child, say, to learn Navajo than it is to learn Samoan.
Otherwise, no one is going to talk it. In Japan and Turkey, where
Looked at from the outside, the difference in difficulty in those lan­
literacy is becoming required, there is a great tendency to moderate the
guages is enormous. One of the very curious things is the fact that any
most difficult properties of the language.
human being will learn any human language if he starts as a small
Gerard: Except that after it has once been learned by the group, it
child. There is no difficulty about it.
does evolve and become more difficult and depend on more cultural
Savage: Didn't you say that in some cultures the older children of
influences.
ten, twelve, or thirteen spend a noticeably disproportionate amount of
Hutchinson: Yes, but if it were really difficult, then still nobody
time teaching the little children the refinements of the language ?
would learn it.
Mead: But that is not a result of the difficulty of the language; that
Rosenblueth: Many centuries have gone by, and learning to read
is a matter of the attitude toward the language in the culture. But, for
Chinese is still terribly difficult.
instance, a great deal of time is spent teaching Manus, which is a very
Hutchinson: Yes, but you don't need to be able to read Chinese to be
easy language, and there is no such amount of time spent in teaching
a Chinese. Presumably it is reasonable to suppose you would have to
Tchambuli, which is a very difficult one. That is another aspect of it.
be able to talk to be a member of the social group.
Klüver: In view of the relations between thought and language, and
Richards: It is true that in some countries illiteracy is almost a crime.
in view of the enormous differences in the structure of languages, it
Rioch: Deaf mutes are one group that we have with us, open to
seems remarkable that such differences should never affect the process
study, and nobody has studied them. That is really the place where one
of language acquisition in young children.
could study something about language.
Mead: Well, it is very striking, if you compare how easy or difficult
Hutchinson: They use some sort of sign language. They can com­
it is for an adult as contrasted with a child. The differential ease for
municate.
an adult doesn't seem to be a function of the language one has learned
88 Cyhernetícs Communication Between Men 89
Teuher: There is some evidence on "language" of deaf mutes, mostly permits them ( a ) to be manipulated, ( h ) to be transmitted, ( c ) t o be
in German—on what can be learned from what they use for a lan­ mutually understood, and (d) to produce an interaction between indi­
guage (23^24). , . r J- • viduals, which interaction gives rise to new experiences. These experi­
Birch: It seems to me that, despite the complexity of our discussion, ences in turn then give rise to new symbolizations, new contexts, and
we are dealing much too simply with the problem of language. Ap­ the development of an increasing and expanding language.
parently, language has more than one function. It has more than a It seems to me very difficult to study language as language. We must
specific symbolic communicative function, and if we restrict ourselves begin, I think, to differentiate, as Dr. Rioch, I believe, suggested, the
merely to a communicative function, then we begin to be able to deal process characteristics involved in a consideration of language, and
abstractly with language and treat language as a logical system for study these in and for themselves for a moment, if we are then going
which another logical system may be developed, for which in turn a to be better able to see the specific aspects and features of this over-all
further logical system may be developed, and so forth. I would submit, end-result phenomenon—not process—that we referred to as language.
however, that in doing this, we lose the richness of language and we In other words, language is not a single unitary event or a homogeneous
lose the richness of the psychological phenomena that are involved in event, but an agglomeration of events which we have in our own
either communication or in language which is a broader area than the linguistic systems categorized under the framework of the term "lan­
area of communication as such. guage." In that sense, to discuss language is to engage in a discussion
Communication really represents but one function of language and not of process but of a normative procedure that we have developed; it
is not identical with language. For example, if we take Dr. Rioch's is to discuss our tool of categorization rather than the processes which
remarks on the question of time and the stopping of time, what I think we have attempted to categorize.
he means is that words can come to take on representative meaning, in MacKay: In a way, that was the point I wanted to make earlier. I
a sense that they can become the equivalent of previous situations; that had been feeling that we ought to ask for a definition of language for
in using a word, then, one can rearouse the psychological processes t h i s r e a s o n ; C o n s i d e r t h e s e r i e s : ( a ) di m a n m a k i n g a s p e e c h ; ( b ) a.
attached to previous situations, and that therefore language in a sense sound-cinefilm showing the whole of the man while he talks; ( c ) í
frees us from the boundaries of specific time scales. But once we say record of his voice; (d) z written transcription. At each stage along
this, we must mean that in communication we must be dealing with that sequence we may say we have a representation of his speech. But
organisms that have a community of representative function in lan­ at what stage do we agree that we have reduced the representation to
guage, and that words are simply not words but are attached to experi­ terms of language?
ences, and that they become the equivalences of these experiences and, In music, we are not above putting in expression marks as additions
as such, become capable of being interchanged. If we then see that, we to the mere symbolic minimum necessary to begin to represent what
see that we not only have the problem of representational function of happened. We put in rises and falls and expressions of all sorts. If we
language, but we have also the problem of the conventional function had equivalent ways of doing that in writing, we would be expressing
of language; which means that the words that an individual learns, if something which is a genuine part of language, in ä sense. But at what
they are to be used in a communicative sense, must be words that other point are we going to stop ? Because essentially the act of replication
people have learned in connection with a communal body of experience that is evoked in the man includes the response to the rise and fall of
that has a certain conventional structure and conventional meaning. the voice, and so on. I know more of what a man means than is con­
Further, we use language in a different way. We may use it symbol­ veyed by the words he uses, if his voice is rising, because it evokes in
ically, we may use it as self-stimulus, we may use it in a variety of me something corresponding to what I would feel if I were to make
manners; so that if we begin to discuss the representative function of my voice rise. How do we define language ? Where do we draw the
language, or the symbolic function of language, or the conventional dividing line ?
features of language, or the contextual features, in which only specific Rioch: Along that line, some work of Jürgen Ruesch (25) in Cali­
representations may be produced, or other representations, depending fornia was very important. When he played a sound-recording of an
upon contextual circumstances, then we begin to deal not with a specific interview which was just a straight-forward interview to an audience
logic but with an abstraction process. In a sense, we deal with a process of medical students, it would have no particular effect upon the audi­
whereby generalized experiences have been converted into a form which ence apart from its content. If he played a record of an interview in
90 Cybernetics Communication Between Men 91
which the one interviewed becomes anxious—and frequently also the by language.? Apparently, we mean a special case of communication.
interviewer—-then the audience listening to the record became anxious; If so, which special case, and would it be sufficient at this point, rather
but they did not become anxious on reading a typed transcription. than after tomorrow night, to try to say what particular aspect of
Therefore, I think we have to separate various functions of language "language" should be used for the purpose of this discussion? I would
which are per se functions of language, but yet have to do with different try to test that definition, then, by asking the additional question: If
performances. that is what you mean by language, would you impute it to animals or
I should like to speculate along the following line: We can assume only to man?
a double integration of the organism with the environment as a starting Mead: Well, I should like to go back to Dr. Birch's point. It seems
point. With cortical function we get differentiation of different parts of to me—^probably this is historical, and I am merely making a historical
the environment. Now, suppose we differentiated a particular part reconstruction for purposes of discussion and not saying that it did
of the space as separate from other parts. We want to subdifferentiate happen—^but it is conceivable that very early in human history, for
that particular part of the space; that is, I want to pay more attention some reason, or accidentally, some people thought of categorizing
to, say, this glass. How do I keep this glass in the space and keep my language as the learnable aspect of other people's behavior. That is
attention on the glass, keep the glass down where it is, instead of the one of the most significant things about it. We have no people so
glass beginning to occupy all of consciousness Ì That is the problem I primitive that they are not able to say, "Those people have a different
am faced with. I am going to pay attention to the glass. How do I language." Sometimes in New Guinea they don't know where the "talk
keep the glass from expanding into all of consciousness, so that there turns" because it modulates sociologically, but they know "that is a
is nothing but the glass Ì different language from ours; I speak it or I don't speak it; I hear it
Actually, I have seen patients who have reported the latter experi­ or I don't hear it." Conceivably, instead of thinking that, one might
ence, for whom the glass comes to occupy all of consciousness, and the have said the postural system of another people is the learnable aspect.
rest of the room only a little fringe around the periphery of their visual I don't know what those sounds they make are. I don't know what the
field. Those patients will then use a word, "The glass is on the table," sounds we make are. That isn't the point. But I can go into their group
and immediately everything is back to normal relations. and I change my tonus, my shoulders, and so forth. Or take the cookery
Klüver: It is characteristic of the effects of mescaline that conscious­ of another people, which is at least as complicated as many languages.
ness may be narrowed down to the experience of one sensory or imagi­ Instead we labeled language. In the whole of human culture, language
nai detail, for instance, to the experience of one scar, one thread, one has been labeled in this peculiar way.
key, or one plate. Such a detail then expands and finally becomes Now, there is a possibility, and it is only a possibility, that the diffi­
"everything." The mescalinized person who has lost all spontaneity culty with language is not this point of using language to talk about
feels that it is only this scar or this thread that he is still conscious of. language, but that historically we have selected language as a segment,
In such a way the boundaries between the world and me, between sub­ and that has been perpetuated throughout human civilization, when it
ject and object, disappear and the individual feels that he is really is not as different from other parts of culture as it appears to be.
identical with the object, for instance, the scar (26).
Hutchinson: In Buddhic mysticism that is a very well-known tech­
nique for demonstrating the unreality of the world. You take an image
of a goddess, blow it up, and fill the whole universe; then you bring it
down to a point infinitely small and then fill it up again. After you
have done this a sufficient number of times, you realize that the goddess
is an illusion.
Teuber: I have a pedestrian concern. We said, somewhere, "com­
munication and language," implying that we are using two different
words for two different referents. I just want to ask—I mean, nobody
in particular has done it, but in general, do we mean to say that we all
know what communication is, but that we need to define what we mean
COMMUNICATION BETWEEN SANE AND Communication Between Sane and Insane 93
INSANE: HYPNOSIS to his session and told me that he had had a most remarkable dream.
In one of its scenes he had been out dancing; and he added, "Just
imagine this—in the dream I met you out dancing." He had not the
LAWRENCE S. KUBIE
slightest idea that this had actually occurred, or that he had recorded
Department of Psychiatry and Mental Hygiene
Yale University School of Medicine the fact of seeing me and then had reproduced it in the dream, with­
out distortion, and all without conscious participation in the process at
any step until the last one.
THE SUBJECT THAT I was asked to discuss here broadened with each Another example centers on an office lamp of green imitation bronze,
successive letter that I received about it. First, it was the problem of with an inverted bell-like shade which opened to the ceiling. I had
communication in hypnosis. Then it broadened to include communica­ owned this for about a month, during which time it stood near the foot
tion in hypnosis and in the psychoses. Therefore, I feel justified in of the couch in my office where no patient could fail to see it during
taking an even broader base; and this despite the fact that as a con­ the analytic hour. One day a patient told me a dream, which was of
sequence what I say will be a collection of fragmentary, scattered ob­ critical importance both in the development of her treatment and in her
servations tied together by a few questions with few answers. These relationship to me, although the manifest content of the dream had
observations are in the main clinical experiences; and since not many nothing to do with either. The dream was of a field full of "strange
of you have had an opportunity to share such experiences, I will de­ green daffodils . . . even the blossoms were green; and instead of droop­
scribe and discuss examples of the type of empirical clinical data which ing over, they pointed straight at the sky." As she finished describing
seem to me to be relevant to the broad question: What happens to the- these daffodils, she said, "I have never in my life seen anything like
processes of symbolic communication in different states of conscious­ them," and with these words pointed at the lamp and exclaimed: "My
ness? ( 1 ) ( 2 ) God, there are my flowers. How long have you had that lamp?" Evi­
I had hoped to begin with one living exhibit; but the patient did not dently "communications" can be recorded unconsciously, both in sleep
show up. He is a key member of our group, and an extraordinarily and in the waking state, and can then be reproduced recognizably in
gifted human being. For me it has been one of the pleasant sideshows subsequent waking thoughts (Norbert Wiener), or in dreams, either
of these conferences to watch him do something which I had often with some degree of distortion, as in this case, or without distortion,
heard described but had never seen before. During the course of a as in the dance-floor dream.
heated discussion among his fellow conferees he will be deeply asleep, I can give a similar example of something which happened to me
snoring, in fact; and suddenly he will break off in the middle of a snore when I was a medical student. On Sunday mornings I used to go on
to come out fighting, literally leaping to speech in the middle of a para­ hikes with friends who lived dowji the street. Our usual signal was
graph. What is more, that paragraph will be relevant to that which to whistle from one window to another. One of the group was musical,
had been under discussion. Evidently you have recognized our good and always whistled some erudite theme. One Sunday morning in
friend Norbert Wiener. This amazing phenomenon is something to spring I heard a whistle; and when I put my head out of my window
give us pause; and it has a relevancy which points in several directions this friend was waving from his. As I walked down the street toward
at once. Before going into them, however, let me give a few com­ his house, a melody was going through my head, a melody which was
parable, but not identical, examples. familiar, but which I could not name. I said to myself: "I must ask
I once had as a patient a likable young college student. One Satur­ David what that tune is. He will know"; and as I reached him, I asked,
day night, I met him out dancing. He was with a crowd of young folk: "What is this tune?" and whistled it. He looked at me in astonishment,
I was with friends. It was a typical New York dance floor, half the and said, "My God, I was whistling that for half an hour, and you paid
size of a postage stamp. I suppose we rubbed shoulders a hundred no attention to me. Then I finally changed to another tune and you
times that evening, as we danced through the crush. He looked me in stuck your head out of the window."
the eye repeatedly, but never batted his; so I assumed that he was Evidently communication can occur effectively without full conscious
embarrassed, and did not greet him, either. I did not want to upset his participation in the process. This point is basic. Even during states of
peace of mind. That was on a Saturday night. On Monday he came sleep and of absorbed attention, an automatic unconscious process can
92 occur by which we record sensory experiences, with equally unconscious

/
94 Cybernetics Communication Between Sane and Insane 95

subsequent consequences. I stress this point in order to indicate that them. This experiment has been repeated many times; and we can
the barriers between the psychological processes which go on in the deduce that we always record more than is readily accessible to us in
sleeping state and in the waking state are relative and not absolute; the fully waking state.
because parts of us are asleep when we are awake and parts awake We do not know whether we record more fully under hypnosis than
when we are asleep. By this I mean that sleeping processes are going in the normal waking state. For instance, in the fully waking state do
on during the waking state, and waking processes during the sleeping we screen out a certain amount of incoming percepts.? In hypnosis is
state. Sleep and wakefulness are relative and not absolute concepts or such selectivity at the intake in abeyance.? We know only that we can
states; and the phenomena of hypnosis (which is the challenge that recall more fully under hypnosis than in the waking state.
was given to me first) probably lie somewhere between. I emphasize all of this because if we should limit our consideration
Moreover, not only is intake possible, but we can also do many things of the problem of communication to conscious speech and our use of
in sleep whose close relationship to what happens in the waking state is and reactions to it, we would be oversimplifying the problem. The data
not always clearly recognized. There is the random activity of sleep given above demonstrate that we use and record and respond continu­
which has never been adequately studied (and never will be until we ously and at all times to an inarticulate or subverbal form of communi­
study infra-red movies of human beings during sleep). Then there are cation, both on conscious and on unconscious levels. Since I am speak­
the more organized phenomena of sleepwalking and sleeptalking which ing to mathematicians, I hesitate to use figures; but certainly it would
deserve study. Finally, there is the form of sleepthinking which we call be no exaggeration to say that our conscious and unconscious psycho­
the dream, which is the only one of these which has been closely stud­ logical processes deal with an enormous number of things simultane­
ied. It is interesting to contrast the emotional discharge which accom­ ously. Therefore, it would seem to me that the first challenge to the
panies these forms of action in sleep and in the waking state. In the mathematicians and to the makers of mechanical models is to arrive at
waking state the emotional processes are least adequately discharged some approximate impression of the numerical limits of this. There
in thinking, more so in talking, and perhaps most completely in action. must be such limits; but these limits have never been established. Cer­
In some measure the reverse relation obtains in sleep. The dreamless tainly the amount of data that is taken in and recorded is infinitely
sleepwalker, the rare pure somnambulist, although he occasionally acts greater than that scanty sample which we can reproduce consciously.
out violent emotions, more often is relatively emotionless. The sleep- The same is true of our psychomotor and emotional responses, to which,
talker comes in between; and, as we know it, the dream is the most in turn, we react automatically, thus compounding our inner psychic
highly charged. processes by geometric progression. Consciousness then represents only
Let me summarize this by emphasizing that a process of communi­ a small fragment of our total psychic state. This fact must be included
cation can occur in sleep, in states of deeply absorbed attention, and in in any study of communication.
the hypnotic state, a process of communication which depends on all It can be studied in the waking state; but as I have indicated, this
of the techniques of communication that we use in ordinary speech. presents special difficulties because of the selective processes which the
Furthermore, although in full sleep such communications are predomi­ waking state involves. With those individuals who are hypnotizable,
nantly a process of intake, they can become a two-way process. Con­ it can be studied in the states of hypnosis; but this brings up its own
sequently, there is such a thing as a communicative sleep, which comes technical difficulty. Nobody has yet been able to determine what it is
close to being what is meant by the hypnotic state. (3) that determines whether or not an individual is hypnotizable, or why
The next point about communication in hypnosis is the fact that some individuals are hypnotizable and others are not. Consequently,
under hypnosis our automatic capacity to record or recall minute-by- the use of hypnotizable subjects introduces a selective factor, the nature
minute experiences is more photographic and phonographic and more of which we cannot evaluate accurately. Communication can also be
inclusive than seems to be true in the waking state. This is the so-called studied in sleep; but this, again, introduces a selective factor, since not
"hypermnesia" of hypnosis. It refers to a well-known phenomenon; all individuals are accessible to either one-way or two-way communica­
namely, that when anyone tries to tell what he has seen or heard, he tions during sleep.
can give at best a limited report; whereas under hypnosis that same Let us carry further our consideration of the study of communication
individual will produce a wealth of additional verifiable memories, in hypnosis, even though it will be seen to be even more complicated.
without his having even been aware of the fact that he had recorded There have been three interesting types of experiment which indicate
96 Cybernetics Communication Between Sane and Insane 97
that under hypnosis individuals can translate things which they can subject's material. When all the reports had been presented and com­
neither translate nor understand when they are not under hypnosis. pared, I presented the analytic material which I had gathered over
This concerns the translation of automatic writing, of automatic draw­ months or years of analytical work.
ing (even doodling), and of the symbols of dreams (4) (5) (6) (7). I was interested to see what different aspects of the personality were
I have a sample of automatic drawing which I will pass around. This delineated by the different tests; but what seemed even more extraor­
is an automatic drawing which in part was translated under hypnosis dinary was the accuracy with which in certain instances the life history
by the subject who had made it, and in another instance by another was described, particularly by the graphologist. For instance, in one
individual under hypnosis (4). The same thing was done with the case, from the study of the handwriting alone of a young woman whom
automatic writing of another patient (5) (6). Then there are the in­ the graphologist had never even seen, the graphologist said: "I think
teresting experiments of Färber and Fisher (7). In these they described this young woman recently had a baby. I also suspect that there has
to Subject A under hypnosis a painful "experience," and then asked him been a recent suicidal attempt." Both of those things were true. Of
to dream about this implanted "experience." The subject thereupon course, a few isolated instances of this kind prove nothing, except to
reported a dreamlike, disguised reproduction of the "experience," trans­ indicate how heavily overweighted are the multiple determining factors
posed into more or less classical symbols, almost as if this naïve indi­ of some seemingly simple and stereotyped acts, and consequently how
vidual had read a dictionary of dream symbolism. Then the experi­ revealing such stereotypes can be.
menters showed Subject A's hypnotic dream to Subject B while under This leads to the problem of universal symbols in general, which in
hypnosis, and asked Subject B under hypnosis what the dream of turn will bring us to the question of communication in the psychoses.
Subject A had meant. Thereupon Subject B translated the dream of Neither problem has been studied or explored systematically. Nor has
Subject A back into the essence of the unpleasant story which had there been any systematic study of the closely related problem of the
originally been told to Subject A. influence of culture on universal symbols. Only fragmentary experi­
That is more than a trick. It is a very important experiment which ences have been noted. These experiences are often amazing, and
has been repeated often enough to make us sure that one human being always interesting and significant, especially when they come through
can on a dissociated level translate accurately the symbolic representa­ the words of small children who have had no opportunity to be con­
tions of the unconscious content of thought and feeling of another taminated by the ideas and the literature of their elders. I see Dr.
human being, thus exposing and uncovering repressed experiences Mead frowning at me already. Perhaps she will agree at least that
which the other has had in the past. This is a critical and conclusive this parental influence is not in a form which is specific enough to
demonstration of the power and specificity of unconscious psychological determine their use of classical symbols. Let . me give you some ex­
processes, and of their importance in the communications which pass amples (8).
between men. A little girl of four, an only child, puts pencils in the drawers of her
I have had a closely comparable experience in another equally inter­ dolls and calls them penises. Has parental influence produced her con­
esting but wholly different form. I must say that it surprised me, be­ flict over whether she wants to be a little girl or a little boy, or has our
cause it was wholly unexpected. Some years ago. Bela Mittleman and culture determined her particular way of symbolizing her problem.?
Ruth Munroe organized a small group which met several times. The The same little girl pins safety pins in the front of her chemise and
group consisted of clinical psychologists who were interested in the calls them her penis. She has hardly been instructed to do this either
meanings of figure drawings, in graphology, the Rorschach, and the by her parents or by deliberate cultural influences in the society in
TAT, plus a couple of analysts like myself. I sent several of my ana­ which she has grown up. Or what of the little boy grown to puberty,
lytic patients for a Rorschach to Ruth Munroe. Each time Ruth asked whose curiosity about the genitals has been stimulated by the fact that
the patient for his reactions to the Thematic Apperception Test, asked he finds his best friend peeking through the windows of the girls'
the patient to make figure drawings (Machover), and then sent a washroom in school. He has a fight with his best friend about this,
sample of the patients' handwriting and the other material to the other and then has nightmares and sleeps badly and comes home from school
participants in the test, with no information except, "This is a twenty- with an eye-blinking tic and worrying about what will happen to his
seven-year-old woman," or "This is a thirty-three-year-old man." Each eye. Then he has a long discussion with his father about the normality
participant was asked to bring an independent interpretation of the of bodily curiosity, after which he dreams about an airplane with
98 Cybernetics Communication Between Sane and Insane 99
strange bulges, whose appearance is so curious that he cannot tell the Finally, these considerations lead us to the problem of psychotic in­
front from the back or the top from the bottom. The day after his sight (or more specifically, to insight as it occurs among schizophrenics)
dream this little boy awakens without his eye-blinking tic. Such com­ and its bearing on the problem of communication. Here we are up
monplace episodes raise issues which are far-reaching in their implica­ against the most puzzling manifestation of what I am trying to bring
tions for communications. What role, if any, does culture play, and together; namely, the fact that schizophrenics (who psychologically are
how does it operate to determine any youngster's use of a flying ma­ among the sickest people we deal with) know most about their own
chine instead of a serpent or a bird for the translation of his instinctual unconscious. Indeed, they can often translate their own symptoms,
conflicts into symbolic language.? their own behavior, and their own symbols. They can say, "I know
Or what of the little girl who night after night has the same pleasant that when I wear a knitted tie, it means that I feel X about my body,
dream; namely, that a little mouse gets into her bed, and snuggles whereas if I wear a silk tie, it means Y," and so forth. The ordinary
down beside her ? She tells the dream each morning. Both the dream­ human being never has this sort of subjective symbolic insight spon­
ing and the telling are meaningful. One day, after about two weeks taneously. Yet where it arises spontaneously in the psychotic, this type
of this, she happens into her father's bathroom just as he has come of insight for some reason seems to be devoid of therapeutic leverage.
out of the bath. He is standing there naked; and she comes to him Let me add another point. It may well be that in sleep there is a
and examines him with close attention. This little girl has seen her continuous flow of active psychological processes. We cannot be sure
brother naked many times, and she knows the names of the male geni­ that this is true. The flow may be intermittent, or it may occur only
tals. Nevertheless, as she backs away from her father's nakedness she during the processes of falling asleep and of waking up; and it may
looks ofi^ into the distance and makes up a language of her own, say­ cease at that level which we figuratively call "deep" sleep; or at this
ing, ""I have just seen a little mouse." What determined the translation level it may become so inactive as to be unimportant. However that
of this small girl's experience into animal terms, and the substitution of may be, we can be certain that even in sleep a symbolizing process
the name of an animal for the known name of a known part of the occurs which can have destructive consequences. The fact that destruc­
body? tive psychological processes can occur during sleep is a daily proof of
We could multiply such experiences many times. I do not under­ the operation of unconscious psychological forces.
stand this process. I know only that it happens constantly, and that it People are naïve about sleep. They think of it as always being a
forces on us the conclusion that it is impossible to talk about anything healing, resuscitative, and constructive force. Yet actually we know
in the outside world without at the same time having it refer to some­ that it can be a destructive experience. People can go to bed well and
thing in the inside world. Thus with every word we utter, with every wake up sick. People can come out of sleep in a paranoid-schizophrenic
gesture, with every expression we talk simultaneously about the inner psychosis. There are patients who waken out of sleep in deep depres­
world and the outer world, with the result that all language units have sions, in acute manic states, in stormy somatic disturbances. These may
multiple meanings in which the relative roles of internal and external be the mild transitory depressions of waking, or they may be the emo­
objects and of conscious and unconscious determinants are in a state of tional states of those more fortunate human beings who waken from
continuous and unstable dynamic equilibrium (9). sleep in transient, hypomanic states. These may not last long, but at
These are demonstrable facts. Furthermore, we find that the mean­ least they get us out of bed. All of this is evidence that much goes on
ings of symbols are not always idiosyncratic, as one might expect, but in sleep. Insofar as we can make contact with this, it is represented in
often are highly stereotyped. Indeed, our ability to make any general the most purely symbolic language that we know; namely, the language
translations of psychological tests depends on this sterotype, whether it of the dream. Further evidence of the activity which goes on even in
is graphology, the TAT, the Rorschach, figure drawing, or the Szondi, sleep is the extraordinary time sense that some people have in sleep,
and so forth. Their meaningfulness is specifically dependent on the but which, for some reason or other, seems to be lacking in others.
stereotyped universality of the conscious and unconscious symbolic im­ I do not want to overburden this. Therefore, at this point I will
plications of their forms. (I must say that of the tests the Szondi interrupt this gathering of fragmentary and puzzling empirical data.
puzzles me most. As with the Wassermann, not one word of the under­ Subsequently I should like to return to the basic issue of symbolic
lying theory makes sense; but for some reason the test works none the thinking. It would be better to interrupt at this point, however, so as
less. I have no idea why.) to open these remarks for discussion and questions. Later, if there is
100 Cybernetics Communication Between Sane and Insane 101

time and interest, I can go on to the other problems, or else reserve brick. The only things you can't check are things on the opposite side
them for another conference. of the brick or an angle off the side of the wall.
McCulloch: May I abuse for a moment the privilege of the Chair­ von Bonin: How many things can the normal person remember ?
man and ask two or three questions ? I spent a considerable amount of McCulloch: The estimate is that the maximum is 10 frames per sec­
time and effort during my senior year in Yale, back in 1921, attempt­ ond.
ing to evaluate how much can be recalled under hypnosis. We were von Bonin: No, no, I don't mean that; the normal person.
by no means convinced that an infinite amount could be recalled, but McCulloch: On the normal person, the best man known on receiv­
only that an amount which was vastly in excess of what is available in ing communication in the United States Navy could give you a hundred
the waking state could be recalled—I would say perhaps a thousand letters in sequence at the end of having received a hundred in 10 sec­
times as much; still it has an upper limit. onds. He had to wait until he had passed through a period during
Pitts: Did you try simple material such as a string of nonsense which he could not recall, and then he would give you all the hundred
syllables Ì letters.
McCulloch: We tried this sort of trick: We took master bricklayers- Bigelow: What kind of communication was it?
who laid face brick and had them recall the seventh brick in the row, McCulloch: Semaphore.
or something else like that, in a given year. They were able to recall Pitts: If you were to hypnotize him—
any one such brick—thirty or forty items at the most. That was a brick McCulloch: That is not more than five bits per letter.
that had been through their hands some ten years before. It is still not Bigelow: But it is essentially controlled by the semaphore process.
an infinite amount. A master bricklayer can lay only a certain number McCulloch: Well, this can be sent as fast as you will. It is sent by
of bricks per diem even when his entire attention is riveted on laying machine.
bricks. The amount is not infinite. MacKay: What was the redundancy in the information about the
Savage: What would you call an infinite amount ? bricks? Or, to put it otherwise, how frequently did the features re­
McCulloch: A thousand bits per minute at the most, or something called crop up normally in bricks ?
like that, that you can get back. It is not infinite. I think that is McCulloch: In bricks, it is rather rare. The kind of things men re­
critical— member are that in the lower lefthand corner, about an inch up and
Savage: Well, who could believe it could be literally infinite? And two inches over, is a purple stone, which doesn't occur in any other
if it is not literally infinite, who could believe it would be as much as brick that they laid in that whole wall, or things of that sort. The
you say? pebble may be about a millimeter in diameter.
McCulloch: There are two things here which are equally important: Bigelow: How could he possibly remember thirty of those features
first, that it is vastly more than one can remember in the waking state; on this one brick ?
and, second, that it is not infinite. McCulloch: Oh, they do. It is amazing, when you get a man to recall
Pitts: You could take nonsense syllables or present a board filled with a given brick, the amount of detail he can remember about that one
letters, and determine how many could be remembered. brick. That is the thing that is amazing. I note, aS a result, that there
McCulloch: It is extremely difficult to get a man to attend to this, is an enormous amount that goes into us that never comes through.
but a master bricklayer looks at the face of the brick when he is lay- Kubie: This is comparable to the experiments under hypnosis, in
ing it. which the subject is induced to return to earlier age periods in his
Mead: Did you find differences between two master bricklayers in life (10).
which thousand items they would notice ? McCulloch: I have never done any of those.
McCulloch: No, not significantly. He mostly noticed three things— Kubie: In these you say to the subject, "Tell me about your seventh
Rosenblueth: How do you go back to check ? birthday," or his ninth birthday, or something of that kind; and he gives
McCulloch: These things are verified by checking the bricks. They you verifiable data about the party, who was there, and so on.
are master bricklayers. That means they are laying face bricks. That McCulloch: I have seen rriy mother's uncle, who was an incredible
means that even ten years later, you can go to that row and look at the person in the way he could recall things—he was law librarian in Wash­
ington for a few years—testify that he had looked at such and such a
102 Cybernetics Communication Between Sane and Insane 103
document some twenty years before, and shut his eyes and read the couldn't separate the groups in terms of the height of response or the
document and the signatures under it. length of response or of any other one thing other than the total breaks
Savage: I don't have absolute recall, but I can remember hearing that in the curve; the general damped character, if you will, of the auto­
story more than once in sessions of this group. nomic response. This is the way they are tending to interpret it.
McCulloch: They said, "But the document does not read that way; it McCulloch: It is extremely damped in the "I" and the recurrence is
reads so and so from there on," and he replied, "Then the document extremely high in the "It."
has been altered and you had better check the ink." And when they Gerard: That's right; those are different from the normals.
checked it, it was a forgery. McCulloch: No, the normals are not distinguishable from the "It's"
Savage: It's the same story, all right. in there. Do you remember ?
Gerard: What about your statement. Dr. Kubie, that a person going MacKay: "There was a difference of 2.5 in 24, or something like that.
into the room under hypnosis can remember still more ? McCulloch: Statistically, it is not distinguishable between the hypno-
Kubie: I am not sure about that. The evidence is not conclusive. It tizables and the nonhypnotizables. The hypnotizables split very sharply
has been hard to quantify these things. into the two groups—the "I's" and the "It's." The "I's" have the very
McCulloch: May I ask one question about the hypnotizable.? And I low number, about 12. The "It's" have about 26 — that is, psycho­
ask it in all humility. There are two groups which, I would admit galvanic wobbles.
right away, are extremely hypnotizable, in the sense that any of us with Pitts: Do you mean fewer of them ?
sufficient effort can hypnotize them. They are very distinct. The other McCulloch: There are very few of them in the "I's" and very, very
day a group of students who had been working on it at the University many in the "It," but the "It" is not statistically different from the non­
of Chicago came to us to discuss this. They had two hypnotizable hypnotizables. Isn't that right?
groups—one of which, afterward, says, "I did this, I did that, I did MacKay: Yes, that was as they described it.
the other"; these subjects perceive the whole experience as something Bigelow: Will you read the reference to this into the notes of this
in which they played a part. They are always ready to come back to be meeting ?
hypnotized again. The other group experiences the world as "this hap­ McCulloch: It has not been published yet, so far as I know.
pened, that happened, and the other happened," completely deperson­ Gerard: These fellows are just working on it.
alized. It is very difficult to get them back for a second trial in hypnosis. McCulloch: Yes; they are trying to get it out now. Now, the next
Bagby and I ran into the same thing at Yale, long years ago, with our thing I want to say about this is communication during sleep. Com­
techniques. Then, we had only about 20 per cent hypnotizable, whereas, munication to somebody who is asleep and communication from some­
according to the group in Chicago, about 30 per cent of college students body who is asleep may be quite different problems.
are. hypnotizable. Do you know what happens with a group of people During World War I, while I was in the Navy, part of my job was
who are not hypnotizable ? Do you know how such people feel about to see to it that every man got out of his hammock when a bugle blew.
the attempts.? Every old sailor, when the bugle blows, throws a leg out of his ham­
Kubie: No. mock. That is known as the "shore leg." When he has shown his leg,
Gerard: Incidentally, Willey and von Borstel, whose work you have it is not permissible either to paddle his bottom with a wooden slab
mentioned, have found an index now that seems to set off those groups. provided for the purpose or to cut the lashings of his hammock, which
McCuloch: Which groups.? drops him on the floor. Those men were as sound asleep by every test
Gerard: The active participators from the nonhypnotizable, and the after they had shown a leg as they were before. There is no question
passive ones. but that they were able to communicate that they were awake in spite
McCulloch: The "I's" from the "It's." of the fact that they were otherwise sound asleep. That is Item No. 1.
Gerard: The number of breaks in the galvanic skin resistance curve. Item No. 2: During World War I, when the Navy went into—yes,
McCulloch: Yes, that is what the criterion was. I think it was only the Navy—a long study of the training of men in
Bigelow: During the process, or what ? dot-dash Morse code while they were asleep, they found that men could
Gerard: No, quite aside from it; that is, under ordinary conditions, be taught the Morse code while they were asleep to such an extent that
or while the operator is trying to hypnotize them. These experimenters although they were never awake, so far as we knew, during the inter-
104 Cybernetics Communication Between Sane and Insane 105
val, when their name was called and they were told to report to station Bigelow: If you don't know any of its specific properties, how can
they understood it in Morse code; so that there is no question but that you be surprised when such properties are exhibited ?
communication to people goes on while they are officially asleep. But I Pitts: If I recall, the EEG of a man under hypnosis is not in the least
think one has to distinguish between communication to and communi­ like that of a man asleep. Instead, it is like the case of a man who is
cation from them. concentrating intensely on something, either a thought of his own or
Rosenblueth: In line with what we are saying, I don't see that there something you tell him to look at fixedly.
is any puzzle here because of the fact that these things occur in certain Bigelow: The same difference in EEG as might exist, for example,
people under certain conditions. The fact that somebody can, at a between a man who is conscious of doing some trivial arithmetic and
given moment, recall many things which he did not recall under other one not conscious of any thought ? It might not relate to his memory
circumstances is not particularly puzzling. The puzzling thing is the properties at all.
problem of memory. I don't think Dr. Kubie's examples differ from Kubie: May I put in a few words before you go too far with this.?
any other type of memory, so far as I can see. The problem that we The whole discussion reminds me unhappily of the ancient discussions
have to discuss, if we are going to discuss it, is the problem of memory. about behaviorism. The behaviorists also hated the concept of conscious
There is one assumption—it wasn't all data that Dr. Kubie gave us— and unconscious psychological processes. They did not even like to use
that we should consider. I may be wrong, but I thought he implied the words "thinking" or "not thinking." But to get away from such
that there must have been impulses registering unconsciously. I think terms, they had to devise a new pair of words: "verbalized" and "un-
this is unacceptable, because it is a contradiction in terms. Either some­ verbalized." These were John Watson's. They also used "laryngeal
thing occurred or it didn't occur. Do we mean by "mental" anything posture tonus" and, I suppose, "laryngeal unpostures." All of this is
a person can report at a given moment.? Then the experiences were plain silly. There are many terms in science, and in human life in gen­
mental. If he cannot, then they were not mental. It becomes mental eral, that are difficult to define, yet this need not mean that we do not
only when he can report it. If that is the definition of the word "un­ know what we are talking about when we use them. We know what
conscious," its use in any other connection, except to say that something we are talking about when we talk about sleep, yet no adequate defini­
has not been mentalized so far, means nothing, in my opinion. tion of sleep can be given. The same is true of awake, or of the process
Rioch: I don't think it is possible to use the word "unconscious," but of hypnosis. We know that these are two states; and we can think of
it is possible to use the word "conscious." I think one can give an them either as two absolutely opposed and absolutely different states,
operational definition of "conscious," but not of "unconscious." or we can look upon them as relative terms, for example, points along
Mead: Can't you think in terms of the steps it is necessary to go a continuous spectrum, as I have suggested. It seems to me that there
through to make something conscious ? are many reasons why the differences between sleeping and waking are
Rosenblueth: Again, on the problem of sleep, I wish to make more relative and not absolute, and that hypnotic phenomena fall somewhere
explicit what Rioch said, because I wouldn't entirely agree with him. in between, with some of the characteristics of one and some of the
What is puzzling is that somebody in sleep registers certain things. The characteristics of the other. This is advanced, however, not as a fact
problem is the problem of sleep itself, about which we know nothing. but as a working hypothesis.
The phenomenon is puzzling only because we don't understand the McCulloch: Can we reserve that question. Dr. Kubie, for when we
problem of sleep. I don't think it is any more puzzling to grasp the come to an attempt to sharpen up the question of hypnosis itself.? I
fact that I can say something to somebody who is awake and he can think you should be in on that later.
repeat it, than the fact that I can say something to somebody who is Kubie: Certainly; but I want to repeat that it is nothing short of silly
asleep and he can repeat it. I don't think there is any important differ­ to waste time in a verbal quagmire over whether we have the right to
ence there. think of "conscious" but not of "unconscious," or of "unconscious" but
Gerard: The problem there is why, under hypnosis, it is possible to not of "conscious." This is like saying, "I can think of white but not
bring these things into awareness. of black." We know that there are processes with full attention and
Rioch: I think it merely makes the situation more complicated. awareness which have psychological consequences, and that there are
Bigelow: I agree with that. others which have equally important psychological consequences and
Rioch: You can have lots of fun with it but you don't know what which operate in the same way, but which are below the level of con-
you are talking about.
106 Cybernetics Communication Between Sane and Insane 107

