Norbert Wiener meets Arturo Rosenblueth

It is difficult to give a date for the birth of the new science and quite impossible to give the credit for the conception to any one individual. Many of the essential ideas had been developing for a long time, often without their importance being fully realised, but some of the first discussion consciously along these lines took place at a monthly dinner party organised by Arturo Rosenblueth in the Vanderbilt Hall of Harvard University in the early 1940s. These meetings were deliberate attempts to further communication between the various sciences. They were attended by scientists from many different specialisms and on each occasion one of them would introduce a discussion on his own topic.

It was in this environment - exceptionally favourable to the development of revolutionary scientific ideas - that Rosenblueth, who was professor of physiology at the Harvard Medical School, and Norbert Wiener, professor of mathematics at the Massachusetts Institute of Technology, first got to know each other. Each had hoped one day to take part in some investigation involving the co-operation of physiologists and mathematicians, but it required a war to bring about the realisation of these hopes.

Prediction and voluntary control

At that time an important military problem was the direction of anti-aircraft gun fire. Aircraft speeds had so increased that they were now of the same order of magnitude as that of the shells used. What was wanted was a method of extrapolating the path of the aircraft to predict its future position, so that a shell could be sent to meet it. Wiener became interested in the problem and suggested some predictor functions, the possibilities of which were then examined by means of the Bush differential analyser. The eventual result was that Wiener and another mathematician, Julian Bigelow, collaborated on the development of a theory of prediction and the design of a computer which could make use of it in the control of anti-aircraft gun fire.

One of the interesting features of the problem was that the whole fire-control system included stages operated by human beings-the gunner and the pilot of the target aircraft. This meant that these mathematicians-very fully informed as they were about prediction functions, control systems, and servomechanisms-had to turn their attention to the study of voluntary control behaviour in human operators. They came to the conclusion that the feedback of errors played just as important a part here as in man-made servomechanisms. A great deal of mathematical theory had already been developed to deal with feedback functions, and it was possible to predict that, if feedback loops did exist in the human nervous system, then, under some conditions of maladjustment, they would give rise to an oscillation in the voluntary movement, or hunting. Wiener and Bigelow consulted Rosenblueth on the point and heard that such symptoms did occasionally appear in human beings.

The birth certificate of cybernetics

From this resulted the collaboration of these three workers on the paper which has been called "the birth certificate of cybernetics" (de Latil, 1953) and which was eventually published in Philosophy of Science under the title "Behaviour, purpose, and teleology" (Rosenblueth, Wiener, & Bigelow, 1943). The main theme of the paper is a classification of types of behaviour with special reference to the concept of purpose. They define 'behaviour' as 'any change of an entity with respect to its surroundings', meaning that any modification of an object, detectable externally, may be denoted as behaviour. Hence the 'behaviouristic approach' to any object is the examination of its output and the relation of this output to the input. They contrast this to the 'fundamental approach', in which the structure, properties, and intrinsic organisation of the object is studied, rather than the relationship between the object and its environment.

Classification of behaviour

Their first classification is into 'active' behaviour, in which the object itself is the source of the energy in the output, and 'non-active', or 'passive' behaviour, in which all the energy in the output comes from the immediate input, or the object controls energy which remains external to it. Active behaviour they further divide into 'purposeful' or 'goal-directed', and 'non-purposeful' or 'random'. Hence they would describe a machine as 'purposeful' only if it has some specific final condition towards which its activity is directed, e.g., a torpedo which contains a servomechanism causing it to seek its target; and they recognise that some machines, e.g., a roulette wheel, are designed to be purposeless.

Purposeful behaviour is then classified into 'feedback' or 'teleological', and 'non-feedback' or 'non-teleological'. Feedback systems are those in which the input is altered by the output in the direction necessary to reduce the discrepancy between the situation so far achieved and the goal situation. Some machinery involves a continuous feedback of the error in this way. The transmission of these signals requires time, so that at certain frequencies the direction of the feedback is in effect reversed, and results in a build-up of oscillation. This is the phenomenon which first invited the comparison of the action of the nervous system in voluntary activity with that of an error-controlled feedback system.

Such feedback behaviour may be further divided into 'extrapolative' or 'predictive', and 'non-extrapolative' or 'non-predictive'. In extrapolative behaviour the path of the target is predicted, and aim taken towards its most probable future position. Such prediction may be of first, second, or higher order. The original problem, in which an anti-aircraft shell is fired at an aircraft, requires second order prediction, because both the path of the shell and that of the aircraft have to be predicted. Perhaps human behaviour may be distinguished from that of other animals by the use of higher orders of prediction.

