The Act of Creation

by Arthur Koestler

  Learning appears then as the adaptation of the innate potential to lived experience. To quote the convergent definitions of one ethologist and one psychologist: 'Learning is a central nervous process causing more or less lasting changes in the innate behavioural mechanisms under the influence of the outer world. . . .' [9] 'Learning is a process by which an activity originates or is changed through reacting to an encountered situation, provided that the characteristics of the change in activity cannot be explained on the basis of native response tendencies, maturation, or temporary states of the organism (fatigue, drugs, etc.).' [10]

  It must be repeated, however, that outside the experimental laboratory it is virtually impossible to draw a precise distinction between the 'innate' and 'acquired' aspects or components in the adult animal's behaviour. Even the discrimination of biologically relevant sign-Gestalten in the environment seems to require a minimum of experience; and one must conclude, with Thorpe, that 'since comparison involves learning, an element of learning enters into all orientation and all perception. Accordingly it is suggested that the difference between inborn and acquired behaviour is of degree rather than kind; it becomes, in fact, a difference chiefly of degree of rigidity and plasticity.' [11] In the terms of our schema, what is inherited is the specific and invariant factor in the native skill -- its code. Its more or less flexible matrix develops through learning from experience. To quote Thorpe again:

  In each example of true instinctive behaviour there is a hard core of absolutely fixed and relatively complex automatism -- an inborn movement form. This restricted concept is the essence of the instinct itself. Lorenz originally called it Erbkoordination or "fixed action pattern". Such action patterns are items of behaviour in every way as constant as anatomical structures, and are potentially just as valuable for systematic, philogenetic studies. Every systematist working with such groups as birds or higher insects will be able to recall examples of the value of such fixed behaviour patterns in classification. [12]

  Tinbergen's Hierarchy

  Thus on the level of instinct-behaviour, 'codes' appear in the guise of fixed action-patterns, which incorporate the rules of the game of courting, nest-building, duelling, etc. Each of these activities is again a hierarchy of autonomous sub-skills. These tend to be more flexible on the higher levels which co-ordinate the drive, and more rigid on the lower levels. The autonomous sub-codes are restrained from spontaneous activity by 'inhibitory blocks', and triggered into activity by patterned impulses from higher echelons. This trigger-sensitive apparatus is called, after Lorenz, the "innate releasing mechanism" -- or I.R.M. for short:

  In all the channels which flow downward from the centre [of a given drive], there is supposed to exist a physiological mechanism which effectively prevents all discharge of activity unless the animal encounters the right environmental situation and stimuli to remove or release this block. Thus there is an innate releasing mechanism (I.R.M.) . . . which is in some way attuned to the biologically right stimulus in the environment . . . and which is, as it were, unlocked by the appropriate releaser, thus allowing behaviour to proceed to the next lower level. These in their turn incorporate blocks and, so long as these remain, action of these lower centres cannot proceed.' [13]

  Tinbergen's famous example of the hierarchic control of instinct-activity is the reproductive behaviour of the male stickleback, which I shall describe in the terminology of the present theory.

  In spring the lengthening of days triggers off the small fish's 'migrating code', while hormonal activities provide the drive or motivation. The fish then migrates into shallow water and swims around until a certain environmental configuration (rise of temperature, combined with green vegetation, etc.) strikes the 'right note', i.e. releases an efferent impulse, which in turn triggers off the sub-code of the nest-building activity. This activity is again subdivided into digging, glueing, etc., each of these skills governed by its autonomous sub-code. The latter are activated by trigger releasers; the order of operations is determined by inputs from the environment and proprioceptive feedbacks. The hierarchy of mating behaviour remains blocked until nest-building is complete; but the 'fighting' hierarchy (with its five different sub-codes) may be called at any moment into action by a trigger mechanism sensitive to a specific sign-Gestalt input: 'red male entering territory'. In this case the fighting code dominates the animal's entire behaviour, and nest-building as well as other activities are inhibited while the emergency lasts: the (functional) part monopolizes the attention of the whole.

  The sub-units of the behaviour-pattern tend to become more specific regarding input and more stereotyped in output on the lower levels of the hierarchy. 'Which one of the five motor responses belonging to the fighting pattern will be shown depends on sign stimuli that are still more restricted in effect.' [14] The nuptial colours of the fish are shining blue eyes and a red underbelly. Accordingly, any crude model which is red underneath will release an attack, regardless of shape and size -- whereas a perfectly shaped model without nuptial colouring will not do so. Apart from colour, behaviour also acts as a releaser. 'When the stranger bites, the owner of the territory will bite in return. When the stranger threatens, the owner will threaten back; when the stranger flees, the owner will chase it; and so on. . . . [15] But fighting is rarer than threat. The threat-behaviour of male sticklebacks is peculiar. Not only do they dart towards the opponent with raised dorsal spines and open mouth, ready to bite, but, when the opponent does not flee at once but resists, the owner of the territory does not actually bite but points its head down and, standing vertically in the water, makes some jerky movements as if it were going to bore its snout into the sand. [16]