sdous attention and awareness. Call them anything you please. Call deal with here, be culturally determined. There are things which make
them "X" and "non-X" if you prefer; but the phenomena remain; and me think otherwise about some aspects of it, but I will come to those
the simplest operational and descriptive characterization of them is as in a minute. I think it is rather crucial that we keep in mind that there
Conscious and Unconscious. may be symbols. For example, a large amount of my dreaming is in
Mead: But, Dr. Kubie, I think you get a complication here by in­ verbal symbols.
cluding under one heading instances of attentiveness, relative attentive- Bigelow: How do you know that ?
ness, under sleep, under hypnosis, and in ordinary states. McCulloch: Because I wake up laughing. Having managed not to
Kubie: This is done on purpose, because the phenomena of attention say something during the day, I then put the people back into their
are closely related to the problem of hypnosis (3). situations and am saying it when I wake up in a gale of laughter.
Mead: Wait a minute. You didn't let me finish my sentence. Those Bigelow: Is this sufficient proof ?
three are paired, or triplicated together, or tied together, or whatever. McCulloch: Yes. I have very clearly the meaning of all the remarks
We get the question of symbolism where, for instance, one subject is that were passing at the time.
able to interpret the artistic product of another subject. I think that it Pitts: Do you mean verbal in the sense of actual speech Ì
is an artefact of our own society and our own methods of exploration, McCulloch: Yes, verbal in the sense of actual speech, but not audible,
at present, that we recognize a particular kind of thinking, if you want or only rarely are they audible. Now, this type of dreaming clearly
to call it that, most prominently in the dream. We also recognize that depends upon puns, and so on, which are only possible in the English
it occurs in poetry and in the arts. We oppose it to rational, conscious, language in many cases. They could not be done outside the English
logical thinking in our culture, and therefore there is a tendency for language. In German they would be impossible; in Latin they would
this sort of thinking that you have talked about to occur more in states be impossible. I think, therefore, that one has to take care of the extent
that are locked off, or in sleep. to which the cultural component is determinate.
If you look at some other cultures, you don't necessarily find that Mead: I should like to make clear, though, that that is not what I
same contrast. We can usefully make a separation in our thinking about was saying. I would regard those things as vocabulary. Surely, if you
this whole problem of consciousness or attentiveness or degree of re­ are English, you dream in English. If you have never seen a certain
ceptivity to imprint and ability to recall—and I agree with Dr. Rosen­ thing, the chance, of your using it as a symbol is low. But the point I
blueth that it is a question of memory and attention in one form or was making, on differences between cultures, was the difference between
which sorts of thinking were the correct kinds of thinking for people
another—putting on the one side the question of types of thinking that
are modeled on the external world and, on the other, the types of to employ. Now, in the Trobriands it is incorrect to recognize what we
thinking that are modeled on the body. We can use this shorthand for regard as logical relations, so that when you are being a rational, con­
considering these two sorts of thinking, one in which you have a whole scious person, you know that copulation has nothing to do with pro­
series of proprioceptive and bodily images which one person can inter­ creation, that food has nothing to do with growth, and that the plant­
pret from another very easily and of course they do it better under ing of seed has nothing to do with the plant growing. But you continue
hypnosis. I have had a subject put under hypnosis and then showed her to act on the assumption that all those things happen. The only possible
films of Balinese trance, and she would pick accurately the place where assumption, therefore, that I can see in the light of our present knowl­
the Balinese went into trance. She can do it perfectly well, and I don't edge, is that the Trobriands make "conscious" something that we do
see why she shouldn't, because under that state she is using a common not, normally, and carry in this other set of imagery the sorts of knowl­
system of thinking, which is what the psychotic uses when he seems edge that it is necessary to plant plants and produce children and all the
telepathic. Now, I don't believe that the association of these two sorts other things they do, in a perfectly accurate and efficient way.
of thinking, although you find it all the time in patients, is necessary in McCulloch: "That is what I mean by the "forbidden" meaning of
this society. If we thought about the two things separately, I think we words. It doesn't seem to me that this is fundamentally different.
Mead: Except that I am assuming that the mechanisms of symbol use
would get further.
McCulloch: Right. That is one of the real questions to my mind. are probably universal. The question isn't: Which symbol? I thought
Dr. Mead. The symbolism required for translation, retranslation, and it was misleading if you just used the content there.
the likeness to the original may very largely, in any population that we Klüver: Dr. Kubie, you repeatedly referred to the fact that items
108 Cybernetics Communication Between Sane and Insane 109
from past experience may be reproduced, for instance in a dream, ate thé structural similarity of psychoanalytical mechanisms and mecha­
without the person, when telling his dream, being consciously aware nisms of eidetic imagery (17) (18). In this connection a truly horrible
that the item in question represents a revival of past experience. It thought has at times occurred to me; namely, that Freud himself, like
seems to me that it is the study of eidetic imagery which has furnished Schilder, was an eidetic. It is true that Zeman published an experi­
the most striking illustrations of such mechanisms. I spent several years mental study in which he insisted that eidetic imagery is commonly
studying eidetic individuals and found again and again that the subject, found in all Viennese children between the ages of eleven and sixteen,
when observing and describing the eidetic image in front of him, was and that 61.5 per cent of all high school students in Vienna—or even
greatly surprised in seeing objects or details he could not recall as hav­ 88 per cent when including "latent" cases—are eidetic (19). I have
ing seen in the picture previously shown to him (11) (12). However, never seen any data on the percentage of eidetic individuals in the adult
it was obvious that the objects or the details seen by the subject were Viennese population. It is also true, of course, that Freud was not
often merely eidetic reproductions of details or stimuli presented by born in Vienna. Whatever the merits may be of this terrifying idea
the experimenter a few minutes, weeks, or months ago. This leads to of mine, at least as an amusing psychological exercise I recommend
a second question— an examination of the edifice of psychoanalysis in the light of the "as
Kubie: May I ask a question.? To your mind, does this provide if" Fiktion that the builder of this edifice was an eidetic.
adequate experi^nental proof that we can register and record percep­ I should like to return for a moment to the problem of "symbols."
tions without conscious participation in that process, and then repro­ Dr. McCulloch mentioned facts of translation, retranslation, and trans­
duce those perceptions without conscious participation in the process ? formation in connection with symbol formation. This reminds me that
Klüver: The relation of eidetic imagery to ordinary visual imagery Storfer once wrote an article on "psychoanalytical animal psychology,"
has been the subject of numerous experimental investigations (13) .and that Imre Hermann described "models of Oedipus and castration
(14). Most investigators in this field agree that visual memory and complexes in monkeys" (20) (21). Hermann, for example, referred
eidetic imagery are, on the whole, independent. The fact that certain to the simian equivalent of the strict father preventing sexual inter­
objects, forms, colors, and so forth, have been consciously perceived course of the son with the mother, or to such observations as the son
and are subsequently remembered in great detail does not necessarily having intercourse with the mother immediately after death of the
imply that the subject can reproduce them eidetically. Furthermore, his father. To be sure, he made the most of such simian demonstrations
intention to revive them eidetically may be entirely ineffective. On the or "models."
other hand, elements or items, for instance, in a picture shown to the I am wondering, Dr. Kubie, whether these psychoanalytical writers
subject may appear in the eidetic image of this picture, although the did not miss a bet. It occurs to me that by an equally strenuous exercise
subject cannot recall such items as parts of the original stimulus. Or of their imagination they could have found far better "models" of
these items may later appear "spontaneously" while the subject is ob­ psychoanalytical mechanisms in the field of bird behavior. I am refer­
serving an afterimage or describing eidetic images of entirely different ring here particularly to the so-called "displacement activities" of birds.
stimulus objects. In eidetic reproduction certain selective factors are In recent years, Armstrong in England has been especially concerned
operative, just as they are in other forms of reproduction. Past experi­ with studying such transposed behavior patterns and the way activities
ence is often eidetically revived by way of very specific spatio-temporal become displaced and ritualized (22). Apparently there is no doubt
fragments. This leads to another problem. that these displacement activities have reached their highest develop­
In analyzing the behavior of eidetic images we often find a trans­ ment in birds and that they play a far less important role in mammalian
location of objects or a transfer of certain properties of these objects, behavior. Armstrong has even considered the possibility that the high
fusions and composite formations, substitutions, the appearance of parts development of the striatum in birds may be related to these ritualized
instead of wholes, the nonappearance or the belated appearance of movement patterns and displacement activities (23). In pursuit of this
objects, reversals of right and left, above and below, and of other delightful speculation, I should like to ask Dr. Kubie or a clinical
directions. Differently expressed, we find condensations, displacements, neurologist whether symbol formation or the character of a neurosis
reversals, and other mechanisms supposedly typical of dream forma­ has special features if the neurotic happens to have at the same time
tion (15) (16). Because Schilder himself was an eidetic, it is not one of the basal ganglia diseases. Has this ever been studied Ì
surprising that he alone among the psychoanalysts was able to appreci- Rioch: Not that I know of. The function of repetition of behavior
110 Cyhernetícs Communication Between Sane and Insane 111
is, I think, a property of all nervous systems; that is, there is a strong The female apparently does not eat this offering, and in some species
tendency to repetition of preceding patterns of function in situations the male just wraps up nothing in silk and hands it over.
of conflicting stimulation. I do not know how far back it goes in the Bigelow: That is a jilting trick.
phyletic series, but certainly to forms simpler than birds. von Bonin: Do they get away with it?
Klüver: It has been often pointed out that displacement movements Hutchinson: The other point I want to make, which may have a sort
and other displacement activities become more easily ritualized or of higher relevance, is that it does not seem adequately to be appreci­
formalized in birds than in any other group of animals. In fact, it has ated that there are certain connections between birds and men which
been suggested that there are relations between speciation and displace­ are not apparent between birds and the dog, or man and the dog.
ment proneness. They include the pre-eminence of vision, particularly color vision, and
Rioch: I can't say. the probable importance of vision in sexual behavior, particularly in
von Bonin: May I put in a word of warning Ì I am quite sure that ourselves and in birds, which we don't share with so many mammals
the striatum in birds, whatever it is, is hypertrophied; it is a huge afl^air that depend more on olfactory stimuli. It may be that we would ex­
that fills up practically everything, but it undoubtedly has an entirely pect the kind of psychoanalytic situations that occur in man to be
difl^erent functional role than the striatum in mammals. In mammals more closely modeled in birds than in lower mammalia, for some reason
the striatum is coupled with the cortex. Birds haven't got a cortex, so connected with that. I don't think that that has been very adequately
it can't be coupled with it. The striata may be much more independent considered, except that possibly Armstrong may have talked along those
and may be a much higher mechanism in birds and may actually replace lines.
to soriie extent the cortex in mammals. Birch: I should like to make a remark or two about Armstrong's ob­
Rioch: I will argue that. servations and general considerations. Essentially, he has great freedom
von Bonin: Well, you can't argue that the birds don't have a cortex. in developing any interpretation, mainly because we have no systematic
Rioch: No, but I will argue that the striatum has this high form of knowledge of the phenomenon under consideration. This is true either
function. in terms of the discussions of display or discussions of displacement or
von Bonin: Oh, I will retract that any moment. discussions of intention movement or what have you. All of these con­
Rioch: I think the large striatum in the bird is probably secondary to cepts developed essentially out of the interpretation of certain kinds of
its large superior colliculus. naturalistic phenomena and out of naturalistic observation. Whenever
von Bonin: I would take exception to that. we begin to have some beginnings or glimmerings of an attempt to
Klüver: I could rephrase the question. examine experimentally the meaning of these behaviors in birds, I think
Rioch: I should like to toss something in here that may be unrelated. that we begin to find that more primitive and less humanistic or an­
We have seen cats and dogs go through running motions in sleep, and thropomorphic or psychoanalytic analogies or interpretations are in
we say they are ""dreaming." Well, with complete transection of the order. I am thinking, for example, of a striking study that appeared
brain at the level of the superior colliculus, cats and dogs make the just a few days ago, or came to my attention a few days ago, a study
identical movements in the identical way; so if those are dreaming by Ramsey, in the course of which he raised some of his birds in an
movements, then dreaming is somewhere around the level of the incubator in which there was a box. Usually, in naturalistic situations,
cerebellum. when the chicks are startled in certain ways, they run to the mother.
McCulloch: Or below. Under these conditions, when they were startled, they ran to the box
Hutchinson: There are cases, I think, in insects that are very similar and huddled in the neighborhood of the box—or a shoe, or in the
to Armstrong's displacement activity where you have presumably a neighborhood of a number of other objects that had been systematically
whole central nervous system which is of the order of magnitude of a introduced into the general home-territory area. What they are re­
very small part of the cord. The only difficulty there is that they appear sponding to is not necessarily what we would tend to respond to, and
to be phylogenetically developed and not things that are alternative may be—and not only may be, but I would submit is—on a very much
for actual behavior within the species. One example I can think of simpler basis, and is only analogically and not homologically or homo-
occurs among the Empid flies (Empis, hilara), where the male normally logously related to the phenomena under consideration.
catches another insect and wraps it up in silk and hands it to the female. Second, I should like to re-emphasize and perhaps extend the point
112 Cybernetics Communication Between Sane and Insane 113

that Dr. Hutchinson made. In looking at these behaviors, whether then the bird or the organism under consideration is not responding to
they be some of the behavior mimicry in birds, in the Australian bower another bird as a bird; but it is so organized and has a nervous system
birds or in any of a number of other bird groupings, we find that so organized that certain activities are produced when it is stimulated
these behaviors represent phylogenetically or evolutionarily determined by any of a number of congeries of external energy organization or of
mechanisms of responding, which have certain positive significances stimulation, which is not true for higher organisms or organisms higher
and are phylogenetically determined rather than ontogenetically de­ in the scale. Thus, two balls are not identical with a head for me, nor
termined. I think that when we begin to develop these analogies, we for you, nor for many other organisms. I think that the study of what
begin to fall into the classical error of an ontogenetic-phylogenetic-re- types of equivalences exist gives us an understanding of the meaning,
capitulation type of theory, and I don't think the mechanisms involved in a sense, of the action of the bird.
in the ontogenetic theory of symbolism at the level of the higher mam­ Kubie: I am going to argue about that. We have moved so far from
mal are necessarily duplicated in the process of the phylogenetic de­ where we were that we are no longer talking about communication at
velopment of certain of these displacement or intention movements or all. There is a reason for this. We do not mind talking about behavior,
any of a number of other phenomena, such as display phenomena, at so long as we stay away from communication for the simple reason that
the level of birds. we cannot talk about communication without facing up to our most
Hutchinson: The whole question about the box and the incubator difficult problem; namely, the problem of symbolic function. Com­
seems to me to be entirely irrelevant. Lorenz, for instance, has a case munication depends upon the symbolic function, and in turn the sym­
in which a parakeet copulated with a ball poised in space some centi­ bolic function is the essential hallmark of man. It is what differentiates
meters from a board. That was undoubtedly some sort of sexual act. It man, sick and well, from all other animals. Therefore, we cannot con­
was not a purely arbitrary affair that happened to be connected with tinue to evade its challenge forever. I purposely have not discussed
a ball, although it might be denied that it had anything to do with re­ the data from lower animals, although I am quite willing to admit that
production. I think it would be extremely silly to insist that it had some of that data are more empirical and clearer than some human
nothing whatever to do with sexual behavior and just happened to be data. Nevertheless, I keep to material from human beings. Why ? Be­
something that occurred in that particular experiment. It seems to me cause man is the only animal with whom we can communicate sym­
that the only problem is: How does the chick recognize the kind of bolically, and who can communicate with us on that level. Therefore,
thing that normally is a function fulfilled by the mother ? And if it is a it is only in man that we can even attempt to estimate in some meas­
box that is big enough and just about the right shape, then we are ure what the symbolic process actually does and how it can become
justified in concluding that from the point of view of the chick the box distorted.
is the mother—just as a chick will react to two balls, provided one, if I Not all symbolic functions are identical. They range themselves in
remember rightly, is not more than one-third nor less than one-half a spectrum. At one extreme are those symbolic processes in which
the diameter of the other. It is merely a question of defining what the what is represented is known in its entirety to the individual who is
sign stimuli are (to use Tinbergen's term), and it merely obscures the communicating. At the other extreme are the symbolic functions in
issues to argue whether it is anthropomorphic to call it the mother bird which what is represented is wholly unknown to the individual who is
or a substitute for the mother bird, or a box. They are all identically communicating. This will vary and fluctuate with every individual,
equal in the language. and from moment to moment. But if we conveniently delude ourselves
Birch: I would disagree with that entirely. I think that what con­ by talking about communication only among lower animals, we turn
fuses the issue is Tinbergen's and Lorenz's tendency to ignore that there our backs on that aspect of the problem which is essentially human.
is a difference between the head and two balls, and that although cer­ We cannot make progress by any such artifice (24).
tain features of stimulation in the environment may release certain be­ Rioch: I disagree with that. Dr. Kubie, for the reason that I think
haviors in the bird, or in other organisms, I think that our understand­ it is only when we get down to simple situations where we can begin
ing of the level of psychological function of these organisms is de­ to get some relatively complete observations that we can begin to work
termined by the kinds of things which are equivalent to the head or to out this problem.
the so-called "mother organism" for the organism under study. If these Kubie: But the lower animal is not a simple situation. Actually, it
are merely signs, stimuli, or releasers, or anything you wish to call them. is in certain ways far more complex to analyze, and far less accessible
114 Cybernetics Communication Between Sane and Insane 115
to observation, precisely because the lower animal's lack of symbolic Bigelow: I think it bears upon the question of whether communica­
speech deprives him of the ability to communicate to us his inner ex­ tion can exist between a human being who is asleep and his outside sur­
periences, his thoughts, purposes, anticipations, memories, and feelings. roundings, does it not?
Therefore, whatever of such experiences he may have on a conscious Teuber: Is it the problem of communication or the problem of sleep ?
symbolic level is inaccessible to us. von Bonin: Exactly what is that experiment ?
Bigelow: May I make a point which I think comes back to the first Bigelow: I am merely exhibiting what I think is a valid example of
round of the speaker's discussion.? I think you were speaking. Dr. communication which exists in the average person, let us say, in that
Kubie, of communication, of whether or not it exists during sleep. state which we call sleep.
There are some very primitive forms of this that one can point to, Teuber: So sleep doesn't have graduations?
which you didn't do, but I am sure everybody knows about them. I Mead: But if you have a low-voiced call instead of a footstep, we
should like to reiterate one of them. I think it is fairly easy to set up avoid the problem of whether it is communication or not. A mother
an experiment in which you show that a person can do operations will wake to her own child's voice, though it is much lower and a long
which are fairly subtle, of the following sort: You put a person in a way off. It can fit in just as well.
room with the Third Avenue Elevated outside his room, and after von Bonin: May I ask exactly how that cat reacts ? If there is con­
a week or so, he gets so that he sleeps perfectly well when a train goes stant noise, does it wake up then; that is, to the slight rustling of a
by. You can put a noise-meter in the room and measure the DB level mouse or something of the sort ?
of noise. You can actually change the schedule of the trains so that Rioch: No.
they are relatively aperiodic, so that the person doesn't really learn von Bonin: Does it react merely to low-pitched noises, not very
to block out his hearing sense during the intervals, and the person will loud noises ?
apparently behave like a normal human being and get his sleep and Rioch: It will give no reaction at all to the people cleaning its cage,
go to work and live perfectly all right. but if you go in while they are cleaning the cage, very quietly, the cat
You can also introduce a noise which is a soft footstep in the room, will immediately prick up its ears and turn its head toward you.
at an enormously lower level, and yet, very often, that person will Bigelow: Is this a question of frequency response of the animal ?
wake up instantly, or almost instantly, upon this cue. That certainly Rioch: You have a frequency spectrum in the room that is far be­
calls for some sort of interpretation process which is certainly, or yond that. You don't get this response to low frequencies.
probably, not of a high intellectual center of the brain; but it certainly Bigelow: Yes, but it is well known that cats have a peak in the fre­
involves something like acoustical triangulation and the recognition of quency response which is in the high region; I mean, that can be easily
some coded message which is not a part of the environment. demonstrated.
I think you can point to many examples like this, in which a person Rioch: You can also do it with a whistle of a thousand per second.
who is in that state which we commonly call sleep, actually is capable Bigelow: But this may constitute a lot more energy input due to
of receiving messages in a very distinct and numerical fashion, in es­ selectivity of the frequency response.
sentially the same way that he can when he is awake. Rioch: Except that there are plenty of those frequencies in the room
Rioch: I can show the same phenomenon in the decorticated cat. at the time.
Bigelow: In a decorticated cat ? Bigelow: But the question is whether or not this sound suddenly
Rioch: Yes. produces a codable message in the animal, independent of its attitude.
Bigelow: Well, I am not trying to raise the question as to exactly That is the question I was trying to raise. The Third Avenue Elevated
where in the nervous system it exists. Is it the opinion of the group that does not wake the man, but a sound that requires a certain amount of
the fact that you can exhibit this in a decorticated cat removes all in­ decoding to know that it is abnormal does wake him up.
terest in this behavior of a human being? Rioch: I will add this to it: You can have a whistle going continu­
McCulloch: Certainly not. ously, and if you add a new whistle to it, the new whistle will evoke a
Bigelow: That seemed to me to be the implication, and I don't see response that the other whistle did not.
why this particular point has a more significant aspect. McCulloch: May I ask one question before we go back, just as a
Rioch: I don't think it bears upon the problem of sleep. matter of fact? Does it make any sense to talk of a midbrain cat,
116 Cybernetics Communication Between Satie and Insane 117
everything above the midbrain gone, sleeping and waking ? because I don't know what to do with them, but once in a while they
Rioch: I think so. I think you've got to make a section at the lower come up anyhow, and the impression I have is this: the severe schi­
border of the midbrain that is behind the inferior colliculus and down zophrenic is very much like the normal in respect to dreams; namely,
to the rostral margin of the pons to get rid of all the rhythmical sleep- in both, the dream life and the waking life are very close replicas. In
wakefulness change in threshold. Now, I am not even sure of that. between there is a wide range, including neurotics in whom there is
It may have to be a lower section. But I am sure you have to cut the this very picturesque type of dreaming which is like conversations in
midbrain out to get rid of it. On the other hand, when the transection is bars and is like the way kids will play when they are not observed. In
made from the anterior margin of the superior colliculus to the pos­ this dreaming there is what may be called conceptual-clang associations
terior margin of the mamillary bodies, the result is a decrease of the (not clang associations referring to sound).
behavior ordinarily interpreted as sleep. The onset is sudden, and the Ritts: Excuse me, but what is a conceptual-clang association.?
threshold for bringing the animal back to what appears to be com­ Rioch: I call conceptual-clang associations phenomena such as using
plete alertness is very low. With the hypothalamus and the central a long or pointed object to refer to a penis in communication. This
nuclei of the thalamus intact, then, to arouse the animal, you sometimes conceptual-clang communication is, I think, universal. Hindu religious
have to shake it; but with the midbrain animal, just a shift in the writings, particularly in describing the goddess Kali as the productive
position of the hind leg is enough to bring it into a very changed state. mother earth and as the force of destruction,'present practically every
McCulloch: Let's go back to a discussion of schizophrenics because psychoanalytical symbol in much the same way as these symbols are in­
the ostensive definitions of schizophrenic terms are patent to anyone terpreted in our culture. This symbolic language is very much limited
who works long enough with a given schizophrenic. to the personal objective view of the child, that is, to anatomical struc­
Rioch: I should like to tell you about a patient I saw briefly in con­ tures of the child and his immediate associates. The communication is
sultation because it illustrates the kind of thing I run into. This patient in a verbal form which presupposes a certain response from the en­
was a boy who had cut off the end of his middle finger and sent it to a vironment in terms of understanding and an answer in a similar form.
friend. He told one observer he was doing an experiment in pain, that This is not wliat one gets in optimal normal performance nor in schi­
he was proving one could disjoint a finger without conscious sensation zophrenic performance.
of pain, which was true. He told another observer that this was an Hutchinson: May I inquire why one doesn't get it in optimal per­
act of getting rid of his mother, getting her out of him. formance ?
Kubie: What was that.? Rioch: Because you see through it and you are interested in some­
Rioch: An act of getting rid of his mother out of him, one of the thing else.
classical schizophrenic forms of communication. I did not, however, Kubie: I can't make out at the moment whether you agree or dis­
find out the other probable significant life experiences which this act of agree with me or want to ask a question.
self-mutilation also symbolized, nor did I find out the implications of Rioch: The question I wanted to ask you was whether this impres­
the formulations which he offered. The practical problem presented is sion I have is correct—whether you have any definite observations on
that of determining the relevant interpersonal experiences and problems schisophrenic dream states and (optimal) normal dream states as con­
symbolized in such a condensed way by schizophrenic patients. trasted with the neurotic form of dream content.
In contrast, there is the form of communication with neurotic patients Kubie: I cannot answer that. A lot of people talk glibly about
and in neurotic dreams. The place where you hear this form of com­ schizophrenic dreams. I have many reservations about them. I have
munication probably in purest culture is in certain moderately low- never been convinced that I could recognize a dream as being, in some
grade bars when the boys are getting a little bit tight. At such times special way, schizophrenic.
everything round in shape is either a breast or a buttock. Everything Rioch: "Those are what I would class with the neurotic dreams. Some
pointed is a penis, and everything that has any depression in it repre­ people claim they can recognize schizophrenic features in a neurotic
sents the female genitalia. The game is to see how picturesquely one dream. I never could make out what they were referring to. I am talk­
can deal with such references and be understood. ing about the kind of dream a severe schizophrenic reports.
Now, there is one thing I would like to ask you. Dr. Kubie, about Kubie: Well, the other implications of what you say really carry
dreams. I warn all the patients I see against giving me any dreams. us far afield from my topic. Nevertheless, it may be worth while to
118 Cybernetics Communication Between Sane and Insane 119
say that in a severe schizophrenic it may be difficult to know whether stages of adolescence, making little model gardens, taking flowers and
he is reporting a dreamed experience or a waking fantasy, because his shells and building lovely little structures with a nice room with a red
waking fantasies are almost as completely overlaid with obscure sym­ carpet. Then one girl takes a leaf and puts it in the doorway so you
bolic implications, connotation, and language as is his dream life; and can't see the red carpet. Then there is a discussion whether she is
in such patients the borderline between waking and sleeping, between going to have the driveway come up to the house or whether motor
experience and fantasy is obscure. Therefore, here again I cannot draw cars are to be allowed in, or whether she should just have stepping-
any hard and fast lines. stones up to the house and the garage down there, and so forth. You
Brosin: I wonder if it would help you at this time, in defining com­ can look at that in formal intellectual terms if you are used to reading
munication in many different states, such as the sleeping state, hypnosis, that kind of material, for instance, in terms of ordinary child analysis,
and the different clinical conditions, to use as a springboard the con­ and see the stages of adolescence of the different girls and what is said
cept of interpretations of many types. These interpretations may vary by that picture. People can do it. Ericson, for instance, could take
from the more rigidly logical to those which we call intuitive, that is, photographs of such play and say: That girl is probably a year before
dependent on nonverbalized learning or past experience. Can interpre­ menarche, this girl is just at menarche, and this one is a year after it.
tation as a method be used to examine these latter phenomena in a Or, after one of these little girls has carefully made a house, and con­
manner different from a method now being more sharply defined, that sidered whether she would let a road come up to the house or keep the
is, the examination of the interpersonal relationship between the partic­ cars down in the corner, you can watch a boy who is a couple of years
ular organism in question and its social matrix ? younger come over and dig a well in the back yard of her house, very
That may not be very clear, but it may be the difference between the carefully. Then she says, "It isn't a well at all; it's an incinerator." At
interpretive methods utilizing models, such as interpreting dreams by that point, you are having communication betw een the two. The boy
various patterns or by looking for the universals in dreams, in hypnotic is certainly saying something very definite to the girl : they may not be
states, in myths, or in symbolisms of various cultures as opposed to— aware of what they are saying, but you can measure it and check it
perhaps they are overlapping—the participation in the direct experi­ against many things. They are using a language which one can handle
ences of a person communicating with his fellows, whether these com­ intellectually, also, so that I don't see that these statements as to usage
munications be internalized or actually outside. of a conceptual model which is mythology or religion or symbolism or
Kubie: I don't get you, Dr. Brosin. poetry, in contrast to the relationships between the therapist and the
Brosin: Well, in its crudest form, I suppose, this is the problem of patient, are exclusive at all. If we hadn't overdone the rational ap­
intellectual interpretations versus emotional relationships. Let's take proach to life, we wouldn't have to make that distinction.
the patient-doctor relationship which has so many ramifications involv­ Kubie: I will go back a bit. I fear that we are losing our ability to
ing problems of communication. I was hoping that we could discuss the communicate with one another here, and are bogging down. I suspect
basis for the types of communication, verbal and nonverbal, intellectual that this is owing in part to the nature of the data, and partly to the fact
and emotional—to use overly simplified terms—to describe these com­ that we are trying to bridge a wide gap in experience. Those of us who
plex acts. As the problem of what language is came up from time to have had clinical experience have no reluctance in considering the high­
time today, we seemed to have multiple definitions available to meet ly symbolic data which, are to be found at one end of this bridge, where­
our needs. Various members referred to these multiple levels of activity, as those others who lack clinical familiarity do not like to deal with
and we know without dictionary definitions that there are many levels data of this kind, because they cannot be corralled in a reproducible
of communication. One is to intellectualize the symbols used, whether experimental situation, as indeed all of us would like to do. There are
they are letters or pictures, and to handle them at various levels, and methods by which this can be done ultimately, and in the not too distant
the other is to examine much more closely the direct relationship of the future, for that matter. But it has not been done as yet; and it never
organism in question to the persons with which it is most concerned, will be done if we continue resolutely to deny the existence of the data
and the emotional relations between them. themselves, merely because they make us uncomfortable by being elu­
Mead: I can give you a model for that, I think. It seems to me sive and hard to pin down.
that the distinction is, again, a cultural one, because we have learned Therefore I should like to return to an elementary and simple restate­
how to intellectualize these things. You take three girls in different ment of my own fundamental working hypothesis (25, 26). The sym-
 Cybernetics Communication Between Sane and Insane 121
120
bolic function characterizes the human being as homo sapiens, setting cause we have been over it several times;, but the discussion makes
him apart from all other animals. By this I do not mean that no other it clear that it is necessary to clarify this before discussing what it means
animal is capable of any symbolic functions, but that between the sym­ for the problems of communication. Actually, it means that every sym­
bolic potential of man and that of lower animals there is an enormous bol we use, whether it is a word or a gesture or an expression or a tool,
qualitative and quantitative difference. Indeed, if it were not for our is multivalent; and that some of the determinants of the relationship
highly developed symbolic function, none of the things which we are of the determinant to the symbol are known and some unknown. Con­
doing here today, and none of the other processes which characterize sequently, if we had an instrument for measuring the relative roles of
human cultural development, would be possible. The human symbolic conscious and unconscious determinants it would show that when I use
function involves the capacity to make abstractions from discrete indi­ such a word as ""chair," on some occasion, 90 per cent of the determi­
vidual experiences, to condense these in various ways, and then to nants of my use of the word would be matters of which I was aware,
represent these condensations of abstractions with gestures, with facial and that 10 per cent would be something I was unaware of. Whereas,
expressions, with the spoken word, and with the written word, that is, on another occasion when I used the word ""chair," 90 per cent of the
through the whole paraphernalia of symbolic language by which we determinants might be unknown to me and 10 per cent of the deter­
communicate to one another our thoughts, our feelings, our memories minants might be known. Therefore, my symbol sometimes represents
of past experience, our hopes and our plans. the known to me, and at other times the unknown; but it is always a
Let me backtrack and say that without this, the highest level of psy­ mixture. If this is true, then in language, in speech, in thought, in
chological experience to which the human being could attain would be poetry, in anything that we are doing, we are dealing with a symbolic
a pallid, sensory reverberation, a sensory afterimage of previous expe­ process in which the relative roles of "conscious" and ""unconscious"
riences, a simple wish-fulfilling dream (such as a little child might determinants, that is, of known and unknown determinants, are con­
dream), or something like a phantom limb. I realize fully that these stantly shifting. That, I think, is fundamental and basic.
words themselves are symbols and therefore subject to distortion, but A few other relevant points concern the relationships which can be
I still think that all of us know what I am talking about and what I established between human beings- in a certain state. It is hard to char­
mean when I say that because of our capacity for abstracting from ex­ acterize this state. Operationally, however, it is a state which results
from a certain type of procedure. We can call it hypnosis, or an in­
perience, plus the ability to represent those abstractions symbolically,
duced, controlled, dissociated state, a communicative state of semisleep,
everything that characterizes human culture becomes possible.
Early in life, however, something of great importance happens to the a sleepoid state, a communicative hypnoidal state—anything you want.
development of the symbolic representation of abstraction. Later in What is important is not the name but the fact that there is such a
state, and that in this state it is possible for one of the subjects to be
life the same thing can occur in the process of falling asleep, in the
more aware of the unconscious determinants of the actions of another
process of waking, in the dream itself, and in the hypnagogic revery to
which we are subject from early childhood. In these states symbols be­ subject (or of himself) than is posible in the fully alert, fully awake
come primarily visual rather than verbal, probably because they had state. Furthermore, something of the same kind may be possible be­
originally been predominantly sensory and visual before they were tween and among psychotics. These interesting points have never been
verbal. These symbols represent more than one thing, however. They studied fully: I mean the ability of one psychotic to understand
are multivalent. Furthermore, under certain circumstances the symbolic another, or of one hypnotic subject to understand another. We know
representative becomes dissociated from that which it represents; and that there are some psychotics who understand themselves; but we
psychopathology starts vi^ith this dichotomy, that is, with the dissocia­ have not investigated beyond this the important question of whether
tion between the symbol and what it represents. It is my working one psychotic can translate another's psychotic products.
hypothesis that the neurotic potential out of which everything that is That is as much of a summary as I can pull together in a hurry, with­
neurotic in human life and human behavior develops arises precisely out trying to tie up every loose string of my own perplexity about this
here. This is universal, because we have not yet discovered how to problem.
avoid or limit this fundamental and basic dichotomy in early life. May I give examples of two other important points? The capacity
What does that mean for communication Ì I did not expect to have for communication is linked closely to the developmental processes of
to reaffirm my fundamental thesis about the neurotic process here, be- the ego, that is, to the ontogenetic process. This can be proved and has
122 Cyhernetícs Commumcation Between Sane and Insane 123
\
been proved in experiments in which regressions to earlier age periods ing about my own feelings on either of these issues. In the first place,
are induced. For instance, an Austrian who learned English when he I do not know whether there are universal symbols. I have always had
was twelve or thirteen was regressed to the age period before he knew a bias against them. But as the years have gone on, I have run into
English, and could no longer speak English. At the same time, he enough disturbing examples of identical translations of elements in
would make figure drawings like those of a child of eight or nine, dreams in subjects of extremely varied ages and cultural backgrounds
write like a child of eight or nine, spell like a child of eight or nine, to make me hesitate to dismiss the idea. Certainly there are some areas
and react to the Rorschach test, to the Szondi, to the Bender Gestalt, of symbolic implication which seem to be, if not universal, at least
as though all of his symbolic processes had returned together to the age broadly distributed in our own cultures and in many different cultures.
period to which he had been regressed (10). I do not think we have any right to speak with confidence of universal
Another interesting experiment, done during the war, illustrates sim­ symbols, any more than I think we have any right to speak with con­
ilar things. (It was restricted, and therefore has never been published.) fidence of inherited symbols; but I must say again that these are two
There was an air crash in which there were no survivors. For various areas that have not been fully investigated (8).
reasons, it was desirable to find out what had happened. As the plane McCulloch: You mean, common to all men, regardless of their
was falling, there had been a brief interchange between the pilot and culture or background, something of that sort Ì
the radio operators at several fields. The pilot's message was an ex­ Kuhie: That is right. I find myself with a bias against such a no­
plosive, excited rush of unfamiliar words, not the usual stereotyped com­ tion, yet unable comfortably to dismiss it as impossible. There may
munication as to direction of wind, altitude, and so on. This was be­ be some types of human experience which are so fundamental that they
fore automatic recordings of such communications were made. After­ are universally represented. For instance, one of the most fundamental
ward, when questioned, every radio operator in every field said, "I don't of all trauma is the business of being a small child in a world of grown­
remember a word he said." We were asked whether we could disen­ ups. This may be represented not only in such fables and myths as
tangle out of these quickly obliterated memories of barely perceived, Jack, the Giant Killer, but also in many other ways.
unorganized, unanalyzed auditory impressions something that would Von Foerster: If there were such universal symbols, could we ever
make it possible to reconstruct what had happened Ì detect them ? Because even if we come up to that age where we are able
I am not sure whether or not we succeeded. The engineers of the to express these universal symbols, we have already merged with the
particular outfit that asked me to make the study and who worked cultural pattern in which we have to think. Therefore, I don't know—
with me were sure it had been reconstructed. All the auditory symbols and I would only raise that question—of any way in which we could
were so charged with emotion that even under hypnosis and with drugs, even detect them, if there are such things.
the radio operators could not talk about it. Therefore I asked them to Kubie: Fair enough.
write about it. The words came along fairly clearly. Then came the McCulloch: May I put the question another way ? Is it that the same
word ""de-icing," and the curve on which they wrote was the fall­ symbol is used by everyone, or is it that the same state is symbolized
ing curve of the plane [demonstrating on blackboard]. It seemed as by everyone? It is a very simple question, first, in my mind.
though they were translating into an arm movement their unconscious Kubie: Dr. McCulloch, I should say—
knowledge of what they believed had happened; namely, that some­ McCulloch: Or is it that there is a lock of the two?
body had tampered with the mechanism to warm the carburetor. Kubie: It is at least conceivable that both may occur; that is, in dif­
McCulloch: I have two fag ends I wish you would clean up for me. ferent individuals a state may be symbolized by one and sometimes by
First, you several times referred to the universal symbols, or the uni­ differing symbols. Your second question arises out of a misunderstand­
versality of the symbols that occur in dreams, and I am not sure what ing. I said specifically that I do not know the relationship of insight to
the terms ""universal" or ""universality" mean in that connection. therapy. I wish I did. One of the most puzzling problems in psycho­
The second thing is that you have twice spoken of insight and therapy is why insight sometimes seems to be powerless and at other
therapy as though they were either necessarily correlative or as though times has extraordinary therapeutic leverage. By ""insight," I mean that
one was necessarily consequent to the other. I did not quite get the a person who has not previously understood the meaning of his own
connection. Would you clear up those two points for me.? behavior, thoughts, or feelings acquires a full knowledge of them,
Kuhie: Gladly, because I would not like to leave any misunderstand- their meaning, their roots, their ontogenetic history in his own life. It
124 Cybernetics Communication Between Sane and Insane 125
is characteristic of many schizophrenics (although not all, nor is this make themselves understandable or establish for themselves a place in
confined to the schizophrenic) that the acquisition of insight seems to the very complicated pattern of the world in which otherwise they
have little therapeutic effect. would not know how to fit.
Pitts: What do you mean by the ""feeling of understanding" of the Bigelow: Isn't self-communication a contradiction in terms ?
subject ? Von Foerster: No, I don't believe so.
Kubie: I mean that the patient has not merely an intellectual realiza­ Rioch: The ""self" being a symbolic social structure and not some­
tion that something may be true, but an awareness which involves some thing inherent in the organism, it is not a contradiction in terms. What
actual memory of the experiences which underly his behavior, plus is actually going on is that the person, using some kind of idea of the
enough emotional participation in that awareness so that in some meas­ ""self," experiments with that until he finds somebody who can under­
ure it is as though the experiences themselves were being relived. Such stand what he is trying to get across. As soon as he finds somebody who
insight is a combination of intellectual understanding, plus memory, can understand what he is trying to get across and who answers ap­
infused with appropriate feelings. propriately, the ""self" mechanism changes and he proceeds to orient to
Rioch: I wish to make a point with regard to the problem of these people in a much more adequate and elaborate form.
universal symbols. I think to a large extent similar symbols are used Bigelow: Does the mechanism display any external manifestations
in many cultures. That doesn't mean that the thing that is being said when this is taking place?
by them is similar at all. Rioch: The one that came to mind did.
McCulloch: Exactly. Bigelow: Then is it really external communication, even at that
Rioch: Especially in that type of communication in which these level ?
symbols are used, the actual communication as to what the situation is— Rioch: Oh, yes.
in terms of what the interpersonal relations have been and what they Bigelow: Then how is it self-communication?
are going to be—^has to be determined not in terms of the content of Rioch: These particular formulations are done very privately, and
the symbols, but by the sequence of the real behavior occurring at the you have to demonstrate with the person that there is an acceptance of
time. Infrequently the content of the symbols due to their particular the situation before he will ever let you in on it. It is like the schi­
cultural significance gives an indication of the predictive value. The zophrenic who goes fifty years before he tells anybody about his hal­
actual communication can be determined only by the sequence of events, lucinations.
not by the general or personal significance of the symbolic content. I Kubie: Mr. Chairman, because we are hearing so much nihilism, I
think that is where, in so far as we differ on this question, the main am going to do what I ordinarily do not do; that is, give an actual
point lies on which we differ. clinical report. It is important to realize the sort of therapy that can
The problem of therapy is so complex that I don't think it can be occur. As I have listened here for the last hour or so, this has kept
used as data. I had one patient, for example, who was cured by Chris­ coming to mind; and I think I will ""spill" it.
tian Science and then broke down again. He described in great detail I have in mind the critical moment, the turning- point in the cure
the ""cure," the break, and the results, and the ""cure" had all the char­ (and I say ""cure" advisedly) of an active homosexual who is homo­
acteristics of insight that I have had in patients whom I have '"cured." sexual no longer. He had been in intensive analytic treatment for only
In one sense, at least, that which is called insight is a matter of having a few months. A couple of years earlier I had sent him to Stockbridge
come to a situation of communicating with another person in a reason­ for several weeks. Before his intensive analytical therapy began, he had
ably reliable, predictive manner. This gives so much relief that those been able to establish himself in a job and had begun to make a few
defenses against anxiety that were used before are no longer necessary. friends of both sexes. He would oscillate between homosexual pick­
""Insight" in this case refers to developing a common content for com­ ups (with all the degradation and risks which that involved) and try­
munication. Whether we ever get anything more than that, I don't ing to establish sound human relationships with young men and women.
know. I think we may get different degrees of it. A moment came in our work when I felt that it was safe to impose an
Von Foerster: Have you ever observed symbols of self-communica­ absolute taboo on any further homosexual contacts which did not grow
tion ? I have the strong suspicion that some people create certain sym­ out of pre-existing sounder human relationships. This blocked his
bols which only they understand, for instance, in the sense that they blind wandering around Central Park, the making of blind pickups, and
126 Cybernetics Communication Between Sane and Insane 127
the danger of his landing himself in extremely unsavory circumstances. Kubie: Oh, yes, constantly. My patient had been able to describe
He accepted this willingly, but he came in with a dream. every last detail of his sexual experiences with men; but until this
This dream was so interesting that I took it down verbatim, and dream opened the door, not even a fantasy about women had been able
have used it in teaching on several occasions since. The dream follows; to enter into his conscious thoughts and feelings.
"I was in a red something or other. I can't remember what it was. Bigelow: However, perhaps his mother was something more sacred
Then I was in a room with some young women and, somehow, I was than all of these.
accepted by the young women more than I was accepted by the young Savage: You would have to be rather elaborate. Dr. Bigelow, not
men; because there were also some young men there. It is funny. to be able to remember this dream, if the dream was explicit. If the
There's a lot more to the dream, but I can't remember anything else dream was a conscious device to communicate with Kubie about some­
about the dream. I am puzzled about it, and bothered about it." Then thing which he didn't want to do, certainly, it was really very elaborate
he talked about a lot of other things, going back to some of the material for him not to remember the dream for twenty minutes.
of the previous day, and to the experiences of the previous evening. von Bonin: Isn't it the Freudian sense that we are talking about
As he did so, it became clear that the dream was a transparent and now.?
literal extension of that evening's events. Then he came back to the Savage: Dr. Bigelow is talking about a very naïve sense in which I
dream and repeated it, puzzled by the gaps; but there was no change call up a doctor and say, "Doctor, a friend of mine has athlete's foot;
in his recital. Nothing was omitted, nothing added, nothing changed. how do you cure it?" and it is my athlete's foot, of course. It is the
I said to him; "You realize, I am sure, that it is quite usual among matter of trying, perfectly directly and explicitly, to trick the doctor,
young girls"—I started with girls purposely—"at some point in their and I do think that is really ruled out fairly well by the evidence here.
lives to fantasy that they might be young boys, so that they would be Mead: Yes. Isn't it worth picking up the last remark Dr. Kubie
free to be with their father or their big brother or a boy cousin in the made when he said that this was an everyday experience to him but
boy's locker room when the boys are dressing and undressing or taking not to some of us ? It seems to me that one of the difficulties we often
a bath or going to the toilet, just to satisfy their natural curiosity. And encounter in this group is between the natural scientists and the psy­
it is equally natural, of course, for young boys to imagine that they are chologically trained. To go back to your word "chair," which in a
young girls for the same reason." I had gone only that far, when he given context is 90 per cent known or conscious and 10 per cent uncon­
broke in; "My God, I remember the rest of the dream! My mother scious. The natural scientists have been trained in this group to pay at­
was there; and she was naked except for some blue lace panties. She tention to the part that is known and to leave out the part that is un­
came up to me and I looked at her and she had a penis; and then sud­ known, to keep it out and rule it out and make a set of rules that make
denly she walked around behind me, from right to left, and when she people more or less sit tight and use the same words and play a game
came around in front of me she was just the way she really is." Then very rigorously.
came a series of memories about his mother, about physical contacts Bigelow: I think we are perfectly willing to conjecture about this
with her in his very early years, about his feeling about her body, and part, but we like to say it is conjecture.
especially about a portrait that had been painted of her, a small copy of Mead: But you are still acting in the conscious realm when you say
which he always carried in his wallet. In this portrait she wears a blue that. When you say it is conjecture, you still want to act according to
lace brassiere, just the color of the panties in the dream; and in the one set of rules. Now, the psychiatrists listen to the other set. They are
portrait the bra is cut so low that one breast was almost completely trained to listen to the other set. Nobody seems to suggest that you
exposed. This started the unraveling of his intricate relationship to his might do both at once, that there isn't any reason in the world, except
mother. Such experiences are our daily diet, but not yours. Never­ the habits of our present discipline in the scientific world, why you can't
theless, you cannot leave this out of account in your approach to this use visual images, kinesthetic images, olfactory images simultaneously.
problem. You can think a great deal faster if you do, as well as using words—
Bigelow: Dr. Kubie, is there a certain element of feeling that it is all which is what probably everybody here does when he is being creative.
right to have such dreams but not all right to have such conscious However, the rules of the game have been, in the natural sciences, one
thoughts, and therefore one actually gets some statements which are set, and in psychiatry a different set, not really different, because they
in the form of dreams but may actually not be such ? split it off too—•
128 Cybernetics Communication Between Sane and Insane 129
Rioch: You are correct, but the rules are not necessarily different. one works on them. But what one tries to do is to get to the point
Why should we take psychiatry and confine it to the operational in­ where one is able to formulate precise questions which are susceptible
terpretation of symbols? Why not deal with events according to the of precise answers. Until we can achieve that, we don't feel that we
rules of the natural sciences ? Dr. Kubie gives a beautiful example of have gotten very far.
the sequence of events in getting into a relationship with this boy. (In­ With other topics it may be more difficult to focus on them in such
tuitively Dr. Kubie knew the time to say, "I, as a person, am going to a way that one is able to formulate precise questions and then to carry
give you support by giving you a rule of thumb.") He does that and out observations that can give precise answers. The only difference I
what happens ? Now comes a flood of memory material which appears can see is that, perhaps because the material with which we deal is
in a relationship which hasn't existed between this boy and anyone else simpler, we manage to select those problems that are amenable to ac­
before. You can study the sequence. When you are treating the schi­ curate analysis and then work on them. We would not be content to
zophrenic, if you don't pay attention to the sequence you are lost. The give an answer to the problem on which we are working when such
neurotic patient is perfectly willing to go along with you. He is so answer turns out to be far too vague, according to our own standards.
tickled at his capacity to see implications that when you find implica­ Rioch: There is another angle to it, and that is the capacity of a
tions that agree, he will go along with you and give up symptoms. But patient or any organism to arrive at a certain state, taking any one of
I think that the same principles hold, that we want to treat psychiatry many possible paths. If we are dealing with an experimental situation,
as a natural science. We have to deal with all the communication in as we don't mind repeating and repeating and testing one part and then
accurate a description of the real setting (which is the sequence of another. When you are dealing with a therapeutic situation, you are
events), and in as extensive a sequence of events, as we can get. A limited in terms of the patient getting what he "wants," because then
great part of the time we don't know just what the relevant sequence he stops. You can't check it again. A large part of the time, the thing
was. we did that we think worked was only some vague reflection of what
Kuhie: I agree with that completely, of course. That is why for the really happened.
last umpteen years I have been arguing that we should reproduce thera­ Rosenblueth: No, let me make this clear by returning to something
peutic sessions in their entirety, so that groups of experienced therapists that Dr. Kubie misunderstood in what I said originally, and which I
could sit down together to study them. Only in this way can we rule was not planning to go back to; but it becomes pertinent in view of
out the artefacts that are introduced by any one individual's isolated Dr. Mead's comment. Within physiology, or within other types of
observations. In the past, the lack of such records has set us apart from science, there are many things we cannot define. We point to them;
any other science that I know about. Every moment is ephemeral. Once we recognize them. Sleep is not only* a psychoanalytical or a psycho­
it passes, it can never be recaptured by the written word or even by the logical condition; it is also a physiological condition. When I said we
tape recorder alone. Too many things are going on at once. It is only did not know anything about sleep, I was really talking as a physiol­
in the last few years that we have had a mechanical means for reproduc­ ogist, but that does not change the comment I made on the distinction
ing the entire phenomena for subsequent study and analysis (25, 26). of certain phenomena when they occur during sleep and when we see
Naturally any such method of reproduction introduces its own arte­ the same phenomena while a person is awake. Granting we all know
facts, which must themselves be studied for their distortions. We will what we are talking about and that it is not necessary to define those
never meet your justified requirements for an objective scieritific process terms. Dr. Kubie's distinction was artificial because he did not give one
until we can reproduce analytical interviews so that we can sit down and single fact that pointed to qualitative differences between the two
study them over and over again, just as a dozen people can look at a phenomena, except for one word, which is the word "unconscious."
specimen under the microscope, or listen to a heart. You may say, "I was totally unconscious when these phenomena, these
Rosenblueth: To come back to Dr. Mead's comment. I don't think things occurred," or else the patient or the subject being examined
there is any real distinction to be made between the language of psy­ may report, "I was not aware." You may use these statements to ex­
chiatry and that of the natural sciences except in the rigor with which emplify the unconscious. But the term has an unfortunate connotation,
we want to do our thinking. I don't think the natural sciences are play­ which is the only reason I objected to it. If we consider the interpret
ing a special game. I don't think we are neglecting any data. There tative connotation that the word has, then it becomes objectionable.
are problems in physiology which are not formulated precisely, and yet Something which is not in the mind cannot be conscious, and I am
130 Cybernetics Communication Between Sane and Insane 131
using "mind" with the quotation marks that Professor Richards used, has ever attempted to penetrate selective memories and thus to re­
and also marks around "conscious," again with the understanding that integrate the dissociated components of human experience (as has been
we cannot define this term, but I think we know what I am talking done in the laboratory and under controlled conditions) cannot think
about. of human psychological processes as though only that small fragment
Bigelow: I should like to add one more remark. I don't think any­ is important of which we can be conscious. Indeed, that is the smallest
body in the natural sciences, as Dr. Mead calls them, would attempt to part of the continuous psychological flux which goes on within us at all
make a rational basis or even come anywhere near such an idea in con­ times. It is a regression to the Dark Ages to hear mature scientists still
nection with something like psychoanalysis. I don't think anybody here haggling over whether to acknowledge that there are levels of psycho­
could object to a description of the processes and the exploration of logical functioning which vary both in their degree of conscious aware­
their vague and ephemeral character. It is only that there is a certain ness and in their accessibility to simple direct introspection, and to
crossing of purpose when it comes to using terms which appear to have which one can penetrate only with special techniques of which the
specific meaning but which are vague to the group, and I think that is pioneer technique, and still the most important, is psychoanalysis.
what Dr. Rosenblueth was also saying. At this point I think it is per­ Heaven knows that it is a clumsy instrument; but it is still the best one
fectly reasonable to get down to brass tacks and say, "I mean to state we have. Therefore, if we are to communicate with one another, we
this pragmatically, my definition of unconscious,' and work with it, so must say at least that this is a fundamental working hypothesis in any
please let me continue." That is perfectly all right. But unless there is study of psychological function.
something specific to hang on to, or unless it is just a useful notion, Mead: Who said it wasn't.? Did anybody say it wasn't ?
loosely tied, which we choose to put up for lack of something better, Kubie: Yes, Rioch and Rosenblueth said it wasn't. They could not
one doesn't know where one stands, so one doesn't know whether to even talk about unconscious processes. They said if it is not conscious,
draw inferences from this or to see a picture. That is the state I find. it is not in the mind.
Kubie: By now we should have gone far beyond this hackneyed issue. Rosenblueth: Excuse me, but I objected to the term. You probably
It has been batted around for fifty years. It has been argued so often are all aware of Poincaré's theory of creative imagination in mathe­
and in so many places and in so many contexts, even here in our own matics. It is one instance in which I think that one of the most intelli­
conferences, that by now we should be able to take this working hy­ gent people who ever existed in the world went thoroughly haywire.
pothesis for granted, as we strive to understand one another. All psy­ It just doesn't make sense. It is not a theory; it doesn't explain any­
chological functions (such as communication) are stratified or layered; thing; it doesn't lead to any further experimentation, merely because
and there are variations both in the degree of awareness of these layers he got bogged down by the word "unconscious." He used it thinking
of psychological processes and in their accessibility to our own con­ he was saying something. He referred to processes going on without
scious self-examination. We must also include in our approach the the person being aware of them; they were unconscious. All the pos­
fact that there are two kinds of unaware psychological functions, which sible solutions were threshed out unconsciously and were ofl^ered to the
have been called by various names. William James spoke of the "fringe conscious only when they satisfied certain esthetic criteria. The sug­
of consciousness." Freud calls this the "preconscious" or the descrip­ gestion has no explanatory and no predictive value. Therefore I con­
tive "subconscious," as opposed to the more dynamic "unconscious" sider that it constitutes a very poor scientific hypothesis, and I think
processes which are walled off from our conscious introspection by that damage was done by the word "unconscious," which is interpreta­
forces which are actively resistant to our understanding of them. Such tive. I don't mean to object to the phenomena. They should be studied.
a term as "walled oif" is obviously an allegorical or figurative term to And certainly I don't want to quarrel with any one of Dr. Kubie's
describe the clinical fact that in order to penetrate to these processes, observations. If I say he uses a word merely to describe them, the use
and in order to reintegrate the various "levels" of psychological func­ is clear, but—
tion, one must overcome certain opposing forces. On the other hand Kubie: I would agree with the way Dr. Rosenblueth uses the word
there are no such barriers to conscious perceptions of our "precon­ also.
scious" or "fringe" material. Bigelow: I think you just did the defining a moment ago that we
Anyone who has any familiarity with pathological psychological would have liked earlier in your talk.
phenomena, who has ever encountered the processes of repression, who Kubie: I am sorry. I thought that because of our many previous
132 Cybernetics Communication Between Sane and Insane 133