Purpose equated with feedback

In the past, the notion of teleology has been discredited because it seems to imply the existence of 'final causes', which in some way occur after their effects. Now that the concept of purpose can be described in terms of feedback loops, the idea of a teleological process no longer presents the same difficulties. Any teleological behaviour which can be shown to be due to control by feedback of error is capable of precise description. It may be useful to investigate 'purpose' without having to be concerned with causality.

One of the first indications of the new science which was to develop occurs in this paper with the sentence "The broad classes of behaviour are the same in machines and in living organisms" followed by "While the behaviouristic analysis of machines and living organisms is largely uniform, their functional study reveals deep differences".

Rosenblueth and Wiener had kept before them the idea of a research program which would use the efforts of many different kinds of scientist to attack subject matter left out by the traditional division of labour. Certainly Wiener regarded this paper as a statement of such a program, and as a possible starting point for their plan for an inter-scientific institute.

Application to cardiac muscle

In 1945, Rosenblueth became head of the physiology laboratories of the Institute Nacional de Cardiologia in Mexico. In the summer of 1945 Wiener went there for a period of ten weeks and the two of them collaborated on a theoretical exploration of the mathematical formulation of the problem of conduction of impulses in a network of connected excitable elements, specifically in cardiac muscle. (Wiener & Rosenblueth, 1946). They considered the main interest of the research to be that conduction in nervous tissue resembles that in cardiac muscle in that the propagation is active, with energy supplied locally, and is of an all-or-none character. In both cases activity is followed in turn by a relatively refractory period, during which the tissues have subnormal excitability. They were particularly interested in the similarities between the tonic, clonic, and phasic contractions in epilepsy and the tonic spasm, beat, and fibrillation of the heart. They discuss the propagation of impulses in a simplified mathematical model, and conclude that it is adequate for a statistically random network of fibres, and the establishment of equations for conduction over such a network, and the consideration of flutter and fibrillation, but that development of the mathematics was necessary before the model could be of further use. This Wiener and Pitts hoped to do.

In the summer of 1946, Wiener returned to Mexico on a grant from the Rockefeller Foundation, for another period of collaboration with Rosenblueth. They experimented with the leg muscles of spinal cats, recording their electrical and mechanical behaviour under conditions which gave rise to periodic contractions. They tried to analyse these by methods taken from servomechanism theory. They communicated their results to the third meeting of the Josiah Macy Conference Group, which will now be described, and to the meeting of the New York Academy of Sciences on "Teleological Mechanisms" (Frank et al., l948).

Josiah Macy, Jr. Foundation Conferences

A very important part in the development of the new science of cybernetics was played by the Josiah Macy, Jr. Foundation, which therefore merits a brief description.

This organisation puts into practice the belief that discoveries in one field of scientific activity can often result from information gained in quite another, that the increasing isolation of the different branches of science is a serious obstacle to progress, and that it is vital to establish channels for the effective dissemination and exchange of information. The Foundation attacked the problem by organising several conference groups. Some of the topics covered by such groups have been: adrenal cortex, ageing, blood clotting, cold injury, connective tissues, consciousness, infancy and childhood, liver injury, metabolic inter-relations, nerve impulse, renal function, shock and circulatory homeostasis. A small number of scientists are selected to be the nucleus of each group, care being taken to include representatives of all sciences which may be relevant. Guests are sometimes invited.

Another principle on which the Foundation works is that the presentation of a formal paper at such a meeting is worse than useless because it tends to give a false air of authority to what is said. To encourage members of the conference groups to challenge each other the atmosphere is made as informal as possible. Each meeting lasts two days and takes place at a secluded inn. To keep the emphasis on discussion there are only two or three main speakers each day, and the others are encouraged to interrupt.

Rosenblueth gives the first talk on cybernetics

In May 1942 this Foundation held a conference on "Cerebral Inhibitions". At that meeting, Rosenblueth gave a talk on the ideas which were to appear in "Behaviour, Purpose, and Teleology". Apart from Rosenblueth, those present on that occasion included: Gregory Bateson, Lawrence Kubie, Warren McCulloch, Margaret Mead, and the Medical Director of the Foundation, Frank Fremont-Smith. All these were very enthusiastic about the new approach and asked that the Foundation should form a new conference group to discuss it further; so McCulloch was appointed chairman of a new group, the nucleus of which he and Fremont-Smith formed by combining some of those who had attended the "Cerebral Inhibition" meeting with those who had been present at a joint meeting of engineers, physiologists, and mathematicians, which had been organized in 1944 by Wiener and John von Neumann, and which had included Walter Pitts, McCulloch, and Lorente de No.

