The Act of Creation

by Arthur Koestler

  NOTES

  To p. 491. The chaffinch song consists of three distinct and well-articulated phrases. Hand-reared chaffinches produce a much simplified, rudimentary variant of the song, where the first and second phrase are often inseparable, and the third partly or wholly missing. But -- and this is the elegant point of the experiment -- if the young bird is left to be reared by its parents, then taken away and isolated in September, that is to say, long before it starts singing, it will nevertheless burst into normal song next spring; Apparently 'these birds have by their first September learned that the song should be in three phrases and that the terminal phrase should contain a more or less elaborate flourish'. Thus 'on the perceptory side the process of recognizing and accepting the specific song as henceforth the normal for the individual (as distinct from the acquisition of the new motor habits involved in performing the song) seems to resemble the original examples of imprinting [the following response] sufficiently close to warrant considering them together' (Thorpe, 1956, p. 375).

  To p. 491. In the human child, processes analogous to Prägung may perhaps be responsible for producing infantile fixations and fetishistic rituals. The cases of boot-fetishism, for instance, frequently reported in works on sexual pathology remind one of imprinting by inanimate objects -- such as a gander's seven-year fixation on an oildrum, reported by Thorpe (p. 365). A further analogy may be found between Gestalt-sign releasers, based on the vital statistics of film-stars, and Tinbergen's simplified laboratory models of the pregnant stickleback.

  To p. 493. Thorpe continues: 'However, the answer may be that the nature of sensory nervous mechanisms is such that to achieve full efficiency at the normal level of stimulation, the threshold must be much lower.' (Thorpe, 1956, p. 283). But other passages which I have quoted show that he does not regard this explanation as satisfactory.

  VIII

  MOTIVATION

  Retrospect

  Future historians will probably regard it as significant that throughout the first half of the twentieth century the dominant schools of psychology -- even schools as far apart as behaviourism and psychoanalysis -- recognized only one basic type of motivation, and that a negative one: the reduction of biological needs and drives, the diminution of tension, escape from anxiety. 'At the level of ego-psychology', wrote Mowrer in his survey on 'Motivation' in the Annual Review for 1952, [1] 'there may be said to be only one master motive: anxiety.'

  The trend seems to have originated in the climate of the Darwinian revolution independently in Germany and America, with Fechner's (1873) 'Tendenz zur Stabilität' and Thorndike's (1898) 'Law of Effect'. Freud (1920), acknowledging his indebtedness to Fechner, postulated his own Principle of Parsimony, according to which 'the course of mental events is invariably set in motion by an unpleasurable tension, and it takes a direction such that its final outcome coincides with a lowering of that tension'. Thus pleasure is derived from 'the diminution, lowering, or extinction of psychic excitation' and 'un-pleasure [Unlust, dysphoria, as distinct from physical pain] from an increase of it'. The organism tends towards stability -- a kind of homeostasis, applied not only to autonomic regulations but also to voluntary behaviour; it is guided by 'the striving of the mental apparatus to keep the quantity of excitations present in it as low as possible or at least constant. Accordingly, everything that tends to increase the quantity of excitation, must be regarded as adverse to this tendency, that is to say, as unpleasurable'. [2]

  Now this is of course true, in a broad sense, in so far as the frustration or satisfaction of primary biological needs is concerned. But it passes in silence a whole class of experiences to which we commonly refer as 'pleasurable excitement'. The preliminaries of love-making cause an increase in sexual tension and should, according to the theory, be unpleasant -- which they are decidedly not. It is curious that in the works of Freud there is no answer to be found to this embarrassingly banal objection. The sex-drive in the Freudian system is essentially something to be disposed of -- through the proper channels or by sublimation; pleasure is derived not from its pursuit, but from getting rid of it. One might argue that in Freud's universe there is no place for amorous love-play because Freud, like D. H. Lawrence, was basically a puritan with a horror of frivolity, who treated sex mit tierischem Ernst.* But arguments ad hominem do not explain the general trend in the first half of the century to interpret motivation as something negative. As Hilgard ruefully remarked, [3] the 'Zeitgeist favoured our seeing incentives not as providing something sought after for what was inherent in the incentive, but something providing relief. The incentive was seen as an avenue of escape from pain, anxiety, tension.' Just as Freud's libido-theory had no room for dalliance, so learning-theory had no room for curiosity or learning-by-play.

