The Act of Creation

by Arthur Koestler

  Thus the function of every component in the organism is determined by two types of 'input'. The first consists of specific trigger signals from its superior controls in the hierarchy; the second are inputs and feedbacks from more or less random events in its environment. But I must stress again that the meaning of the word 'environment' depends on the hierarchic level to which it is applied. The environment of John driving his car is the traffic stream around him. The environment of John's right foot is the brake-pedal on which it rests. Let us call the former an environment on the t-level, where t stands for the top of the hierarchy controlling the various sensory, motor, and cognitive processes which constitute the skill of driving; then the brake-pedal will be an environment on the, say, t-minus-4 level. Now John approaches a sign which reads 'Halt -- Road Works Ahead'. This input is analysed and relayed by various stations of the perceptual and cognitive hierarchy, and is eventually re-coded into a 'sign releaser' which triggers off the pre-set patterns of slow-down-to-a-halt behaviour on lower echelons of the motor hierarchy. Thus an environmental input on the t-level has been transformed into a specific trigger signal in a 'vertical' hierarchy; in other words a 'category c' input on a higher level has been translated into a 'category b' input on a lower level, activating the autonomous pattern of the slowing-down skill. This consists in several sub-skills: braking, steadying the wheel, going into neutral gear at the proper moment. The foot on the brake-pedal is not responding to the 'Halt' signal from the environment; it is responding to a specific 'excitation-clang' travelling down John's spinal cord. But the foot's pre-set response is modified by feedback from the environment on its own level: the 'feel' of the pedal's elastic resistance governs the 'strategy' of braking -- neither too abruptly nor too softly. Similar feedbacks influence the automatized motions of the hands on the wheel, etc.

  We may perhaps speculate that the digitally coded signals: 'brake', 'steady wheel' have been converted into analogue-computing servo-mechanisms. But speculations apart, we can confidently say that an action-pattern of a general nature has been initiated on the top level, and that the details were successively filled in by feedbacks from more and more restricted local environments on the lower echelons. Similarly, the future adult is 'roughed in' in a summary way in the morphogenetic gradients of the zygote, then 'sketched in' in the organ-Anlage of the uncouth embryo, and so on, until the last detail is elaborated by the joint action of a cell-group's self-differentiation potential and its local environment. The performance of a skill means executing a general order by a series of progressively differentiated action-patterns, each controlled from above, and adjusted by local feedbacks.*

  We thus arrive at a synthesis between the principles of hierarchic organization and feedback adjustment. To the hierarchy of autonomous sub-wholes must be added the complementary 'hierarchy of environments' and 'hierarchy of feedbacks' -- of loops-within-loops, from those embracing the personality as a whole in the 'total field', down to the molecular level. An elementary but engaging model for a hierarchy of servo-mechanisms is the 'TOTE unit' proposed by Pribram. [1]

  Let me dot my i's on this rather important point by a martial analogy. The commander of the N'th Army has decided, taking into account all available intelligence reports on the enemy's dispositions, to capture tomorrow at dawn the vital hill No. 607. He sends his orders to his six divisional comanders, broadly outlining their tasks. The commander of the 3rd division is assigned the task of occupying hamlet X. There are several approaches to the hamlet; to decide which is best, he sends out some reconnaissance aircraft which feed informtion to him. He then communicates his orders to his battalion commanders. Each of these will send out patrols to get the lie of the land allotted to them before giving orders to his company commanders; and in the end each individual soldier will have to make the best of his own small environment of protective hedges and ditches as he moves forward in obedience to the sergeant's orders.