discussions of this problem another systematic review of it would be who are used to dealing with another section, and we are not accus­
superfluous by now. tomed to using them both at once. Now, that is not saying that we
Rioch: I find that definition quite inadequate. There are a large are prejudiced. The scientific method is certainly a way of behaving
variety of phenomena one has to differentiate. and of being very rigorous and listening to everything.
Kubie: May I say one further word.? I would agree with you. I said Rioch: The scientific method is an attitude, not a way of thinking.
that simply to characterize one of the outstanding features. I am McCulloch: Gentlemen, we won't settle it here tonight. We are
wholly uninterested in definitions. adjourned.
Rioch: My objection to "unconscious" is that we talk about layers
and we have no idea what sort of layer it is. Is it a longitudinal layer.?
Is it a horizontal layer? Is it an interlacing layer? I don't think that
the methods that have been used to investigate the so-called "uncon­
scious" have been adequate. There are phenomena in the schizophrenic
which are nonverbalizable, which are totally different from phenomena
in the neurotic. Again, if you go into animal experimentation, it is
possible to formulate a set of categories of behavior which one can
also apply to the human. Then one can define central nervous system
activities which get down to simpler levels than those we have been
considering in this, and also define more complex forms. I think what
we need is some kind of an attempt at an operational definition, so that
the operation can be described so that somebody else can do it—
McCulloch: Hear, hear!
Kubie: This is precisely what I have been attempting to give you; an
operational definition in terms of a procedure to meet certain concrete
clinical phenomena. In doing so, I use a terminology which describes
in recognizable terms both the clinical human challenge and the pro­
cedure accurately enough to make it possible for others to duplicate
the experience. This provides a constant frame of reference for further
investigations of the subject by other observers, and this is precisely
what I have offered you here. Now, for heaven's, sake, let us get on
with the job and end this childish hugger-mugger over words.
Pitts: I would like to say that I maintain the scientific method is
something more than a cultural prejudice, along with Margaret Mead—
Mead: I didn't say that.
Pitts: —that is, if the methods which the psychoanalyst uses in deal­
ing with this material are not scientific, it is up to him to make them
so, not for us to admit that his methods or modes of dealing with them
are just as good as ours, if we are scientists. There is a terrific differ­
ence between the—
Mead: That isn't what I said at all. I said that one group here is
used to dealing with a certain kind of rigor. They use the word "game"
with approval. I object to that out of my background, and Dr. Rosen­
blueth objects to it out of his; but it is an evaluative term, not a
devaluative term, when I use it. There is another group of people here
COMMUNICATION BETWEEN ANIMALS Communication Between Animals 135
the case of Oliver Twist under propitious circumstances, both organ­
isms obtain food and are sustained by this event, but the amoeba is
HERBERT G. BIRCH
responding now not to food as such but to a weak chemical stimulation,
Department of Psychology
City College of New York
and it will so respond to almost any of a wide variety of weak chemical
stimuli. Thus, if the minute food particle is replaced by a minute
particle of potassium permanganate, what happens, as in the case of
WHEN I WAS asked to speak to this meeting on the problem of com­ Hyman's and Child's experiments (l) (2), is that the amoeba ingests
munication in animals, I had, at the very beginning, a negative reaction the potassium permanganate in the same manner and with the same
to the topic, and that negative reaction, instead of abating in the course alacrity that it ingests the ordinary food particle in its environment.
of preparing my talk, has grown ever more intense, until at this point However, the consequence for the amoeba is not sustenance but a vital
I would say that I am unwilling to discuss the problem of communica­ staining process which produces a dead amoeba, beautifully stained.
tion in animals for a variety of reasons, but I am willing to discuss the I would further say, in the case of a human organism reaching for
various ways in which animals interrelate one with another. The reason food, it is food that is being reached for. The organism has some
for that kind of resistance and for that differentiation stems from a notion of what it is that it is responding to. In short, the equivalences
general difficulty that applies not only to the problem of communication in the environment for the human organism are not at all the same as
but applies also to any general category in animal behavior that can be are the equivalences in the environment for the unicellular organism.
used or tends to be used in a consideration of the problems of the way If we take the problem of migration, we will find that the salmon
in which animals adapt to the world or to one another. migrates, it is true, but basically the migration of salmon rests upon
I would lump the problem of communication with such problems as certain endocrinological changes which occur in it, which endocrino­
feeding, mating behavior, migration, or what have you. These behav­ logical changes produce certain pigmental changes, which pigmental
iors are really not behaviors at all, and we are not discussing behaviors changes are related to a sensitivity to light, which sensitivity to light
when we are discussing communication. We are, rather, discussing produces certain behaviors at a later point; then further endocrino­
certain phenomenal similarities that may appear in highly different logical changes create a change in the metabolic rate of this organism,
forms of behavior. We are concentrating not upon the behavior proc­ increase its need for oxygen, and drive it in the direction of fresh-water
ess as such but, rather, upon the end result of any of a number of streams once again; so that the behavior of this organism is based in
different kinds of behavior processes. When one concentrates upon large part upon biochemical temperature and endocrinological changes
similarity in end results, it is especially important to concentrate next in its own internal structure.
upon the differences in underlying processes, both physiological and The behavior of the human nomad, for example, as our anthropo­
psychological, that may be producing the end result. Otherwise, we logical colleagues have sometimes told us, is not based upon such
fall into the pervasive trap in science; namely, the trap of analogy changes primarily but, rather, is based upon the ways in which the food
rather than the method of understanding that is known as the method substances or the means of sustenance of the community are produced,
of homology or the examination of processes which have a systematic and the behavior of the nomadic individual is based upon these tech­
relation, one to the other. nological problems of his existence rather than upon the specific nature
Perhaps I can illustrate this point best by taking a few examples of biochemical changes.
from the problem of feeding behavior and of so-called "migratory" Now, then, if we start from a broad, general, end-result category,
behavior in animals. There is no doubt that an amoeba feeds. There we must be especially careful to seek out and to enunciate the dis­
is also no doubt that a human being feeds. There is no doubt that a similarities and the discontinuities of process which may underlie the
salmon migrates. There is also no doubt that the human nomad mi­ described behaviors. With this caution in mind, we can perhaps formu­
grates. But I would submit that there is a fundamental difference be­ late the problem of communication between animals. Let us start with
tween the amoeba, extending its pseudopod in the direction of a small a broad definition of "communication" as the effect of the behavior of
food particle, and Oliver Twist, extending his porridge bowl to the one organism upon the behavior of another organism. It is clear that
cruel dispenser at the orphanage. In the case of the amoeba, and in this category is so broad as to become almost meaningless and almost
134 useless as a means for differentiating communication function from so-
136 Cybernetics Communication Between Animals 137
called ""other" functions of organisms. But I think we can use it because mosquito may be elicited; in fact, it tends in nature to be elicited by
it leads us to a consideration of the ways and the levels at which ani­ certain tonal vibrations or vibration frequencies that are transmitted by
mals interrelate, one with another. the female mosquito of its species or of certain related species. Now,
I should like to start with the story of the scallop and its relation to the mating behavior of the male mosquito is affected, that is to say, it
the starfish. The scallop, as you know, is one of the food sources for is brought into play and into action by the noises which are transmitted
the starfish. If a starfish is placed in the environment of a scallop, it by the female mosquito, but would we say that the female mosquito is
rather quickly elicits a flight reaction on the part of the scallop. Is the communicating with the male? Perhaps some of us would. I, for one,
starfish communicating with the scallop Ì Well, most certainly the scal­ would not, because in Roth's experiments he showed that if you present
lop is fleeing because of stimuli that have impinged upon it and which frequencies of vibration by means of tuning forks or other arrange­
have emerged from the starfish as another organism. The behavior of ments, you elicit the mating behavior of the male mosquito now to the
the scallop has been afl^ected by the behavior of the starfish. The star­ tuning fork or now to certain other objects located in space, none of
fish has now come within a given range of the scallop, and within the which has any specific relation to the female mosquito, except in that
framework of this broad definition we would be forced to call such the tonal frequencies transmitted are those characteristic of the wing-
behavioral interrelations communication. beat frequency produced by this female mosquito.
However, if we do what several investigators have done and take a Most recently, we have been amazed and entertained and enlight­
starfish and boil it and make a soup, a nice rich starfish soup, and then ened by a series of researches on the communicative interrelations in
extract this soup just a little bit longer and have a very strong broth of insects that have been reported as the result of the work of Professor
starfish and place this starfish broth in the environment of the scallop, von Frisch on the behavior of bees (4). Dr. Griffin at Cornell, in his
the scallop at once exhibits the flight reaction. Under these conditions, introduction to von Frisch's volume on the bees, claims that no sensible
are we justified in saying that the broth of the starfish now is com­ scientist or no competent scientist ought to believe what von Frisch has
municating with the scallop, and how does soup communicate with the reported, upon first reading. He goes further and implies in his fore­
individual.? I would submit that under certain social conditions, an word that these phenomena as reported by von Frisch reveal a kind of
individual may communicate with soup, but that is, again, another complexity of relationship between insects that should force the stu­
problem. dents of animal behavior to a complete revision of their conceptions.
I think that what we find here, then, is that at the level of the lower This idea is echoed in a more explicit form by Dr. Thorpe in England
invertebrate organisms there have evolved in the course of the history (5). He, upon viewing von Frisch's work, claimed to have been
of animals certain sensitivities, which sensitivities permit of the survi­ amazed, astounded, entranced, and a number of other things, and even
val of organisms within their customary environments. went so far as to visit von Frisch's field station to have reproduced
This should not surprise us. It should not surprise us at all to find for him certain of the behaviors that von Frisch reported for the bee.
that the scallop has certain adaptive characteristics. It should not sur­ Upon his return to England, he wrote an extremely enthusiastic article,
prise us, because with these characteristics not present there would to­ in which article he claimed that the behavior of the bee was so complex
day be no scallops. What we find here, then, is that such a variety of that it should revolutionize our conception of the evolution of behavior
animal as the scallop has survived because it possesses such sensitivities as such. He felt that this complex behavior was so complicated that it
that throw it into flight reactions when chemical characteristics, such as could only have been subserved by something that we might call intelli­
the chemical characteristics emitted by the starfish, come into contact gence; and, again, I shall use Professor Richard's inverted commas, ex­
with its sensory apparatus. cept that I use a different direction. Thorpe always puts this in commas,
There we have, in an elementary sense, a variety of interrelation be­ as does Griffin. But putting it in commas is not the important thing.
tween animals, but I think we will see that there are a number of things The important thing is, what are these people getting at ?
absent from this that we customarily assume as existing in such phe­ The assumption that underlies their thinking is expressed in other
nomena as communication, particularly at the human level. I should articles of Thorpe's in which he gives the animals at the level of the
like to give another example. insects the capacity for insights; that is to say, complete reorganizations
In several recent studies on the mating behavior of mosquitoes (3), of their experience and an understanding of the world in which they
Roth, among others, has found that the mating response of the male live—really a kind of conscious understanding.
138 Cybernetics Communication Bettveen Animals 139
Now, I should like to point out that whenever a complex interrela­ base of that flower, but also some of the aromatic substances in solution
tionship among insects is described, and before it is fully explored, that characterize this flower; in other words, the nectar itself is scented,
there are always individuals who will leap upon these phenomena and and the secondary bee now is not merely being stimulated to a higher
use them to advance a notion of high levels of intelligence and high level of activity by this food exchange, but also by a second kind of
levels of psychological function in insects. This is not unique to Griffin activity that the finder bee engages in, a dance which is a kind of figure-
and Thorpe. The same situation existed in regard to the interpretation 8 type of dance, when the food is near the hive. When such secondary
of the extremely complex behaviors of the army ant, before Dr. bees leave the hive, they will tend to fly around, and if there is a
Schneirla (6) (7) performed a whole series of investigations that variety of different flowers present, for example, if you have phlox in
revealed that these complexities of behavior are not based upon any one place and another kind of flower in another place and the finder
high-level intellective function but, rather, are based upon very simple bee has been drinking at the phlox, then the secondary bees will move
insectlike features of psychological function — more about which a toward the phlox patch and will there ingest the nectar, after which
little later. they will return; so that two things now occur: first of all, the finder
Briefly, I wish to tell you about von Frisch's experiments. I am sure bee excites other secondary bees to go out; second, the finder bee pre­
that many of you are familiar with them, but perhaps a brief review sents certain chemical stimulation to these secondary bees, which
would be in order. W^hat von Frisch found—and this is what is most chemical stimulation leads to selectivity in the subsequent behavior of
important for communication—is that the activity of the finder bee, the bee.
that is, the bee that first finds food, when this finder bee returns to the It has been shown by von Frisch that you can perform this sort of
hive, influences the amount of activity of secondary bees who go to thing experimentally, so that if you have two dishes of sugar water,
the food in a variety of ways. First of all, it influences the distance one of which is scented with lavender, for example, and the other of
which these bees will tend to fly in their search for food; second, it which is not scented with oil of lavender, and you get the bees from
influences the kind of food materials that they will tend to be respon­ one hive to feed upon the lavender-scented materials and from another
sive to, third, and most astonishingly, the behavior of the initial finder hive upon the unscented materials, the finder bees that drank from the
bee influences the direction in flight—and this is what is apparently lavender-scented material will transmit to the secondary bees in that
most astounding—of the secondary bees. neighborhood the scent or the odor of lavender, and there will be pre­
The first problem is relatively simple to deal with and to describe. eminent feeding of such bees at the lavender dish rather than at the
The finder bee, when it is eating food, may accumulate certain of the unscented dish.
chemical characteristics of this food, either on its abdomen or in its Savage: Do you mean literally unscented.? Can bees discover a liter­
social stomach. Insects are peculiar in having a "social" stomach. I ally unscented dish?
am using Forel s designation there. By "social" stomach we mean a Birch: Well, you can do a number of things. You can shellac over
stomach that regurgitates easily. It is a stomach that regurgitates to the sensory organs involved in olfaction, both at the tarsi and at the
antennal contact upon the part of another insect. Thus, when the finder antennae tips, and under those conditions you can guarantee they can­
bee comes back with its "social" stomach gorged with nectar, it enters not be responding to the olfactory materials. There is a variety of ex­
the hive and begins to have antennal contacts (which have a history of periments here, and I don't think there is any real question on that point
their own) with other bees, other worker bees in the hive. These young at all. I had thought initially that Professor von Frisch was going to be
ladies in turn stimulate the antennal receptors of the finder bee. The here to discuss this, but that was impossible.
stimulation of these antennae now produces regurgitation almost re- While these things are interesting enough in and of themselves, cer­
flexly in the finder bee, and a droplet of nectar is transmitted from the tain further things happen to the behavior of the secondary bees and to
finder bee to the secondary bee in the hive. This pattern of behavior is the behavior of the finder bees in connection with the foraging act. It
characteristic not only of the bees but also of a variety of insects, such has been reported by von Frisch, and both Griffin and Thorpe have
as the ants and others. said, that they have repeated the following kinds of observations: If
The nectar itself, I should point out, has chemical characteristics. the food dish is within a hundred meters of the hive, then the behavior
Thus, for example, if the nectar has been obtained from a given kind of the finder bee is characterized almost entirely by this figure-8 type
of flower, it contains not only the sugars that are in solution at the of ground dance. However, if the food source is more than a hundred
I4ü Cybernetics Communication Between Animals - 141
meters away from the hive, then a peculiar change appears in the activ­ McCulloch: None.
ity of the finder bee as it returns to the hive. It engages now not only Von Foerster: May I refer at this point to a formula I tried out once
in this round dance but also, between the two segments of the round and which correlates the wiggle-frequency f in wiggles per second with
dance, it engages in what von Frisch has called the "waggle" dance. It the distance S in meters of the feeding-place from the hive. The for­
is a relatively straight-line run which has a given directional orientation, mula has a hyperbolic form and reads:
and then the other half of the round dance is engaged in. (S-S„).f = c
A second point appears: the rate at which this combination round It fits astonishingly well for different experimental results as far as I
and wiggle-waggle dance is engaged in, that is, the number of times it could find those in some publications. The first constant seems to
is engaged in per second, is roughly inversely proportional to the dis­ be the minimum distance for which the bees decide to use the wiggle-
tance of the. food source from the hive; that is to say, that if the food dance communication-technique. The other constant c has the dimen­
source from the hive is at two hundred meters—or, rather, if the bee is sion of a velocity. If you evaluate from the experimental results the
two hundred meters from the hive, it makes some twelve or fourteen numerical value of this constant, you come to a very surprising figure:
of these per unit of time. If it is at a distance of a kilometer or five c becomes about 330 meters per second which is very close to the
hundred meters, half a kilometer, from the hive, the number of turns velocity of sound in air. It seems to me that the velocity of sound may
in its activity goes down; so that per unit of time it may now make only be for the bees the same universal constant as is the velocity of light in
four, five, or six of these activities in the time period. man's electromagnetic philosophy.
Bigelow: What is the method of observation.? Photography? Savage: The experiment should be carried out in different atmos­
Birch: No, the method of observation is direct observation on a pheres or something, to see if there is any truth in that.
glassed-in hive, laid out so that all of its combs can be seen through von Bonin: How do you choose the S„ ?
glass. Von Foerster: In the formula above, S is the distance from the hive
Von Foerster: And also motion pictures. to the feeding-place. But actually the bees begin to interpret distances
Klüver: Perhaps you should point out whether these observations of the feeding-place with the wiggle-dance when the feeding-place is
refer to dances on vertical honeycombs. a minimum distance away. This minimum-distance I call and has
Birch: Well, some of the observations, yes; that is why I didn't say the magnitude of about 100 meters, as it was pointed out before.
that, because under certain conditions that we shall come to in a min­ Birch: While von Frisch, Griffin, and Thorpe have all said that this
ute, these hives are placed vertically, but under other conditions, the waggle dance is simply appearing full-blown at a hundred meters,
hives are placed horizontally. which gives you your S„, C. G. Butler (8), for example, of the Depart­
McCulloch: May I describe the setup for a moment? I saw it not ment of Bee Behavior in England, has made thousands of drawings of
long ago. He uses a glass-sided box which can be rotated from the the bees' dances after having fed them at different distances from the
vertical to the horizontal position, so that he is able to observe and to hive, and he finds that this waggle dance does not appear full-blown
photograph the behavior of the bees in the hive. out of nothing but, rather, that as the distance increases from 0 to 100
Bigelow: Does he actually take a long string of motion pictures or meters, there is a greater and greater separation between the circles and
something ? they become more and more distinct, so that gradually what emerges is
McCulloch: Yes. two separate circles, which more and more are joined by a straightaway,
Bigelow: And take a given bee and put an X on it and follow its so that the does not represent a true starting point, but simply repre­
path and gather statistics on it? sents that nodal point at which the wiggle-waggle phase of the dance
McCulloch: Oh, yes. is most unmistakably present. So there is no question about the validity
Birch: Yes, he labels each of the bees in the feeding situation by put­ of using So at that point, except for purposes of computation at this
ting a small drop of quickly drying lacquer on them, or pigment and moment. Therefore, we have the formula (S - S„). f == 330 meters per
shellac, and under those conditions it is possible, through a numbering second.
system that he has developed on the basis of dots on different portions That is just one phase of the problem that requires much fuller
of the insect, to identify the insects that are involved. investigation, of course.
Bigelow: In other words, there is no reasonable doubt about his While these things are amazing enough, I think that a consideration
data ?
142 Cybernetics Communication Between Animals 143
of the wiggle-waggle dance yields certain other interesting results. It is toward the place at which it has been fed, in relation to the polarized
found usually that the hive of the bee is in the vertical plane, so that characteristics of the light, and it is then transforming this into activity
when a bee is standing on the hive it is standing vertically. In flight the in relation to a gravitational field.
bee is moving in the horizontal plane, in the main. The interesting When the finder bee engages in its wiggle-waggle dance, it excites
point is that these bees indicate direction or provide a basis for direc­ other bees in the colony, and these bees, too, now begin to go into
tionality in horizontal flight to the secondary bees on the basis of the activity. They begin to engage in their so-called "dancing," and they
direction of the waggle dance itself. This is done in terms of a conver­ tend to orient toward the abdomen of the finder bee and to direct a
sion of horizontal directionality which is related to the position of the straight-line movement of their own in relation to the wagging abdo­
sun into a system of vertical directionality which is related to the gravi­ men of the finder bee itself. Thereafter, when they reach a certain
tational field, where the gravitational pull apparently becomes the stage of excitation, they leave the hive; and now, rather than moving,
equivalent in direction indication to the sun itself. around in a random distribution, in all directions around the hive,
You can begin to examine the bees' relation to the sun, as von Frisch which is the case when merely the round dance is engaged in, the bees
has done, by tilting the hive from the vertical to the horizontal, under now take off in the direction pointed to by the,wagging bee. Well, cer­
which conditions there is no gravitational component to which it may tainly, here we have, in a complex animal social organization at the
orient. Under these conditions, if no light is permitted to enter the hive bee level in the society, interrelations between bees, and we have the
and the bees are observed through red glass, that is, red light is per­ behavior of the secondary bees consistently affected by the behavior of
mitted to go in, we can see it, but it is outside the visual spectrum of the finder or primary bee.
the bee, then you find that these dances simply become randomized and But does this necessitate a revision of all our concepts of animal
have no. specific direction. However, if sunlight is permitted to enter behavior? Does this mean that we now have to consider bees as
the hive, then these dances at once are oriented in direction and begin geniuses ? Does it mean that we now have to impute certain high levels
to be pointed toward the food place as it is related to the direction of of psychological function like abstraction or what have you to the bee
the sun itself. itself ? I would submit that we do not.
But what would happen on a cloudy day or on a day when the condi­ Klüver: We should have more information on the mechanisms in­
tions are such that no direct vision of the sun's position in the heavens volved in imparting information in the dark hive, that is, in the absence
itself would be observed.? Well, von Frisch finds that the bees are still of polarized light and under vertical conditions.
capable of maintaining this kind of directional orientation as long as a Birch: Yes, under vertical conditions.
small patch of blue sky is visible to them, and by "blue" he simply Klüver: Perhaps you have some ideas as to what really happens.
means clear. I don't know whether it is blue for the bee. Birch: That is what I meant when I said that the polarization field
This led him to extend certain notions that had been initially devel­ can be converted and is converted into the gravitational field.
oped in connection with the study of ant behavior; namely, a notion Rioch: I wonder if there are data on several other points. Does Dr.
that the bee may not simply be responding to the directionality of the Birch have data upon the number of bees it takes to find the food after
sun's rays as such, but may now be responding to the polarization of the finder bee comes back? That is, can one other bee respond to the
light as such, and there is a relationship between the directional posi­ finder bee? Then, another question: Does the finder bee do the dance
tion of the sun and the polarization matrix which may exist; so that if there are no other bees in the environment, that is, if the hive has
apparently the bee now is responding not to the sun itself, but may be been .emptied ? Is it the hive or the presence of other bees that evokes
responding to the polarization characteristics of light. This can easily the dance, or is it something like the distance back from the food that
be tested. evokes the dance? And also, are there any data that give an idea as to
If one were to take a piece of polarized glass, for example, as von how long the finder bee can remember the direction? That is, if one
Frisch did, and place it over this opening, and rotate it in this open­ were to put a tarpaulin over the hive so that the bee had to fly without
ing, then, if the bee is responding to the polarization characteristics of any polarized light to give him direction for any length of time, how
the light, the direction of pointing in the waggle phase of the dance long could he remember that? You see, you have this fantastic busi­
should be changed; and this is exactly what does happen. The bee then ness in insects of types of memory. Your bee knows its own hive on
is responding and maintaining its orientation toward the food place or the basis of the one experience of coming out of it just after he has
144 Cybernetics Communication Between Animals 145
been hatched, or at least so I understand. engages in certain behaviors that excite certain subsequent behaviors
McCulloch: That's right.? upon the part of the other bees. Now, I would be proud of that, but I
Birch: That is not so, but go ahead. don't know about the bee.
Rioch: Well, then some time I should like to hear some of the data Savage: Well, does it not come to pass that so many are coming back
on it. that the pattern is broken up, that there are so many dances going on
Birch: I will agree that it is fantastic. at once that nobody has an audience ?
Rioch: But if this bee has to fly under trees or something for a while, Birch: A hive is a pretty big place, and there are many dances that
what distance can he fly and still remember which way the polarized can go on at any point in time. There may be some fifty thousand
light was Ì Those are the questions I had. individuals present, or more, and certain groups of them will orient
Bigelow: My questions are very much in the same category. Mine with regard to one of these finder bees, and others to another one.
concern exactly what response the other bees in the hive give when Now, may I answer these questions as they have been given ? In the
you perturb the directional axis of the dance by polarized light or what­ first place, about the memory of insects and orientation toward the
ever means you wish, whether the bee himself shows any disturbance nesting place itself: you find that this orientation is something whkh
of his normal pattern in finding his way back; and whether anybody the bee or the ant or any of the other social insects has to learn after it
has ever carried out an experiment of keeping a bee in the air by some emerges from its incubation state. When it appears as a nymph, it
subterfuge for a certain period of time corresponding to a different makes small migrations and it begins gradually to increase this distance
distance of flight, to see whether or not this affects his dance régime of activity from the hive; it begins to orient toward specific features of
into Mode 1 or Mode 2 ? the environment about the hive or about the nesting place. For ex­
Brosin: I have an extension of Dr. Rioch's question. Is there more ample, there are many studies at the level of the ant, where you can
information about the social organization or the "needs" of the bees take a tree, for instance, that previously bore a given relation to the
that alter the relationship between the finder bee and the subsequent nest and move it, dig it up and move it a certain distance, or establish
behavior of the secondary group ? an artificial potted tree, so that you can move it more easily. Under
Pitts: I think there is one factual point I might make about Dr. those conditions, you find that when the insects begin to come into the
Rioch's question. If you look at any square centimeter of sky and region of the hive itself, or the nest itself, they begin to orient in rela­
measure the direction and the degree of polarization in it, you can tell tion to those landmarks and those features rather than to the specific
where the sun is. In very rare cases there may be two positions for the position of the nest itself; so that they do learn to respond to specific
sun that would be compatible with the amount of polarization in any features of the environment. Once they get within the range of those
given place; but those are rare cases. In general, you can tell where environmental fields, they will tend to return.
the sun is by looking at a piece of sky, however small, in any direction. Rioch: Do you have any data on that, as to the number of experi­
Birch: If you are sensitive to polarization, yes. Certain humans are, ences that are involved ?
in small part. Birch: It has not been studied in a sufficiently systematic way to do
Pitts: No, I mean the data are determinable in the apparent field. that. I can extrapolate, of course, from some of Schneirla's researches
von Bonin: There is one more question. You talk'about the finder on learning in the ant itself, in a maze situation in which the animal
bee coming back. Suppose the next bee goes out; does it also go must traverse a given kind of maze before it arrives at a feeding place
through the dance when it comes back? Does it become a finder bee and then traverse a different kind of maze in the return from the feed­
by virtue of— ing place to the nest itself. Under these conditions you find that the
Birch: Oh, yes. learning of the ant is an extremely slow process; that is to say, if you
von Bonin: And some more come up ? control the laying down of chemical pathways by removing the linings
Birch: Oh, yes. The finder bee is not a special individual. It is of the maze itself, then it takes the ant many, many trials—forty, fifty,
simply the bee that has been to the food. or sixty trials—to begin efficiently and relatively errorlessly to traverse
von Bonin: It doesn't matter that it is the first one, but anyone that a relatively simple maze, and this learning, further, goes through at
comes back is so proud of it that he has to tell about it ? least three stages: first, a provisional orientation stage, in which the
Birch: Well, I wouldn't put it that way. Anyone that comes back ant learns not to walk on the glass cover of the maze but, rather, on
146 Cybernetics
Communication Between Animals 147
sections of the floor and things of that sort; then, a phase in which McCulloch: Will you go ahead to the next point. Dr. Bircli ?
blinds are omitted in a very piecemeal way. Blind alleys here are not Birch: Yes. Are other bees necessary ? Although there has been no
eliminated in the manner that you find in many animals. systematic examination of that problem, there is some incidental evi­
Rioch: I didn't want to get off into learning of insects, except in dence, I think, that other bees are not necessary for the dance to occur.
this— . , Thus, for example, a finder bee may alight on the entrance portion of
Birch: That was the question you asked. the hive and there be stimulated in various ways, and begin to go into
Rioch: No, no. I asked a question about a particular learning. You a dance on the platform in front of the hive rather than in the midst
see, we have been dealing with the problem of things that are learned of the other bees.
with one experience, and things that are learned that will require mul­ Pitts: Be stimulated in what way?
tiple experiences, and things that require particular situations, internal Birch: For example, you can tap its antennae.
situations in addition to the experience. This is the question I asked: Pitts: That is, be stimulated in a way that is normally done only by
Is there in the insect a certain type of experience which, with one other bees?
experience, produces a durable change? We know that in the bird Birch: Certain of the stimuli may be given, or are usually given, by
there are such situations. other bees, but under certain circumstances, too, there can be this spon­
Birch: Yes. I would say—^well, a durable change in terms of how taneous activity. It would be an interesting question to examine a bee
long ? For example, I think that the chemical stimulation of the insect, going into a hive that has previously been emptied. It has not been
or of the bee in this situation, produces a change which will persist for studied in any detail. Like many of these processes, at this moment,
a finite period, long enough in a number of cases to lead the organism what we have is a description of the complex behavior and not any
to a given kind of feeding place. But whether or not this is something thoroughly detailed analysis of the specific actions and of the specific
which it has learned, or something which provides a residuum of con­ interrelations between the individual organism and the environment; in
tinued stimulation on the basis of the nectar which it has ingested, on other words, we are at a preliminary stage in the investigation of this
the basis of residual chemical effects after its eating, is another ques­ problem.
tion. We don't know whether it has learned that in one trial or whether Rosenblueth: The behavior of the bee is not very different from that
it is now simply responding to a continuing situation of a stimulàting of a warrior who gets home and begins to dance by himself.
kind. Birch: I would say the bee is very different.
McCulloch: May I in part answer what I believe is your question? Rosenblueth: The bee behaves similarly but looks different.
I think the best data on learning from a single experience by any of Birch: The warrior behaves differently and looks similar. I would
the hymenoptera is Baerends's work on wasps, the solitary wasp. I re­ say there is a phenomenal similarity but a difference in—
fer to the work of G. P. Baerends of Groningen. There is no question Rosenblueth: They are both excited and stimulated and dance by
that a single sight of a landmark is enough to enable them to find it themselves.
again. MacKay: I think a closer analogy is that of a man singing in his
Rioch: Can they forget it? bath.
McCulloch: If that landmark is moved, they don't find it again. This Birch: It is again an analogy and, as such, I will accept it—as long
is a matter of something which persists over days or even weeks. as it stays an analogy.
Rioch: But can they forget it? That is, can they learn another land­ Mead: But the present data do suggest that you need something of
mark which means forgetting the first one ? I mean an organization of the order of the stimulation of the antenna to set off the dance ?
the central nervous system, whatever the— Birch: No, the present data do not indicate what you need to set off
McCulloch: The answer to your question is "Yes, I think so," but I the dance. It sometimes apparently occurs spontaneously. Under other
would have to look up the data. conditions it is a further consequence of subsequent excitation and
Birch: But with difficulty, if the other landmark is still present. things of that sort.
Rioch: Because the forgetting is much more important a problem in Mead: Although the usual pattern is the stimulation ?
the central-nervous-system organization. Birch: Well, the usual pattern is the stimulation because a bee com­
Rosenblueth: The important thing is to find another landmark the ing into a hive with fifty thousand other bees present is inevitably in
next day.
148 Cybernetics Communication Between Animals 149
contact with those other bees. I will grant that. It's awfully hard; it's complex human behavior, and I think Dr. Birch is anxious to tell us
like not touching your neighbor in the subway rush hour. how they have been broken down experimentally. I am sure many of
Kubie: The danger of superficial analogizing is very great. Let us these casual remarks were made partly in fun, to keep the thing alive,
suppose that a college lad goes into church and starts to dance— but I should like to hear what the facts are.
von Bonin: Church Ì Rioch: They were not made in fun.
Kubie: Yes. Perhaps he is doing this to pay an election bet, or as a Gerard: Well, let's hear the rest of the answers, and then we can
hazing stunt; or perhaps he is doing it because of a delusion that he argue about them.
has had a message from Rome which directs him to do so. Perhaps Rioch: There is one other question to answer; that is, if there are any
he is doing it because of the pressure of an unconscious obsessional- data upon the duration for which the bee will remember.
compulsive drive which must be obeyed, even though he knows it is Birch: I shall try to answer it. The problem next is: What happens
foolish. Evidently the psychological significance of the behavior derives if you restrict the activity of the bee over a period of time Ì Does it still
neither from its superficial nature nor from its effects, but from that maintain its orientation ? There only the sketchiest experimentation has
subtle balance of conscious and unconscious processes which have pro­ been done at the bee level, but extremely interesting experimentation
duced the act. has been done over a long period, a long time ago, at the level of the
Bigelow: Don't we all grant that the analogies now being made are ant—another insect. Santschi, for example, in some of his early experi­
superficial ? ments, and some investigators working with the driver ants in Africa,
Kubie: Just a second. We may or may not. The point is that— as well as Goetsch (9) in many of his investigations of the ant have
Rioch: No, these are absolutely fundamental. shown that if an ant is following a trail—if this ant is a visually ori­
Kubie: Viewed superficially, the behavior may be sensible or foolish, ented ant rather than a chemical trail follower—if you put a box over
useless or useful, creative or destructive, good or bad; but in human the insect and keep it in this box, away from the sun, that is to say,
beings, at least, we know that it always serves several concomitant pur­ away from light stimulation, for a period of time, and then remove the
poses, some of which may be manifestations of deeply unconscious box, it takes off at an angle which is related to its previous orientation
forces. When, on the other hand, we turn to such an animal as the to the sun as such. Apparently a continuing process persists, and the
bee, we can know nothing about the relationship even of his hypotheti­ ant now is orienting toward the sun and maintaining this kind of
cal conscious purposes to his activities, much less of his unconscious orientation over a period of several hours.
purposes. Therefore when we make any simple analogies between Pitts: He doesn't allow for the motion of the sun ?
human behavior and that of the bee, we are in danger of getting in Birch: He doesn't know about it. He is now simply responding in
hot water. terms of his previous orientation to the sun, so that I think this is more
Bigelow: It is all right to do it, though, isn't it.? important for Dr. Rioch's purposes, I believe, than the notion that the
Kubie: I doubt it. Let us consider so simple a human function as ant may or may not have heard about the movement of the sun in rela­
eating. We know that we eat to serve innumerable conscious pur­ tion to the earth; so that you do have the persistence of such effects over
poses—out of fear, anger, sorrow, loneliness, or to celebrate. On top a relatively long period in insects. I am glad you raised that question,
of that, we also know that we may eat for a wide variety of unconscious because it is especially important, I believe, in regard to the beginning
purposes, in the service of which eating itself becomes a purely sym­ of an investigation of the directional orientation of these bees.
bolic, symptomatic act, as much a symptom as a hand-washing compul­ But before I get to that, let me try to answer a couple of other
sion. If we could not talk, we would know nothing of this, all of which specific questions that were raised. There was something about time
can be represented in the simple hieroglyphic or shorthand behavior of of flight. Who asked that?
compulsive eating, phobic eating, delusional eating, and so forth. What Bigelow: I did.
access do you have to any such data for any one of the subverbal animal Birch: What was the question?
forms Ì Bigelow: The question concerns the following conjecture: Possibly
Bigelow: Yes, but it is all right to do it if you know you are doing the bees changed from one mode of dancing to another on the basis of
something risky. some physical fatigue, not distance but fatigue.
Gerard: All these questions have been directed to different aspects of Birch: Oh, yes, I remember the question. Well, there is some evi-
150 Cybernetics Communication Between Animals 151
dence on that. For example, if we were to assume that the bee is in the next bee, it gives the same result ?
merely responding to distance, then, independently of the time required Birch: But, after all, the finder bee now has a tail wind to the sec­
to traverse that distance, it should give the same temporal sequence of ondary bee, so you begin to have compulsory action.
dances. But von Frisch himself has shown that when there are winds Rosenblueth: I want, for my information, to know whether you think
present, for instance, the bee behaves as though the distance were "excites" belongs to a lower category than "informs."
greater—in quotes now. When there are tail winds, under those condi­ Birch: Oh, yes.
tions, the bee behaves as though the distance were less; in other words, Rioch: The bee must be able to make allowance for head wind and
the bee is apparently responding to its own effort and to the metabolic tail wind. In Maine an old .method of finding a beehive in the woods
expenditure that is involved in the movement from the feeding place is to use a trap with two compartments in one of which is a bowl of
to the hive. sugar. After getting a bee in the trap, the hunter walks in the direction
Bigelow: You would grant, then, that his energy output is in fact from which the bee came, and then releases it. When it returns it is
his distance indicator? again trapped and the process repeated. In this way the bee leads the
Birch: His energy output tends to be his distance cue. hunter to the hive. The bee must make some judgment as to distance
Pitts: But which trip-—from the flower to the hive or from the hive on each trip.
to the flower ? Bigelow: This method would result in finding the beehive if the bee
Birch: From the flower to the hive. just knew which of the two 180° sectors of a circle is correct.
Pitts: That is what he measures in time ? Birch: The other factual question was: What are the needs of the
Birch: Yes, because in arriving at the flower—for instance, if you bees? The problem of what different bees would begin to do under
were to follow the flight of any bee, you would see that it will meander controlled differences in any motivational state or in food-deprivation
in many, many different directions and eventually wind up at a feeding state is a question yet to be investigated. It has not been fully investi­
place. Now, let's say you take a place that is denuded of flowers, a big gated. We find some small amount of evidence. We know this, though,
football field or something like that, and at the end of it establish a that the giving of nectar by one bee to another functions to excite it,
feeding place for bees, and at the other end of it (as I myself once did and to prod it into activity.
as a sort of check experiment) have a hive. Under those conditions you Klüver: This is perhaps the time to say a few words about the be­
will find that the bee traverses a huge amount of distance from the hive havior of bees in New York City. In the experiments reported by von
to the feeding place. Frisch, it seems remarkable that the upward movement during the
Pitts: It certainly is more intelligent for him to do it the other way. straight component of the wagging dance corresponds to the direction
Birch: And his response now tends to be oriented toward the place of flight toward the sun, and that the bees, aroused by the dance on
where the bee filled its belly rather than toward the hive itself. There the vertical honeycomb in the dark hive recognize the angle of the
is a relation— dance relative to gravity. For instance, they recognize, if the straight
von Bonin: And then does it make the proverbial beeline home ? portion of the dance is pointed 60° to the left of the vertical, that this
Birch: Yes, depending upon the availability of orientation cues. means that the feeding place is located 60° to the left of the sun's posi­
von Bonin: The football field ? tion in the sky. This implies a truly remarkable interrelation of gravita­
Birch: Yes. He proceeds relatively straightly. tional and optical factors. More than twenty years ago. Selig Hecht, in
Savage: From what you say, I gather that in the presence of a wind the Laboratory of Biophysics at Columbia University, wanted to meas­
the finder bee systematically deludes all the other bees ? • ure the visual acuity of the honeybee, but decided not to make use of
Birch: That the finder bee then gives—^no, I'm not going to fall into such conditioning or training methods as had previously been used with
that. That the finder bee excites the secondary bees in such a manner such great success by von Frisch and others. He looked for a method
that there is a lesser accuracy in their activity than is the case under involving no training of animals. It is known, of course, that many
more neutral conditions. animals respond to a sudden movement in their visual field. If the
Bigelow: I'm sorry, but isn't it relatively accurate? Because he as­ visual field consists of a pattern of dark and illuminated bars of equal
sumes they are replicas, and he is using a yardstick whose value changes size, an animal will respond to a displacement of the whole field only if
with energy output, but when applied to the same flight-energy process. it is able to distinguish the components of the pattern, that is, if it can
152 Cybernetics Communication Between Animals 153
resolve the black and white bars. If it cannot do so, that is, if the whole for picking up mechanical motion or accelerations.
field appears uniformly illuminated, it will not respond to a movement Klüver: Dr. McCulloch, since you have just been to Groningen,
of the field. In Hecht's setup the experimental bee was confined in a I am wondering whether you can report on any progress made by
glass compartment. Light reached this creeping department from the de Vries, one of the physicists in Groningen. In 1844 Haidinger dis­
bottom after first passing through two pieces of glass held in a frame; covered that the human eye can detect the direction of polarization of
namely, an opal glass plate and a transparent plate with opaque and linearly polarized light. In October, 1950, de Vries told me that he
transparent bars. In this setup it was possible to compute the visual was engaged in some experimental work on the properties of the
angle from the dimensions of the bars and the distance between the human eye with respect to the so-called "Haidinger brushes." I wonder
bars and the center of the bee's eye. Hecht hoped, of course, that a sud­ whether he has any new data bearing on the interpretation of these
den movement of the black and white stripes of the visual field below polarization brushes
the glass compartment in which the bee was crawling would induce a McCulloch: No, and I don't think they were far enough along with
sudden change in the direction of the crawling. Unfortunately, this did it then to be ready to say anything. That was just a few months ago.
not happen at all, and I well remember that Ernst Wolf, the collabo­ Birch: I would like to deal analytically with some of these fascinat­
rator of Hecht, became almost desperate when he tried in vain, month ing phenomena that have been revealed by von Frisch. In the first place,
after month, to elicit a response of the bee by moving the visual field. I should like to deal with the problem of how the insect maintains a
However, all troubles were over one morning when he hit on a slight kind of orientation with the polarization characteristics of light rather
modification of the experimental setup. This modification consisted in than to the specific light intensity directions themselves.
simply tilting the bee's creeping compartment and the movable black This question von Frisch answered (4) by analyzing the nature of
and white stripes of the visual field at an angle of about 30°. The the eye of the insect, and by pointing out that this compound eye rep­
moment the whole system was tilted at an angle of 30°, the bee in the resents an ideal kind of analyzer for a polarized system; so that if you
glass compartment, because of its negative geotropism, tended to crawl were to take polaroid glass and arrange a model which would, in a
upward in a straight line, and deviations from its linear progression simplified way, represent the directions to which these various eye units
were easily observable each time the visual field below the glass com­ of the bee are pointing, you could then begin to analyze the way in
partment was moved. In other words. Wolf finally found an elegant which a given bodily orientation toward the direction of the sun could
way of measuring the visual acuity of the bee by taking into account produce different effects that are the result of the polarization char­
both gravitational and optical factors (10). acteristics of the light. If you take such a model, as von Frisch did,
Birch: This shows the importance of the gravitational factors, and made out of about six units of polaroid glass and then rotate these in
other directional indicators, and one that in some way is reciprocal and the direction of light, much as an artificial eye may be rotated, then you
equivalent to the visual system. Just what that relation is will have to find that what is projected by such a system is a pattern of brightnesses
be determined. characteristic of the direction in relation to the polarization phenome­
Klüver: I am sure it will be necessary to study not only certain aspects non; so that if the bee now turns in a different direction to the sun, a
of the behavior of the dancing or sender bees in greater detail but very different brightness pattern is established in its visual field. Its
also—and this is probably even more important—to analyze the be­ maintenance of visual orientation, then, is based in large part, or may
havior of the receivers or of what you have called the secondary bees. be based in large part, upon the existence of such brightness patterns,
Rosenblueth: What kind of labyrinth, or the equivalent, do bees and differential brightness patterns, depending upon the orientation
have Ì toward the polarized features of light.
McCulloch: May I answer that.? The bee has a material which is ap­ The big and key question is : How in the world does the bee translate
parently piezoelectric, and which is capped with a smallish weight. this kind of orientation, based upon the polarization features of light,
Work on similar piezoelectric devices, particularly in fish, is now under into the gravitational field itself? When it is in the hive, the hive it­
way in Groningen, in deVries's laboratory. self is quite dark; there is no light entering the hive, and you can ob­
Rosenblueth: With electric signals.? I mean, is it converted into serve only if you create artificial conditions in which you look at the
electric signals? bees through red glass. Under these conditions, it is found that these
McCulloch: It is probably not very dissimilar from all of our devices animals translate the direction maintained toward the sun into a set of
154 Cybernetics Communication Between Animals 155