  Thorndike's 'Law of Effect' was essentially a stick-and-carrot theory; the reward (and to a lesser degree, punishment) is the factor which stamps in the correct responses in learning, and stamps out the incorrect ones. In the extreme behaviourist systems of Watson and Guthrie, the mechanization of the living organism is complete: contiguity is the basic factor in producing associative S.-R. bonds, and motivation has virtually disappeared from the picture. Nor is any theory of motivation allowed to enter into Skinner's concept of 'operant behaviour'. His system is by programme confined to the description of experimental operations, preferably in quantitative terms. The effects of different rates and sequences of positive and negative reinforcers are counted and plotted; the entity on which they act is the 'operant strength', which in turn is measured by the rate and number of responses during extinction; but the motivation of the animal is represented by a single, crude variable: the number of hours in which the rat had been deprived of food. Optimum learning results from the combination of the appropriate number of hours of deprivation with the appropriate rate of applying positive reinforcers, i.e. stimuli of the type, one might say, which are apt to deprive the organism of its deprivation. Differences between the learning abilities of various species, or of age groups within a species, are not considered to be relevant to this type of 'functional analysis' of behaviour. By the same method of selective reinforcement, by 'baiting' each step in a series of steps, pigeons can be trained to describe a figure-eight with their heads held high, and students can be trained to select the right answer among several alternatives and to punch it into the tape of the learning machine -- the reinforcement in this case being that the tape moves on to the next question. Since each reinforcer is a drive-reducer, learning becomes a process of progressive de-motivation.

  Hull did not share Skinner's rigidly positivistic, hypotheses-non-fingo attitude. He kept elaborating and modifying his theory until his death in 1951; the system has been described as the last and most impressive attempt to build an edifice on S.-R. foundations. His emphasis gradually shifted from primary to secondary drives and secondary rewards; and from need-reduction to drive-stimulus reduction (eating eliminates the stimulus of the hunger drive but not the biological need -- which will be satisfied only later by digestion). This made the system more elastic, yet in spite of these refinements, the primitive drives of hunger, sex, avoidance of pain, were considered to be the only motivational factors in learning. To quote Hebb's (1949) summing up of Hull's theory: 'Its weakest point, and clearest departure from the facts, is in the treatment of motivation as biological need. According to the theory, the rat in the maze should learn nothing about it until one of his responses is accompanied by a decrease of hunger or thirst, or escape from electric shock, or some similar reward. In actual fact, when he is allowed to run in the maze without reward or punishment, the rat learns a good deal about it. It is clear of course that the primitive drives of pain, hunger, and sex are often of overwhelming importance. We need an approach to motivation that neither minimizes these things nor fails to provide for the unrewarded learning that also occurs when the animal's belly is full and his sex drive satisfied.' [4]

  If we turn to the opposite camp -- Tolman and the Gestalt psychologists -- th
e emphasis shifts from the need-reducing to the goal-seeking aspects of behaviour. In classical Gestalt theory, motivation by rewards, usually in the form of bananas, is taken for granted; it does not question the effect of reward on learning, the dispute is about whether this effect is achieved by stamping in or by insight. Similar considerations apply to Tolman's sign-learning theory, although he has progressed a considerable step further by his explicit rejection of reinforcement theories, by his emphasis on 'expectancy' and 'purposiveness', on latent learning and 'creative instability'. Lastly, Kurt Lewin's 'psychological field theory', with its complex and changing motivation, its concepts of 'ego-involvement' and 'levels of aspiration'; above all with its notion of striving after 'success' (which is subjective and relative in contrast to reinforcement by tangible rewards), played an important part in promoting that change of climate which has been increasingly noticeable since about 1950.

  Decline of the Reflex

  This new orientation seems to be the cumulative effect of independent developments in several fields, such as: (a) disillusionment regarding the utility of the reflex-formula both in neurology-and psychology; (b) rediscovery of the fact that organisms are not passive masses of software reacting to environment, but 'open systems', feeding on 'negative entropy', engaged in spontaneous activities on all levels, and (c) that animals are capable of 'latent' learning in the absence of tangible rewards, motivated solely by their exploratory drive.

  (a) The physiological concept of the reflex arc, which even Sherrington considered as no more than a 'useful fiction', has become an anachronism.* The Pavlovian conditioned reflex was another useful fiction which exercised at first a stimulating, then a paralysing effect -- a phenomenon frequently met in the history of science. In Hebb's words: 'Pavlov has deservedly had a great influence on psychology, and his theory has not been rejected because it is too physiological but because it does not agree with experiment.' There is no need to recapitulate the evidence which has led to this rejection. [5] 'Conditioning' is still a useful term when applied to induced changes in glandular and visceral reactions, but leads to confusion when used in a loose, analogical way for other types of learning.

  The last blow to the reflex-arc concept came with the discovery that it was impossible to make a precise distinction between 'stimuli' and 'responses'. As already mentioned (p. 435) not only motor units, but also sensory receptors display constant spontaneous activity in the absence of external stimulation. [6] External events alter the pattern of this spontaneous activity, but this in itself does not yet constitute a stimulus. The receptors are under efferent control from higher levels of the central nervous system; the acceptance, suppression, or modification of the input starts on the periphery, and the centre decides what shall constitute a stimulus and what shall not. Even the stretch-sensitive receptors in muscle spindles are controlled by efferent fibres from the centre. In other words, 'stimuli' and 'responses' are not one-way processes, and cannot be isolated: 'because stimulus and response are correlative and contemporaneous, the stimulus processes must be thought of not as preceding the response but rather as guiding it to a successful elimination of the incongruity. That is to say, stimulus and response must be considered as aspects of a feedback loop. . . . [7] These properties are a far cry from the ubiquitous S.-R. reflex arc diagrams that grace (more appropriately one wants to say "disgrace") today's texts' (Pribram). [8]