  6. It ought to be evident by now that the terms 'matrix' and 'code' are not meant to refer to separate entities, but (like 'structure' and 'function') to complementary aspects of a unitary process. The code is the invariant pattern of the process; it is not affected by environmental input. The matrix is the ensemble of part-processes, or 'members' potentially capable of being activated by the code; it thus represents the total repertory of alternative (equipotential or equifinal) variations in carrying out the process, according to feedbacks from the environment. The code is the fixed, the matrix the adaptable side of the process; the former determines the rules of the game, the latter the actual course of the game. The matrix, therefore, represents the more 'alert' or 'articulate' or 'explicit' side of the unitary process -- the side turned towards the environment The twenty-letter alphabet of protein-synthesis in the cell-matrix is more explicit than the three-letter alphabet of the genetic code; it 'spells out' what the latter implied. The articulated motions of the limb spell out the compressed message of the excitation-clang, as the pianist's fingers spell out the tune. In the perceptual and cognitive hierarchies, the codes which govern performance (e.g. grammar and syntax) function on lower levels of awareness than the performance itself.

  7. The control of the whole over the parts is exercised, as in our military hierarchy, through 'regulation channels'. The genetic code does not interfere with the details of ATP synthesis: it activates the sub-code of the mitochondria. The centre coordinating the motion of the limbs -- in newt or man -- does not deal directly with individual muscles; it activates the proper sub-centres. The battalion commander does not issue orders to individual soldiers, or even squads; he signals to Company headquarters: 'D Company will advance at 1800 hours.' The Company, the limb, the mitochondria are complex sub-wholes; but they are activated from the next-higher level as units, through their codes; and they in turn activate their members as units through their sub-codes.

  To put it in a different way: each part-process is a pattern of relations; but it is manipulated from the next-higher level as a unit -- a relatum . We shall see that as a general rule, when we ascend in any hierarchy, relations turn into relata, which enter into new relations, and so on. The code can be said to represent the invariant pattern of a relation; the matrix the ensemble of the relata. But one step up, and the code itself becomes a relatum; one step down, and the members of the matrix are seen as complex relations. We may thus add one more pair of complementary terms to characterize the Janus-faced entities in the developmental hierarchy: part whole; structure function, regulative mosaic, autonomous dependent, relation relatum, matrix code.

  8. The stresses set up between the organism's inner and outer environment are matched by active adaptations on various scales. The term 'dynamic equilibrium' indicates adaptative processes which do not entail major changes in the pattern of the whole; 'regenerative span' refers to the organism's capacity for 'adaptations of the second order' to challenges which can be met only by a reshaping of structures or a reorganization of functions; while 'routine regenerations' occupy an intermediary position, and overlap with both.

  'Equilibrium' in this context refers not to relations between parts, but between the excited part and the controls which represent the whole. Under conditions of dynamic equilibrium, the stresses between the self-assertive tendencies of the excited part and its integrative controls are of a transitory character. Paranormal challenges may lead to the phenomenon of'physiological isolation', owing to over-stimulation of the part or blockage of communication with its normal controls. In lower organisms, the isolated part tends to develop into a new whole. If it was segregated ab ovo, as sex cells and regeneration cells are, this development follows a straight course; if isolation occurs at later stages, as in fissure, budding, and organ-regeneration, it involves a temporary regression of the part to an embryonic or more juvenile phase of development, and the liberation of genetic potentials which are normally under restraint. It is a safety device which enables the organism to cope with traumatic challenges, and correct faulty integrations; it fur
thermore confers on it a super-flexibility which plays an important part in biological and mental evolution.

  On higher levels of the evolutionary scale, regenerative processes are predominantly reorganizations of functions. These range from the repair of neuro-muscular co-ordination to the compensation of cortical damages, and to the re-structuring of perceptual and conceptual patterns in the reculer pour mieux sauter of the creative process.

  During the regressive, catabolic phase, the part tends to dominate the whole through the reversal of axial gradients and hierarchic controls. This may lead to irreversible changes of a pathological nature (malignant growths, idée fixe). To avoid snapping of the loosened ties, the isolation of the part must be temporary and not complete: after the routine-controls have gone out of action, the organism as a whole must assist the regenerative process.