directions related to the gravitational field. ment under given ecological conditions produces the complex kind of
Frankly, I have no specific answer to this question, nor do I feel that behavior that we find in a swarm-raid formation and in bivouac forma­
von Frisch does or Griffin or Thorpe. However, that we have no tion and other things, and, under different ecological conditions, pro­
specific answer to the question is not, to me, a sufficient basis for assum­ duces just this same kind of self-destructive behavior. We find this
ing that such a translation could only be able to occur on the basis of happening not only in the laboratory but also in our own kitchens.
some higher-level intellective process wherein the bee understands the I had this happen with a group of ants that was closely related to the
relationship of gravity as a directional feature to polarization as another Ecitons. I was living at that time in a town in Florida. To the discom­
directional feature, and so on. I think that such a position can be taken fort of my wife, a circular mill was established on the ceiling of our
only if we ignore all the other kinds of finding about the essential kitchen. She wanted to break it up, but I spent a whole day trying to
simplicity of the behavioral process in insects. observe this mill, so we had a rather difficult condition for a little while.
I think that a productive line of inquiry here would be the examina­ But the features that produce this milling are the same features that
tion of what the intersensory equivalences are in an organism such as -produce the stereotyped kind of performance that really characterizes
the bee. To what degree does an orientation process which is estab­ the insect as an organism.
lished by chemical means or by light means or by gravitational means With this, I should like to move away from the insects for a little
become one which the organism can now engage in independently of bit and begin to deal with some of the problems at the vertebrate level.
the specific kind of receptor system which is being stimulated ? Of course, we could stay with arthropods all day if we wanted to, and
In that sense, then, I would begin to ask the question: How equiv­ begin to discuss such things as the so-called "display phenomena" in
alent for the bee is a visual stimulation with a chemical stimulation spiders and things of that sort, but I prefer to make this presentation
with a gravitational pull of various kinds ? "What is the way in which synoptic rather than inclusive.
such different sensory inputs have for the bee the same effect upon it Ì At the vertebrate level we have a number of interesting phenomena
That would require a detailed and minute investigation of the sensory in the lower vertebrates, that, under a general heading, could be classi­
physiology of the bee, which investigation has not yet been engaged in fied as communication. Of course, almost any of the mating behaviors
in this form. There are certain items of investigation, but our informa­ of the vertebrates represent a way in which the characteristic activity of
tion certainly is incomplete. Animal A affects the characteristic activity of Animal B, but let us try
The reason I say that I would move toward such ah examination of not so much to find examples of instances in which the activity of Ani­
the problem rather than toward an interpretation based upon the no­ mal A influences the activity of Animal B, but, rather, to find the kinds
tion that the bee now knows what the light direction is and then knows of examples which permit us to understand what the basis for the ac­
what gravitation is like and then tells the other insects what is going tivation of B by A happens to be. I should like to deal with a couple
on, is that when any complex insect phenomenon has been investi­ of phenomena in fish, for example.
gated—and I refer here particularly to Schneirla's research on the raid We have the phenomenon of schooling in fish, and the phenomenon
of the army ant, which is an extremely complex pattern involving the of milling in fish. "Various fish will begin to establish schools, and still
development of an expanding feeding front and the development of other fish will begin to establish mills, like the herring or the mackerel.
ancillary flanking movement, so called, all of which have an extremely In the schooling behavior of fish, the investigations of Breder and
useful function in the survival of the army ant—^when these are ana­ Nigrelli (11) (12) (13) (14) primarily have indicated that this be­
lyzed, they are found to be based upon the characteristics of sensory or­ havior is related to the visual system of the fish, as such; that it is re­
ganization that cause this ant to do the most "stupid" kinds of things lated to the kind of visual angle which the fish has, and to its directly
under other conditions of environment. Thus, for example, if, in the determined responses to certain kinds of visual stimuli. There is a cer­
laboratory, you take a bell jar and turn it upside down and let a group tain optimal position of visual fixation on objects between Fish A and
of army ants begin to walk around it, and then lift up the bell jar, what Fish B and Fish C, such that a change in the distance between them
happens is that you have a circular mill established, which continues produces a distortion of image, and the fish then tend, to maintain
until either these insects die or until dessication factors force a break­ relative positions, which are positions of relative optimal fixation.
up. Therefore, in the army ant, the enslavement of the ant, in a sense, If you have these fish in an aquarium and have a school going and
to the chemical features of stimulation is such that this same enslave- turn out the lights, the school breaks up. If you turn the light on again,
156 Cybernetics Communication Between Animals 157
you see the fish scattered here, there, and everywhere, and then you be­ used? Well, according to Parr's hypothesis, what you would expect to
gin to see them responding to the visual stimulation of the movements find would be mills going in one direction from the left-eyed fish and
of other fish, and they begin to establish a spatial relation of a neat, mills going in the other direction from right-eyed fish, and if you were
orderly kind, a repetitive kind, in a sense, which we refer to as a school. simply to use shellac or other things to blind them, and then reverse
Therefore, the schooling in fish in itself is a relatively simple phenome­ the blinding, if Parr's.hypothesis is again correct, then you would expect
non in which fish whose behavior is almost completely directly deter­ to find a reversal in the mill formation. It would be a neat experiment
mined in a stimulation sense—a given stimulus is presented and the and easy to do. It just hasn't been done. It suggests itself quite natural­
fish will tend to respond, if it has given intensity characteristics—now ly from the phenomena and from the proper explanation of the phe-
form these schools on the basis of their sensory situation rather than nomenà.
upon their liking for other fish or anything of the sort. You can es­ A second kind of behavior in fish is one that customarily is referred
tablish schooling formation or positional relations, therefore, by putting to when we speak of the way in which behavior of Fish A affects the
other kinds of objects in the aquarium, and by moving those along so behavior of Fish B. This is what happens in some of the fish that
that equivalent stimulation may be given to the fish. "establish" nests or regions where they deposit eggs, and so on. Some
W,ith regard to the phenomenon of milling, the establishing of cir­ of these fish, as Tinbergen has pointed out, and as others have pointed
cular mills in fish. Parr (15) developed a hypothesis which I think is out, the male fish, by its orientation toward a given place, now causes
still the best for consideration of circular mill formation in fish. He the female to respond and to thrust her head into the given region.
refers it to the problem of fixation, again the problem of distance, and This occurs even when females and males have been entirely separately
to the problem of the panoramic visual field of most fish, so that in the reared, and it also occurs when such organisms have never had any
schooling fish you have one eye that looks to the right and another eye chance to learn specifically to respond to a nesting place as such.
that looks to the left, and the twain do not encompass the same portions You find that this can be done with artefacts, too. If you take a small
or very much of the same portions of the visual field. This fish [indicat­ piece of wood and point it in the direction of the nesting place, the stim­
ing on board] is responding to a visual object here, and this fish to a ulus orients the female fish and she again begins to thrust her head in
visual object here, and the mill is established in the sense of maintaining there. It need not be a nesting place, I should like to point out to you;
a continuous relationship in space and in fixation, again, between Fish you could take the stick and point it at a side wall of an aquarium,
A and Fish B; so that the phenomenon of schooling and the phenome­ and the fish would begin to bang its nose against the wall of the aqua­
non of milling simply represent two different forms which can occur as rium. Or you can point it down or you can point it up, and the orienta­
a result of the same characteristic features of the sensory organization. tion that is established through this visual stimulus is persisted in with­
Pitts: When you talk about the position of maximum fixation, I am in limits by the fish, independently of what the specific environment is.
not quite clear about that, as to exactly what pattern of stimuli on its Of course, under natural conditions this has positive survival value.
visual field-—• It has survival value because it does tend to orient the animal toward a
Birch: I didn't say "maximum fixation." I spoke of optimal fixation given kind of nesting place where certain spawning will take place.
and clarity. But I think it is clear from these little experiments that the male fish
Pitts: Well, what does it do to try to maintain itself in its visual is not doing this because it wants the female to look there. Rather, it is
field? doing this and the female looks at it. What we have here is a very in­
Birch: Apparently it maintains, or functions to maintain, a sharpness teresting example of the way in which, in the course of evolution, sur­
of vision. vival has been promoted by the appearance of certain characteristic be­
Pitts: Of a single fish, or of several ? haviors rather than an example of the nice intentional behavior of a
Birch: Of single ones, usually. lower vertebrate with regard to its mate.
Pitts: It tries to get the next fish in a given position in its visual field I should now like to jump away from fish for a minute. Because of
and in focus? the lateness of the hour, I will not deal with bird display—song and
Birch: That's right. A very interesting experiment suggests itself. things of that sort, and I will skip the reptiles and the amphibia for the
It would be a direct test of Parr's hypothesis. What would happen if moment, although I should like to deal in detail with Armstrong's
the fish were blind in the right eye and the left eyes of all fish were i analysis, for example, of bird display (with which I disagree heartily),
and with Tinbergen's analysis of bird display (with which I disagree
158 Cybernetics Communication Between Animals 159
even more vigorously, if that were possible). I don't think I will take single chimpanzee to pull. Under those conditions, the first animal
the time to do that here, because it would be expressing more a pet pulled at the box and didn't get it in. The second animal pulled at the
peeve of my own than dealing with the problem of communication in box and didn't get it in. The first animal returned and pulled at it.
the vertebrate series. Then, over a long and complex course of interbehaviors, the animal
If we skip all the way from the lower vertebrates to the mammal, and learned to develop a solicitational relation with the second animal,
particularly to the primate mammal, we begin to find certain very real and the second animal developed a solicitational relation to the first
changes in the basis for interbehavior in animals. In the first place, animal, in the course of which, when such a. circumstance arose,
mammals in their relation one to the other generally have a character­ Animal A would wave to Animal B, or go and grab Animal B by the
istic which is not present in many of the lower vertebrates, although it hand and bring him to the situation. In some cases, he took his hand,
begins to appear in birds a little bit; that is, that the behavior of the put it on the rope, and then the two of them would pull. Under con­
sender animal or the transmitter animal is aimed or directed. Second, ditions in which A would be pulling and B wouldn't pull, A would
it is aimed at Animal B under certain rather specific conditions. Third, reach back and wallop B, under which conditions B would begin to pull,
the activity of Animal A in regard to Animal B is something which does too. Therefore, in a kind of co-operative-work situation, a set of ges­
not simply exist in the structure of the organism; that is, it is not some­ tures began to take on not only sign characteristics, that is, directional-
thing that is simply built into the organism or wired into it or whatever producing activity features such as we have in the lower vertebrate, but
term you want to use, but it is something which the organism has to they also began to take on significant characteristics. They began to
acquire in the course of its relations with other organisms or with its have meaning in terms of the interrelation between the two animals.
environment. In other words, we begin to move from what I would We find that in field observation of the primates, too. In a study both
call a phylogenetic kind of determination or an evolutionary kind of of the howling monkeys and of the red spider monkeys of Panama,
determination, in the historical sense now, of mechanisms of interrela­ Carpenter (18) has shown how, in the course of the life of many of
tionship, to what I would call an ontogenetic development of a com­ these primates, you have the development—the ontogenetic develop­
municative relationship; that is to say, when we begin to deal with the ment now, the learned development—in the specific life history of the
higher mammal, and particularly with the primate, we begin to deal given organism, or a relationship between experiences and gestures,
with learned patterns of activity that have been acquired by the given between features of the environment, let us say, and activity, which
animals in the course of their existence. We begin to move toward a begin to take on conventionalized significance. By "conventionalized
condition which is more homologous with the kind of thing we mean significance" I mean that these gesture-environment relations begin to
by communication between individuals, such as ourselves. be the property, not of any specific individual organism, but become the
What we mean, essentially, is this: the animal's behavior tends to joint property and the joint "learned" features of life in more than one
become directed; second, in this activity, certain content is transmitted; organism.
third, this content predictably elicits for the organism sending it certain Given this as a general sort of situation, we can also see this kind of
kinds of activity in other organisms, that is to imply, the organism that behavior in the feeding situation of young primates in captivity, who
is now sending has an expectation. It anticipates a behavior upon the establish a play group. We had one chimpanzee at Orange Park who,
part of the other organism. Further, certain of these activities become when feeding time came around, would always stay in the far corner of
conventionalized, and they begin to take on meaning in the specific life the enclosure and would not come up for his milk. This went on for a
histories of the specific organisms that are being dealt with. Thus, for period, until a very good social interrelation had been established
example, if we examine some phases of the behavior of primates, I among the animals, including food interchanges as part of the social
think the studies that were performed by Crawford and by Nissen on relationship. Under these conditions I observed that the chimpanzee
co-operative problem solving in the chimpanzee are extremely interest­ who tended to stay in the corner had developed a very close relation
ing in this connection (16) (17). with another highly active chimpanzee. When this had occurred, when
First, the two chimpanzees learned to pull rather light boxes and ob­ feeding time came, this partner chimpanzee would now dash off to the
tain food from them. The next task that they were given, after a varied far corner, grab his pal by the hand, and drag him or carry him, practi­
training series—I am skipping a lot of steps here—^was one in which cally, across the enclosure to the feeding place, and push him up against
they were placed together and given a box that was too heavy for a the wire mesh at the time of feeding; so that you had not merely the
160 Cybernetics Comniunication Betiveen Animals 161
way in which an animal's activity, occurring for its own reasons and tained for myself, at any rate, the very deep-lying prejudice that the
independently of another organism, affects the behavior of another initial world for these organisms, the perception and the interrelation
organism by modifying the general environment, but the interrelations of forms and space, of directions and space, are not the same. We have
between two individuals as well, where there is an interaction now be­ a whole series of problems that we must investigate if we are going to
tween the organism sending the signal, the organism receiving the sig­ look at communication. We want to find out what it is that an animal
nal, and then, reciprocally, between the receiver and the sender. is perceiving. We cannot assume that because Hayes's chimpanzee
I would submit that in a legitimate sense the kind of communication down at Orange Park can say "Papa" that that now means "Papa" or
that we usually refer to as communication at the human level represents anything of the sort. Personally, I think the whole direction of that
the extension and the elaboration of such ontogenetically developed study involves a failure to understand chimpanzees, but that is neither
components rather than the extension of behavior of the phylogenetic, here nor there. If we study the development of communication in chim­
stereotyped kind of components that we find either in the invertebrate panzees, it would be far more profitable to study the development of
series or in the lower vertebrate series of organisms, and that actually gesture symbolization, for example, and a few other things.
true communication, in my sense at any rate—not necessarily in the Pitts: Could you characterize, even in a vague way, the differences be­
sense of the communications engineer or in the sense of the physicist, tween these perceptual worlds Ì
but in the sense of a student of animal behavior—would be represented Birch: I will be very vague, because my data are entirely preliminary.
by this kind of level of interdependent communication that has direc­ This is work conducted from five o'clock in the morning until seven-
tion, that involves the process of anticipation, and that involves the thirty in the morning, through the summer. The problem is that vis­
process of conventionalization of sound. itors come into the zoo and begin to interfere with your legitimate
This means, then, that the study of the evolution of the communica­ activities, or you with theirs, after a certain hour of the day. But at
tion process at the mammalian level requires not the examination of least you can get some general notion.
communication itself, but the study of the learning process, of the per­ For example, in patterned string problems, in the subanthropoid or­
ceptual process, and of the process of social interrelationship. It means, ganism, there appears to be a greater tendency to follow the direction of
then, that we have not one problem but a series of problems which the relation of the animal's body motion to the food than to follow the
must be investigated. visual line which existed between the food and the animal. You see,
What kind of learning is necessary for an organism to generalize, to you had food out in cylinders, and string patterns. Food would be at­
abstract from concrete experience, certain gestural relations which are tached to one of the strings, no food to the other, and then there would
relevant to these experiences? So we have, then, the problem of the be cross-strings.
study of the evolution of intelligence in organisms; that is to say, the In the lowçr orders of primates, there appeared to be a tendency to
study of the development of modifiability. Further, we should like to begin to reach rather directly for the grape or the cherry or whatever
know what is the perception of the world by these different organisms. food it was, and to establish a linear relation between itself and the
How do they see it 1 food, rather than to begin to respond to the lines that I had introduced,
I was engaged, while working as a Behavior Research Fellow for the the lines of connection that I had introduced into the environment.
New York Zoological Society, in a series of studies involving an ag­ That kind of process, whether the animal is responding in terms of its
glomeration of primates at the Bronx Zoo. One thing in which I was own bodily orientation and spatial distance from the food or whether
very much interested was this: Is the perceptual world of the gibbon, it is going to respond to the organization of visual field as you present
for example, like the perceptual world of the chimpanzee Is the per­ it, became one of the problems that I wanted to investigate, and one
ceptual world of the young orangutan like the perceptual world of that emerged out of these preliminary activities. Now, such a study,
young chimpanzee What does it see Ì What does it respond to Ì What or such studies, would have to be done, and I think they would be
is the organization of the world } Can we analogize from our own per­ highly profitable, if they were carried through. At the moment I don't
ceptions to the world of these animals ? I thought we could not, so I know of any place where they are being carried on.
started to test them on patterned string problems of various kinds, and Pitts: How about the differences between the chimpanzee and the
I got the most astounding differences. Unfortunately, the zoo is not a orangutan and possibly the young gorilla.?
good place to do systematic research, but in a preliminary way I ob­ Birch: I don't have enough data. I just have suspicions.
162 Cybernetics Communication Between Animals 163