  It is historically interesting that an independent but parallel softening-up process of the hard and fast S.-R. concept took place at the same time in psychology, e.g. in Skinner's and Hull's systems. Skinner was careful to state that he used the word 'reflex' not in an anatomical or neurological sense but as a purely psychological, descriptive term for the 'unit of behaviour'. But his definition of the unit was constantly shifting and changing. 'A reflex is not, of course, a theory. It is a fact. It is an analytical unit, which makes the investigation of behaviour possible. [9] The appearance of smooth curves in dynamic processes marks a unique point in the progressive restriction of a preparation, and it is to this uniquely determined entity that the term "reflex" may be assigned.' [10]

  As Miller et al were to comment, to define the reflex in terms of the smoothness of curves is a 'somewhat odd approach'. [11] Yet even so it did not work: 'Skinner's "unit appropriate for experimental study" turns out, in fact, to have a measure of arbitrariness about it. . . . Sometimes the functional unit is a simple response, sometimes a complex act, sometimes a rate of responding. The unit no longer has the clean dimensions of a correlation between a class of stimuli and a class of responses as implied in the original concept of a reflex. The atom of behaviour proves to be evasive.' [12] In the later versions of Skinner's system the stimulus no longer even precedes the response: in operant behaviour the organism emits responses in search of a stimulus as it were. The reflex as a unit of behaviour has evaporated like the physicist's hard little lumps of matter.

  Skinner's experimental work had some lasting merits. He was among the first to demonstrate that 'intermittent reinforcement' -- where only some correct responses are rewarded -- can be as effective as the consistent rewarding of all correct responses. Humphreys [13] then showed that random rewards are actually a superior (more extinction-resistant) form of training -- the rat thus trained is less discouraged when the reward is withheld, than the rat trained by the consistent-reward method. From this there was only one logical step to Tolman's theory of motivation by expectancy -- a step which Skinner never took.

  In Hull's case the 'softening-up' process took a different course. In his later years, Hull's attention shifted more and more from primary biological drives to secondary drives (from the 'need' to the 'taste' or 'appetite'). These secondary drives he saw manifested in anticipatory events -- 'fractional antedating goal-responses Rg', and 'fractional antedating goal-stimuli Sg. 'The fractional antedating reaction (Rg) with its proprioceptive stimulus correlate (Sg), provides for the "automatic (stimulus) guidance of organismic behaviour to goals".' The great importance Hull attached to this postulate is illustrated by his comment: 'Further study of this major automatic device presumably will lead to the detailed behavioural understanding of thought and reasoning, which constitutes the highest attainment of organic evolution. Indeed the Rg -->Sg mechanism leads in a strictly logical manner into what was formerly regarded as the very heart of the psychic: interest, planning, foresight, foreknowledge, expectancy, purpose, and so on.' [14]

  From a strictly logical point of view, the postulate makes no sense -- as Hilgard has pointed out in a careful analysis -- because Sg acts at the same time as a producer of the secondary drive Sd and as a secondary reinforcer which reduces Sd. [15] Hilgard put down this confusion as a sign of logical weakness in Hull;* yet he did not seem to realize that Hull was intuitively on the right track, that the contradiction is merely an apparent one, and vanishes if one stops thinking in terms of need-reducing motivation. 'That stimulus associated with reinforcement could become at once both a drive and a reinforcing agent' sounds like a contradiction, but makes eminent sense if, getting rid of the dreadful terminology, we translate it as follows: 'A rewarding experience can at the same time be both an incentive and a reward'; or even simpler: 'Some pursuits are self-rewarding'. That is the implied conclusion of Hull's eighth and last postulate which he regarded as the crowning achievement of his system. That he himself did not realize this implication only shows that the once useful S.-R. formula had by that time become a strait-jacket to thought.

  Hunger, Fear, and Curiosity

  It took natural philosophy nearly a thousand years to rediscover that the earth is round; it took experimental psychology nearly fifty years to rediscover, after its Dark Ages of need-reducing S.-R. theories, that rats and men are pleasure-seeking creatures, that some activities are pleasurably self-rewarding, and that exploring the environment, solving a chess problem, or learning to play the guitar are among these activities.

  An interesting reflection on the spirit of the times was the lo
ng, impassioned controversy which followed the earth-shaking discovery that rats who were allowed to familiarize themselves with the maze by running around in it without reward, got quicker to the food-box when this was eventually put in than the control rats who ran the maze for the first time. How could the rat profit from its previous experience in the maze without being rewarded by food or punished by electric shock? As Berlyne put it: 'There are plenty of experiments to show that latent maze-learning can occur in the rat, which is embarrassing for those whose theories are not built to assintilate it. . . . Where does the reinforcement for these responses come from? Several writers have considered the possibility that it comes from the reduction of curiosity.' [16] Other writers suggested that it came from the drive-reducing diminution of boredom. One might as well say that composing a song is a silence-reducing activity.