  This of course is an exceptional example -- nest-building is a rarity among fish. But the rigidity of fixed action patterns in certain classes -- such as birds and insects -- remains nevertheless a striking phenomenon. The ritualized rules of the game of courtship and display, of threat and danger signals, of tournament fighting and social behaviour, sometimes reminds one of the ceremonious observances at Byzantine courts, at other times of the obsessive rituals of compulsion-neurotics. And the process of 'ritualization' does indeed suggest the 'emancipation' -- or isolation -- of a behaviour-pattern from its original context, accompanied by intensification, stabilization, and rhythmic repetitiveness of the pattern; the reasons are as yet hardly understood. [17]

  Appetitive Behaviour and Consummatory Act

  In spite of the relatively stereotyped nature of fixed action-patterns -- of which ritualization is an extreme example -- it would be entirely wrong to regard the hierarchy of instinct behaviour as a one-way affair, in which a plastic, general drive (the 'appetitive behaviour') discharges downward along pre-formed and discrete alternative channels into the completely rigid and mechanical, fixed-action-patterns of the 'consummatory act'. This conception of the organism as an automaton whose 'adaptability' is reduced to that of a kind of automatic record-changer or jukebox, with a choice between a few dozen fixed records appropriate to the occasion, seems to have originated in a misunderstanding of the distinction made by Wallace Craig between 'appetitive behaviour' and 'consummatory act'. This point must be briefly discussed as it is of some importance for the sections which follow.

  Appetite (or 'appetance') was defined by Craig (1918) as a 'state of agitation', a striving for an absent 'appeted' stimulus (conversely, a striving to escape from a noxious or disturbing stimulus); whereas the 'cousummatory act' was meant to bring the activity to a close by attaining (or escaping from) the appetitive stimulus, 'after which the appetitive behaviour ceases and is succeeded by a state of relative rest'. [18] More generally, 'the term appetitive behaviour is used by present-day writers on ethology to mean the flexible or variable introductory phase of an instinctive behaviour pattern or sequence'. [19]

  Thus 'appetitive behaviour' became a more refined and noncommittal name for the old, shop-soiled concepts of 'need', 'drive', 'instinct', and 'purpose'.* So far all was well; it was the 'con
summatory act', which led instinct-theory into a cul-de-sac. The trouble started, rather inconspicuously, when first Woodworth [20] then, independently from each other, K. S. Lashley and Konrad Lorenz became impressed with the stereotyped character of certain 'consummatory acts' (animal rituals and automatized habits in humans), as compared with the more general 'appetitive behaviour' or drive.** Eventually the focussing of attention on such fixed patterns of behaviour led to a distortion of the whole picture: Lorenz and Tinbergen made a rigid distinction between appetitive behaviour which was supposed to be flexible, and consummatory acts which were supposed to be completely fixed and automatic. Thus Tinbergen:

  It will be clear, therefore, that this distinction between appetitive behaviour and consummatory act separates the behaviour as a whole into two components of entirely different character. The consummatory act is relatively simple; at its most complex it is a chain of reactions. . . . But appetitive behaviour is a true purposive activity, offering all the problems of plasticity, adaptiveness, and of complex integration that baffle the scientist in his study of behaviour as a whole. . . . Lorenz has pointed out . . . that purposiveness, the striving towards an end, is typical only of appetitive behaviour and not of consummatory actions. . . . Whereas the consummatory act seems to be dependent on the centres of the lowest level of instinctive behaviour, appetitive behaviour may be activated by centres of all the levels above that of the consummatory act. . . . [21] The centres of the higher levels do control purposive behaviour which is adaptive with regard to the mechanisms it employs to attain the end. The lower levels, however, give rise to increasingly simple and more stereotyped movements, until at the level of the consummatory act we have to do with an entirely rigid component, the taxis, the variability of which, however, is entirely dependent on changes in the outer world. This seems to settle the controversy; the consummatory act is rigid, the higher patterns are purposive and adaptive. [22]

  But what exactly, one might ask, constitutes a 'consummatory act'? A glance at Tinbergen's diagram on page 479, for instance, will show that all actual manifestations of the reproductive instinct are classified as 'consummatory acts', whereas 'building', 'fighting', etc., are merely abstract, classificatory terms in which longer sequences of consummatory acts are bracketed together. Where, then, is the 'appetitive behaviour'? In the stickleback's spring migration in search of a nesting site? But that action-pattern was, judged by its dependence on specific releasers (temperature, verdant vegetation, etc.) on the same 'consummatory' level as, for instance, 'testing of materials'. The nearest Tinbergen gets to a definition of the consummatory act is in the following passage:

  The activation of a centre of the lowest level usually, perhaps always, results in a relatively simple motor response: biting, chasing, threatening, etc., in the case of fighting. . . . actual eating, actual escape, actual coition, etc., in other instincts. . . . These relatively simple responses are, usually, the end of a bout of prolonged activity, and their performance seems to "satisfy" the animal, that is to say, to bring about a sudden drop of motivation. This means that such an end-response consumes the specific impulses responsible for its activation. Fighting, eating, mating, "playing the broken wing" etc., are, as a rule, "self-exhausting". [23]