The new conference group forms

The first meeting of the new group was held in March 1946, with the title "Feedback Mechanisms and Circular Causal Systems in Biological and Social Systems". Two more meetings were held that year: in September, one on "Teleological Mechanisms in Society", and in October, another on "Teleological Mechanisms and Circular Causal Systems", at which Rosenblueth and Wiener described experiments which they had been doing on clonus in cat leg muscle.

The New York Academy of Sciences was stimulated to ask the same group to hold a symposium on "Teleological Mechanisms" which was subsequently published (Frank et al., 1948).

The third Josiah Macy Conference was held in 1947, again on teleological mechanisms, and the fourth and fifth, which was chiefly concerned with the structure of language, were in 1948.

Cybernetics is given its name

By the summer of 1947, the science of control and communication had developed to such an extent that it was beginning to be inconvenient not to have a name for it, and so the term 'cybernetics' was coined. So much confusion has arisen over the meaning of this word that the following passage is worth quoting in full from the book which Wiener wrote in 1947 and dedicated to Rosenblueth. It is not only of some historic interest, but is quite clear as to the meaning that was intended for the new word.

"Thus as far back as four years ago, the group of scientists about Dr. Rosenblueth and myself had already become aware of the essential unity of the set of problems centering about communication, control, and statistical mechanics, whether in the machine or in living tissue. On the other hand, we were seriously hampered by the lack of unity of the literature concerning these problems, and by the absence of any common terminology, or even of a single name for the field. After much consideration, we have come to the conclusion that all the existing terminology has too heavy a bias to one side or another to serve the future development of the field as well as it should; and as happens so often to scientists, we have been forced to coin at least one artificial neo-Greek expression to fill the gap. We have decided to call the entire field of control and communication theory, whether in the machine or in the animal, by the same 'Cybernetics', which we form from the Greek kubernetes or 'steersman'" (Wiener, 1948a, p. 19).

Further justification for the new term is that kubernetes is the root of the Latin verb gubernare, 'to govern', and that one of the earliest forms of automatic control mechanism was the speed governor of the steam engine. Incidentally, the word cybernétique had been used, in something approaching the present sense, when Ampère used it as a name for his science of civil government (Ampère, 1834).

The new group holds regular meetings

'Cybernetics' was immediately chosen as the title for all the subsequent conferences of the Josiah Macy group, the original title being used as a subtitle. From that point on the character of the conferences changed. No record is available of the first five meetings, but from the sixth onwards a stenotype record was taken of everything that was said. The group met regularly every year, from 1949 to 1953, and the stenotype script was edited by Heinz von Foerster (1950, 1951, 1952, 1953, 1955) and published by the Josiah Macy, Jr. Foundation.

The editorial policy was to interfere as little as possible with the text of the actual discussion. The policy is a brave one - it often results in a record that is muddied, repetitive, or impossible to follow - but it is amply justified by the admirable success of the published proceedings in reconstructing the atmosphere of the meetings. Fremont-Smith (von Foerster, 1953), "By preserving the informality of our conferences in the published transactions, we hope to portray more accurately what goes on in the minds of scientists and to give a truer picture of the role which creativity plays in scientific research". These aims have been achieved - all the obscur ity is there, so is the feeling of frustration caused by the difficulty of putting specialist ideas across to workers in an entirely different field. Often a sort of retroactive inhibition is evident - the participants are irritated by an exposition of some of their own ideas in the terminology of, or from the viewpoint of, another science, and seem to find them more difficult to grasp than they would have done if the ideas had been entirely new ones. Sometimes the inter-group loyalties and jealousies show up very clearly.

Because the participants in the Josiah Macy meetings represented the confluence of several different streams of thought, it was inevitable that, although they started with the consideration of feedback situations, their interests soon widened. By 1949, when the transactions of the Group began to be published, its members had come to believe that the usual physical scientists' approach - that chiefly in terms of energy transformations - was not the most appropriate to problems involving feedback. It had become clear to them that the best fundamental concept to use was not energy, but information.

