The Act of Creation

by Arthur Koestler

  It seems to be wrong, however, to go to the opposite extreme and stretch the meaning of the word 'play' so as to cover all manifestations of the exploratory drive -- as Berlyne seems to do when he says that 'in human beings, ludic behaviour includes everything that is classified as recreation, entertainment, or "idle curiosity", as well as art, philosophy, and pure (as distinct from applied) science. . . . ' [2] This, of course, is a matter of definitions, but I think it more expedient to use the word play in a more precise and restricted sense, which is closer to its colloquial usage. A small child, kicking a ball about, plays; a professional football 'player' works hard for a living. When the monkey takes the puzzle apart and puts it together again, he 'plays'; when there is food inside the puzzle he 'strives'. Two chess masters may play a friendly café game; in a tournament they compete. The examples show how fluid the borderlines are, yet the principle is clear: the degree of 'playfulness' in an action decreases in proportion as the exploratory drive is adulterated by other drives; or, to put it differently: as the self-arousing and self-rewarding nature of the activity, characteristic of the exploratory drive, yields to striving for specific rewards. This foreshadows a similarly continuous, graded relationship between the dynamics of latent and reinforced learning, to be discussed later.

  The Ludic and the Ludicrous

  It follows from the above that play can only arise at an evolutionary level or in such special situations, where the organism has been partially liberated from the tyranny of 'primary needs' in the traditional sense, and can afford to 'take time off' to play. This happens among animals where the young mature slowly and enjoy prolonged parental protection and care; under the sheltered conditions of domestication and captivity; and in human history, of course, with the increase of security against the hazards of the natural and social environment. To quote Thorpe again: 'The prolonged childhood of the human species [has] been of prime importance in the process of freeing appetitive behaviour from the primary needs. This and man's growing mastery of his environment have been the essential first steps not only for play but for all those activities which transcend mere maintenance and which underlie the mental and spiritual development of man; activities which, though originating in "play", have produced real advantages in knowledge and comprehension, of the scheme of things. . . .' [3]

  A related process of emancipation, namely the detachment of reasoning from emotion, gave rise, as we saw, to humour. Man's emergent ability to perceive a thing or event simultaneously in two incompatible mental contexts enabled him to take the step from the 'ludic' to the 'ludicrous'. The historic link between the two is probably reflected in the word 'ludicrum' -- stage play. The actor's or bard's pretence of being himself and somebody else at the same time was at the origin of tragedy and epic; a similar act of magic -- carving or painting a thing which is meant to be something else -- was the origin ofrepresentational art. These, of course, are activities on an incomparably higher level than the play of kittens and birds; yet as Lorenz has pointed out, both imitation and pretence occur already on the animal level. When puppies fight in play, they do not hurt each other or their masters; they conform to certain 'rules of the game'. Whether these have their phylogenetical origin in the ritualized fights of their wild ancestors or whether they are acquired by social learning, the fact remains that such fights are 'not in earnest' and necessitate 'bringing in the higher or more psychological concept of pretence.' [4]

  Equally suggestive is the so-called 'sub-song' of birds. As distinct from the true or full song which is fixed and species-specific, the sub-song is 'a somewhat amorphous, rambling utterance'. [5] Birds indulge in it when their 'primary needs' are not pressing -- before the mating season or in captivity -- as a kind of vocal play which might either represent practice for the true song, or else a form of 'pretending': some birds which never imitate alien species in their true song do so in their sub-song. Thorpe compares this vocal imitation to the process by which human infants learn to speak -- from 'amorphous' babblings to the imitation of sounds produced by their elders (echolalia). Thus imitation, pretence, as well as art, seem to have their precursors in the playful activities of the higher mammalia and birds.

  To sum up: 'exploratory behaviour is motivated by the exploratory drive'. In play, its purest form, it is generalized, non-specific, and indiscriminate -- a puppy let into an unfamiliar room rushes to and fro, sniffs at every corner, picks up any object, beside itself with excitement under the incentive of novelty. On the other hand, when the exploratory drive is canalized towards more specific targets, it results in latent learning and, still higher up, in problem-solving. While play is self-rewarding, in problem-solving the search itself may also be self-rewarding to varying degrees, but the principal reward is finding the solution. In this broader sense, of course, the law of effect remains valid; but the reward, the pleasure derived from success, is specific to the exploratory drive -- its 'consummatory act' as it were -- and not a premium extraneous to its nature. If the problem was an easy one, the solution may be both a reward and an incentive to have a go at another problem at once; if the drive was obstructed, involving stress, the solution is tension-reducing. And, of course, the solution may carry supplementary rewards -- the carrot of satisfied ambition, for instance.