  However, neither 'digging' nor 'leading female to nest' is an end-response, or self-exhausting, or leads to a 'drop in motivation'. 'Testing of materials' is not a consummation, but a part-activity in the flexible, i.e. 'appetitive', pattern of building. And the building activity is not a one-way affair in the sense of higher levels in the hierarchy discharging along fixed conduits into the lower level of consummatory acts, along irreversible gradients. On the contrary, the control of operations oscillates all the time between different levels; the operational units responsible for one kind of 'end-response' carry on until a centre on a higher level, informed by feedback, switches to some other 'consummatory act'. It is at this point that the concepts of 'hierarchies of environment' and 'hierarchies of feedback' become important. In a complex activity like nest-building, even the relatively stereotyped operations on subordinate levels are under the dual control of their fixed codes and variable environment; and furthermore, information about their activities is constantly fed back (by proprioceptive and exteroceptive channels) to higher centres, so that the whole always remains in hierarchic control 'through regulation channels' of all of its parts. Tinbergen's schema does not really represent an organismic hierarchy, but a mechanical one -- rather like an automatized telephone net-work where the subscriber making a trunk call first dials the code-number of the whole town ('appetitive behaviour'), then the code of his fiancée's local exchange (semi-appetitive, semi-consummatory?), and lastly her personal number (consummatory act).

  Thorpe, arguing on similar lines, has given an inventory, which fills two printed pages, of the eighteen releasers and fourteen distinct action-patterns in the Longtailed Tits' 'consummatory acts' of building a nest -- ending with the exclamation: 'So much for simplicity!' He concludes:

  It seems, then, that in much of appetitive behaviour the animal's own activities . . . must be self-rewarding and self-stimulating. . . . In other words, much appetitive behaviour is also in a sense the consummatory act. . . . [24] Hinde concludes that appetitive behaviour and consummatory act differ only in degree, and that no absolute distinction can be made between them. Both are to some extent "spontaneous" in that they show evidence of external activation, and both are stereotyped to some degree and show some rigidity. Thus the classic examples of appetitive-behaviour and consummatory act can be regarded as the two ends of a series ranging from extreme variability and plasticity on the one hand to almost complete fixity on the other. [25]

  Many patterns of instinct-behaviour are of course cyclic: hunting ---> capturing ---> ingesting ---> digesting ---> hunting, etc.; and all one can say about the sub-activities in the cycle is that they are both 'appetitive' and 'consummatory', but that some are more appetitive and some more consummatory than others.

  What really matters in our context is the continuous scale of gradations between rigid and flexible action-patterns. Somewhere near the middle of the scale we find the common spider, whose web-making I have already used as a paradigm for an invariant yet adaptable built-in code (Book One, p. 38). It will suspend its familiar net from three, four, or more points of attachment, according to the lie of the land; yet the centre of the polygonal web will always coincide with its centre of gravity and the radial threads will always intersect the lateral threads at equal angles. We thus have a simple fixed code, yet a highly flexible strategy. Moreover, if some of the garden spider's legs and claws are amputated, it will still construct a more or less normal net -- the code remains unaffected by the elimination of some members of the matrix.

  Leerlauf and Displacement

  Towards the 'rigid' end of the scale we find reflex-like matrices, exemplified in the so-called Leerlauf activities. This term, too, was coined by Lorenz; the current English translation is 'vacuum activity' -- but 'freewheeling' would perhaps be more appropriate. Seagulls, reared in isolation, will perform on the stone floor of the laboratory their characteristic 'tap-dance' which, under normal circumstances, would serve to bring small animals to the surface of the tidal mud. Cats will go through the motion of burying their faeces on the kitchen tiles; and hand-reared young flying-squirrels 'when given nuts, would go through all the motions of burying them in the bottom of the wire-cage, and then go away contented, even though the nuts were exposed to full view'. [26] The same author describes the behaviour of hand-reared tawny owls 'which, after being fed, would act as if pouncing upon living prey though it had never had the experience of dealing with a living mouse'. [27]

  Such examples of 'stupid', automaton-like behaviour are the strongest evidence for innate codes of action. At the same time they are also additional evidence against the chain-reflex theory of instinct-behaviour: the owl, which has never seen a mouse, pounces after being fed, and without any visual stimulus; in the
gull's case, the hard floor of the laboratory is a stimulus quite different from the soft mud -- hence the 'chain-reaction' ought never to start, or to break off after the first unsuccessful attempt at 'digging' the tiles, Instincts are purposive and flexible, but their flexibility is limited to conditions more or less within the experience of the race. In a crassly unnatural environment the performance degenerates into 'freewheeling' and loses its purposive aspect.

  According to Lorenz's rather controversial theory, the motivation of Leerlauf activity is derived from a 'damming up' of the animal's 'specific action potential' (SAP) which lowers the threshhold of the innate releasing mechanism so that the action will go off even in the absence of appropriate stimuli. Hence also the term 'overflow activity'.