Pitts: What sort of suspicions ? to the male. The male just sat in the corner and crossed his legs. The
Birch: What is important is whether it is visual stimulation as con­ female moved a little bit closer to the male, presented again, assuming
trasted with olfactory stimulation, whether it is taste as a basis for so­ this so-called "subjugated" position. The male ate a piece of food that
cial relationship or touch, or contact. Do you see what I mean ? What was lying on the floor and ignored her. The female presented a third
is the basic structure, the nature of this organism, out of which these time, and the male ignored her further, at which point she turned
complex ontogenetically developed patterns are developed? Those are around and beat the living daylights out of the male. She just tore into
problems that we still have to investigate. him, bit him, scratched him, and kicked, so we had to turn the hoses
Kubie: Have you any data on facial expression ? on them to separate them.
Birch: There are some data on facial expression but, again, not sys­ Savage: "Hell hath no fury."
tematized. Birch: Exactly. At any rate, this gesture, as such, may have mean­
Kubie: That seems so primitive in the human infant. ing, ontogenetic meaning, as a submissive gesture, under certain social
Birch: Yes. Well, it appeared to be a fundamental problem for circumstances for the chimpanzee, but it has significance now as a de­
Darwin (19), but it has not been picked up and carried through, in his mand gesture at a different point in the relationship.
whole initial study or rather, discussion of the relation of expression Klüver: In connection with problems of "comparative" psychology
to emotional states and feeling states which you certainly begin to see and, more particularly, the problem of comparing the perceptual worlds
in the chimpanzee. You have species-type gestures, for example, spe­ of different animals, it is pertinent to recall that most investigations in
cies-type expressions and general bodily-type orientations which have the field of sensory psychology have been concerned with threshold de­
signaling value to other organisms. terminations, for instance, with determining visual, auditory, gustatory,
I should like to end with just one further precautionary note in deal­ and so forth, acuity. Such studies have shown, for example, that there
ing with some of the materials on primates. It is very easy to find cer­ are no great differences in the sensory equipment of man, apes, and
tain gestures that are attached to specific situations in the case of pri­ monkeys. There is no reason to believe that even the sense of smell is
mates and then to make the assumption that these gestures then refer more highly developed in monkeys than in man. And yet, a careful ob­
specifically to these situations rather than to understand that a specific server, in watching the reactions of different primates to sensory stimuli,
gesture for the primate is interpreted by a second primate and has will often gain the impression that different species or genera differ
meaning to the first primate itself only in terms of the surrounding cir­ markedly in their reactions to the stimuli of the external environment.
cumstances and the situation, in a sense, the environment in which the Qualitatively, there seem to be some striking differences in the reac­
gesture is being made. I should like to highlight this with one example tions of a marmoset, squirrel monkey, and chimpanzee, for example, to
from my own experience. stimuli of the visual environment, and yet, quantitatively, these differ­
Students of the social behavior of primates have customarily referred ences are often not caught by existing methods. Why?
to the "female presentation gesture," the presentation of the genitalia If we examine the methods of sensory psychology, we discover that
on the part of either the male or the female, to another dominant or­ most of them are concerned with measuring the animal's ability to cope
ganism as a sign of submission, and that once the animal does this, with the discriminable aspects of the external environment. In effect,
then it is clearly a submissive animal, and the animal that does the the intent of such methods is to answer the question of how small we
mounting is clearly the dominant animal. It seems to me that that is may make certain stimulus differences and yet obtain a differential re­
an extrapolation into biological theory of a mid-Victorian or perhaps sponse. And in trying to answer this question, the classical procedure of
earlier conception of human sexuality, in which the assumption is that keeping all variables except one constant is generally being followed.
females engaging in sexual behavior, particularly if they are primates, However, the net result of employing such methods seems to have been
are really suffering some kind of abnegation. There is the assumption that no significant differences between the perceptual worlds of differ­
that some degradatory phenomenon is involved. ent primates, let us say, between the visual world of a squirrel monkey
Well, in studying hormonal effects on social relationships of the and that of a chimpanzee, have been discovered. To detect such differ­
chimpanzee, I paired a large male chimpanzee with a very actively ences we must choose, I believe, an entirely different point of departure.
mating female. The male had never had any mating experience. The Instead of testing the ability of the organism to cope with discriminable
female was put into this cage, and the first thing she did was to present aspects in otherwise homogeneous stimulus situations, we should con-
164 Cybernetics Comrnunication Between Animals 165
cern ourselves with identifiable aspects in heterogeneous situations, that zontal is to give himself an opportunity under controlled conditions to
is, we should analyze the ability of the organism to identify or isolate a observe the way in which changes in polarization may affect the orienta­
certain factor in stimulus situations exhibiting widely different proper­ tion.
ties. Instead of testing the ability to discriminate small differences in Gerard: Well, I agree, if it occurs naturally, it is not the same sort of
brightness, color, size, shape, and so forth, we should inquire whether problem. It still seems to me rather a neat problem that the bee can do
large, numerous, and complex stimulus differences do or do not inter­ it in either of the modalities. That takes a bit of doing.
fere with the animal's ability to identify or isolate a given brightness Another, unrelated, question is with respect to the fish-school story.
or color or a particular brightness or color relation. Differently ex­ I once spent a moderate amount of my time lying on the pier at Woods
pressed, to discover differences between the perceptual worlds of dif­ Hole, watching the fish schools go round, and one of the problems that
ferent animals, the most promising approach, as far as I am able to has always intrigued me is the cue and the timing involved in a sudden
judge on the basis of work done so far, is to study, not the animal's shift in direction. As far as I could tell from observing them visually,
ability to attend to one aspect in an otherwise nondifferentiated stim­ there was no wave of change from a leader or from any other one fish.
ulus situation, but to study the ability to attend to one or even several They all moved simultaneously.
aspects in multidifferentiated stimulus situations. It is for the latter I was once able to check that in the case of birds. A flight of birds
purpose that I developed "the method of equivalent and nonequivalent was going along parallel to my car, so I could time them. I happened
stimuli" and, when employing this method, had no great difficulties in to be watching them as they all veered away, and I would certainly
finding differences, for example, between durukulis, marmosets, squirrel have seen one bird go forward or drop back relative to the others if its
monkeys, spider monkeys, and rhesus monkeys in their reactions to timing was off. As I remember, I calculated there was less than five
stimuli of the visual environment (20). milliseconds possible time for cueing from one to another. I would like
Gerard: I have been waiting for the floor for a long time. May I any evidence on that question that you have.
have it ? I want to ask three questions of the speaker that are relevant And let me ask my third question now, so it doesn't get lost. This
to the things he has developed. The first one is immediately in line turns to primate behavior and its relation to human beings. I suppose
with what Dr. Klüver was just pointing out. everybody is worrying about it because it is the basic problem. Is there
While I would confess freely that the detailed physiological mecha­ any valid evidence in animal groups, not of learning, that certainly oc­
nisms of the translation of visual orientation to gravitational orienta­ curs, nor of teaching, as a man can teach a chimpanzee or as one chim­
tion, or vice versa, are not known, it somehow doesn't bother me partic­ panzee can teach another, but of socially propagated teaching ?
ularly because I think it is equivalent to other situations with which we Birch: Socially, from generation to generation ?
are familiar, and therefore don't worry about, or, in another sense, Gerard: Primarily that.
worry about all the time. It would, in principle, be no different from Birch: You mean a culture?
the reaction of posture to vestibular impulses; it is built in, in some Gerard: That's right. In other words, social inheritance of habit
way. But what does worry me is one of the experimental aspects of patterns. Is there any positive evidence of that ?
this, unless I misunderstand you; namely, that when the hive is hori­ Birch: I will leave the first question, which we discussed a moment
zontal, and appropriate polarized light is allowed to enter, the bee then ago, about the naturalness. Let's deal with the problem of schooling
makes his dance, not by translating his previous visual cues into gravita­ and cueing in schooling. I left out one factor that has been referred to,
tional ones, but now in terms of a visual cue ? at least by Breder and some of his associates, in the development of
Birch: That's right. schooling itself. He once took a group of young jewel fish, which were
Gerard: And that is not normal. This is, therefore, something that schooling, from one pond environment and brought them into a
seems to me to present a very real problem. museum or aquarium environment. Under those conditions, the school
Birch: It is normal because the bee, when landing at the entrance was re-established. He inferred from that that he had a very interesting
to the hive, for example, will not infrequently go into a dance and in­ problem—which I must point out to you has not been systematically in­
dicate direction that is related to the polarization characteristics of light. vestigated. All we are doing today, probably, is opening up areas for
It is now on the horizontal. It is not vertical. That is natural, in a investigation more than we are answering questions. But apparently
sense, therefore. All that von Frisch has done by putting it on the hori- there are environmental features, as well as interindividual-stimulated
166 Cybernetics Communication Between Animals 167