  In laboratory experiments the animal's exploratory possibilities are restricted, and artificial motivations replace the drive as it operates in freedom. At the same time, animals in the laboratory are induced to pay attention to, and discriminate between, stimuli which under normal circumstances would be biologically irrelevant to them; or else to perform motor actions (e.g. Skinner's ping-pong playing pigeons) which are outside their natural repertory. Tricks of this kind can be taught only by stamping in; and attempts to build a universal theory of learning on such methods carry the danger of confusing a travelling circus with Plato's Academy.

  NOTE

  To p. 509. The Concise Oxford Dictionary gives no less than thirty-four meanings of the word.

  X

  PERCEPTION AND MEMORY

  I must now switch from animal to man, and later back again. The manner in which animals learn holds important lessons for man; but in order to interpret the data the experimenter must make certain minimum assumptions regarding the animals' experiences; and, whether he is aware of it or not, these assumptions are based on his own human experience. We talk about the animal's pain-reaction or fear-reaction because we have experienced pain and fear; we interpret certain signs as meaning that the animal is alert or apprehensive by inferences which are often unconscious and contain an unavoidably anthropomorphic element. Even Lloyd Morgan's canon acknowledged this; it merely said that one should not be more anthropomorphic than one could help.*

  Now, learning involves perception and memory; and since we know incomparably more about both in man than in cats or rats, we must discuss some aspects of man's perceptual and sensory-motor skills before we turn to learning in animals. Instead of the over-worked province of visual perception, I shall start, for a change, on hearing.

  Screening the Input

  It has been said that visitors to Stalin had to go through seventeen successive screenings: at the outer gate of the Kremlin compound, at several inner gates, and so forth, until the last corridor and the last door leading to the inner sanctum. 'Inputs' which aspire to become 'stimuli' apparently suffer a similar fate. Where hearing is concerned, the brain's stimulus-screening activity starts in the ear. Efferent, inhibitory fibres from higher centres to the cochlea of the cat were discovered by Galambos in 1956. In a series of remarkable experiments [1] the cat's auditory nerve was tapped and wired to an amplifier, so that impulses passing from ear to brain were directly recorded. The impulses were caused by the clicking of a metronome. But the moment a mouse in a glass jar was shown to the cat the firings in the auditory nerve were diminished or ceased altogether: the cat was turning a 'deaf ear' on the metronome. The point of the experiment was to show that the process of stimulus-selection is centrally contro
lled, but sets in at the periphery -- the outer gate of the Kremlin compound.

  Attitude and expectation -- the pattern of the behavioural matrix to which the organism is attuned at the time -- determine what shall constitute a stimulus and what shall not. On a happy family evening, when people are talking while the radio is playing, junior is crying, and the dog is begging to be let out, each of these simultaneous inputs may be perceived as 'signal' and the rest as 'noise'. In audition, at least, the 'figure-background relation' seems to be more complicated than the Gestalt school suggests; it is not something innate in perceptual organization, but dependent on past experience and present state of mind. Women were known to sleep soundly through an air-raid but to awake at the slightest cry of their babies; people deeply asleep show sharp EEG reactions when their own name, or the name of a girlfriend, is read out in a list of other 'background' names. [2]

  The point has also an indirect bearing on the controversy whether discrimination is based on the 'absolute' or relational properties of stimuli. [3] The answer seems to be, briefly, that absolute stimuli do not exist -- short of sticking a knife into somebody. Yet even on the primitive level of pain, the matrix influences perception -- as witnessed by the General in the American Civil War who, in the heat of battle, did not notice that his middle finger was shot away; not to mention anaesthesia by hypnosis in dentistry and child-birth -- or the even more remarkable phenomena of hysterical conversion blindness.

  Thus the higher centres exercise a selective influence on sensation and perception; those aspects of the input which are irrelevant will be treated as noise, and forgotten 'without leaving a trace'. But the criteria of relevance depend on the 'rules of the game' which the organism is playing at the time.

  Stripping the Input

  Selective control of the input is the first stage in the process of extracting information from the chaotic noises and other sensations which bombard the organism's receptors; without it, the mind would be in a kind of Brownian motion. This first stage is followed by the processing of the input in a series of relaying operations, each of them designed to strip the input of what appears to be irrelevant -- according to the criteria of relevance which operate along that input-channel. One might call this a process of "de-particularization". It is a clumsy word, but it conveys what is really implied in the terms 'generalization' or 'abstraction', with their multiple connotations.