  'Routine repairs' were seen to range from the regeneration of tissues lost through wear and tear, to the restorative effects of sleep. Dreaming could be described as a de-differentiation of reasoning-matrices and even, up to a point, of personal identity.

  9. These periodic fluctuations from the highest level of integration down to earlier or more primitive levels and up again to a new, modified pattern, seem to play a major part in biological and mental evolution. Their universality is reflected in the myths of death and rebirth, the 'dark night of the soul', etc. The 'magic' of organ-regenerations, and of unconscious guidance in creativity, both owe their striking character to the sudden re-activation of (morphogenetic or psychogenetic) potentials which are normally under restraint in the adult individual. The period of incubation may be compared to the catabolic phase in organ-regeneration: the former releases pre-conceptual, intuitive modes of ideation from the censorship imposed by the conscious mind; the latter triggers off embryonic growth-processes equally inhibited by the mature organism. The contact-guidance of nerves towards their end-organs and the revival of other pre-natal skills, provide enticing parallels to the unconscious gradients and ancient 'waterways' which mediate the underground rendezvous of ideas.

  Summary

  To the hierarchy of sub-wholes in the development and behaviour of organisms, we have now added a complementary 'hierarchy of environments', and a third hierarchy of (exteroceptive and proprioceptive) feedbacks -- of loops-within-loops which connect the first and the second on every level. Certain homologue principles of organization. were seen to operate on all levels, such as: (a) the dichotomy of self-assertive and participatory tendencies derived from the dual character of each part as a 'sub' and a 'whole'; and the related complementarity of regulative and mosaic development, of equipotentiality and fixed pathway, of relations and relata. (b) Control within the organic hierarchy is exercised by 'regulation channels', i.e. high centres do not normally have direct dealings with lowly ones, and vice versa. (c) Trigger-releaser devices seem to be the general rule in the activation of pre-set, autonomous patterns. (d) The releaser signals (excitation-clangs, frequency-modulation sequences?) from higher echelons were found to be of a more implicit, generalized order than the actual performance 'spelled out' by the addressee. (e) The pattern of the performance is determined by its invariant code, but sub-wholes have varying degrees of freedom for adaptable strategies (equipotential variations) dependent on feedback from their local environment. (f) Under normal conditions these flexible strategies are sufficient to restore dynamic equilibrium between the whole and its excited parts. (g) Traumatic experiences may cause irreversible, degenerative changes in the exposed part, but under favourable conditions may initiate superflexible adaptations of a second order -- regenerations of structure or reorganizations of function, which are capable of redressing faulty integration, and also play an important part in biological and mental development.

  The reader may consider some of these conclusions trivial, others perhaps as rash generalizations. In the following chapters their validity will be tested in the light of instinct-behaviour, learning, and problem-solving.

  NOTE

  To p. 470. 'Feedback' is used here in a broad sense, to include all exteroceptive and proprioceptive inputs relevant to the ongoing activity.

  VI

  CODES OF INSTINCT BEHAVIOUR

  The Genetics of Behaviour

  The phylogenetic origins of instinct-behaviour are among the blackest black boxes found in the sciences of life. The causative mechanism responsible for the evolution of species in their morphological aspect is perplexing enough; regarding the origin of specific behaviour-patterns, the darkness is almost complete. As one eminent ethologist laments: 'The backward position of ethology is striking. Owing to the difficulty of tracing genetically determined behaviour components, geneticists have nearly always used morphological characters as indicators of gene-function. . . . A genetics of behaviour still has to be developed.' [1]