features, which produce these characteristic school or flock relation­ Bigelow: There would be an appropriate electrical pickup on every
ships in these organisms as well. person.
Our attention has been directed, because we have been interested, in Savage: Perhaps within a millisecond of each other, not within a
part, in the problem of communication, at the way in which activity of millisecond of the gunshot.
animals within the school influences the behavior of other animals with­ Gerard: Yes, within a millisecond of each other. That is just what I
in the school. An equally legitimate question would be the way in doubt.
which identical behaviors, which would then maintain a relationship Savage: How long does it take a man to respond to a shot like that ?
within a school, could be produced and are produced by environmental How long from the gunshot to his response ?
changes having identical effects on the different organisms, whether it Rioch: Two-fifths of a second. t
be your car or any other feature. Savage: Well, a couple of hundred milliseconds, so it implies a syn­
Gerard: Well, say one millisecond. chronization of a half per cent or so.
Birch: If one of them could respond within one millisecond, there Gerard: That is so, and that is an extremely abrupt and vigorous
is no reason why another one could not respond within one millisecond. stimulus, which apparently was not happening with the birds.
What I am talking about now is that each may be responding individu­ Klüver: If I remember correctly, the observations on birds you men­
ally; that is to say, to the same feature of the environment rather than tioned a minute ago were described by Gerard in Science several years
to one another. ago. Is that correct?
Gerard: Yes. All I am saying is that the simultaneity must be there. Gerard: That is correct.
I don't know the time of response. Klüver: Some ornithologists have studied the way birds flying in
Birch: Simultaneity can be the function, and usually is the function, flocks take to wing synchronously. Apparently the taking to wing is
of a given environmental change which is simultaneously affecting the preceded by a great deal of mutual stimulation, that is, by certain initi­
different animals, under which conditions the simultaneity of response ating or "intention movements" which are "understood" by all fellow
or the lack of simultaneity of response is a function of the reaction members of the species.
time of each of the individuals therein. Gerard: That is right; so you are suggesting that there must have
Gerard: I would question—and I could be completely wrong; in­ been some further communication?
deed, I don't know what other answer there is beside the kind you are Klüver: As Lorenz undoubtedly would express it, the finely graded
giving—but I would question whether, when you have trained a group intention movements and movement patterns preceding the flight are
of like organisms to respond to a particular signal, you would find the "releasers" switching on the flying reaction in social birds.
response of the whole group so beautifully synchronized as that. Gerard: Which is what I would suspect.
MacKay: I am not quite sure why you feel it must be a millisecond, Birch: I would raise the question of your perceptions. I think part
or as short as that, because, suppose you have a statistically scattered of what Dr. Klüver is saying is this: if you take photographs or mov­
response, you have at the same time continuous feedback from all the ing pictures of bird groups starting off, and then do frame by frame
positions of all the neighbors, and any slight lag in one creature would analysis, you find that what you viewed perceptually as a whole group
very soon be fed back. going off simultaneously shows you birds in many, many stages of ac­
Bigelow: You still need terms like "millisecond," though, not to tivity, all of which would indicate that there is a wider range of reac­
deviate too far. There must be a lag-correcting operation there which tion times in those organisms.
is very close to this sort of magnitude. May I answer the last question on culture ? This question has in­
MacKay: In a nonlinear system of this sort the rate of change is the trigued many people, and Yerkes as you may know, for years made a
important thing. It is really a question of how many bits of information systematic effort to find a culture or a prototype of culture in the higher
you need, and how fast. primate. He ended up with one instance that I can think of; that is,
Bigelow: But to get back to the point, isn't it true that if you fired off that in the wild, the chimpanzee tends from generation to generation to
a gun in this room, everybody would jump within a millisecond ? sleep in a sort of bower-type hut, in which branches are bent over and
Gerard: We wouldn't within a millisecond. I doubt if we should all sat down upon, but males raised in captivity do not have this, and it
jump. does not apparently develop spontaneously when they are left in wild
168 Cybernetics Communieatton Between Animals 169
situations. However, nobody has ever studied the captive-raised chim­ on which leads the human to that goal, and that this mechanism is
panzee in a vi^oods situation under conditions in which it could spon­ very different from the mechanism used by the amoeba—I find quite
taneously develop this kind of activity over a long-enough period to impossible to sustain. Now, if you can give me some criterion by which
make it meaningful. Further, the captive chimpanzee is not at home in I can apply that distinction, I should like to hear it, because I cannot
the woods. find it myself. I made my first point, which was that it seems to me that
Rioch; How do you regard the training of the young primiparous some distinctions are rather artificial. I don't see how one can give
chimpanzee by the multiparous chimpanzee in the acceptance of the first them any sense unless one introduces into one type of behavior notions
baby, and the training of the young male by the old female to copulate ? that cannot be used or measured. That brings me to my second com­
Birch: First of all, I would deny the training of the primiparous by ment. Dr. Birch spent a lot of time trying to prove to us that animals
multiparous. I know of no good evidence whatsoever that that takes do not have "intelligence"—in quotation marks—in the sense in which
place. I have, in my own observations of chimpanzees, observed that he was using this term. He even used the word "genius," which needs
when an old multipara or any other chimpanzee comes near a mother more quotation marks than the word "intelligence." Nobody can prove
chimpanzee with its infant, the response is not one of training but is a or disprove that an animal or a man does or does not have intelligence.
fighting kind of response. It seems to me absolutely irrelevant to the problem being studied—the
Rioch: This was at the time of birth that you observed them ? problem of communication. The question of intelligence is something
Birch: Oh, yes, I have made observations at birth, where animals that is going on in the mind. I don't see what bearing it can have on
were in adjacent cages. Some were primíparas and some were multi­ the problem of communication. The only way we can get together with
paras, and there is not that tutelage relationship that has been some­ other people and observe other people and make ourselves come in con­
what romantically described. In the learning to mate, all you have to tact with them and receive contacts from them is in terms of their be­
do is watch a young male begin to mate. What the female does is to havior. That is the only thing we can see, that we can judge; and that
maintain a position in response to the male stimulation and keep orient­ is going to make an impact on us.
ing toward the male. What the male learns, he learns on the basis of Dr. Klüver mentioned the danger of gross analogies. Well, I don't
his own errors and his own activities rather than on the basis of the know; I personally feel that the problem of other people's minds be­
specific tutelage. I don't think you can read tutelage into it. longs in that same realm. It is one of the grossest analogies. It is an
Rosenblueth: I find myself in disagreement with many of the state­ indispensable one. I think we all assume these minds; that is why we
ments that Dr. Birch made because I think they are irrelevant to the are here. On the other hand, it doesn't belong with the problem that
problem he was considering. There is a false distinction, I think, in we are considering. We can dismiss it entirely.
his introduction, in the criteria which he adopted. The distinction be­ As to the question of whether communication among animals is in in­
tween the feeding behavior of the amoeba and that of the human being telligent communication, first, it cannot be discussed because nobody
I find quite impossible to maintain. He decided that they were different can either prove or disprove intelligence. I don't know of any operation
types or modes of seeking food, because the amoeba could be fooled by which we can judge whether or not such a thing exists. Second, it is
into seizing a particle of potassium permanganate. Of course human quite irrelevant to the problem. When we describe the behavior of
beings are also fooled into seizing material which is not food. The no­ lower organisms or of machines (as has been done very often in this
tion that feeding is something specific, that it is not just a movement in group), we use terms which can be qualified as mentalistic. I don't
relation to something in the surroundings of an organism, I don't think think there is any special objection to this if it is understood that the
can be upheld. A child can eat things that are not food, and even adults terms are used merely for convenience. When Pavlov tried very serious­
have been poisoned because of eating material that was not proper food. ly to dismiss any terms with a psychological implication in his descrip­
Both the amoeba and the human being respond to certain stimuli which tion of the studies he was carrying out on animals, there was one point
may have not been identical in the two cases. That is not surprising. at which he broke down, at which he violated the law that he had es­
We are dealing with different organisms. All they were doing was re­ tablished, and which led him to become separated from several of the
sponding to certain stimuli and then reacting with the acts that corre­ collaborators that had joined him in his work, because they were not
sponded to their own organization. But the distinction in terms of seek­ able to avoid the use of terms of that sort, and they became very upset
ing food—and I take it that means there is a conscious process that goes because they were not used in the study. He broke down when he used
170 Cybernetics Communication Between Animals 171
the term "experimental neuroses." But, of course, it is quite clear from on that of another, B. By that, I take it we can mean any organisms we
Pavlov's writing that all he meant was a verbal shortcut. wish. With either definition, it may be that we will find there are dif­
Kubie: He even used the word, "unconscious" (21). ferent types of communication, and that some are inborn and some are
Rosenblueth: Then he broke down twice. He may have broken down learned. I am sure that in the human being, among human communica­
many times, but he could have done it quite consciously and there tions, there are many which are not learned, which are inborn. Certain­
wouldn't have been any objection. ly, some of the sexual behavior reactions, some of those exhibited at
If we say that a machine has a memory, what we mean is that in so moments of emergency, and many other reactions belong in that cate­
far as we can describe in objective terms what we mean by memory, gory. That is why I said that the restriction to learned behavior is quite
that can be put into a machine. When we say the machine learns, what arbitrary.
we mean is, again, that in so far as we can state in objective, accurate It may be interesting from one standpoint to know what may be the
terms, independent of our own personal and private experience, what responses to inborn messages as opposed to learned reactions, and it is
we mean by learning, that can be put into a machine. There is no ob­ an important distinction from many standpoints. But it is not a partic­
jection to using those terms so long as it is understood that they are go­ ularly important division from the standpoint of the problem of com­
ing to have a particular realm of application and a particular set of munication. The group of messages studied should be quite inclusive.
meanings, so that actually— And then we may perhaps classify it, but if we do that, it is not de­
Bigelow: May I add to that when you are through? sirable to adopt a very sharp and fine distinction made largely on the
Rosenblueth: Go ahead. basis of something like "anticipation" or "direction" or "content."
Bigelow: There is one further point I should like to air a little bit, Those are terms which I don't think Dr. Birch or anybody else can
on the story about putting a soup made of a starfish near a scallop. This define.
story is very amusing, and it seems to reduce the question of communi­ Gerard: I should like to comment on all three points. I simply want
cation to an absurdity. On the other hand, I don't think that one can to say, on the first point, I am inclined to agree with Dr. Rosenblueth
argue that the fact that communication is reduced to a mechanically that there is a good deal of unnecessary verbalism in the distinction. On
ridiculous process means that it is not communication. The definition the second point, I should like to defend Dr. Birch. I don't think he
lies elsewhere than in that fact. did any of the things you accused him of. Dr. Rosenblueth, and when
Bigelow: A phonograph record is certainly a mechanical device he spoke of a genius in the chimpanzees, he was talking figuratively,
which can elicit communication from a human being. just as when he said the male went in the corner and crossed his legs.
Rosenblueth: I think also that the definition adopted by Dr. Birch is I don't believe that actually happened, either.
much too restricted and too narrow. It would be inapplicable even to Birch: It did.
human beings. If he is going to restrict it to that, he is going to elimi­ Gerard: Well, then as one of the other things the chimp did, not in
nate certain things which we all want to include in the group of inter­ the sense that a man would.
relations. I t-hink "intelligence" was used behavioristically and that it is a per­
If we should adopt his criterion, the fact would remain that there are fectly good word to use to describe such behavior. There is no nec­
other types of messages sent by organisms of the same or of different essary imputation of what is going on inside.
species which are usually included under the general term of "communi­ Rosenblueth: I think that is a hopeless task. You can neither prove
cation," and which are worth studying; we should not eliminate them. it nor disprove it.
We know the physiology of some reactions, and if we are going to Gerard: You said in your last sentence or two that which I think
postulate that this knowledge implies that the reactions are no longer invalidates much of what you said in the rest of your comment. It is
communications, but belong to some other category, I don't think we purely a semantic matter—whether or not one wants to call something
gain anything. It seems to me that the proper way to approach the communication. What is important is that we recognize various cate­
problem would be to take a very broad definition of communication. gories, and that is what Dr. Birch was trying to do. Whether one calls
If one wishes, one might adopt one of the expressions that Dr. Birch them all different phases of communication or calls one communication
used, such as "interrelations between organisms." That is one possibility. and one not communication is, I think, a trivial matter.
Or we might define it as the influence of the behavior of an organism A Birch: I think that the last remark of Dr. Gerard is the most perti-
172 Cybernetics PRESENTATION OF A MAZE-SOLVING
nent one. What I was saying did not mean that there were no un­ MACHINE
learned behaviors in the human being. It simply meant that at different
levels of the evolution of animals, we have pre-eminently present meth­
ods of interrelation with other animals which are not the same, and that CLAUDE SHANNON
there is a difference between the ontogenetic acquisition of communi­ Bell Laboratories, Murray Hill, N. J.
cative devices and the phylogenetic emergences of those. Now, what
I was trying to do was to give a picture of the way in which interrela­
tions between animals could take place in a variety of ways, and not to
deny that such interrelation was taking place. If the human is fooled, THIS IS A maze-solving machine that is capable of solving a maze by
he is fooled in a different way, and because of different mechanisms trial-and-error means, of remembering the solution, and also of forget­
from the way in which an amoeba is fooled. If Dr. Rosenblueth cannot ting it in case the situation changes and the solution is no longer appli­
distinguish between these ways, that is unfortunate; nevertheless, these cable. I think this machine may be of interest in view of its connection
ways are different. with the problems of trial-and-error learning, forgetting and feedback
Pitts: Not always. systems.
Rioch: But that is not the problem. As you can see (Figure 8), there is a maze on the top panel of the
McCulloch: Well, let's get Bowman's data before us now. machine which has a range of 5 x 5 squares. The maze can be changed
Bowman: There is a black beetle which is found in fairly large—I in any desired manner by rearranging the partitions between the twen-
won't say social groups, but in large colonies—under rotten stumps. ty-five squares. In the maze there is a sensing finger, which can feel the
You will find at least ten and sometimes up to several hundred of them, partitions of the maze as it comes against them. This finger is moved
if you find one. Along with the adults, there are always various im­ by two motors, an east-west motor and a north-south motor. The prob­
mature stages. The beetle itself probably has very few enemies. It is lem facing the machine is to move the finger through the maze to the
very heavily armored and has extremely strong mandibles, but the goal. The goal is mounted on a pin which can be slipped into a jack
larvae and pupae are practically defenseless. If you open a stimip that in any of the twenty-five squares. Thus you can change the problem
contains these beetles, you will hear a rather high-pitched hiss, a sort any way you choose, within the limits of the 5x5 maze. I will turn it
of -whistle, at which time all of the soft-bodied forms will head toward on so you can see it, in the first place, trying to solve the maze. When
the center of the stump, and all of the adult beetles will face outward. the machine was turned off, the relays essentially forgot everything they
The adults themselves can make that sound. If you just wait and watch, knew, so that they are now starting afresh, with no knowledge of the
they will soon resume normal activity, but then you can, without further maze.
disturbing them, give that same note and they will perform that same Savage: Does than mean they are in a neutral position, neither to the
protective defensive act. right nor the left ?
Pitts: Which family of beetles is this Ì Shannon: They are in a kind of nominal position. It isn't really a
Bowman: A species that falls in a family all to itself—^passalus. neutral position but a meaningless one.
They are big black things. You see the finger now exploring the maze, hunting for the goal.
Birch: What is the frequency? When it reaches the center of a square, the machine makes a new deci­
Bowman: It is around 6,000 cycles, I should guess. You can make sion as to the next direction to try. If the finger hits a partition, the
it with your mouth. It is quite sharp. If you whistle a slide over a motors reverse, taking the finger back to the center of the square, where
range, you get response at that one pitch. a new direction is chosen. The choices are based on previous knowl­
edge and according to a certain strategy, which is a bit complicated.
Pitts: It is a fixed strategy ? It is not a randomization ?
Shannon: There is no random element present. I first considered
using a probability element, but decided it was easier to do it with a
fixed strategy. The sensing finger in its exploration has now reached
173
174 Cybernetics
Maze-Solving Machines 175
the goal, and this stops the motors, lights a lamp on the finger, and
rings a bell. The machine has solved the maze. I will now run the
INDCXCO
finger, manually, back to the starting point, and you will see that the
TLEXIBCE CABLE
BL- 3734 7A machine remembers the solution it has found. When I turn it on, it
CARALA&C WHEELS
TO CARRIAGE \
goes directly to the goal without striking the partitions or making side
excursions into blind alleys. It is able to go directly to the goal from
any part of the maze that it haS visited in its exploration. If I now
move the finger to a part of the maze that it has not explored, it will
fumble around until it reaches a known region. From there it goes di­
rectly to the goal.
^ \ EAST-WEST
Now I should like to show you one further feature of the machine.
CENTERING
SWITCH I will change the maze so that the solution the machine found no longer
works. By moving the partitions in a suitable way, I can obtain a rather
O O o o interesting effect. In the previous maze the proper solution starting
from Square A led to Square B, then to C, and on to the goal. By
changing the partitions I have forced the machine at Square C to go to
O O
PARTITIONS
\ O •o o a new square. Square D, and from there back to the original square, A.
SENSING
TINGER When it arrives at A, it remembers that the old solution said to go to
B, and so it goes around the circle A, B, C, D, A, B, C, D. . . . It has
O o o o o established a vicious circle, or a singing condition.
Gerard: A neurosis.
JACKS rOR GOAL
Shannon: Yes.
O o o (/ \) Savage: It can't do that when its mind is blank, but it can do it after
it has been conditioned?
Shannon: Yes, only after it has been conditioned. However, the
o o o o o machine has an antineurotic circuit built in to prevent just this sort of
situation.
Mead: After it has done it a number of times ?
INDICATING LAMPS
Shannon: After it has gone around the circle about six times, it will
o o ''' ^o ^
UOVE 1
1=) O O 0 0 o o o o o o o break out. The relay circuit includes a counter which stops this be­
COUNTER
o o o havior at the twenty-fourth count.
AUTOMATIC O O
Frank: How many relays are there in it ?
OO
STEP BY STCP
©O
© o o o Shannon: All told, there are about seventy-five relays.
Savage: It doesn't have any way to recognize that it is "psycho;" it
BUTTON just recognizes that it has been going too long ?
Shannon: Yes. As you see, it has now gone back to the exploring
C. R.
DTA..FCB.I3JMI
strategy.
Teuber: Now, does it have to relearn the entire maze, or can it still
FIGURE 8
utilize some form of it?
Shannon: No. As it stands, it can't utilize any information it had
before.
Savage: But it is trying to utilize it, I suppose. It is moving as it
would move.
176 Cybernetics Maze-Solving Machines 177
Shannon: As a matter of fact, the old information is doing it harm. a blind alley must be left by way of the same square through which it
Bigelow: I think it's getting to it. was entered, the direction D retained for that square will necessarily
Shannon: Yes, it is gradually working over toward the goal. I lead to the goal directly rather than by way of the side excursion into
should like to spend the rest of my time explaining some of the things the blind alley. In a similar way, if the machine follows a circular or
which are involved in the operation of the machine. re-entrant path in exploring its way to the goal, the direction retained
The strategy by which the machine operates can be described as fol­ for the last fork in this path must be that going to the goal rather than
lows: There are two modes of operation, which I call the "explora­ around the side loop. As a consequence, the machine follows a fairly
tion strategy" and the "goal strategy." They are both quite simple. direct path to the goal after it has first found its way there.
The exploration strategy is used when it is first trying to find the goal. The final feature of forgetting is obtained as follows: After reaching
For each square in the maze, there is associated a memory, consisting the goal, suppose we move the sensing £nger to a different point in the
of two relays. These are capable of remembering one of four possible maze and start it operating. The machine then starts counting the num­
directions: north, east, south, or west. The direction that is remem­ ber of moves it takes, and if it does not reach the goal within a certain
bered for a square is the direction by which the sensing finger left the specified number of moves, which happens to be twenty-four in this
square the last time it visited that square. Those are the only data the case, the machine decides that the maze has been changed or that it is
machine remembers about the course of the finger through the maze. in a circular loop, or something of that sort, and that the previous solu­
There are some other memory functions in the computing part of the tion is no longer relevant. The circuit then reverts to the exploration-
circuit, but these remembered directions are the data which allow it to type strategy which is mathematically guaranteed to solve any finite
reproduce its path at a later time. solvable maze. ' •
Now, let's call the remembered direction for a particular square, D, There are a few other points about the machine which may be of
considered as a vector. In exploration strategy, the machine takes the some interest. The memory is quite undifferentiated in the sense that
vector D and rotates it 90° as the first choice when it comes into a I can take the group of wires leading from the rest of the circuit into
square. For example, suppose it left a square in the easterly direction the memory, shift them over either in the north-south or east-west di­
at the last visit. If it comes to that square again, it will try the north­ rections, and the machine will still operate correctly, with no significant
ern direction as the first choice. If it hits a barrier and comes back, it change, although the data corresponding to a square are then stored
again rotates 90°, because it has just put this northern direction into in a different part of the memory.
the memory, and, advancing 90°, it tries the westerly direction, and so Another point is that there are, of course, a large number of feed­
on. The choices progress around counterclockwise, starting with the back loops in this system. The most prominent is the feedback loop
direction by which it left the square last time—^with one exception: it from the sensing finger through the circuit to the driving motors and
also remembers the direction by which it came into the square at the back to the sensing finger, by mechanical motion of the motors. Nor­
current visit, and on the first rotation of the vector D, it skips that di­ mally, if you have a feedback loop and change the sign of the feedback,
rection of entrance. This is to prevent the path repeating too much. Be­ it completely ruins the operation of the system. There is ordinarily a
fore that feature was installed, there was a tendency to explore up to a great difference between positive and negative feedbacks. This maze-
new square, go back through the entire maze, and then go one square solving machine, however, happens to be such that you can change
further, and so on; and it took a very long time to solve the maze. It either or both of the signs in the feedback connections, and the machine
required about three times as long as it does now, with this skipping still operates equally well. What it amounts to within the circuit is
feature added. that the significance of right and left is interchanged; in other words,
When it hits the goal, a relay operates and locks in, and the machine the effect on the strategy if one of the feedback loops is changed is that
then acts according to the goal strategy, which is also based on this the advance of 90° counterclockwise becomes an advance of 90° clock­
vector D. wise. If both of them are changed, the strategy is not altered.
In the goal strategy, the machine takes as its first choice direction D, Von Foerster: If there are two different ways to reach the target, cer­
which is the direction by which it left the square on its last visit. This tainly the machine is only able to find one. Does the possibility point
is very simple to do, and it has many convenient features for maze to its making a choice of the better way Ì
solving, because it cancels out all blind alleys and circular paths. Since Shannon: No, it does not necessarily choose the best way, although
178 Cybernetics Maze-Solving Machines 119
the probabilities are in favor of its choosing the shorter of two paths. of different solutions of the maze as well as additional computing re­
Incidentally, the exploration strategy of this machine will solve any lays to compare and evaluate them. It surely could be done, but it
maze whether it be simply or multiply connected. Some of the classic would be more difficult; it would mean a much more complicated ma­
solutions of the maze problem are satisfactory only in the case of sim­ chine than this.
ply connected mazes. An example is the method of keeping your hand Savage: And it would have to decide when to invest the effort to
always on the right-hand wall. While this will solve any simply con­ seek a new path. That is really a very important problem in any kind
nected maze, it often fails if there are closed loops. of real human learning. If you can already peel a potato, why should
Savage: This cyclical feature that you illustrated occurred because you take the trouble to find a better way to peel it? Perhaps you are
the machine was not then in really searching condition ? already peeling it correctly. How do you know ?
Shannon: No, it was in the goal strategy rather than in the explor­ Von Foerster: What happens if there is no goal ?
atory. Shannon: If there is no goal, the machine establishes a periodic path,
Savage: A goal strategy is to go the way you last went, but what are searching for the goal; that is, it gradually works out a path which goes
you to do if the attempt to do that is frustrated Ì through every square and tries every barrier, and if it doesn't find the
Shannon: Then it returns to the center of the square and advances goal, the path is repeated again and again. The machine just continues
90° and tries that direction. But it still remains in goal strategy. looking for the goal throughout every square, making sure that it looks
Savage: I see. When it gets into the next square, it tries to go ahead at every square.
in the accustomed direction ? Frank: It is all too human.
Shannon: That's right. The purpose of this is that it may have Brosin: George Orwell, the late author of 1984, should have seen
learned most of a maze in its first exploration, but not quite all of it. this.*
If we put it into a square it has not visited, it explores around by trial Von Foerster: And after that ? For instance, if you put a goal into
and error until it reaches a familiar square, and from there goes direct­ the path after the machine has established such a periodic motion, what
ly to the goal. The previously unknown squares have by this process happens then ?
been added to its previous solution. Shannon: When it hits the goal, the machine stops and changes into
Bigelow: You can then put new loops on any known path; it will the goal strategy, and from there on it goes to the goal as placed there.
learn those new loops immediately and not get into trouble. Is that Incidentally, it is interesting to think of this—if I can speak mathe­
right ? matically for a moment—in the following way. For each of the twenty-
Shannon: That's right. five squares, the memory of the machine retains a vector direction,
Bigelow: Because when you come back to the main stream, the north,- east, south, or west. Thus, as a whole, the memory contains a
search goes in the right direction, if it recognizes that square. vector field defined over the 5x5 maze. As the sensing finger moves
Shannon: I am not sure I understand what you mean. through the maze, it continually revises this remembered vector field in
Bigelow: It forms a single-directional path. Now, then, if you in­ such a way that the vectors point along possible paths of the maze
troduce a new path which brings it out of the known path into strange leading to the point currently occupied by the finger.
territory, back into the known path again— Teuber: If you rotate the field through 180°, would it continue to
Shannon: Such a side path is completely canceled when it has gone function ?
into the goal strategy. McCulloch: Suppose you reverse the connections and leave the
Bigelow: But once you start it around that circuit, then the procedure motor, so that you reverse your direction of rotation; can it still find
is correct after the starting point. its way?
Shannon: If it is in goal strategy, yes, but not in exploratory. Shannon: Only if I reverse some switches within the machine which
Bigelow: What would you have to do to minimize running time—in tell it what square it is currently occupying. If I reverse the motors, I
order to make it learn on repeated trials eventually to take the shortest must change these switches to compensate. Otherwise, it would think
possible path in a more complex maze ? it was moving one way and put that in the memory and actually be mov­
Shannon: I think that would require a considerable amount of ing in a different direction.
memory in the form of relays, because of the need to store up a number
^Orwell, G.: 1 9 i 4 . New York, Harcourt, Brace & Co., 1949 and Signet Books, 1950. No. 798
180 Cybernetics IN SEARCH OF BASIC SYMBOLS
Gerard: That would be like cross-suturing the motor nerves of ani­
mals and getting flexion when you want extension. DONALD M. MACKAY
Bigelow: Have you considered how difficult it would be to have a King's College, University of London
circuit which, instead of forgetting everything, goes back to the origin
and remembers what it did at the first square but tries something else,
say, the opposite search sequence? When that produces no new solu­ I DON'T KNOW whether you people feel agreeable, but if there are any
tion, go back where it was, in the second square, but try the opposite,
questions which you believe are not absolutely vital, such as matters of
therefore asking for the possibility of replacing each square in its mem­ error, I should be very grateful if you would make a note of them and
ory as it goes systematically through. In other words, this would re­ bring them all up at one time rather than interrupt the flow, because I
quire a very small addition of memory because it need only remember
think it may be difficult to get across enough to make sense of what I
the entire past pattern once, but then, having reached the state where
want to say at the end.
goal behavior is no longer a solution (which it knows by exceeding "N" What I want to do first is to present a way of looking at the problem
trials), then, instead of erasing its entire thinking, you have a switching
tackled by general information theory which finds a place for, and
technique where it goes back to the origin, and then tests each hypothe­
shows the relationship between, different concepts which have been
sis in turn, and finds the particular one to replace. labeled "information" by different people, which finds a place for con­
Shannon: I haven't considered that, but I think it would be rather
cepts such as meaning, and which I think links on to the domain of
slow, because there is a great deal of backtracking in that procedure,
symbolism and language.
back to the origin, as it tries out different hypotheses.
In common speech we say we have received information, when we
Bigelow: If it knows how to get from the origin to the target, does
know something now that we did not know before; when the total of
it not always know how to get from the target back to the origin, by
"what we know" has increased.
a very simple reversal of the switches? If then we were able to measure "what we know," we could talk
Shannon: No. You see, this vector field, if you like, is unique in
meaningfully about the "amount of information we have received, in
going in the direction of the vectors, but going backward, there are
terms of the change it has caused.
branch points, so it does not know where it came from.
General information theory is concerned with this problem of meas­
Savage: Does this vector field flow into the target from every point ?
uring changes in knowledge. Its key is the fact that we can represent
Shannon: Yes, if you follow the vectors you will get to the goal, but,
what we know by means of pictures, logical statements, symbolic
going in reverse, you may come to branch points from which you may
models, or what you will. When we receive information, it causes a
go in any of various directions. You can't say where the sensing finger
change in the symbolic picture, or representation, which we would use
came from by studying the memory.
Savage: It is not organized around any particular initial point; and to depict what we know.
We shall want to keep in mind this notion of a representation, which
that is one of the features of it, that once it has learned the maze, if
is a crucial one. Indeed, the subject matter of general information
you start it anywhere where it has been on its way to the maze, it con­
theory could be said to be the making of representations—the different
tinues; if you start it where it hasn't been, it finds one of those places
ways in which representations can be produced, and the numerics both
where it has been, and then continues.
of the production processes and of the representations themselves.
McCulloch: Like a man who knows the town, so he can go from any
By throwing our spotlight on this representational activity, we find
place to any other place, but doesn't always remember how he went.
ourselves able to formulate definitions of the central notions of infor­
mation theory which are operational, with more resultant advantages
than that of current respectability. In any question or debate about
"amount of information," we have simply to ask: What representa­
tional activity are we talking about, and what numerical parameter is
in question?" and we eliminate most of the ground for altercations
181
 182 Cybernetics Basic Symbols 183
or we should do so, i£ we are careful enough ! listening for a signal which each knows will be either a dot or a dash.
We can cover, I think, all technical senses of the term "information" A dash arrives. A has made various measurements, represents what
by defining it operationally as that which logically enables the receiver happened by a graph, and remarks that there was "a good deal of
to make or add to a representation of that which is the case, or is be­ information" in the signal. B says: "I knew it would be either a dot
lieved or alleged to be the case. or a dash. All I had to do was to make a single choice between one of
When, on the other hand, we come to measure "amount of informa­ two prefabricated representations. I gained little information."
tion," we may expect to find ambiguities. We shall expect two people A and B, of course, are not in disagreement. For lack of a vocabu­
to differ as to whether A or B has given them more information, unless lary, they are using the phrase "amount of information" to refer to
both have the same representational activity in mind, and are estimating different measurable parameters of the different representational activi­
the same parameter. Our expectation is not disappointed: few topics ties in which they engaged.
can arouse stronger debate. The problem is simply one of a deficiency A (to whose activities we shall return) was concerned with repre­
of vocabulary, and we have the same reasons for confusion as we senting what had actually happened, as a new, never-before-described
should have if we lacked the linguistic means to distinguish between spatio-temporal pattern of relations. B was concerned with replicating
volume, area, and length as measures of "size." I am afraid, therefore, the sender's representational symbol—a dash. For him, what happened
that our first concern must be to make enough distinctions and provide was merely a determinant of a choice between preconceived possibilities.
ourselves with an adequate vocabulary to avoid major trouble. To help Preconceived possibilities: that is the key phrase in communication
supplement what now must be a very condensed presentation, I shall, if theory. The communication engineer assumes that the receiver possesses
I may, append to our published proceedings an integrating survey of a filing cabinet oi prefabricated representations, so that for him a signal
the nomenclature of information theory prepared for last year's London is an instruction to select one from the assembly or "ensemble" of pos­
Symposium on Information Theory* sibilities already foreseen and provided for. His representational ac­
Representations commonly can originate in two distinct ways. The tivity is not a constructional but a selective operation.
difference between these is the essence of one of the most important You are all familiar, I expect, with the way in which "amount of
distinctions in information theory, between the theory of communica­ information" is defined for a selective operation. We imagine ourselves
tion on the one hand, and what, for want of a better term, we may call playing a game of "twenty questions," in which every question may re­
the theory of scientific information on the other. Both a communication ceive only the answer Yes or No; and we define amount-of-selective-
process and a scientific observation process result in the appearance of information (the adjective, I think, is essential) as the minimum .num­
a representation in the "representation space" of the receiver or ob­ ber of such questions logically necessary to determine the selection. To
server. But what distinguishes communication, I suggest, is the fact identify one out of N possibilities, for example, we require at least
that the representation produced is (or purports to be) a replica of a logaN independent yes-or-no answers.
representation already present to (with, in the mind of) the sender. When some possibilities are more likely than others, we keep propor­
Communication is the activity of replicating representations. tionately more replicas of them in our filing cabinet, so that (with the
This is to be contrasted with the typical activity of physical scientific optimum selection mechanism) it takes us fewer questions to hit on
observation of which the goal is the making of a new representation, those which are more often required.
representing some additional knowledge of that-which-is-physically-the- If one of these, say, the ith, occupies ( l / N ^ ) of the filing cabinet,
case concerning some unique space-time tract not heretofore represented it will require roughly loggNj questions and answers to locate it.
anywhere. We say that its selection has required (or provided) logaN^ "bits"
We might put it crudely as the distinction between the replication of selective information. The average number of bits per selection will
and the formulation of knowledge. The problems raised in the two evidently be the weighted mean of log^N (as defined above) over all
cases are in some respects quite different, and give rise to different possible selections. But we have supposed that each possibility occupies
"measures of information." space in the cabinet proportional to its own frequency of occurrence.
An example will illustrate this point. Two people, A and B, are 'The weighted mean is thus simply 2 (l/N^) logaN^, or in terms of
probability (pace the rigorists) : 2 /'(logg/'j.
*See Appendix I So much for selective information in communication theory. Claudç
 Cybernetics Basic Symbols 185
184
Shannon and others, of course, treated the whole matter in greater We are thus clearly faced with a twofold problem; First, we must
detail some years ago (l) (2), and I have given this outline only to be able to define distinguishably in operational terms the blank state­
help us to see where it fits into the general picture. Amount of selective ments which we want to prepare. In other words, something in the
information is evidently a measure of the statistical rarity of a repre­ design of the experimental apparatus or procedure must enable us to
sentation and has no direct logical connection with its form or content, identify and distinguish between observations if we want to call these
except in cases where these affect its statistical status. One word which observations "independent" or even "distinct."
was unexpected could yield more selective information to a receiver Then, second, we must collect evidence for our statements by observa­
than a whole paragraph which he knew he would receive. tion of events. We "plug in" observed data, so to speak, into the blank
Now it is evident that in any situation in which what is observed is spaces which we have for them in our previously prepared propositional
thought of as specifying one out of an ensemble of preconceived possi­ structure. If we boil a typical statement down to the oversimple form,
bilities, the amount of selective information so specified can in principle "Value X relates to interval Y," then our two problems are the opera­
be computed. The concept has, therefore, a much wider domain of use­ tional definition of Y and the collection of evidence for X .
fulness than that of communication theory. The point is that it is al­ I think there is a fair analogy of the first problem, by the way, if you
ways a relevant parameter of a communication process, because suc­ imagine a fly crawling on a perfectly blank, infinite white sheet. If you
cessful communication depends on symbols having significance for the want to make a description of the movements of the fly, you are word­
receiver, and hence on their being already in some sense prefabricated less unless you have some means, by projecting the co-ordinate system
for him. The practical difficulty, of course, is to estimate the propor­ of your eye, or some way or another, of identifying co-ordinate points
tions of the appropriate ensemble, when these are determined by subjec­ on the blank sheet. In order to utter your description, in order to make
tively—and even unconsciously—assessed probabilities. any scientific statement at all about the movements of the fly, you must
But now let us turn to this other problem, which faces, say, the somehow or other have means of labeling the fly's position. In a sense,
physicist; namely, making a representation of that-which-is-physically- therefore, it would be defensible to say that that which enables us to
the case concerning some tract of space-time. This I have discussed at name, to formulate our propositions, is "information." Or at any rate
length elsewhere (3) (4), and I want now only to indicate the different we can define a measure of information content in a certain sense as
and complementary senses of the term "information-content" to which the number of independent propositional functions which we are en­
it gives rise, and to outline the kind of formalism which is useful to abled by a particular experimental method to formulate. To distinguish
represent the processes concerned. this from other senses of the term, we shall call it the structural in­
Here we are not usually in a position to select from a filing cabinet formation content of our representation. This could be described as the
of preformed representations; we have to produce our representations number of logically distinguishable degrees of freedom of the repre­
ab initio. Our scientific representation is in general compounded of sentation. Each of the blank statements we were talking about a mo­
elements asserting certain relations between the magnitude of a voltage ment ago represents ideally one independent respect in which the repre­
and a particular point on a time axis, or between the intensity of trans­ sentation could be different.
mitted light and a particular co-ordinate intersection in the field of view I don't think that in this gathering it would be appropriate to go too
of a microscope, for example. "We say "the voltage was 10 volts at far into technicalities, but Í do want to mention that Gabor (5), in the
time /j, 10.5 at t^," and so forth. field of communication, defined what he called the "amount of informa­
Our ability to name operationally a certain number of distinct co­ tion" in a signal in such a way that he was essentially talking about the
ordinate values such as t^ and t^, enables us to prepare in advance the number of independent propositions necessary to define its amplitude
same number of distinguishable, independent "blank statements" of over a given period of time. Let's look at an example:
the form: "The magnitude had the value such and such at co-ordinate Suppose we want to represent the voltage of a signal coming through
point (or or what have you), or rather, "The magnitude a channel of a certain band width, as a function of time. At certain
had the average value such and such over the co-ordinate interval intervals, we want to take "new" readings to provide "new" ordinates
around q^ {q^-v 1n+i)^ so on." The blanks in these statements for our graph. Obviously, however, if we take two too close together
or "propositional functions" we then fill in as a result of our observa­ they are practically the same reading, since the inertia of the system
prevents very rapid changes. Gabor showed by an elegant method for
tions.
186 Cybernetics 187
Basic Symbols
the ideal case that there is a minimal separation in time between read­
noise amplitude and square it (the variance being the square of the
ings, below which (according to a certain criterion of independence)
noise amplitude), you get something which we can call the amount of
they cease to be "practically independent." This minimal separation—
metrical information, which certainly increases as the reliability of your
let us call it A/—is related to the band width A/ by a very simple
measurement increases. If you have a voltage of 10 volts with noise
relation of the form Af-At'^K, where K is a constant depending on
of 1 volt, then that gives a more reliable reading than the measurement
convention, but of the order 1/2. There is incidentally a rather in­
of 10 volts with a noise of 2 volts; and it is also, intuitively at any rate,
triguing way of looking at this (3) that I can't go into now, which
more worth while than a measurement of 5 volts with a noise voltage
brings out the fact that the size of the interval At is limited really by
of 1 volt. The signal-noise ratio, of course, is familiar to electrical
our inability to name a smaller interval in the language whose terms
engineers as related, at any rate, to the notion of reliability; and I think
are operationally defined by the apparatus we are using. So the "un­
if we pass from this particular illustration, we can agree that, in gen­
certainty principle" here is in essence a logical truism.
eral, this is a legitimate and distinct use of the phrase "amount of in­
But the point now is that in time t, apparatus with a band width f
formation" to represent the amount of evidence we have for the state­
enables you to formulate jiist about 2f X t independent propositions
ment we are making about a reading. Passing further from the descrip­
about the signal amplitude, no matter how you chop up your frequency- tion of readings to the general notion of the assertion of propositions,
time area, so to speak. You could either have a lot of channels of nar­
we can say that in a representation which we have been enabled to make
row band width, in which case each signal would take a long time to
by certain observations, by devising an experimental situation in which
be succeeded by its next practically independent signal, or you could
a given number of structural propositional functions are provided, then
have a wide band width; then you would have many independent read­ we can define a measure of our total evidence for the propositions which
ings close together in time; so that his definition of "amount of infor­ we eventually formulate, as the total amount of metrical information
mation" was, again, a measure of the number of labels or blank state­
provided by the experiment.
ments with which his experimental method provided him, a priori, be­ We can symbolize this in quite a simple way. Fisher's measure, or
fore the performing of the experiment. It is the structural information my modified form of it, which in this case is the square of the signal-
content of his ultimate description of the signal. A given band width, noise ratio, is additive in the sense that the metrical information content
which is to be available for a given time, provides him before the of a combination of readings (such as their mean) can equal but never
experiment begins with knowledge that 2 ft practically independent exceed the sum of their individual information contents. Assuming for
propositions (as he defined independence) could be formulated about the sake of argument that we are measuring a steady voltage, then if
amplitude. we take two readings and combine them and calculate the amount of
Let's go on now to the complementary problem; namely, the collec­ metrical information in the mean, we shall find that we have just twice
tion of evidence by performing the experiment, or, if you like, the what we had in either of the two individually. Each structural propo­
acquisition of a measure of confidence in the propositions which we are sition adds its contribution to the metrical information content of the
going to make, in this case about our signal voltage over a given period
resultant summary statement.
of time. Here we make contact with the thinking of R. A. Fisher, who, Now a set of independent propositions can be represented or sym­
back in the early thirties or before, but in particular in his book. The bolized by a set of perpendicular axes in a multidimensional hyper-
Design of Experiments, in 1935 (6), defined what he called "amount space. So we can represent this additive process by a convenient geo­
of information" in such a way that, in the simplest case, it is measured metrical vector model in which for each new independent proposition
by the reciprocal of the variance of a statistical sample. In other words, we add one dimension to our hyperspace—our "information space."
if we take the case of communication. Fisher made "amount of informa­ We then can take distance in each of those dimensions to represent
tion" depend on the amount of noise present in a signal—to be precise, some function of the amount of metrical information associated with
the "noise power," or at any rate the variance of the amplitude—taking each corresponding proposition. If each structural proposition is repre­
the reciprocal of it as his measure. sented by a vector whose length is the square root of its metrical infor­
Well, this quantity is not dimensionless. Again, I don't want to mation content, then the total information content, strucural and metri­
bother going into detail too much, but it is true for a certain class of cal, is represented by the vector sum of the individual components.
measurements that if you take the ratio of the magnitude itself to the For example, if we had just two propositions, we could define their
 Basic Symbols 189
188 Cybernetics
in prefabricated components at least) and the problem of scientific
total information content by drawing a single vector whose two perpen­ description, where your own procedure must provide you by ostensive
dicular components are the square roots of the amount of metrical definition with the symbols that appear in your representation
information in each. In the particular case of voltage measurement In communication between human beings and possibly between ani­
the two propositions concern two successive independent readings, and mals the problem is ultimately the production in one reasoning mech­
these vector components are actually proportional to the signal-noise anism of a representation—a pattern—already present m another rea­
voltage ratio. In that case, of course, the square of the length of the soning mechanism. In the human case communicahon theory is in­
resultant is the sum of the squares of the lengths of its individual com­ terested especially in the most economical way in which we could con­
ponents and is proportional to the total energy, and so we get a repre­ duct the selective operation that evokes the appropriate pattern; and
sentation in which additivity is preserved. , •. • since we can do this by coding, we are always prepared in principle to
How does all this relate to our initial notion, the one which is take the logarithm of the total number of possibilities as our measure o
familiar to all of us, the definition of information which Claude has the amount of information given, irrespective of the properties of the
given in communication theory, in terms, roughly speaking, of the
statistical rarity of a representation? Well, I would suggest that if we ^^fnThe'?Íífef process, the process of scientific description in which
had to communicate to somebody else a representation such as the one vou are confronted by a situation about which you are mitially wordless,
we have developed, we could think of our activity as instructing him to your experimental method, your mode of approaching the situation,
select out of a certain number of possible positions for the information provides you with {a) the conceptual possibility of formulatmg-giv-
vector, one representing the result that we have obtained. The tip o fng distinguishable significance to—a certain number of propositions
the vector can be represented as occupying one of a number of cells and {b) as a result of observing events, the ability to adduce evidence
into which the space is divided or quantized. In that case (on the
for these propositions. c i.
assumption that each position is equally probable for the sake of argu­ Now, what about the concept of meaning? Suppose we forget to
ment), we can take the logarithm (base 2) of the number of possible the moment about signals, which are very often symbols for something
positions out of which our result has selected one, as a measure, first, else and just take the case of two propositions. In ordinary mathe­
of the number of binary decisions to which this selection is equivalent matical logic, one could say that if you asserted two independent propo­
and, hence, as a measure of the amount of information in Shannon s sitions A and B, you have said somethmg which is equivalent to the
sense, which you remember we distinguished by calling it the amount- logical combination of these two, which could be symbolized, therefore
oi-selective information. From this standpoint, when we are talking as a point in a diagram with four possible positions (Figure 9) -
about information content, we are now thinking of the problem o position 1, you have said both. In position 2, you say A but not B
replicating by a certain procedure, a procedure in which we have a ^nd so on. Ld one can say in a rough way that the statement you make
filing cabinet of all possible representations of this sort; and we have could be defined by a vector—the vector hnking these points to the
to pick one out of it, and we assume for the moment that these are all origin—which has four possible quantal positions.
equally likely and all equally represented in our filing cabinet. In that
case, the logarithm of the number of possibilities represents, you re­
member, the minimum number of successive questions in a game of
2
twenty questions by which we should arrive at the point which has been
specified. I won't bother to go into the case where the probabilities are
not equal, beyond saying that if you like to picture the cells as deform-
able, then the logarithm of the number of cells will still give you the
selective-information content if you warp your space so that all cells
remain equally likely to be occupied.
Recapitulating, we have seen that information can be defined general­
ly as that which enables us to make or add to a representation. We
then distinguished between the problem of communication (which is
the production here of a representation already in existence somewhere,
 Basic Symbols 191
190 Cybernetics
It is common experience that we do not define many of our concepts
in terms of a set of unique propositions. Someone mentioned, I think
yesterday, that the concept of a chair is not definable simply by enumer­
ating a certain number of characteristics, because we all know that if
a chair doesn't have a leg, we may still judge that it is a chair which
has lost its leg, or something like that. I have heard of some work go­
ing on in Cambridge, England, on this point with, I believe, the con­
clusion that the most you can do may be to enumerate a set of possible
characteristics of chairs, of which any adequate subselection constitutes
a chair where found. What we would like, it seems, is some way of
symbolizing the partial participation of one or more characteristics in a
definition. The black and white of atomic yes-or-no components is too of conceptually infinitesimal gradations of orientation of the vector.
coarse for our everyday terms. Quite precisely, what we mean by the meaning of a given term which is
Well, now, can we sharpen up this notion of "partial participation?" definable in terms of this space of basic vectors (elementary component
What could we mean by saying that a certain term means "a little of characters) is the orientation of its representative vector—the direction
A and a lot of B," instead of accepting the four-way choice offered to which defines the proportion in which those elementary components
us by conventional logic in Figure 9? Suppose we try to say that the enter into our experience of the ostensive definition of the term.
meaning of a term represents, not a discrete selection from a set of Pitts: You are assuming that it is an ellipsoid ?
yes-or-no characteristics, but a selection of each in a certain proportion. MacKay: Let me say it again. Given two propositions, A and B, I
What reasonable meaning could we give to that ? am suggesting that a proposition which is neither A nor B nor A and
What we could mean, of course, is that our definitions are osten­ B equally, or, if you like, a word which is defined neither by the char­
sive—that their meaning is defined in terms of experience. Then, in acter A nor the character B alone nor by both equally, but by the two
the past over which a given term has acquired certain associations—to in certain proportions statistically, can be defined by a direction in this
which we point, implicitly or explicitly, in order to give our ostensive space (Figure 10).
definition—^we have found that A was associated only 10 times, say, Pitts: They combine additively ? That is what I wondered.
and B, 100 times, so that the term means "a little of A and a lot of B." MacKay: Yes.
If you like to picture the elementary characteristics as spread out along Pitts: That is the point.
a scale, the meaning of a term becomes a kind of spectrum, a spectral MacKay: All right. Now, you can see that the concept—
distribution, over the scale, the relative frequencies of occurrence of the McCulloch: Might I make this clear for myself? "A" here is a line
different elements being symbolized by the height of the spectrum over which represents one item of structure on a logon principle, essentially,
the corresponding points on the scale. and "B" is another proposition. The distances out on each of these are
I wonder if I am making that clear ? The idea is that you think of the metrical strengths?
chairs as things sometimes having four legs, backs, and what have you; MacKay: Yes, the square root of metron content, to be precise. It
then, again some chairs are backless, some chairs do not have four legs, is possible, of course, to use different metrics here. You could arrange
and so on. So, if we accept these (for the sake of argument) as simple it so that as the probability of inclusion goes from 0 to 1, so length
yes-or-no characters, then over a long experience of the word "chair," only goes from 0 to 1, as in quantum theory; but if you are dealing in
we should build up a concept of "chairfulness" which could be defined terms of metrical information you make no restriction, and the distance
by the proportions of different characters in the ensemble of all chairs could go to infinity (which would mean in theory an infinite number
experienced. of past experiences).
The interesting thing is that introducing this possibility would corre­ The choice of metric for our purpose is not particularly important.
spond in our vector model (Figure 9) to attributing significance to all The point is that with a given metric you can give a precise significance
orientations of the vector. Instead of having merely the possibility that to the meaning of a statement for a given individual, remembering
the vector is vertical, horizontal, or at 45°, we now have the possibility
192 Cybernetics Basic Symbols 193
that each individual receiver will have his basic vectors defined by his resistance at temperature T, so that the "noise power" is given by
own receiving apparatus and a few other things—which we shall come kTAf, where k is Boltzmann's constant and Af, as before, the band
to in Part II. width. Per unit of metrical information, therefore we shall require
The notion of amount of meaning, therefore, is scarcely meaningful, signal-power dissipation at least equal to kTAf, and therefore energy
and certainly not precise. So I suggest that we ought not to talk, as dissipation of kTAf At, where Ai is the minimal duration appropriate
Wiener does, about "amount of meaning," but that we ought to keep to the band width Af. If we now ask what the entropy increase cor­
the concepts of information and meaning quite distinct. I would sug­ responding to this transfer of energy at temperature T is to be, we find
gest that we can, in terms of this diagram, for any proposition or any its lower bound, at least, by dividing the energy transferred by the
body of data or any representation, define quite unambiguously what temperature at which the transfer occurs—if this is in the range for
we mean by (¿Ï) the amount of structural information, dimensionality, which classical Boltzmann statistics hold, of course. Since Af-At is
number of degrees of freedom, if you like; (¿) the amount of metrical of the order 1/2, we see that a physical entropy increase is required
information, the degree of confidence in, or the amount of evidence for, which cannot be less than about 1/2 ^ units. 'The amount of entropy
the assertion represented by the naming of the propositions or the change in general is proportional to the amount of metrical informa­
making of the statement, and (f) the amount of selective information^ tion, so—
measuring the statistical rarity of the selective operation specified by Pitts: This is the minimum amount of entropy ì
the naming of the total proposition. From the last point of view, when MacKay: Yes.
this (Figure 10) comes up, I think of it as an instruction to me to make Pitts: But the actual—this does not establish a connection between
a selection from the repertoire of my past experience to represent it. the actual entropy and metrical information.
To take this very simple case, if there are only two possible choices, and MacKay: This establishes the actual entropy change and the actual
if one of them has in the past been elicited ten times more frequently metrical information content, if you extract all the metrical information.
than the other, then one could imagine that over the course of time I Pitts: No, the entropy change can be greater than that.
would evolve a means of selecting "B," which involved fewer binary MacKay: True, but it can be equal to it under optimum conditions.
choices than the means of selecting "A," so that in a sense I should say This, you see, equates the minimal entropy change to the square of
that receiving the message that elicited "B" gave me less information the length of our information vector. On the other hand, the amount
than receiving the message that elicited "A." Then I should be talking of selective information, as we saw earlier, is definable as the logarithm
about selective information; I should essentially be referring the in­ of the number of distinguishable positions of the information vector,
coming representation to my ensemble or assembly of past-experienced which is not the same thing.
situations in which the same representation arose. This might appear to be a paradox because, of course, when you
McCulloch: Will you tie up which of these is now related to the define amount of selective information in terms of probabilities, you
concept of entropy.? arrive at something which has the same form as the definition of en­
MacKay: It is rather intriguing, really, because it shows how be­ tropy in statistical mechanics. The point is that we have begun by as­
wilderment arises if one does not define one's ensemble. I made it suming each of thôse positions of the vector to be equally likely, in
clear, I think, that selective information is defined only relative to an computing our amount of selective information. We have done so be­
ensemble—to a filing cabinet. It is a measure of the amount of trouble cause we are prepared to operate in such a way that at our receiving end
it would give you to pick out the thing you are talking about in the we can regard each of those signals as equally likely. Consequently,
filing cabinet if the filing cabinet is designed according to optimum we are referring our question as to the amount of selective information
principles. to an ensemble appropriate to the assumption that all of those states
Now, suppose that in the case of voltage measurement we take the are equally probable—in which, if you like, all possible states are equal­
definition of metrical information content as signal-noise power ratio, ly represented.
and suppose we inquire how much entropy change is involved in meas­ When we calculate the amount of physical entropy, on the other
uring a single independent reading, using the definition that one inde­ hand, we are referring to the ensemble appropriate to a physical system
pendent reading requires an "uncertainty product" Af- At of the order in equilibrium at temperature T, for which not all possible states are
of 1/2. We can suppose that we have loaded our source with a matched equally probable. And I think that all the debates and paradoxes
194 Cybernetics Basic Symbols 195
which keep cropping up as to the relation between Shannon's amount pings of a space into itself. Concentrating on the distribution function
of selective information and the concept of physical entropy disappear of the ensemble, which is the primary concern of the selective informa­
if one asks precisely what assembly is being used for the computation tion, it is invariant under any such measure-preserving mapping, be­
of the amount of selective information. (I should be interested to know cause it is defined just in terms of the probability measure.
whether you agree with that, Dr. Shannon). You get the physical MacKay: Yes. You would agree, though, that from the point of
measure if you use an assembly defined for thermodynamic equilibrium; view of the experimenter, who does not think so much in terms of map­
and you get quite a different measure, of course, if you use the artificial pings as of design, an experimenter who knows that he can play with
assembly (the filing cabinet of the receiver) that regards all states as chopping up band width—multiplexing band width and multiplexing
equally likely. In that case it is the metrical information content and time—the first measure is invariant with respect to his ingenuity in
not the selective information content that correlates with physical en­ design. It represents the upper limit to the number of independent
tropy increase. Does that answer your question ? physical readings you can get by utilization of a given frequency band
McCulloch: That answers my question. Are there any other ques­ width in a given time.
tions ? If there is nothing from the floor, then I think we had better Shannon: It seems to me it depends on how much chopping you are
develop the question of symbolism and then throw it open to discussion. allowing; that is, there are these one-to-one mappings of a three-dimen­
I believe there is one question from Von Foerster. sional space into a two-dimensional space, which correspond to ways
Von Foerster: Can the two vectors A and B, which are also to be of compressing band width or enlarging band width, either way, which
considered as unitary vectors, also be considered as results of two other preserve everything—I mean, if your experimenter is rugged enough
vectors, let us say, alpha and beta, and so on ? with his equipment, he can actually do that, so it is not invariant under
MacKay: Yes, indeed. That does bring up a point I should have anything he can do but only under relatively continuous things that he
made. One naturally asks with respect to what are those different meas­ can do.
ures invariant? Well, the first measure, the structural measure, we saw MacKay: But it represents an upper limit, surely; basically it means
was invariant with respect to subdivision of a given band width and that you require 2 ft independent ordinates to specify a signal (within
time tract. No matter how you chop up your frequency band or time a precisely rectangular band width /) which persists for time t; and
tract, you can never get more than 2 ft independent signal ordinates. there is no question but that you cannot get more ordinates out of that
(You can certainly get more distinguishable values of each ordinate, particular specification, because the signal is defined when you set that
but that is a different matter.) The second measure, the metrical in­ number of ordinates.
formation content, we saw was invariant with respect to coalescence of Shannon: Yes, that's right.
independent readings, say by averaging—^which the third is not. The Bigelow: That isn't clear to me. In general, that is true, but is it
third measure represents the invariant upper limit to what you can do always absolutely true?
by ingenuity in coding. Is that fair. Dr. Shannon ? Shannon: I think it is, but aren't we talking about a technicality here
Shannon: Well— which has no bearing on the main point ?
MacKay: Perhaps you could put in a sentence what the third is in­ McCulloch: The point is that in one way, the first way, it preserves—
variant with respect to. what ?
Shannon: There are two things I might argue a bit here. In the first MacKay: The first way preserves the number of independent propo­
place, the structural measure is really a measure of dimensionality and, sitions you can make—the number of independent readings of ampli­
as such, is not invariant under anything you do. For example, you can tude you can take—against your ingenuity, within wide limits any way,
map a two-dimensional region in a one-to-one manner into a one-or in chopping up the band width or time. TThe second preserves the meas­
three-dimensional region. You cannot do it with a topological mapping, ure of total statistical reliability, roughly speaking—the total amount
so I would prefer to characterize the structural measure as being in­ of evidence—against coalescing of two adjacent readings to become
variant under all topological mappings. These do preserve dimension one. That is where your question comes in. Dr. Von Foer^er, because
as a basic theorem of dimension theory states. we can think of any information vector whose length is \/n and whose
It seems to me that another way of characterizing selective informa­ amount of metrical information is therefore n, as the resultant of, or
tion is that it is invariant under all one-one probability-preserving map­ as equivalent to, n unitary components. In other words, if you have a
196 Cybernetics Basic Symbols 197
space, an «-dimensional space, and a unit vector along each axis, then which amounts to, or is defined by, a different proportionality of those.
the square of the length of the resultant vector is n. Then you can ro­ Pitts: Well, have you defined how you determine the meaning which
tate your basis so that this vector lies along one of your new axes, where projects it along the core of axes ?
you can think of it as representing just one unit of structural informa­ MacKay: Yes, it is just another way of representing the spectrum
tion carrying n "metrons" (as one calls them)—the individual units of meaning that we were talking about. The height, or some function
of metrical information which have gone to make up the total evidence of height of the spectral lines at each point representing an independ­
for the structural proposition. These n metrons can be thought of, if ent elementary component proposition, defines the orientation of the
you like, as what might have been, alas, the "stuff" of n independent vector.
propositions, but they have all been run into one; the point being that Von Foerster: Are the components negations ? Because they are cer­
you are no longer able to distinguish order in this group of n units tainly independent from each other; for instance, red and nonred or
combined to form one resultant vector. You can't distinguish one from something like that.
the other, and that is why they represent only log n units of selective MacKay: Those two? [indicating two perpendicular components]
information. They are not bits. They only represent log n bits because Oh, no.
you have "folded up the umbrella," so to speak, by combining the con­ Von Foerster: They do not have to be negations, because with a
tents of n dimensions into one resultant vector. negation you could certainly divide the whole world into two different
"Well, this business of symbolism is a long story, and time is short. parts.
Perhaps I can best introduce it by presenting to you the bare bones of a MacKay: They are independent characters which are not necessarily
possible reasoning mechanism (7), which we must think of as dealing present in unit quantity, because in the statistical ostensive history of
not only in black-and-white logic, but as capable of handling or trans­ the definition they may not have occurred with equal frequency.
ducing or reacting appropriately to the kind of stuff we have been talk­ Savage: What does "independent" mean? Just that one can occur
ing about—information in the general sense. In terms of this concept, without the other, or that they are statistically independent?
you remember, we find precise interpretations for terms such as mean- MacKay: The assertion of one does not affect the probability of the
ing—even shade of meaning (both related to the orientation of our other.
vector) and relevance. The last, I perhaps ought to mention, is quite Savage: How can you tell ?
precisely the square of the cosine of the angle between two directions MacKay: Oh, let's say they're defined to be so. Right? Let's get on,
in information space, in this sense; if you ask how much metrical in­ then, to consider the kind of thing which could handle what we have
formation (evidence) is afforded to a dependent statement by a given called information; that is to say, essentially a probabilistic mechanism
body of information, the answer is found by squaring the projection of in which trains of thought (if you like the term) correspond to suc­
the information vector on the ray which defines the statement. cessive transformations of information vectors, or changes of state in
There isn't time to expand on this, but you can see that it does corre­ which each state represents, or can be thought of as representing, a
spond fairly well to our notion of the relevance of data to a statement. particular total proposition symbolized by certain information vectors—
Pitts: What space is this, incidentally? Von Foerster: Excuse me, but if some concept is 20 per cent area and
MacKay: Co-ordinate Cartesian. 30 per cent length and 50 per cent volume, and so forth—
Pitts: I know, but when I say "space," I mean, what do the axes MacKay: Oh, not volume!
represent ? Von Foerster: I am simply picking random ones, any one you like.
MacKay: Independent propositions. If another one could be tied in in the same way with different percen­
Pitts: But here you have just a number of them. You were talking tages, then in what sense can you talk about an angle between meas­
about two dependent events. uring the conditional probabilities of one on the other? You see, I
MacKay: No, I'm sorry. These are not events. I am assuming that don't know whether you are dealing with concepts, statements, or
you have made a certain observation or statement which has a certain what, exactly.
metrical reliability indicated by the length of this vector, and a certain MacKay: A thing can't be 20 per cent volume. You may make an
meaning indicated by its orientation relative to the vector basis, the assertion about its volume which has 20 per cent reliability, or some­
basic component propositions. And then you want to make a deduction thing like that.
198 Cybernetics Basic Symbols 199
Von Foerster: Oh, but this is an unspecified object, of which we tion probabilities are higher than they would have been originally. We
know the probability, that it has these characters, and that is what we think, in other words, of some form of mechanism which, beginning
are plotting. with random attempts to scratch and having a success signal«, or more
MacKay: That's right. generally an evaluating mechanism, which controls the probability of a
Von Foerster: If we are talking about two different objects, how can given sequence being tried the next time, arrives at what is essentially
the angle between them say anything about the probability that those a symbolization of the incoming stimulus; namely, the sequence by
statements are true?
which it is successfully replicated.
MacKay: If you are talking about two different objects, then the Suppose now that we have several of those stimuli—say, three form­
statements you are making about them are independent, and then their ing a triangular pattern—and suppose that we have a mechanism whose
vectors are at right angles. I don't think there is any obvious fallacy. constant activity is to "doodle" in speculative movement from one to
In any case, the essential things we are going to need are, first, some the other; in other words, it is built to respond with initially random
means of symbolizing probability, of introducing probability into our attempts at movement from one to the other. We will equip it with
mechanism; and, second, some means of identifying the basic vectors, means of determining that it has successfully finished all its scratching.
the elementary components, the ^sic symbols, out of which such a It goes from here to there, from there to there, and gets the sensation
mechanism could build its representation of a universe of discourse. that it has no more scratching to do. And we will suppose again that
I think, perhaps, since my chief concern today is with the second triangular patterns recur with sufficient frequency in its incoming stim­
problem, we might leave the probabilistic question to the last, and con­ uli. Then just the same process of natural selection, which I think you
sider first the problem we are confronted with in asking how an or­ can easily envisage, is all we need invoke in order to make a device that
ganism reacting with its environment derives its vocabulary of ele­ automatically discovers for itself, and names, an abstraction—a device
mentary symbols. As I see it, we have a choice between (a) the use which, when stimulated in this case with a triangular pattern, would
of the incoming stimuli or filtrates thereof as elementary symbols, and raise the probabilistic status of the sequence, "make a line, turn, make
(¿) the use of the elementary ("atomic") acts of response to received a line, turn, and so forth," so that we could think of the sequence as
stimuli as the symbolic components for the description of what is per­ becoming one of the elementary symbols of the "experience" of this
ceived. So this is our question: Does one's internal representational device—its name for "triangularity." Out of the welter of all possible
mechanism describe a concept by simply asserting that it means so much symbolizations, this one has acquired the dignity of a universal through
of this received stimulus and so much of that received stimulus? Or recurring with sufficient frequency among the successful responses to
are the elementary symbols going to be something else ? the flux of incoming stimuli.
Again for the sake of brevity, I will jump straight on to suggest a One can go on from this to think of any incoming pattern (which is,
mechanism on the second principle, without discussing now the reasons again, suflSciently recurrent or which persists for long enough at a
that have led me to favor it in preference to the other. given time to evoke a satisfactory sequence of transitions from one ele­
Suppose we have a device that reacts to incoming stimuli by an act mentary component act of response to another). We can consider any
of symbolic replication. At the very simplest level, let's consider one invariant in this way as gaining the status of a "universal" in the world
which has an irritable surface and is designed so that if it is excited here of discourse of this device. You see, I am going behind the problem of
[indicating palm] at a single point, its reaction is initially a random devising a machine to reason deductively, which we can take for granted
hunting motion which eventually elicits a success signal. Call this, if well enough, to the problem of transforming the information carried by
you like, a scratch-response mechanism. It is easy to see that if this incoming "sensory" stimuli into a symbolic linguistic form suitable for
stimulus recurs often enough, you can devise a self-molding statistical such a machine.
mechanism (I can go into details later on, if you like) which will in­ Forgetting the probabilistic aspect of it for the moment, and just
crease the probability that on the next arrival of such a stimulus the supposing it is a question of symbolizing propositions, our typical prob­
scratch mechanism will tend to hunt in its neighborhood. You can see lem is this: How are you going to insure that a deductive mechanism
that if any particular irritation is a sufficiently consistent feature of the presented with, say, a triangle in the field of its optical receptors will
flux of incoming stimuli, it will quite soon elicit the successful scratch always make the deductions appropriate to the presence of triangularity,
response by a series of elementary operations between which the transi- irrespective of size, shape, or orientation ? What I have been describing
200 Cybernetics Basic Symbols , 201
is one possible way of solving the problem which in detail, I suspect, sponse which is invariant with respect to the transformations that leave
may not be very realistic; but you can see what will happen if you the abstraction invariant, out of the random trial of such responses, by
design the machine properly. Its definition of a particular complex a self-guiding process in which the statistical configuration for each
universal—its name for it—is a compresence and/or a sequence of the next attempt is dependent upon the success of the last. Experience
elementary acts involved in responding to it. elevates the statistical status of certain response sequences, and these
Now comes the probabilistic aspect. These components will not al­ can then appear as definitive elements in the internal logical vocabulary.
ways recur in association in equal proportions. Therefore, a complex And then second-order abstractions (which we call hypotheses), or even
universal becomes defined not by simply enumerating the elementary nth order abstractions, are in principle just a repetition of this process
acts which go into its symbolization, but by enumerating those plus the at the next level or at a higher level.
relative frequencies with which they have had to go into its replication. This kind of approach to the concept of reasoning might, I think,
And here, of course, our old information space and the cloven hoof of find a place for most of the concepts that we arrive at from the other
probability appear again; because manifestly what we have done, in direction. Consciousness, for example—if I dare stick my neck out—
effect, is to define our universal by a vector in a space which is not might be introduced in this way; We might say that the point or area
quantal (at this macroscopic level at any rate), but has the possibility "of conscious attention" in the field of view—^in a field of data—is the
of representing practically continuous gradations or shades of meaning. point- or area under active symbolic replication, or evocative of (in­
The meaning of this universal is defined by the orientation of the in­ ternal) response. When a man speaks to another man, the "meaning"
formation vector, by the statistical spectrum, if you like, over the ele­ of what the man says is defined by a spectrum over the elementary acts
mentary acts of response which exemplars of this universal have evoked of response which can be evoked in the hearer. If the hearer is such
in the organism in its past. that when the man raises his eyebrows something inside the hearer hap­
Well, the story goes on and on. Obviously, this merely takes us past pens which coresponds to imitation—the internal initiation of part of
the first step. We have discussed an artificial organism which can rec­ the sequence that would normally lead to the raising of eyebrows, and
ognize pattern in the flux of experienced stimuli—^which can abstract so on—then, clearly, the meaning of what the man says can only be
from among its received data those relationships whose recurrence or fully symbolized in terms of the full vector basis (basic-symbol com­
invariance gives them the status of universals. plex) defined by all elementary responses evoked, which must, for ex­
The next step is to consider the possibility of recognizing pattern in ample, include here the initiation of the "internal" command to raise
the flux of perceived universals—of abstracting from among the evoked eyebrows, and may include visceral responses and hormonal secretions,
acts of symbolic response those relationships whose recurrence or in­ and what have you.- So, along these lines, I think one would say that
variance give them the status of universals. an organism probably includes in its elementary vocabulary, its cata­
This, I think, is the essence of the making of hypotheses: to predicate logue of "atomic propositions," all the "atomic" acts of response to
the pattern of a group of abstractions. Can an artificial organism spon­ the environment which have acquired a sufficiently high probabilistic
taneously formulate hypotheses? I think it can, certainly in the sense status, and not merely those for which verbal projections have been
that we can devise a second-order probabilistic mechanism analogous to found.
our first, but one in which each individual basic symbol now represents Now, I shall have to stop somewhere, only briefly throwing out the
one complete abstraction; and the orientation of the representative suggestion that such concepts as emotional bias, and the like, would
vector in the corresponding space of basic vectors indicates the relative have obvious analogues in such a mechanism, in the shape of the al­
frequency of occurrence of total response sequences (in past experience teration of the thresholds, or, if you like, the distortion of the probabili­
if this hypothesis is to be based on experience). I believe this would be ty amplitudes, appropriate to the basic vector components affected by
at least the simplest way to do it. It's evident that we have here the the "bias." The recognition of things like this as equivalent to things
possibility of a hierarchy of abstraction, which could itself become the like that [indicating letters d and p on board], mentioned yesterday by
subject of discourse; but perhaps I've said enough to show you the Dr. Richards, would be expected in such a device if, before it reached
possibilities. the level at which these were meaningful as letters, the reasoning mech­
Here, our device, designed probabilistically, makes its abstractions by anism had already developed the habit of doing that [drawing a semi­
a process of natural selection. It chooses a means of (internal) re- circle adjacent to a line as in d and p} by some self-guiding process
 Basic Symbols 203
202 Cybernetics
cabinet-coded key. If the fact of the particle being on the left means
leading to satisfactory replication, and regarding the corresponding se­ something to an intellect by previous arrangement, by education, then
quence of internal commands as definitive of a universal. Because, you we have a system here that distinguishes quite sharply between entropy
see, if you think of a servo mechanism set dov^^n on this white line (on and information.
letter d), and if you stand outside and listen to the commands that the MacKay: What kind of information?
servo mechanism gives itself in following the line, the command se­ Bowman: Let me use the term in quotes first, then gradually build
quence will be, "Go straight ahead; turn (right) with a certain curva­ up a definition of the term "information," as I am using it now. Per­
ture until you're back on the line." That sequence of commands defines haps I am using it in an anticipatory sense.
the universal for that particular simple mechanism; and so you would If we have two things in the box, they may be on the left side or right
expect that a mechanism that used this method of naming (recogniz­ side, or one on each; and, again, a change in their position will not af­
ing) patterns would not distinguish between this, that, and the other fect the entropy but will affect information, if you have a proper code.
variants {d, b, p, q) unless some experience indicating their nonequiv- With a third particle, I believe you get something different for the first
alence caused a search so that discrimination of direction came in and time, in that there appears a sort of objective information. There is a
added a new dimension, if you like, to the space in which the meaning fundamental difference in the probability of an arrangement like this,
was defined. and one where you have one of the particles over on the other side.
The same sort of principle would apply to the recognition of Chinese Pitts: Excuse me, but the two halves of the box are not distinguish­
faces which we were discussing earlier. If the act of recognition in­ able. We can't tell if they are turned around ?
volved active response by symbolic internal replication, then a suf­ Bowman: You can't tell.
ficient experience of the nonequivalence of samples of Chinese faces Pitts: Oh, that is the point?
would evoke elementary operations for—^would raise to the status of Bowman: Yes.
symbolic elementary operations—the drawing in of those identifying MacKay: Could you define the person making the choice? That is
features which are peculiar to Chinese faces. Once that had happened, very important. You get an answer depending on which you take.
then either never again, or certainly not for some time afterward, would Bowman: May I go on for a bit ?
you cease to be able to recognize them; because although they would MacKay: I'm sorry. I think that was Pitt's question.
not be internally "named" by you in faces which did not have the fea­ Bowman: I should like to see if I could introduce the idea now of
tures, your "descriptive space" for faces would include the necessary useful information, perhaps as a brand-new kind, as distinguished from
dimensions, and you could be conscious of their absence. what we have discussed before as a number pair. The usefulness of the
Along these lines, I think one could go a very long way toward information is a measure of coincidences between two measures. A
simulating what appears to be the ordinary conscious behavior of book in your native language can convey information to you that is
human beings. On the other hand, of course, if one were to ask whe­ potentially useful. A book in a language you do not know does not, and
ther such a mechanism could ever be built, I would take refuge for the yet may be a translation and contain inherently the same information as
moment in the blessed phrase "in principle," and say that in principle the book you can read. For the usefulness of information, then, we
I see no reason why it shouldn't; but I would not myself be surprised must look to coincidences between number pairs—the one, the stored,
were one to attempt to devise a probabilistic mechanism with the same the filing cabinet, the education, if you like, and the other, what we ord­
mobility, and so on, as Homo sapiens, if one would have to go in for inarily think of as a code which stands for something.
mechanisms in protoplasm instead of mechanisms in copper. That Now, you can easily set up some extreme examples of that. If I gave
seems to me to be one implication of some of the very neat tricks that you a book in very fine print consisting all of zeros, it wouldn't mean
we find in the central nervous system. anything, and yet, under conceivable codes, that might be a very re­
Bowman: I should like to return for a little bit to the entropy and markably large amount of information. On the other hand, if I had
the filing cabinets of the first half of Dr. MacKay's talk. If we start made an elaborate tentative plan with any one of you and said, "Well,
with a box that has an imaginary partition in it, and we know that now, I don't know at this time whether or not this will go through, but
somewhere in that box there is a particle, the entropy of that system here is the plan," and then on the following day, I said one word,
is the same if the particle is on the left or on the right. It makes no "Yes" or "No," that one bit of information, in fitting into a coin-
difference. The information, however, is different if we have a filing-
204 Cybernetics Basic Symbols 205
ddence with a whole subsequence previously stored and matched, con­ New interest in random strategy has risen from the theory of games.
veys a great usefulness. When one is faced with an opponent, a relatively clear case may be
I should like to see if we can't recognize a double algebra: the filing made for the use of them, but our dealings with nature are a different
cabinet (as Dr. MacKay termed it), or, as I would term it, the code or matter.
the education, on the one hand, and the communication or replication I would therefore suggest that in your automata, random strategies
on the other hand. The coincidences between those two, and only those, play a superficial role except, perhaps, as a little gloss or decoration
are of value. I do not believe that we can speak in a purely objective tending toward realism. Thus, you might object to Dr. Shannon's ma­
way of information as such. Information must have meaning. It must chine on the grounds that it never makes a mistake, but it doesn't seem
be understood to be useful. Information without a code, or at least to me that any really deep function is necessarily served by incorporat­
with some code that is so deeply innate in us as to be unrecognizable, ing random strategies into the automata. We should, at any rate, give
is perhaps what we do have in the entropy in the sense of Clausius—the fundamental thought to what role random strategies would, or do have,
simple "dq over t." There, you need no code. That is as objective as in behavior before assuming glibly that the automata which best mimic
any physical quantity is. A simple statement of what it is in other ob­ human behavior do necessarily rattle around inside.
jective terms would enable one skilled in the art to measure the entropy MacKay: There is nothing glib about this. If you consider that the
of something. I don't think that that objectivity can be applied to in­ organism is designed to interact with a statistically fluctuating environ­
formation. There is a subjective half, half of the number pair, that we ment, in which it is precisely because two events are never similar that
have in our head. the organism must only have a probabilistic response, then I think you
Savage: I should like to applaud what you have just said, and, now can see that the thing can be more efficient as a business proposition if
that I have the floor, I should like to speak about the last portion of Dr. you design it to have a spectrum of probability of operation which
MacKay's talk; but first I must know clearly whether he did or did not matches the spectrum of frequency of demand for that.
mean to say a certain thing. Did you. Dr. MacKay, intend that the au­ Savage: No, I cannot see that at all. I, and others, have carefully
tomata you discussed should contain in themselves random gadgets, or considered the behavior of an ideal statistician embedded in just such
did you mean only to say that they are in some sense capable of doing a statistical environment as you allude to, and, according to our con­
inductive inference ? Do they operate on random strategies in the sense siderations, he has not the slightest incentive to rattle around. The
that Shannon's mechanical rat does not.? point is a technical one and would have to be sketched out technically,
MacKay: Oh, yes, that was one of the bits I had to miss. They are but, as I say, the conclusion is not altogther a casual one.
essentially statistical in their operation; that is to say, they contain de­ MacKay: Yes. Well, there are two points, of course. The first ques­
vices of which it is meaningless (unless you go down to the molecular tion is whether you are trying to produce something that resembles
level) to predicate predictability. You simply have a set of statistically human behavior, and that certainly is one of the reasons why I think we
defined transition probabilities as, for example, in a network of thyra­ ought to make our mechanism a statistical one. But the second question
tron tubes (gas tubes) whose bias is very near the threshold, or—actual­ is whether a statistical mechanism could pursue an optimum strategy.
ly a much more likely example—a nervous network. Now, if your mechanism had to be prepared for one of two possible
Savage: That, then, is what I should like to make a few remarks situations, of which one occurs just slightly more frequently than the
about. Modern mathematical statistics are largely concerned with the other, then it seems to me that to arrange that the device always adopt
critique of inductive inference. Statisticians, as such, do not ask how an one of those, instead of now trying one and now the other in appropri­
animal behaves, but how he should behave. More specifically, they ask ate proportions, is going to make it more difficult for it to operate in
how a human being should find things out. Some statisticians, apparent­ situations where it can only learn by sometimes making the other deci­
ly beginning with Fisher, have advocated the use of random strategies sion.
in finding things out. Indeed, this is the position of almost all statisti­ Savage: I can only reiterate that I believe the situation to have been
cians today, and I myself am inclined to share it. Nonetheless, it is my explored in some technical detail, and that it has been established that
opinion that no one has yet clearly shown how the employment of ran­ the most efficient results can be achieved by following deliberate non­
domized strategy can help a person isolated from others in his search randomized strategies. Randomized ones can be at best as good as the
for truth. most efficient nonrandomized ones, but not better.
206 Cybernetics Basic Symbcjls 207
MacKay: That is right; they can be as good. Bigelow: That is to say, that the outside world actually plays a game
Savage: But not better. against him?
MacKay: They can be devised much more simply, much more eco­ Savage: No, it isn't to say that the outside world plays a game
nomically. They certainly can be as good. You have shown that, in against him. It would be absurd and ruinous for him to assume that the
effect. Dr. Shannon. outside world is aiming at his downfall and destruction, for him to re­
Savage: That they can be as good is obvious, because nonrandomized gard it as a competitor.
strategies are by definition special instances of randomized ones. Bigelow: But the point I make is that, essentially, if you use random
MacKay: Therefore, if you are trying to produce something which strategies in the game theory, you somehow make use of a type of deci­
has any parallel with the mechanism of human thinking, it would be sion which is unknowable to your opponent each time you use it, and
rather foolish to start with a complicated deterministic one. yet you yourself know the event of the particular decision on each oc­
Gerard: I should like to point out that, if I have followed this ele­ casion that you use it.
gant and useful formulation, the point I raised last year about the syn­ Savage: Well, as I said before, in the theory of games you sometimes
apse itself being analogical and not digital, and therefore operating have a special reason to adopt a randomized strategy, for you are strug­
near the threshold of response, is relevant. This is exactly what is gling against an opponent who, in some sense, might know what you
needed to get this kind of behavior out of the nervous system, so per­ are going to do unless you incorporate randomness in your behavior.
haps the nondigital performance of the synapse is what makes imagina­ Bigelow: Unless you get something of which he has zero informa­
tion possible. tion, essentially.
Pitts: I should like to ask Savage a rather obvious question because, Shannon: There are a couple of remarks I wanted to make, if I can
well, you can see why it is obvious by what I am going to say— remember what they were. In connection with this last discussion, it
Savage: Is the answer also obvious ? Otherwise, I would rather not seems to me that a random element in the machine, in a theoretical
try it. sense, may be necessary because of such results as Church's theorem in
Pitts: No; that is, just a superficial inspection of why the statistician symbolic logic, that there are some mathematical theorems that you
uses random behavior in planning an experiment is simply, of course, cannot prove according to any given determinate strategy of proof you
this; he wants a mode of procedure which he can be fairly sure is un- set into the machine, although a human mathematician might go di­
correlated with some given variable, and he feels more sure of being rectly to the result. If the random element were in the machine, it
able to secure that requirement if he uses the random arrangement than leaves the possibility open of arriving at a proof in a similar manner.
if he uses any systematic one. At least, isn't that the usual reason given ? Actually this is a rather theoretical question, but on the practical end
Savage: "The usual reason given by statisticians is that they, unlike I don't think there is too much difference between a very complicated
MacKay, don't believe in an entirely probabilistic world. They believe determinate machine and a truly random one. It is really a matter of the
in a kind of unknowability and an unknowingness which is not proba­ complexity of the determinate part of it, compared to the length of life
bilistic. They believe in, and often allude to, a sort of absolute ignor­ of the machine. If we are constructing random numbers, for example,
ance which defies description in probabilistic terms. In such an un- as Dr. Bigelow has suggested, he might, in the Maniac computer, by
probabilistic context, the strategy proposed by Bayes, which would multiplying a pair of ten-digit numbers together, and by taking away
otherwise be the ideal one, is simply not meaningful, so here statisti­ the first and last five digits of the product, leaving the middle ten, and
cians look around for something else. In particular, they sometimes then repeating the process, get something which looks more or less like
find an advantage in introducing mixed strategies. This enables them a random sequence, although it is perfectly determined, and you could
to inject probability into situations in which, from their point of view, calculate the entire sequence, knowing the first element. It will look
it would not otherwise occur. random over a period comparable to 10^°, if you are working with ten-
Bigelow: Does your conclusion disallow any possibility that the in­ digit numbers. After that, it begins to repeat. If the numbers are be­
dividual who is operating against the outside world produces a change ing used only for that length of calculation, it doesn't matter whether
in his environment by his action? they will repeat at some later point or not.
Savage: No, his action cannot influence the facts of nature; but his "There is one other point I want to make while I have the floor. In
fate depends, typically, not only on the facts of nature but also on his connection with Dr. Bowman's remarks, it seems to me that "we can all
action.
208 Cybernetics
Basic Symbols 209
define "information" as we choose; and, depending on what field we
with noise or perhaps similar distortions—distortions, say, of a word
are working in, we will choose different definitions. My own model of
beyond a certain amount result in some loss of intelligibility, and re­
information theory, based mainly on entropy, was framed precisely to
sult in its effective conversion from a signal into noise—
work with the problem of communication. Several people have sug­
Bowman: I would make a big distinction, though, between a word
gested using the concept in other fields where in many cases a com­
all hashed up with noise and a word unknown to the recipient. There
pletely different formulation would be more appropriate. In the com­
is a big difference there.
munication problem, entropy is the precise concept you need, because
McCulloch: Hold it a minute. Is this point fairly clear, that one of
the particular problem you are interested in is, "How much channel do
us was thinking about a sender and the other about a receiver? If you
I need to transmit this information?" and entropy is a quantity which
take both into account, then you are going to deal with a number pair.
measures or determines the amount of channel required. So long as you
Pitts: Certainly both must be considered, but the question as to how
ask that question, that is the answer. If you are asking what does in­
you want to calculate the information numerically depends upon the
formation mean to the user of it and how is it going to affect him, then
exigencies of the particular situation, of course. I am sure we are all
perhaps such a two-number system might be appropriate.
agreed on the necessity for considering both factors. With respect to
Bowman: I believe that the use of something like entropy as a meas­
Church's theorem which Shannon mentioned earlier, just for the record
ure of information is a perfectly" valid basis for communication prob­
I should like to say one thing about that. It is not impossible to make
lems where the recipient of the information is assumed to have had in­
a machine that will prove provable theorems; but what Church's
finite education; that is, if you are talking to a person on the telephone
theorem asserts is that it is impossible, given the theorem, to set any
and the person talking to you uses words that you know, without excep­
upper boundary to the time it may take. It is very easy to show that you
tion, then I would say that the use of the entropy function, or some­
can make a machine to print all theorems because you can write out the
thing like it, as an information measure is perfectly good. If, on the
axioms in a finite list as they are generally constructed, and you can
other hand, you are working with a digital calculating machine that
reduce the rules of procedure to be applied to those to a small finite
feeds information back into itself and has a very limited built-in sys­
number. Then you can simply classify all the theorems as those which
tem of codes, a small filing cabinet of education, then I think you have
result from one application of the rule, those which result from two ap­
to regard the usefulness of the information not as a number but as a
plications of the rule, and so forth; that is, the machine can print all
number pair.
theorems in order, starting from the axiom. The only point is, if you
Pitts: Perhaps you could simply alter the definition of "noise" slight­
are given a theorem, you don't know how long it will be before that
ly and call everything which is not in the receiver's filing cabinet "noise."
particular theorem shows up. A random process doesn't help because
Bowman: It is, to that particular receiver, in a subjective sense, but
there, again, although you may be able to be sure with Probability 1
it might not be noise to a different receiver or a different recipient.
that every theorem may occur sooner or later, still you can place no up­
Pitts: Well, when you have the information, you are always count­
per limit to the bounds which it may take for a given theorem, so it
ing it as transmission of information; that is, you are always counting
doesn't help you there. But that wasn't one of your important points.
it between one definite place and another place, and to say you transfer
it to another receiver or a different channel so that it might not be the Shannon: Perhaps I misunderstood the theorem, but I didn't have
that impression of it.
same amount of information is perfectly possible. It is a practical ques­
tion, of course. Pitts: Well, you see, in the case which I mentioned—in the sense
that the common systems to which Church's theorems apply can be
Bowman: You could take that point. I would prefer to look upon
listed that way—since the theorem is defined as the end result, and
the information as transmitted as something measurable. The informa­
since the single steps are each of a mechanical character, of course, all
tion from a radio program as broadcast is a measurable number of bits.
can be obtained.
The meaning of that to the receiver is something that he takes two
McCulloch: Dr. Klüver, do you want to speak of Shannon's point?
numbers to specify: first, the number of bits that were broadcast and,
Because Shannon has two more points to bring up.
second, some measure of the comprehension capacity of the listener.
Klüver: I want to make sure that I got a certain point straight. You
Pitts: Well, you see, I was speaking exactly from the listener's point
talked about the important problem of how concepts, abstractions, and
of view. If one considers nonlinear loss of information or combinations
hypotheses are arrived at. Did you wish to imply that such mechanisms
210 Cybernetics Basic Symbols 211