  The most familiar examples of 'de-particularization' are, of course, the visual constancies. The triangle, or the letter 'W', is stripped of the irrelevancies of retinal position, size, etc. Thanks to colour constancy the accidents of light and shadow are discarded; thanks to size constancy, my moving hand does not seem to shrink or grow -- changes in perspective size are 'dis-regarded' by the regard. Yet the criteria of relevance and irrelevance depend, even in these cases of apparently spontaneous perception, to a considerable degree on interpretative frames -- on perceptual matrices acquired by past experience. When an object of the appearance of a tennis ball is inflated against a homogeneous background, it will be seen as if it were retaining its size and approaching the observer. [4] This is different from size constancy because in this case the observer has to accommodate his eyes and make them converge at a closer range so that the ball gets out of focus and should be seen as a blurred double image. Yet the knowledge that tennis balls behave reasonably and do not grow into footballs somehow manages to compensate for this, and to discard the anomalies in the situation as irrelevant noise. To quote Bartlett once more: 'Even the most elementary perceptions have the character of inferential constructions.' The Baconian ideal of observation without theorizing is undermined by the mechanism of observation itself. Perception is polluted by implied hypotheses. To look, to listen, to taste, means to ask questions; and mostly they are leading questions.

  To obtain a more detached view of the living organism's methods of coding and storing its experiences, let me make a naïve comparison with a typical' engineering procedure. Examining a modern gramophone record with a magnifying glass, you see a spiral curve with lateral oscillations of varying amplitude and spacing -- a curve where the abscissa represents time, and the ordinate the amplitude of the needle's oscillations. And yet this two-dimensional curve, with a single independent variable, can reproduce any sequence of sounds, from the Sermon on the Mount to the Ninth Symphony performed by orchestra and choir, including the buzzing of a fly and a cough in the audience. In fact the entire range of human knowledge and experience could be expressed by the function of this one independent variable, so that one is tempted to ask why the nervous system does not produce engrams in this simple type of code, instead of the incomparably more complicated methods it uses. The answer is, that a 'linear' memory trace of this type would be completely useless for the purposes of analysing, recognizing, and matching new inputs, and for working out the appropriate responses. It would merely represent the 'blooming, buzzing confusion of pure sensation sans organization' which, in the words of William James, is the new-born infant's world. Before it can be more or less permanently stored, the input must be processed, dismantled, and reassembled in various ways, which the following examples may serve to illustrate.

  Dismantling and Reassembling

  Let the input be fifty instruments and fifty voices performing a choral part of Beethoven's Ninth. On the gramophone record, and in the air-waves which make the ear-drum vibrate, the pitch, timbre, and loudness of the individual voices and instruments have all been superimposed on each other -- scrambled together into a single variable pulse. The individuality of soprano, flute, viola, is lost in the process; it requires a human nervous system to reconstitute it.

  The pulse is transmitted and amplified by the bones of the middle ear and enters through the oval window into the cochlea. Here the basilar membrane, based in viscous fluid, starts the process of unscrambling the acoustic omelette. This is done, partly at least, by a kind of Fourier-analysis of the oscillatory curve, which breaks it down into its spectrum of basic frequencies.* The parallel fibres of the basilar membrane form a kind of spiral harp; each fibre responds to a specific frequency. This analysing mechanism operates over a range of twenty to twenty thousand cycles per second, and auditory discrimination varies from about 0.05 at low frequencies to 0.025 at 2,000 c/s (Piccolo flute). Each frequency has its separate 'place' on the spiral membrane. Each 'place' is presumed to be connected by a separate group of fibres, running through several relay stations, to a presumably fixed location in the primary reception area in the auditory cortex -- area 22.

  But this mechanism of transmission by fixed pathways and non-specific impulses is only half of the story; the other half is transmission of the lower frequencies by 'volleys' in a bundle of fibres firing in turn at the specific frequency of the input. [5] The details of both theories are still controversial, but the available evidence indicates that they complement each other. We have, then, here one more instance of the complementary character of two types of nervous function: conduction by specific pathways, and conduction by specific signals over equipotential pathways.

  We now have our fifty singers and fifty instruments decomposed into a constantly changing mosaic pattern of excitations in area 22 where each point (or region or circuit) [6] represents the frequency of one pure tone, and in some form also its intensity -- regardless of the instrument or voice in which it originated. This state of affairs bears no resemblance to any conceivable neural model based on S.-R. theory -- or on the Gestalt physiology of Köhler and Koffka.* In fact the whole physiological theory of Gestalt, and many of its psychological postulates, break down when we come to audition. This is not surprising since Köhler concentrated entirely on visual perception; and in the seven hundred-odd pages of Koffka's "Principles of Gestalt Psychology" exactly one page (p. 200) is devoted to 'other (than visual) senses'. Even on this one page, the only reference to audition is the statement that 'sound' and 'stillness' have a reversible figure-background relation.