  Evolutionary genetics lies outside the scope of this book, but a brief remark in passing may be excused. If, apart from a few tentative studies, [2] the genetics of behaviour is still an uncharted territory, the reason may perhaps be an unconscious reluctance to put the already strained theoretical framework of neo-Darwinian genetics to an additional test. To quote a very trivial example: an individual songbird or jackdaw or sparrow, on spotting a predator, will give an alarm call, warning the whole flock. 'These alarm calls', Tinbergen points out, 'are a clear example of an activity which serves the group but endangers the individual.' [3] Are we really to assume that the occulo-vocal 'wiring diagram' in the sparrow's nervous system which releases the alarm call in response to a sign-Gestalt stimulus of predatory shape, arose by random mutations and was perpetuated by natural selection in spite of its negative survival value for the mutant? The same question could be asked concerning the phylogenetic origin of the ritualized tournament fights in such various animals as antlers, iguana, wolves, and fish. Wolves sprawl on their backs as a token of defeat and surrender, exposing their vulnerable bellies to the victor's fangs. One is inclined to call this a rather risky attitude; and what is the individual survival value of not hitting (or biting, goring) below the belt? Or if it comes to that, of the digger-wasps' nerve-racking maternal activities?

  A female of this species, when about to lay an egg, digs a hole, kills or paralyses a caterpillar, and carries it to the hole, where she stows it away after having deposited an egg on it (phase a). This done, she digs another hole, in which an egg is laid on a new caterpillar. In the meantime, the first egg has hatched and the larva has begun to consume its store of food. The mother wasp now turns her attention again to the first hole (phase b), to which she brings some more moth larvae; then she does the same in the second hole. She returns to the first hole for the third time to bring a final batch of six or seven caterpillars (phase c), after which she closes the hole and leaves it for ever. In this way she works in turn at two or even three holes, each in a different phase of development. Baerends investigated the means by which the wasp brought the right amount of food to each hole. He found that the wasp visited all the holes each morning before leaving for the hunting grounds. By changing the contents of the hole and watching the subsequent behaviour of the wasp, he found that (1) by robbing a hole he could force the wasp to bring far more food than usual; and (2) by adding larvae to the hole's contents he could force her to bring less food than usual. [4]

  Let me repeat: the reason why 'a genetics of behaviour still has to be developed' seems to be that it cannot be developed with the existing theoretical tools without reducing the whole attempt to absurdity. It may still be possible, and even respectable today for a geneticist to state that: 'The hoary objection of the improbability of an eye or a hand or a brain being evolved "by blind chance" has lost its force.' [5] But are we also to assume that the behaviour-patterns of the digger-wasp, or of the courtship and fighting rituals of various species have all evolved 'by pure chance'? This assumption is implied in the doctrine of contemporary genetics -- though rarely stated in explicit form. Similar assumptions have been made by extreme behaviouri
sts in the field of learning theory; there is, in fact, a direct continuity between the doctrine of natural selection operating on random mutations, and reinforcement operating on random trials. Both grew out of the same philosophical climate. But while learning theory is in full retreat from that extreme position, and has a variety of alternative suggestions to offer, nothing the like is in sight in the genetics of instinct-behaviour.

  Instinct and Learning

  Learnt behaviour is built on the foundations of innate behaviour, though it is often difficult, if not impossible, to tell where the 'foundation' ends and the 'building' starts. But the absence of fool-proof delineations between 'inheritance', 'maturation', and 'learning' need not prevent us from recognizing the existence of distinct patterns of animal behaviour which are (a) stereotyped, (b) species-specific, (c) unlearnt in the sense that they can be shown to appear, more or less completely, in animals raised in isolation. It has been objected against this view that 'innate' behaviour, e.g. the pecking of chicks, may partly be due to prenatal influences, [6] that 'isolation' is never absolute, [7] and that learning may be practically instantaneous (as in imprinting). Such arguments are valuable in showing that pure heredity sans environment is an abstraction; but they do not alter the fact that each animal is born with a hereditary potential to feed, hoard, court, nest, fight, and care for its young in certain specific and highly characteristic ways which are as much part of its native equipment as its morphological features, and which can be modified by, but are not derived from, imitation and learning. Only the unbalanced claims of some extreme behaviourists could temporarily obscure the obvious fact that 'if the physical machinery for behaviour develops under genetic control, then the behaviour it mediates can scarcely be regarded as independent of inheritance'. [8]