of concept formation as you have described are really involved and oc­ mind; and by talking about variable threshold, I do not mean just
cur in human thinking ? variation at a synapse, but anything such as facilitation and inhibition
MacKay: I am suggesting that this possibility is present in the com­ due to proximate volleys, anything which afl^ects the probability of
ponents which we have to play with in speculative models. transmission at branch points, and various mechanisms of that sort,
Klüver: For instance, the formation of the concept "chair"—at least, which affect probabilities of excitation.
the way I understood it—comes down to a sort of statistical considera­ I am sorry I had to condense this so very much. I probably expressed
tion of its components ? it very badly. The idea is, if you follow this through, that you can de­
MacKay: Yes, in the shape of the modification of threshold ampli­ velop an evolutionary theory of perception, evolutionary in the sense
tudes. that it relies on self-molding statistical processes to abstract complex
Klüver: I should like to make one general remark. In listening to percepta from this activity of elementary replication. There will be two
physicists and engineers, I am generally impressed by their optimism; complementary possibilities: first, quite conceivably, that what you
but in hearing Dr. Bowman today, I believe that I detected a somewhat might call the natural choices (in the way of activities of replication)
pessimistic note. are those which, because of the hereditary structure, have a high proba­
Bowman: It was intended. bility of excitation as responses; and then, second, that those will be
Klüver: It looks as if the human organism is often viewed here as molded and modified by the actual success evaluations and threshold
merely a marvelous device for registering incoming stimuli, for re­ feedback evoked by successful replication. That is really why I gave
ceiving and coding of information, and for doing a large number of the example of the triangle arrived at by "doodling." A mechanism
equally remarkable things. For the psychologist, the picture is un­ which begins by doing nothing but doodling could evolutionarily evolve
fortunately more complex; unfortunately he cannot see such simple a symbol for triangularity by this mechanism. I don't see that there
outlines. To be sure, in this picture we have the influence of past ex­ should be any great difficulty in going from that to higher things,
perience, we have the storing of items in filing cabinets. But experi­ particularly since one can go to the next level of abstractions about ab­
ence can enter the picture only because we are able to catch it by means stractions, and on up.
of certain schemata. What are these schemata? And must we assume Shannon: I should like to add a word to Pitts which just occurred to
that the very schemata by means of which we catch experience can in me with regard to the random element in a machine for proving mathe­
turn be influenced by experience? At best, for the psychologist, the matical theorems. What I really had in mind was a finite state machine,
picture resolves itself into the formulation of a large number of un­ that is, a machine with a finite number of possible internal states. In
solved or only partially solved problems. I am glad that Dr. Bowman such a case the machine can only go through a periodic sequence from
introduced a somewhat pessimistic note when considering the proper­ one state to another because there are only a finite number of states
ties of the human receiver. that it can be in, and each subsequent state will be definitely deter­
Bowman: It ties in, perhaps, with some of the things we discussed mined by the previous one, if it is a determinate machine. On the other
yesterday, in the sense that a phenomenon observed depends on the way hand, if you have a true random element inside, it will not in general
in which it is observed. The instrument used to observe it is an ex­ be a periodic sequence and could presumably arrive at any possible
tension of the observer and is limited by his education, so that the ob­ theorem.
servation made and the conclusions reached depend upon the instru­ Savage: Well, conceivably, you would have technical work to do, to
ment and the education of the observer in a sort of, well, as we said show that, considering that some theorems, for example, have very long
yesterday, complementary way. statements, let alone proofs, and the machine would presumably have
McCulloch: I think MacKay is shaking his head. When Bowman is to bear the whole or large parts of a proof in mind in order to com­
through, you can go ahead. plete it.
MacKay: Well, I wonder if I shouldn't make one point clear about Shannon: I think it probably couldn't go through everything—
the old business of engrams. I think the only stable threshold modifica­ Savage: But the scope could be enlarged.
tion one can envisage would be a variation in the diameter or length of Shannon: It could come out with any sequence of output symbols,
fibers or something like that. I am not suggesting chemical traces or presumably. It may not have proved that it was a true theorem.
anything similar. I am not sure whether that question was in your Pitts: But that is the point. First, if the theorem is too long, the
 Basic Symbols 213
212 Cybernetics
manipulation of the design of a physical method. If you can devise an
memory of the machine, clearly, could not contain it. Second, if any of experimental method by which a given frequency-time domain can give
the necessary steps on the way to the machine is too long for the ma­ you more than 2 ft data on amplitude, you are contradicting the theory
chine, it could never contain it. which Shannon and Gabor and others have worked out. I think we are
Bowman: Isn't it quite possible for a finite determinate machine to constantly confusing this physical question with what you can do to a
compute a transcendental number to any desired number of significant logical structure by way of mapping without loss of selective informa­
digits ? That is certainly not a periodic output. tion. There are any number of ways in which you can code something
Pitts: No, that is not possible, unless it refers to the numbers which to get back from one representation with few dimensions to another
it had previously calculated. with more, but that is a problem of communication, of specifying a
Bowman: Oh, you exclude that? selective operation, not of constructing a representation ab initio.
Pitts: Yes. I am trying to exclude this infinite or potentially in­ The problem you run into concerning structural information is our
finite memory which it can use for calculation or for reference. problem of the fly and the need for a co-ordinate system. That's typical
Bowman: Oh, yes, I see. of all. You have your crawling fly. Somehow or other, you have to be
Pitts: It can only refer to its internal memory, so it must certainly able to utter words about its position. It is the old business of Wittgen­
be a periodic output. stein and, "Vovon man nicht sprechen kann, darüber muss man schwei­
Bigelow: The question of the invariance of this vector system he had gen" (If you can't name something, you can't talk about it). There­
on the board— fore, you must somehow be able to call out co-ordinates that label or
McCulloch: Were you interested in the problem of the invariance, identify a statement uniquely. You are just using the same "blank
as to what was invariant under what transformation ? statement" twice if you are not doing anything to distinguish two state­
Bigelow: He had a propositional vector system, and then said he ments except in the trivial sense that you make them at two different
had put all these vectors into it, or put in something which was a map­ times. For example, if I say, "The height is 6," "The height is 5," I
ping from one direction to another. am simply contradicting myself by using the same propositional func­
Shannon: The point I was raising there was, perhaps, that old result tion twice. But if I am able to say "The height at x = 1 cm. is 6," and,
of Kantor's that you can map a square, for example, into a line in a one- "The height at x = 2 cm. is 5," all is well. Do you see what I mean?
to-one way, so that what MacKay calls the structural information, Now, in a given band width and a given time, it can be shown that you
Gabor's notion, is not invariant under anything you can do. have a finite number of independent propositions which you can make
MacKay: You mean, not invariant under ¿Ï// things you can do? and must make to describe the amplitude of the particular signal you
Shannon: Yes, under arbitrary remappings. You can take two vec­ observe during that period.
tors, or two numbers given to an arbitrarily large number of places, and Pitts: Not if the noise level is zero. If the noise level is zero, then I
condense them into one number which contains all the information of don't think that is true.
the two which you are able to construct from this one, and vice versa. MacKay: But surely what I'm saying is an old story ? What you are
Bigelow: Yes, that contains all but the selective information. Your saying is that in that case the number of cells in the information space
measure of selective information isn't at variance at all if you consider is not finite.
also as information the details of the condensations. Pitts: Well, let us ask Shannon. Does his theorem hold when the
Pitts: You have lost the structural information, in a sense, from the noise level is zero ?
composite number. Suppose I have two ten-digit numbers and I make MacKay: We are not talking about Shannon's theorem. This is a
the contention that I can make a twenty-digit number out of them, the question of independent measurements of amplitude, not a question
even ones being taken in order from the first and the odds from the of what you can code. I think Shannon's theorem surely holds with
second. If I am given the composite number, I know not only what the respect to that because he takes "2 ft" as the power of his brackets in
two ten-digit numbers were, but also which way they were combined. deducing capacity. Let's put it this way. All I am saying is that the
I can retranslate them without loss in both directions. dimensionality of the space is finite and determined by / and t. The
Bigelow: If you notice another additional factor, of course. number of distinguishable points in the space, on the other hand, is a
MacKay: I have said it twice, but I have said it badly each time. function of the noise. The noise goes down, and though the dimen-
This structural information-content is invariant with respect to a priori
214 Cybernetics Basic Symbols 215
sionality doesn't change, you get more distinguishable points for a some conventionalized code of representation of something which is
given amount of power or energy; therefore, of course, you can sym­ already namable, in such a way that the man from the other end can
bolize (in communication) a larger selective operation; and if you reproduce it. There are two distinct problems which have given rise to
know your code you can always map from the received representation different concepts of information, and both, I think, are relevant. They
to another which shows no sign of the dimensionality of the signal. can all be covered by the definition of information as "that which en­
That, I take it, is what you are saying. But, you remember, I began ables one to make a representation." But when you talk about amount
by distinguishing between the problem of communicating representa­ of information, you get the same kind of variation (though it is not a
tions, in which you already have the representation in file at the good analogy) as you get between area, length, and volume, as meas­
receiver and you want to make some sort of code wiggles on a commu­ ures of "size." It is a very poor analogy, but one would expect or could
nication line which will instruct him to draw from the filing cabinet— expect that kind of diversity in the concept. As for perception, remem­
Pitts: When you are talking about dimensionality of the space being ber I said, "unless you project the optic co-ordinate system in some way
finite, you are using "dimensionality of space" in a rhetorical sense. onto the screen." What you say is exactly to my earlier point, because
What space is this ? the perception of the movements of a fly (for which we all know per­
MacKay: It is the space of which the axes are defined by the inde­ fectly well you don't need a grid on the screen), the naming of it to
pendent readings, or the necessary definitive amplitudes, of a function ourselves, would on this theory consist in the internal act of imitative
limited within a frequency band f, and limited to a time period t. or symbolic response to the movement of the fly. You see? The "basic
Pitts: If the band width is finite, then no two readings are strictly words" of the organism are formed automatically by the acts of re­
finite, in the rhetorical sense. sponse.
MacKay: Let's put it this way: What Shannon and Gabor and others Pitts: Let me ask this again. It seems reasonable to me at the mo­
have established is that a given band-limited fvmction, persisting for a ment, although I don't recall it, that what comes out is something with
time t which has passed through a band width f, requires 2 ft ampli­ a rectangular, well, with a strictly rectangular impedance function, that
tudes to determine it and no more. Now, it is quite true that there is, as a periodic function, as a superposition of approximately 2 ft, or
may be instances in which the amplitudes you measure are not inde­ distinct periodic components. However, the first remark is that physi­
pendent, in the sense that there is "a bit of one contributing to the cally it is impossible to realize anything with an impedance function of
reading of the other"; but between them, they do define a 2 /^dimen- that kind.
sional space. MacKay: I agree.
Bigelow: The accuracy of the observation here must enter somehow. Pitts: It must tail off at the ends, and, as a matter of fact, it must
The accuracy with which you are able to determine each of these dis­ tail off not too fast at the ends, in a certain well-defined sense. Now, as
crete components— soOn as it tails off slowly enough at the ends to be physically realizable,
MacKay: No, I am sorry, but the 2 ft theorem is a truism, as I under­ then no longer would any finite number of amplitudes be specifying
stand it. It arises from the definition of what you mean by "frequency."
what will come out.
It is the same thing which gives you the quantum error. MacKay: Well, what you are saying, in effect, is that you have to
Pitts: You say 2 ft amplitudes will define what? consider noise when you want to determine effective band width if you
MacKay: The 2 ft equally spaced ordinates will define a function for
don't have the ideal case I specified. I was deliberately simplifying this
you, which has passed through a band width f, over time t. so that people who weren't particularly interested in the technicalities
Pitts: Well, suppose I increase / by a number; suppose I increase the would get a concept of structural information without worrying about
band width by some small part of the epsilon, what happens to it ? Do which approximate number we take to measure it in nonideal cases.
I get one more amplitude or not? Suppose 2 ft is not an integer? Pitts: It is not relevant.
MacKay: Then it is a limit. The point is that you can't have more MacKay: Noise is completely relevant if you ask how many of those
than that. But the essential thing, you see, is to distinguish between amplitudes are far enough above the noise to specify one quantum of
the problem of devising physical means of labeling a proposition; that
metrical information.
is to say, confronted with initially blank experience, acquiring the ability Pitts: No, that is not the point. It has to do with structural informa­
to label uniquely something in experience, and the problem of making tion. You would define "structural information" strictly for the case
 Basic Symbols 111
216 Cybernetics
which is not physically realizable in principle; namely, where you have
a sharp cutoff. What the engineer commonly does, of course, when he
concerns himself with realizable cases and inquires what the band width
is, is simply to select a certain low level of the impedance function on
both sides, cut it off there, and say that that is the effective band width.
But then you have an infinite number of amplitudes.
MacKay: I did not define ""structural information" in terms of fre­
quency or time at all. I instanced that particular theoretical case to illus­
trate what I meant by "structural information." I'm not at all interested
at the moment in defining frequency band width for the nonrectangular
case, which is a technical red herring.
Gerard: What are / and t in terms of ordinary propositions ?
MacKay: I don't quite know what you mean, but if you mean, what
are the definitions of / and t, f is the frequency band width over which MacKay: Yes. Disagreements between two people as to shades of
meaning, I suspect, could often be described and clarified in those
an idealized receiver is flatly responsive and cuts off at the edges, and
terms.
t is the time during which you make the observation. But, you see,
Pitts: The probability points to ""pip."
any —
Bigelow: Frequencies from 100 to 10,000, say, are absolutely com­ MacKay: We want a language with two dimensions, really.
Savage: I wanted very much to say a word or two about Bowman's
pletely passed through, with no distortion, and with unit-multiplication
second concept of amount of information. This theme has come around
factor, and everything below that and above that, say, is absolutely
again and again. Dr. Shannon has said, ""Well, I have a kind of infor­
deleted ? I am putting this in a terribly crude fashion. The band width
mation, but I invented it for a special purpose and it would be fool­
is how wide.? Is it from 1 to 100,000?
hardy to suppose that it will really work for every purpose." Bowman
Gerard: I did understand that, but I thought MacKay generalized
has hinted that there must be a kind that somehow tells what the infor­
this specific illustration to propositions in general; the number of items
mation is worth to the person who gets it. He says, ""Well, it depends
that you add up with experience to make the word ""chair." What
on what the person already knows." I think he might also have said,
happens to these ?
"'It depends on what the person wants to do."
MacKay: You don't have to think of frequency and time. Perhaps I
If I tell you something, however intelligible, but something you don't
should never have brought them in.
want to know, as when I tell you exactly how some industry is carried
Gerard: But what are their equivalents in your structural proposi­
out in some far corner of the world, supposing you are not curious
tion?
about such things, you may get all I say but you find it worthless. I
MacKay: Well, consider an organism which has only two possible
think, in the last analysis—and I do not mean this facetiously, though
modes of response, by way of replication, doing this and doing that.
I am saying it in a compact form—the value of information is its cash
Let's call them '"pip" and "pop." Well, then, the evocation of pip
value, understanding cash in a rather metaphorical sense. If you want
corresponds to the perception of pip in the stimulus, and the evocation
to talk seriously about what information is worth to the receiver, you
of pop corresponds to the perception of pop; and something which has
must have in mind what it is the receiver wants to do, and you must
in the past evoked each at one time or another, 60 per cent pip and
have in mind some quantitative description of his well-being. Then you
40 per cent pop, you see, is represented by something which is not set
must ask yourself, ""How does this information contribute to his well-
by merely saying ""pip-pop," but saying it with modulation (Figure 11).
being?" "There are two questions to be distinguished here. I may be
Now, we can't do that verbally, and that is the diflSculty, I think, in a
about to get information; for example, I may be about to do an experi­
great deal of discussion—I am not thinking of our discussion now but,
ment. Then I can say, ""Well, what is it worth to do this experiment?
in general, of human discussion—the inability to modulate the metron
What would I pay to do the experiment?" Again, the experiment being
content (the metrical information content) of those propositions.
done, I now occupy a different situation, am poorer or richer. If, for
McCulloch: You mean you should say something like "pip-^o^" or
'"pip-^0/7" to do it?
218 Cybernetics Basic Symbols 219
example, the experiment consists of opening a letter from the Fuller discuss a communications channel and determine that it has a certain
Brush Company, to whom I have recently applied for a job, if, on open­ information capacity, then, in various ways you can produce paradoxes
ing it, I find I've got the job, well, then I am ten times as rich as I was. resulting from attempts to think about information in abstracto as ac­
Therefore, there are these two: There is the cash value of an oppor­ tually flowing through it; for example, if the channel is in any static
tunity to get information, and there is the shift in your cash assets, or state, one is at a loss to know whether, as a matter of fact, information
in your personal evaluation of your cash assets, on receiving the par­ is continuing or not. Every state of the channel constitutes information
ticular information. There is no shortcut avoiding such analysis to that, transmission and every state of the channel may constitute full use of
I believe. Statisticians have sought one for over fifty years now, and in the information capacity.
the last eight or ten years they seem to have been settling down to the Actually, what is done here is to define a capacity rather than to
view which I am expressing; namely, that only a detailed economic define the concept of information in any concrete or abstract sense.
analysis of the situation can properly appraise the value of information. Would you comment on this point Ì
McCulloch: I am afraid "value" in your sense would turn out to be McCulloch: Will you hold your answers to that and let's take Pitts's
a multidimensional affair, with very little chance of its being simplified question; then we will answer all of them.
to a measure. Pitts: I was just going to make one very short remark. I think the
Savage: No, it is simply one-dimensional. The values to which I missing character comes not between the transition from capacity to
have referred as "cash" may be measured in actual cash, or lives saved, past information and the amount of information, but in supposing that
or whatever it is in the given situation which properly measures the we can divide the compound phrase "amount-of-information," into
well-being of the individual concerned. This value is, in fact, to refer "amount" and into "information." What Shannon has defined is
you to a theme familiar to us, the Von Neumann-Morgenstern util­ "amount of information." I think that should be regarded as a specific
ity (8). phrase, and possibly there should be some Greek word invented to
McCulloch: Very familiar and very illusory. convey it by itself.
Pitts: The El Dorado. Shannon: I agree with what Pitts has just said, and I should like to
Savage: In the two hundred years since the concept of utility has add that we should also remember that this kind of information is an
been introduced, it has been known by several names, but surely, espe­ ensemble concept. It is not a statement about a proposition, if you like,
cially since Von Neumann is one of our members, we will continue to or a fact, but a statement about a probability measure of a large en­
call it by the name preferred by him. semble of statements or propositions or facts. It is a measure of a kind
There is one misunderstanding. I said "cash" because that is a short, of dispersion of that probability distribution.
vulgar word, but the Von Neumann-Morgenstern utility theory is so­ I think perhaps the word "information" is causing more trouble in
phisticated enough to capture other things, including even the joy of this connection than it is worth, except that it is diflîcult to find another
discovery, and, in principle, there need not be a multidimensionality to word that is anywhere near right. It should be kept solidly in mind
the situation; that is, in principle, in so far as the Von Neumann- that it is only a measure of the difficulty in transmitting the sequences
Morgenstern theory really works, it captures and compares in one that are produced by some information source.
measure, and in a way successful for this context, cash, admiration, a Now, there is one other point I want to bring up in connection with
satisfaction of discovery, the saving of life, and all the sort of things the ideal filter. It is not possible to obtain a sharp cutoff over a limited
that we do value. It is an elaborate, rather farfetched construct, but band width with zero phase, that is, with no delay. Bode's loss-phase
there is some satisfaction in it, too. relations show that this is impossible. But it is possible to obtain this
Bigelow: I should like to raise a point to see if MacKay and Shannon sharp cutoff if you are willing to allow delay. As you allow larger and
would comment on it. It has always seemed to me that the concept of larger delays, you can come closer and closer to a square gain function
information that is used in these theoretical analyses of communication over a limited frequency band. This means that we cannot determine
does not really ever speak about information in abstracto nor define it these co-ordinates that we are talking about, or perform these ideal
in abstracto, but is concerned with something like the capacity of a operations instantaneously, but to do so we must wait for some time
• channel or the capacity of a system or the capacity of a structure; situ­ until these tails for that delay have been reduced to a tolerable amount.
ations which are conceptually somewhat concrete. But if you once Pitts: Except that keeping the same rate of decline at the edges, you
220 Cyhemetics Basic Symbols 221
can increase its absolute rate by increasing the delay, but with any finite because the machine presumably is working into a whole host of differ­
delay, of course, the approach is fixed independent of the delay. For ent ensembles corresponding to the expectations of the different people
example, you can't realize the function of a square with any amount of who are reading it. There would, I think, be no more reliable way of
phase. getting a measure than by totting up the propositions. My own theory,
Shannon: No, but we can realize a nearly square function which has incidentally, in so far as it has been different from what has been done
a total area outside its square function less than an arbitrarily small by other people, is characterized by attending to this notion of number
epsilon, which means that only epsilon power is getting through from of atomic propositions as suitable measures of each of those three
outside the frequency range in question. senses of "amount-of-information." I do agree very much that the ex­
Pitts: But you can't make the rate of decline at the edges much pression should be hyphenated. I suggested. (3) (4) logon content
faster than— and "metron content" as names for "amount-of-information in the
McCulloch: All right; we get the point. Will you close. Dr. MacKay.? structural and metrical sense to avoid using the mischievous word. One
MacKay: Well, when I called my talk, "In Search of Basic Symbols," could talk about "number of bits" for Shannon's measure, but there
I meant "search," and I suppose it is partly my own fault that we did doesn't seem to be any obvious word, as he says. I would make a plea
not pursue our search very far. I should like, however, to suggest that that if we are going to talk about information, we should use adjec­
if those who are more competent can give some thought to the problem tives—perhaps mine are not good—but some adjectives such as selec­
of identifying the basic symbols in the human thought mechanism, or, tive," "structural," and "metrical," to distinguish what we are talking
rather, the basic physiological response acts in the human being which about. It is my experience that as soon as one becomes accustomed to
might conceivably form and define the basic vectors in his symbolic thinking in those terms, the majority of arguments one hears about
information space, it would be a matter of some interest, as representing "information" dissolve; and it is sad if clarity is lost in any way through
or suggesting possible tests. not using distinguishing symbols for different concepts.
As to the discussion about information, let me say first that I was not The only other thing I should like to say is that thinking in those
presenting in any way an alternative theory to Shannon's, but a wider terms, I believe, provides one with a sharp definition of "communica­
framework into which I think his ideas fit perfectly. In fact, as I tried tion" in terms of the replication of representations, which I feel ought
to make clear, what I was doing was to show how Shannon's ideas were to be used as a razor in situations where people attempt to talk about
validly applicable, I think, universally, though such an analysis is not the physical world "communicating with you" in an experiment. I
necessarily appropriate to all problems. I think he is too modest in his think there is an inevitably anthropomorphic connotation in the word
disclaimer, because as long as you define your ensemble, then his defini­ "communication " which bedevils discussion, and that such a use of the
tion of the entropy or selective-information content of the ensemble, it term would certainly be invalid in this case, and should apply only in
seems to me, is always applicable without any contradictions. The cases where a representation is already present at the sending end of
trouble people get into, I think, is always due to the fact that they do the process. That, of course, is only a personal plea, but I feel that
not define the proper ensemble; and that is particularly difficult when we are in danger of getting this subject into disrepute among those
the ensemble is defined by the internal statistical structure of the man sketchily acquainted with it if we ourselves do not make as many new
receiving information; that is to say, where only subjectively appre­ symbols, in the shape of technical terms, as the diversity of our concepts
hended probabilities define your ensemble. But if you do define your requires. I don't think we should be frightened off from doing so by
ensemble, I can only say that I haven't come across any cases where the fact that it may take some time to be understood.
that concept cannot be validly applied, and usefully applied if the
problem is one of communication. Naïve equation of selective informa­
tion content with physical entropy, however, is generally unwarranted,
as we have seen.
When dealing with digital computing machines, and so on, the
problem of defining the ensemble is particularly difficult, and it is pos­
sible that the best interpretation of amount of information is obtained
by simply totting up the number of atomic propositions in a situation,
 APPENDIX I Appendix I 223
Right at the start the term "information" takes on two different kinds of
meaning in answer really to different kinds of question. An example will illus­
THE NOMENCLATURE OF trate the point. Two people, A and B, are listening for a signal which they
know will be either a dot or a dash. A dash arrives. A makes various measure­
INFORMATION THEORY* ments, represents what has happened by a graph, and asserts that there was "a
good deal of information" in the signal. B says: "I knew it would be either
DONALD M. MACKAY a dot or a dash. All I had to do was to choose one or other of these prefabri­
King's College, University of London cated representations. I gained little information."
A and B are not, of course, in disagreement. For lack of a vocabulary, they
are using the term amount of information" as a measure of different things.
THE EXPLANATORY survey of nomenclature which follows is not a glossary in A is using the term in the sense of what we call "Scientific Information Con­
the sense of an agreed list of standard terms in Information Theory. No such tent. This in itself has two aspects, relating roughly to the number of inde­
agreement yet exists in this new subject. Its purpose is rather to collect and col­ pendently variable features (structural information content) and the precision
late as many as possible of the terms which are in current use, and to define or reliability (metrical information content) of the representation he has made.
tentatively the ways in which they are related and the senses in which they may The knowledge which he says has increased is knowledge of what has actually
be interpreted without conflict. Only time will show whether thesé interpreta­ happened and been observed.
tions will be adequate or acceptable; but if it succeeds only in demonstrating the What of B? He was not waiting to observe everything that happened. He
complementary and noncompetitive relationship of different current approaches already knew that for his purposes only two kinds of representation would be
to the problem, this somewhat Augean task will have been worth attempting. needed, and he had prefabricated one of each. The knowledge he acquired was
The glossary is not of course meant as an "introduction for beginners," so knowledge of which representation to select. B was therefore using "informa­
much as a logical framework in which terms may be seen in perspective as they tion in the sense of "that which determines choice," which we may call
arise. The proportions of space devoted to different aspects of the subject are selective information.
therefore no indication of their relative importance, being dictated merely by the One word which was unexpected would yield B more selective information
exigencies of exposition. Although the approach is bound to be a personal one, than a whole message which he was already sure he would receive.
a considerable effort has been made to base it on logical consideration of what A s approach is typical of the physicist, who wants to make a representation
has actually been meant by various authors, in consultation as far as possible with of physical events which he must not prejudge. B's is typical of the communica­
other speakers. The particular terms ""metrical," "structural," and "selective" tion engineer, whose task is to make a representation, at the end of a communi­
information content may possibly be thought unsuitable; but the logical distinc­ cation channel, of something he already knows to be one member of a set of
tions for which they stand appear to be essential. Confounding of these is standard representations which he possesses. His concern is therefore not with
believed to be the source of much unedifying debate. the size or form of a representation, but with its relative rarity, since this will
In everyday speech we say we have received information when we know some­ govern the complexity of the "filing system" he should use to identify it. Each,
thing that we did not know before: when "what we know" has changed. If however, may on occasion find both approaches relevant to different aspects of
then we were able to measure "what we know," we could talk meaningfully his work.
about the amount of information we have received, in terms of the measurable To sum up, if we ask how much information there is in a given represen­
change it has caused. This would be invaluable in assessing and comparing the tation, we may mean: "How many distinct features has it? How many elemen­
efficiency of methods of gaining or communicating information. tary events does it describe?" in which case we require answers in terms of
Information Theory is concerned with this problem of measuring changes in scientific information content; or we may be ignoring questions of the size and
knowledge. Its key is the fact that we can represent what we know by means of complexity of the representation, and thinking instead of the complexity of the
pictures, sentences, models, or the like. When we receive information, it causes selection process by which it was identified, meaning: "How unexpected was it?
a change in the symbolic picture or representation which we would use to depict How small a proportion of all representations is of this form? In how many
what we know. It is found that changes in representations can be measured, steps were you able to identify it in your "filing cabinet' of possibilities?" In
so "amount of information," actually in more than one sense, can be given this case our question refers to selective information content. Rarity here is the
numerical meaning. It is as if we had discovered how to talk quantitatively touchstone, as against logical structure in the first case. It will be realized—and
about size through discovering its effects on measuring apparatus. We should at this niay be an important help to the understanding of the subject—that the
once find that it had the quite different but complementary senses of volume, term ""information" means something quite distinct from "meaning." If the
area, and length—if not others. The analogy is potentially misleading, but may reader begins by divorcing the two completely, he may find it easier to trace
show us what to expect. the connections in any subsequent reunion.