The theory of information

By this time a theory of information was already in existence, having been constructed chiefly by engineers striving to design efficient communication devices. One of their main problems was the achievement of a favourable 'signal-to-noise ratio' and, to tackle this effectively, they had to be able to measure the information content of a message and the information capacity of a communication channel, and to compare the efficiency of different coding processes (Shannon, 1948: Shannon & Weaver, 1949). A main theme of the conferences was the consideration of ways in which the introduction of these theoretical notions could further the understanding of human communication, the structure of language, how they are learned, and how they are understood. For example, the work of G. A. Miller (1951a; Miller & Frick, 1949; Miller & Selfridge, 1950, and of E. B. Newman (1951a, 1951b) was discussed. Another topic was the application of information notions to social science (Bavelas, 1948) and to anthropological questions (Bateson 1949).

Automatic computing machinery

These discussions naturally led to the consideration of devices which store information, and of automatic computing machinery. Judged particularly important were the differences in performance between machines employing digital coding and those using analogue means of handling the information. Because some features of nervous systems act on an all-or-none principle, the digital computer is an attractive model for nervous activity. Discussion along these lines led the Group to look at several mathematical systems - in particular the theory of the universal Turing machine (Turing, 1937) and the logical calculus due to McCulloch and Pitts (1943). From these they went on to consider the possible ways in which information might be stored in the brain (von Foerster, 1950; Young, 1953).

References

BATESON, G. (1949). Bali - the value system of a steady state. In Meyer Fortes (Ed.), Social structure - studies presented to A.R. Radcliffe-Brown. Oxford: Clarendon Press.

BAVELAS, A. (1948). .A. mathematical model for group structures. Appl. Anthrop. 7, 3, 16.

de LATIL, P. (1953). La Pensee artificielle -Introduction a la cybernetique. (Artificial thought - Introduction to cybernetics.) Paris: Gallimard.

FRANK, L. K., HUTCHINSON, G.E., LIVINGSTON, W.K., McCULLOCH, W.S., and WIENER, N. (1948). Teleological mechanisms. Ann. N. Y. Acad. Sci. 50, 187-277.

McCULLOCH, W.S. and PITTS, W. (1943). A logical calculus of the ideas immanent in nervous activity. Bull. math. Biophys. 5, 115-133.

MILLER, G.A. (1951). Language and communication. New York: McGraw Hill.

MILLER, G.A. and FRICK, C.F. (1949) .Statistical behaviouristics and sequences of responses. Psychol. Rev. 56, 6.

MILLER, G.A. and SELFRIDGE, J.A. (1950). Verbal context and the recall of' meaningful material. Amer. J. Psycho1. 63, 176-185.

NEWMAN, E.B. (1951a). The pattern of vowels and consonants in various languages. Amer. J. Psychol. 64, 369-379.

NEWMAN, E.B. (1951b). Computation methods useful in analysing series of binary data. Amer. J. Psychol. 64, 369-379.

ROSENBLUETH, A., WIENER N. and BIGELOW, J. (1943). Behaviour, purpose and teleology. Phil. Sci. 10, 1, 18-24.

SHANNON, C.E. (1948). A mathematical theory of communication. Bell Syst. tech. J. 27, 379-424 and 623-657.

SHANNON, C.E. and WEAVER, W. (1949). The mathematical theory of communication. Urbana, Ill: Univer. of Illinois Press.

TURING, A.M. (1937). On computable numbers with reference to the Entscheidungsproblem. Proc. London math. Soc. 2-42, 230-265.

von FOERSTER, H (1950). Quantum theory of memory. In H. von Foerster (Ed.), Cybernetics - circular, causal and feedback mechanisms in biological and social systems. Transactions of the sixth conference. New York: Josiah Macy, Jr. Foundation.

von FOERSTER, H. (Ed.), Cybernetics - circular, causal and feedback mechanisms in biological and social systems. Transactions of':
(1950). Sixth conference, March 24-25, 1949.
(1951). Seventh conference, March 23-24,1950.
(1952). Eighth conference, March 15-16,1951.
(1953). Ninth conference, March 20-2l, 1952.
(1955). Tenth conference, April 22-24, 1953.
New York: Josiah J,Macy, Jr. Foundation.

WIENER, N. (1948b). Cybernetics; or control and communication in the animal and the machine. New York: Wiley.

WIENER, N., and ROSENBLUETH, A. (1946). Conduction of impulses in cardiac muscle. Arch. Inst. cardiol. Mex. 16, 205-265.

YOUNG, J.Z. (1953) .Discrimination and learning in octopus. In H. von Foerster (Ed.), Cybernetics - circular, causal and feedback mechanisms in biological. and social systems. Transactions of the ninth conference. New York: Josiah Macy, Jr. Foundation.