^Prepared for the Symposium on Information Theory held in London in September, 1950.
Revised in April, 1951.

222
 Appendix I 225
224 Cybernetics
Information
EXPLANATORY GLOSSARY
The foregoing definition of the scope of Information Theory pro­
Information vides the necessary background for the definition of Information.
1. The Scope of Information Theory

2.1 In all its senses, the term can be covered by a general "operational"
Representations l.l Information Theory is concerned with the making of representa­
definition (i.e., a definition in terms of what it does: as [e.g.]
tions—i.e., symbolism in its most general sense.
force is classically defined in terms of the acceleration which it
causes or could cause). The effect of information is a change in a
1.1.1 By a representation is meant any structure (pattern, picture, representational construct.
model), whether abstract or concrete, of which the features pur­
port to symbolize or correspond in some sense with those of some
other structure. 2.1.1 Information may be defined in the most general sense as that
•which adds to a representation.
1.1.2 The physical processes concerned in the formation or trans­
formation of a representation are thus distinguished from other 2.2 This leaves open the possibility that information may be true or
physical processes by the element of significance which they possess false.
when conceived as representing something else.
2.2.1 When a representation alters, we define the new information
1.1.3 For any given structure there may be several equivalent as true if the change increases the extent of correspondence between
representations, defined as such through possessing certain abstract the representation and the original.
features in common.

2.2.2 The information is said to be false if the change diminishes

1.2 It is these abstract features of representations which are of interest
the extent of this correspondence.
in Information Theory. Its aims are (a) to isolate from their par­
ticular contexts those abstract features of representations which can
remain invariant under reformulation,* (b) to treat quantitatively 2.2.3 Strictly, the truth or falsehood is an attribute of the result­
the abstract features of processes by which representations are made, ant representation, but it is customary to attribute it to that which
and (c) to give quantitative meanings to the several senses in has given rise to the change in the representation.
which the notion of amount of information can be used.
3. Measurement of Information Content
1.3 The scope of Information Theory thus includes in principle at least
three classes of activity:
1.3.1 Making a representation of some physical aspect of experi­ Two quite different but complementary approaches are possible to
Scientific ence. This is the problem treated in Scientific Information Theory. the measurement of Information Content, and have given rise to
uses of the term in quite different senses.
Information
Theory 1.3.2 Making a representation of some nonphysical (mental or
ideational) aspect of experience. This is at the moment outside our 3.1 From a quantitative analysis of what a representation portrays, we
concern, being the problems of the Arts. can isolate fundamental numerical features common to all its equi­
valent representations, and can say that they constitute the "corpus
of information" which it contains or represents.
Communication 1.3.3 Making a representation in one space B, of a representation
Theory already present in another space A. This is the problem of Com-
Communication munication Theory, B being termed the receiving end and A the 3.1.1 "Information Content" in this context is a numerical index
Channel transmitting end of a Communication Channel. (in one sense or another) of the "size" of the representation itself.

Space 1.3.3.1 By a space is meant any physical or abstract mathematical 3.2 But if instead of asking "How many elements, and so forth, are
co-ordinate framework or manifold, of any number of dimensions, there in this representation?" we ask, "In how many stages, and in
in which the elements of a representation can be ordered. what way, has it been built up?" we arrive at a different kind of
measure. This becomes clear if we consider that the same represen­
1.4 These classes are not, of course, exclusive. The problem of com­ tation could be constructed in a number of different ways according
munication in particular is seldom separable from one or both of to the amount of préfabrication used.
the first two. In its present state of development. Information
Theory is concerned mainly with (1.3.1) the problem of represent­ 3.2.1 "Information Content" in this context is a numerical index
ing the physical world, and (1.3.3) the problem of communicating of the complexity of the construction process.
representations (of any kind). It is communication theory (1.3.3.),
however, with its immense practical importance, which has received 3.3 In the following paragraphs 4 and Î, these two approaches will be
the greatest attention; and it is only the logical priority of the other discussed. The first has given rise to two complementary definitions
two which prevents it from coming first on the list. of what has been called "amount of information," and the second
to a third. These, however, are not rivals, but are autonomously
"^This aspect of the subject is already an established branch gf miithçmîitics undçr thç namç of valid measures, appropriate in answer to different questions.
Representation Theory or Abstract Group Theory,
226 Cybernetics Appendix I 227
4. Analysis of Representations Logon Capacity 4.1.4 The number of logons provided by apparatus per unit of co-
synonym: ordinate space (centimeter, square centimeter, second, etc.) is
4.0.1 Representations communicable in a two-valued (yes-or-no) Logon Density termed its logon capacity.
form are necessarily quantal in structure, since an "imperceptible
change" in a two-valued logical form is by definition meaningless.
All the changes are discrete, therefore the elementary concepts of 4.1.4.1 For example, a channel whose band width permits of
logical representations are discrete and enumerable. X independent amplitude readings per second has a logon capacity
of X; in a microscope, logon capacity is a measure of resolving
power; and so forth.
4.0.2 In fact, a large class of such representations can be reduced
to a form made up only from identical elements, so simple that
their only attribute is existence (1). This fact provides a basis for Structural 4.1.5 The reciprocal of logon capacity is termed the structural
the quantitative analysis of such representations (2). (Representa­ Scale Unit scale unit for the apparatus.
tions not amenable to precise logical description have not so far been
considered in the theory, though a large class of these might be
handled in terms of approximate quantal equivalents representing 4.2 Metrical Information
"upper and lower bounds" to their logical content.)
Metrical 4.2.1 The number of (indistinguishable) logical elements in a
4.0.3 In general a pattern reduced to such fundamental terms will Information given group or in the total pattern is termed the Metrical Informa­
contain a certain number of distinguishable groups or clusters of Content tion Content of the group or pattern.
elements, the elements in each group being indistinguishable among
themselves. There are thus two numerical features of interest:
(1) the number of distinguishable groups or clusters of indistin­ Metron 4.2.2 The unit of metrical information, one metron, is defined as
guishable elements in a representation, and (2) the number of that which supplies one element for a pattern. Each element may
elements in a given group or cluster. be considered to represent one unit of evidence. Thus the amount
of metrical information in a pattern measures the weight of evidence
to which it is equivalent. Metrical information gives a pattern its
4.0.3.1 The number of groups, and the numbers of their elements, weight or density—the "stuff" out of which the "structure" is
may be thought of as respectively analogous to the number of col­ formed.
umns and the number of entries per column in a histogram.

4.2.3 In a scientific representation each metrical unit may be

4.1 Structural Information thought of as associated with one elementary event of the sequence
of physical events which the pattern represents.
4.1.1 The number of distinguishable groups or clusters in a repre­
Structural sentation—the number of definably independent respects in which
4.2.4 Thus the amount of metrical information in a single logon,
Information it could vary—its dimensionality or number of degrees of freedom
Metron Content or its metron content, can be thought of as the number of elemen-
Content or basal multiplicity—is termed its Structural Information Content.
synonym: tary events which have been subsumed under one hand or "con-
Information Content densed" to form it. For example, in the case of a numerical param­
Logon 4.1.2 The unit of structural information, one logon (3) (2), is eter, this is a measure of the precision with which it has been
that which enables one such new distinguishable group to be defined determined.
for a representation. Thus structural information is not concerned
with the number of elements in a pattern, but with the possibility
of distinguishing between them. 4.2.4.1 Notice that these elements are indistinguishable, so that
their number is not the number of binary digits (5.1.3) to which
the logon is equivalent.
4.1.2.1 For example, if we are counting identical sheep jumping a
gate and have no sense of time, our result can only be represented
by a certain total number; but if we have a clock, we can now 4.2.5 When we come to represent the results of physical observa­
define what we mean by "the number in the first minute" and "in tions, we are often interested in magnitudes which are not directly
the second," and so forth, and represent our result by a set of distin­ proportional to the metron content. The representations we use do
guishable subtotals. The clock has provided Structural information. not then show the metron content explicitly. It must be clearly
In a similar way the ability to distinguish (e.g.) spatial position realized that metron content as defined is a measure of the number
along the gate would provide distinguishable subtotals and hence of elements appearing when what is believed to have happened is
increase the structural information content of our representation. represented in its most fundamental physical terms.

Logon Content 4.1.3 Logon content is a convenient term for the structural infor­ 4.2.6 Thus, if an estimate is made of a parameter from a statis­
mation content or number of logons (number of independently vari­ tical sample, the elementary events concerned are the arrivals of
able features) in a representation (e.g., the number of independent "unit contributions" to the sample. These could be represented by
coefficients required to specify a given wave form over a given Conceptual the number of intervals occupied on a conceptual scale proportional
period of time). 5cale to metron content.
228 Cyhernetics Appendix I 229
mate would be proportional to the number of sheep per group, and
4.2.7 On the other hand, the usual representation shows the mag­ the probable error in each estimate would be inversely proportional
nitude of the parameter concerned, and not generally the metron to the square root of this number. Hence the number of proper
content, on a linear scale, graduated in elementary intervals which scale intervals occupied by the estimated parameter would be pro­
in the useful limit are just large enough to give the representation portional to the square root of the metron content of each group.
of the magnitude (scale reading) a probability of Yz. Such a scale
Proper Scale is termed a proper scale. Now the probable error, in a normal 4.2.14 The term "amount of information" in a sense analogous to
population, is inversely proportional to the square root of the size this was first used by R. A. Fisher (4), who defines it as follows:
of the sample. Hence the magnitude in such a case would be shown Suppose we have a probability distribution function / (x^xo) show­
as occupying on the proper scale a number of elementary intervals ing how in a given population a variable x is distributed about a
proportional only to the square root of the number of elementary
parameter Xo; e.g.,
events. The number of metrons is (in this special but common
case) the square of the number of occupied intervals shown in this
less fundamental physical representation.
The "amount of information" in n samples from the population
Numerical 4.2.7.1 In connection with radar information, the term numerical is defined as n times the wcigbtiul mean of
Energy energy has been used to represent what is essentially the metron
content of a signal. It is the ratio (Total Energy)/(Noise power
per unit band width).
0
4.2.8 In general, then, a clear distinction exists between (a) the
fundamental representation on a conceptual scale showing the in­ over the range of oo » i'C-»
variant number of logical elements, and (h) the representation of
the magnitude which is of practical interest. In fact, the connec­ CO
tion between the two is little closer than that between the precision
with which a given variable can be measured, and its magnitude. n / ( y \ dx
Precision increases monotonically with metron content, but few
quantities are linearly related to metron content. Power and energy
J '0^0
— 00
in the classical case are among the few exceptions; this accords with
Equivalent forms are
their apparently fundamental status among physical concepts.

Scale Unit 4.2.9 The scale unit of a magnitude is the minimum interval in r ^ 9/
terms of which the scale can usefully or definably be graduated. n f — ( )' äx
J /f 9^0
4.2.9.1 For a magnitude imprecisely known, it is defined as above — 00
(4.2.7) to be equal to the probable range of error. A magnitude and
supported by a single metron occupies just one interval on such a 00
scale. In practice larger units are often used—e.g., range of stan­ 3'log/
dard error.
J ?IXo'
Coincidence 4.2.9.2 But it should be remembered that in theoretical representa­ — 00
Relations tions the size of the scale unit is generally limited by our inability
to define a smaller unit in terms of the coincidence relations out of
which physical representations are constructed. In the case of a normal distribution, this reduces to the reciprocal
of the variance, provided that the range is independent of Xo. It is
4.2.10 The number of metrons per unit of co-ordinate space is thus a direct measure of precision, though it is not dimensionless
Metron Capacity
synonym ; termed the metron capacity or metron density of a physical observa­ unless suitably normalized.
Metron Density tion system (<"/. 4.1.4).
4.3 Representation of Information
Conceptual Unit 4.2.11 The co-ordinate interval over which one metron is acquired
undesirable is termed a conceptual unit or (undesirably) a metrical scale unit
synonym : of co-ordinate. The Information Content of a given representation is specified by
Metrical Scale Unit setting down the metron content of each logon. This may be repre­
sented in various ways.
4.2.12 It will be noted that metron content is necessarily positive.
Information Vector 4.3.1 One convenient method is to use a multidimensional infor-
4.2.13 Returning to the example of the jumping sheep, let us now Information Space -mation vector in an information space of which each axis represeius
suppose that we are trying to determine a figure for the average one logon. The squares of the components of this vector are the
value of some parameter of a sheep in each group which we are metron contents of the respective logons. Thus the square of the
able to distinguish. Assuming for the sake of illustration that the length of the vector itself is the total metron content, the sum of
parameter is normally distributed, the metron content of each esti­ the individual metron contents.
 Appendix I 231
230 Cybernetics
5.1.2.1 The base 2 is chosen because a selection among a set of
4.5.1.1 In this representation the angle between two directions has n possibilities can be carried out most economically by dividing
a direct interpretation as a measure of relevance. A dependent state­ the total successively into halves, quarters, eighths, etc., until the
ment is defined by a ray in the space, and the metron content af­ desired member is identified. The number of stages in this process
forded to it by the information is found by squaring the projection is then the integer nearest to, and not less, than logs n.
of the information vector on the ray.

5.1.2.2 The information measure so defined equals the number of

4.3.1.2 A new complete representation may be set up by supple­ independent choices between equiprobable alternatives which would
menting this dependent statement (4.3.1.1) by a set of others repre­ have to be determined before the required representation could be
sented by orthogonal rays. This process amounts to a rotation of identified in the ensemble of which it is a member. (The prior prob­
axes, which leaves us with a new total metron content equal to the ability measures the fraction of the members of the ensemble which
old. are of the required kind.)

4.3.2 The same processes can be represented in terms of matrix 5.1.2.3 This, like the first two measures (Paras. 4.1.1 and 4.2.1),
algebra, if the metron contents of logons are set out initially as the represents a number of logically "atomic" propositions; but this
information Matrix elements of a diagonal information Matrix. Dependent statements time they define, not the representation, but the selection process
now define vector functions, and their metron content is found by leading to it.
forming scalar products of the form • /$, where I is the infor­
Trace mation matrix, ^ a vector function, and its transpose. Under
synonym: all complete ("unitary") transformations, the trace or sum of the Selective 5.1.3 Information in the above sense of that which determines
Spur, Characteristic, diagonal elements of 1 remains invariant, being the total metron Information choice may be termed selective information.
Diagonal Sum content.
Binary Digit 5.1.4 The unit of selective information, one binary digit or bit, is
4.3.3 An alternative geometrical representation suitable for some synonym : that which determines a single choice between equiprobable alterna-
particular cases employs a three-dimensional histogram having a co­ Bit
ordinate and its Fourier transform (e.g., time and frequency) as its
two basal axes. Since the number of logons provided by given 5.1.5 In a long sequence of different representations of which the
apparatus is proportional to the product of band width (q.v.) and i'th kind has a prior probability pi (and hence an average fre­
conjugate co-ordinate, the base is divisible into equal cells each quency of occurrence /»»), the average amount of selective informa­
representing one logon. On each cell is erected a column having a tion per representation is evidently the weighted mean of —log p
height proportional to the logarithm of metron content. This gives over all kinds of representation, or H = —2 pi log pi, which,
the total volume of the histogram the same qualitative significance
apart from some ambiguity of sign in the literature, is also the
as the logarithm of the volume spanned by the information vector
Entropy standard definition of the entropy of a selection.
of 4.3.1.

5.1.5.1 Where representations take the form of continuous func­

5. Communícatíon: Replication of Representations tions, H takes the form —5/ (x) log f (x) dx, where / (x) is the
probability distribution of the representative variable x. It is thus
5.1 The problem of communication usually concerns representations of the weighted mean of [— log /] over the range of x.
which all parts exist already in the past experience of the receiver.
.In other words the receiver already possesses prefabricated compo­ 5.1,6 In practice the receipt of a communication signal disturbed
nents of the representation. by noise merely alters the form of f (x) (generally narrowing it),
and does not specify x uniquely. The amount of selective informa­
5.1.1 In fact it is generally assumed to be known that the corn- tion received is then defined as the difference between the values of
Ensemble plete representation to be replicated is one member of a finite en­ H computed before and after receipt of the signal.
semble {(I.V.) of possible originals, some of which have in the past
been communicated more often than others. We may then say that 5.1.7 When two representative variables, x and y, say (discrete or
these more common messages "give less information" than the oth­ continuous), are in question, knowledge of the value of one may
ers, using the notion of "information content" now in the important affect the prior probability of the other. In the above notation,
sense of Para. 3.2.1. f (y) depends on x, so that the entropy H oí y will also vary
with X. If we picture an ensemble in which values of x occur in
5.1.1.1 Our message is here thought of as telling us to select one Conditional their expected proportions, we define the conditional entropy of y,
prefabricated representation. Entropy Hx(y) as the average value of the entropy of y (calculated for each
value of x) over all members of this ensemble.
5.1.1.2 We are not now asking "How big is it?" or, "How much
detail has it?" but rather, "How unusual or unexpected is it?" 5.1.7.1 This may also be described as the "weighted mean entropy"
"How much trouble will it take to find it in my ensemble?" of y, weighted by the probability of getting the different values of
x. It therefore measures our uncertainty about y when we know x.

5.1.2 A convenient measure of information content in this sense is 5.1.7.2 An analogous conditional entropy Hy(x) can be defined
the negative logarithm (base 2) of the prior probability of the for x when y is known.
representation concerned (5) (6).
232 Cybernetics Appendix I 233
Î.1.8 Where x and y represent respectively the input and output 5.4.1 For example, if all n distinguishable voltage-levels of a trans­
of a noisy communication channel, the conditional entropy liy{x) mitted signal are regarded as equiprobable, the selective information
Equivocation is termed the equivocation. It is a measure of ambiguity. content per logon is proportional to log n. On the other hand the
physical entropy increase is proportional to, or must exceed n
Î.1.9 The number of bits per second which a channel can transmit (2) (7).
Capacity is termed its capacity. For the case (5.1.8) it is defined as the
maximum of \.H(x) - Hy(x)^. 5.4.2 Here the correlation is between metrical information content
(Para. 4.2.7) and physical entropy increase. In fact, metron con­
5.1.10 The ratio of the entropy of a source to the maximum value tent can be thought of here as the number of unit increases of
which it could have while using the same symbols is called its physical entropy—i.e., of elementary events—which have been sub­
Relative Entropy relative entropy. sumed under one head, thereby losing their distinguishability and
potentiality of serving as "bits."

Redundancy 5.1.10.1 One minus the relative entropy is termed the redundancy.

5.2 These considerations suggest a more economical method of commu­

nicating a representation.

5.2.1 Instead of transmitting a physical representation of the rep­

resentation itself, we may transmit a representation of the selection
process by which it may be identified in the ensemble of possible
representations which is assumed to exist at the receiving end.

5.2.2 A system whereby a representation is defined by a selection

Code System process is termed a code system.

5.2.3 The corresponding representation of the selection process

Code Signal transmitted is known as a code signal.

5.2.3.1 As a physical sequence the code signal will itself have

metrical and structural features as discussed in Para. 4, and will be
definable by a vector in an information space. But its structure need
not have anything in common with that of the representation which
it identifies.

5.2.3.2 On the other hand, the ordinary case of making physical

representations could be thought of formally as a special case of
coding, one-for-one.

5.3 It follows that the result of an experiment, as well as a communica­

Selective tion signal, could be analyzed in terms of its selective information
Information content.
Content
5.3.1 This is a relative measure, depending on the number of dis­
tinct results which were regarded as equally probable by the ob­
server. The result observed is thought of as specifying one of a
number of possibilities already contemplated by the observer as
forming an ensemble in defined proportions.

5.3.2 The amount of selective information derived from the ex­

periment can then be computed in the same way as for a message,
(5.1).

5.4 The entropy of selective information content of a selection should

not be facilely identified with the physical entropy of thermo­
dynamics. The two are equivalent only in the particular case where
the ensemble from which the selection is made is a physical one
defined for a state of thermodynamic equilibrium.
234 Cybernetics Appendix I 235

ALPHABETICAL INDEX OF TERMS USED IN INFORMATION Information matrix: A matrix in which the metrical and structura! information content of an
THEORY AND RELATED COMMUNICATION THEORY experiment are specified.

References are to Paragraphs in the Glossary Information space: The space in which independent logons are represented by orthogonal
rays, and their metron contents by the squares of distances along these
rays. (4.3.1)
Band width: In general terms, the region of Fourier space (q.v.) to which the output
of an instrument is confined. In particular, the effective frequency range
Information vector: The vector whose components in information îpace are the distances just
(conjugate to a given co-ordinate) to which it responds.
mentioned.

Binary digit: (S.1.3.) Unit of selective information-content.

Bit: Logon: Unit of structural information content (q.v.).

Capacity: Logon capacity:

(5.1.9) Number of bits transmissible per second.
possible (4.1.4) Number of logons per unit of co-ordinate space,
synonym :
Code system: (5.2.2) Logon density:

Code signal: (5.2.3) Logon content: (4.1.3) Number of independently variable features,
synonym :
Conceptual unit: Structural infor- ,
(4.2.11) The co-ordinate interval in which one metron is acquired. Re­
ciprocal of metron density. mation content

Ensemble: A set of possibilities each of which has a defined probability. Metron: (4.2.2) Unit of metrical information content (q.v.).

Entropy; Metron capacity: (4.2.10) Cf. logon capacity,

(5.1.5) (a) In statistical mechanics, the weighted mean of the (nega­
tive) logarithm of the probabilities of members of an ensemble, (h) In
synonym :
Metron density
thermodynamics that function of state of a body or system which in­
creases by
2 Metron content: (4.2.4) Measures the amount of evidence to which a representation is

/
synonym : equivalent.
Ag/r Metrical infor­
mation content

in a reversible process between two states 1 and 2, where AQ is tíi« heat Numerical Energy: (4.2.7.1) Ratio of (Energy)/(Noise Power per unit band width).
taken up by the body or system at temperature T. [Definitions (a) Analogous to metron content.
and (h) are equivalent.]
Proper scale: (4.2.6) A representational scale on which equal intervals are equi­
Conditional: (5.1.fi) probable.

Relative: (5.1.10) Redundancy: (5.1.10.1) One minus relative entropy.

Equivocation: (5.1.8) Representation: (!•!) A symbolic picture, model, statement, etc.

Fourier space: The space whose dimensions represent variables which are Fourier trans­ Scale unit: (4.2.9) The minimum interval in terms of which a scale can definably
forms of co-ordinates (e.g., the frequency of conjugate to the time or usefully be graduated.
co-ordinate).
Metrical: (4.2.11) Undesirable equivalent of conceptual unit. Reciprocal of
information: That which alters representations (2.1). metron density.

Metrical: (4.2.) Specifying the number of elements of a representational pattern. Structural: (4.1.5) Reciprocal of logon capacity.

Selective (5.1.3) Specifying the unforeseeableness of a pattern.

Structural: (4.1.) Specifying the number of independently variable features or

degrees of freedom of a pattern.
 Appendix II 237
APPENDIX II 14. JAENSCH, E. R. : Über Raumverlagerung und die Beziehung von
Raumwahrnehmung und Handeln. Ztschr. f. Psychol, u. Physiol,
d. Sinnesorg. 89, (1922).
REFERENCES 15. BATESON, G.: Pidgin English and cross-cultural communication.
Tr. New York Acad. Sc. Ser. II, 6, 137 (1944).
COMMUNICATION PATTERNS IN PROBLEM-SOLVING GROUPS 16. STENZEL, J.; Über den Einfluss der griechischen Sprache auf die
1. MILLER, G. A.: Paper read at Psychological Round Table, De­ philosophische Begriffsbildung. Neue Jahrb. f. d. klass. Altert.
cember 3, 1950, Elizabeth, N. J. (Not published) 47, 152 (1921).
2. BIRCH, H. G.: The relation of previous experience to insightful 17. ; Metaphysik des Altertums. Handbuch der Philosophie.
problem-solving. /. Comp. Psychol. 38, 367 (1945). Bäumler, A., and Schröter, M., Editors. Munich, Oldenbourg,
3. WARDEN, C.J. : Animal Motivation. New York, Columbia Univ. 1934 (Sect. L D- 1-196).
Press, 1931. 18. SAER, D. J.: The effect of bilingualism on intelligence. Brit. J.
4. VON NEUMANN, J., and MORGENSTERN, O.: Theory of Games . Psychol. 14, 25 {1920).
and Economic Behavior, (2nd ed.). Princeton, Princeton Univ. 19. SULLIVAN, H. S.: Conceptions of Modem Psychiatry. Washing­
Press, 1947. ton, D. C., W. A. White Psychiatric Foundation, 1947.
20. POPPER, K.; Indeterminism in quantum physics and in classical
COMMUNICATION BETWEEN MEN physics. Brit. J. Phil. Sc. 1, 173 (1950).
21. HEBE, D. O.: Organization of Behavior. A Neuropsychological
1. STEVENSON, C. L. : Ethics and Language. New Haven and Lon­
Theory. New York, Wiley, 1949.
don, Yale Univ. Press and Oxford Univ. Press, 1945 (p. 80).
22. NIELSEN, J. M.; Agnosia, Apraxia, Aphasia. Their Value in
2. RICHARDS, L A.: Emotive meaning again. Philosoph. Rev. 57,
Cerebral Localization. (2nd ed.) New York, Hoeber, 1946
145 (1948).
(p. 292)
3. OPPENHEIMER, J. R.; The age of science 1900-1950. Scient.
23. ISSERLIN, M.: Die pathologische Physiologie der Sprache. Erg.
Am. 183, 20 (1950).
Physiol. 33, 1 (1931).
4. KROEBER, A. L.: Anthropology. Scient. Am. 183, 87 (1950).
24. REUSCHERT, E.: Die Gebärdensprache der Taubstumrnen. Leip­
5. NEWMAN, S. S.: Behavior patterns in linguistic structure: a case
zig, H. Dude, 1909.
study. Language, Culture and Personality. Spier, L., Hallowell, 25. RUESCH, J., and PRESTWOOD, A.: Anxiety; its initiation, com­
A. L, and Newman, S. S., Editors. Menasha, Wise., Sapir munication and interpersonal management. Arch. Neurol. & Psy­
Memorial Publication Fund, 1941 (p. 94).
chiat. 62, 527 (1949).
6. KANTOR, J. R.: An Objective Psychology of Grammar. Bloom- 26. KLÜVER, H.: Mescal. London, Kegan Paul, 1928.
ington, Indiana, Indiana Univ. Book Store, 1936.
7. RICHARDS, I. A.: How to Read a Page. New York, W. W. Nor­ COMMUNICATION BETWEEN SANE AND INSANE
ton & Co., Inc., 1942 ( p. 66).
8. WHORF, B. L.: The relation of habitual thought and behavior 1. KUBIE, L. S., and MARGOLIN, S.: Therapeutic role of drugs in
to language. Language, Culture and Personality. Spier, L., Hal­ process of repression, dissociation, and synthesis. Psychosom.
lowell, A. L, and Newman, S. S., Editors. Menasha, Wise., Sapir Med. 7, 147 (1945).
Memorial Publication Fund, 1941 (p. 75). 2. KUBIE, L. S.; Value of induced dissociated states in therapeutic
9. STOUT, G. F.: Analytic Psychology. New York and London, process. Proc. Roy. Soc. Med. 38, 681 (1945).
The Macmillan Co., Swan Sonnenschein & Co., Ltd., 1902 3. KUBIE, L. S., and MARGOLIN, S.: Process of hypnotism and nature
(Vol. II). of hypnotic state. Am. J. Psychiat. 100, 611 (1944).
10. RAYSOR, T. M., editor: Coleridge's Shakesperean Criticism. Cam­ 4. ERICKSON, M. H., and KUBIE, L. S.: Use of automatic drawing
bridge, Harvard Univ. Press, 1930 (Vol I). in interpretation and relief of state of acute obsessional depression.
11. TINBERGEN, N. : The hierarchical organization of nervous mech­ P s y c h o a n a l y t . Q u a r t . 7, 4 4 3 ( 1 9 3 8 ) .
anisms underlying instinctive behavior. Symp. Soc. Exp. Biol. 5. ; Permanent relief of obsessional phobia by means of
4, 305 (1950). communications with unsuspected dual personality, ibid. 8, 471
12. VON MONAKOW, C., and MOURGUE, R.: Biologische Einführung (1939). . .. ^
in das Studium der Neurologie und Psychopathologie. Stuttgart 6. : The translation of the cryptic automatic writing of one
and Leipzig, Hippokrates-Verlag, 1930. hypnotic subject by another in a trance-like dissociated state.
13. STERN, W.: Über verlagerte Raumformen. Ein Beitrag zur Psy­ ibid. 9, 51 (1940).
chologie der Kindlichen Raumdarstellung und Auffassung. Ztschr. 7. FÄRBER, L. H., and FISHER, C.: Experimental approach to dream
f. an gew. Psychol. 2, 498 (1909). psychology through use of hypnosis. Psychoanalyt. Quart. 12,
202 (1943).
236
238 Cybernetics Appendix II
5 THORPE, W. H.: Orientation and methods of communciatwn of
8. KUBIE, L. S.: Body symbolization and development of language.
honey bee and its sensitivity to polarization of light. Nature
Psychoanalyt. Quart. 3, 430 (1934).
9. : The neurotic potential, the neurotic process, and the
neurotic state. U. S. Armed Forces M.J. 2, 1 (1951).
6 SneiÌÌI'^T.' C.: Social organization in insects as related to
10. SPIEGEL, H., SHOR, J., and FISHMAN, S.: Hypnotic ablation tech- • individual function. Psychol. Rev. AS, 465 (1941). . ,
7 MAIER, N. R. F., and SCHNEIRLA, T. C.: Prmcples of Ammal
nique for study of personality development; preliminary Med.
report. Psychosom. Med. 7, 213 (1945). PwÄoWy. New York, McGraw-Hill, 19p. ^
8. BUTLER, C G.; Bee behaviour. Proc. Roy. Inst. Gr. Brit. 34,
11. KLÜVER, H.: The eidetic child. A Handbook of Child Psychol­
ogy. Murchison, C., Editor. Worcester, Clark Univ. Press, 1931
(p. 643).
I
9 G jetsœ.^W.: Die Staaten der Ameisen. Berlin Springer, 1937.
10. HECHT, S., and WOLF, E.; The visual acuity of the honey bee.
12. : The eidetic type. Research Nerv. & Ment. Dis. Proc.
14, 150 (1934). J. Gen. Physiol. 12, 727 (1929). , , • r u f ficVipc
13. : Studies on the eidetic type and on eidetic imagery. Psy­ 11. BREDER, C. M., JR.: Certain effects in habits of
chol. Bull. 25, 69 {1928). as based on observation of Jenkinsia. Am. Mus. Novit. 382,
14. ; Eidetic phenomena, ibid. 29, 181 (1932).
and HALPERN, F.: Innate and acquired behavior
15. : Fragmentary eidetic imagery. Psychol. Rev. 37, 441 12.
(1930). affecting aggregation of fishes. Physiol. Zool. 19, ^-54 (19 )•
and NIGRELLI, R. F.: The influence of temperature
16. : Mechanisms of hallucinations. Studies in Personality 13.
and other'factors on water aggregation of sunfish, Leponts Aun us.
Turman, L., Editor. New York, McGraw Hill, 1942 (p. 175).
17. SCHILDER, P.: Psychoanalyse und Eidetik. Ztschr. f. Sexualwis- Ecology 16,^33 j . Comparative behavior of various
sench. 13, 56 (1926). 14.
18. : Mind: Perception and Thought in Their Constructive fishes under different conditions of aggregation. Copeta 1, ¿V
Aspects. New York, Columbia Univ. Press, 1942.
19- ZEMAN, H.: Verbreitung und Grad der eidetischen Anlage. 15.
PARR^A E.: a contribution to the theoretical analysis of school­
Ztschr. f. Psychol, u. Physiol, d. Sinnesorg. 96, 208 (1924). ing behavior of fishes. Occ. Pap. Bingham Oceanogr. Coll. 1,
20. STORKER, A. J. : über psychoanalytische Tierpsychologie. Psycho­
analytische Bewegung. 3, 40 (1931). 16. &A
16 S
FORD,M.P.: The cooperative solving of problems by young
21. HERMANN, I.: Modelle zu den Odipus — und Kastrationskom­ chimpanzees. Comp. Psychol. Monogr. 14, No. 2 (1937).
plexen bei Affen. Imago 12, 59 (1926).
Í
17. YER ES, R. M., anä NISSEN, H. W.: Pre-linguistic sign behavior
22. ARMSTRONG, E. A.: Bird Display and Behaviour. London, in chimpanzee. Science 89, 585 (1939)- • i i
Drummond, 1947. 18 CARPENTER C. R.; A field study of behavior and social relations

S howHng monkeys (Alouatta palliata). Comp. Psychol. Monogr.

23. : The nature and function of displacement activities.
Symp. Soc. Exp. Biol. 4, 361 (1950).
19. C^-^The Expression of Emotions in Man and Antrnals.
24. KUBIE, L. S. : The concept of normality and the neurotic process.
Practical and Theoretical Aspects of Psychoanalysis. New York,
International Univ. Press, 1950 (p. 12). 20 KLÜvm'^^H.t^Functiónal significance of the geniculo-striate sys-
25. , et al.: Problems in clinical research, Round Table, 1946. (-iew). Psychoanalyt.
Am. J. Orthopsychiat. 17, 196 (1947).
26. : The objective evaluation of psychotherapy. Round table, Quart. 10, 329 (1941).
1948. ibid. 19, 463 (1949).
IN SEARCH OF BASIC SYMBOLS
COMMUNICATION BETWEEN ANIMALS 1. SHANNON, C. E., and WEAVER, W : Mathematical Theory of
1. HYMAN, L.: Metabolic gradients in amoeba and their relation to Communication. Urbana, Univ. 048
the meehanism of amoeboid movement. /. Exper. Zool. 24, 55 2 WIENERIN.: Cybernetics. New York, Wiley, 19 -
3 MACKAY, D. M.: Quantal aspects of scientific information. Phtl.
2. CHILD, C. M.: Physiological Foundations of Behavior. New ' Mag. (Ser. 7) 41,289 {\9^Q)-
York, Holt, 1924. 4 : (See Appendix I, this volume) c • oa
3. ROTH, L. M. : A study of mosquito behavior. Am. Midland 5. GABOR, D.: Theory of communication. /. Inst. Elee. Engrs. 93,
Naturalist 40, 265 (1948).
4. VON FRISCH, K.; Bees. Ithaca, Cornell Univ., 1950. 6. S sher,^K^A.: The Design of Experiments. Edinburgh, Oliver
& Boyd, 1935.
 Cybernetics
1. MACKAY, D. M.: Mind-like behavior in artefacts. Brit. f. Phil.
Sci. (Aug. 1951).
8. FRIEDMAN, M., and SAVAGE, L. J.: The utility analysis of choices
involving risk. /. Polit. Econ. 56, 279 (1948).

THE NOMENCLATURE OF INFORMATION THEORY

1. WITTGENSTEIN, L.: Tractatus Logico-Philosophicus. London,
Kegan Paul, and New York, Harcourt Brace, 1922.
2. MACKAY, D. M. : Quantal aspects of scientific information. Phil.
Mag. (Ser. 7) 41,289 {19^0).
3. GABOR, D.: Theory of communication. J. Inst. Elee. Engrs. 93,
429 (1946).
4. FISHER, R. A.; The Design of Experiments. Edinburgh, Oliver
& Boyd, 1935 (p. 188).
5. SHANNON, C. E. : Mathematical theory of communication. Bell
Syst. Tech. J. 27, 379 (1948).
6. WIENER, N.: Cybernetics. New York, Wiley, 1948.
7. MACKAY, D. M.: Entropy, Time and Information. Information-
Theory Symposium, London, September, 1950.

75Ì9 58