The Act of Creation

by Arthur Koestler

  Master-Switches and Releasers

  Motivation has been discussed in Chapter VIII. It determines what kind of game the subject will engage in, and activates the proper codes. If he feels the need to build castles in Spain, day-dreaming will replace the routines of planning ahead. The rules of day-dreaming impose a minimum of restraints and leave a vast choice of strategies to reach the desired goal, wish-fulfilment. Thus even day-dreaming is 'goal-directed', but the direction of thought is determined by emotional gradients, not by a concrete target.

  At the opposite end of the motivational spectrum are activities like problem-solving, governed by complex and precise rules. The goal to be reached is sharply defined, but has in itself no emotional significance; the reward is not contained in the target but in the act of reaching it. Some textbooks make a distinction between 'associative' and 'directive' thinking; but directiveness in the sense described is present even in the daydream, and controlled association enters into problem-solving; the difference is one of degree. Similar considerations apply to other classifications: abstract -- concrete, realistic -- autistic, etc.

  Most ordinary thinking is of a mixed kind; it may pursue a set directional course for a while, according to strict rules, then go off at a tangent and drift along, until some higher centre enters into action, and discipline is restored. We have discussed these general aspects of ideation before; it remains to consider briefly some specific patterns of verbal thought.

  At the base of all hierarchies which enter into our universes of discourse operate the implicit sub-codes of grammar and syntax; 'implicit' becanse they are automatized and we are not aware of their functioning. Lashley's dictum on perception is equally applicable to speaking and listening: 'We are aware of an organized structure; the organizing is never experienced.' The rules which determine how thoughts are put into words cannot themselves be put into words -- except by the patient labours of logicians and semanticists to 'break the code'.

  Next come the rules of common sense or common-or-garden logic, which are also empirically acquired, abstracted relations -- codified modus operandi which the majority of people are no more able to define than they can define how they ride a bicycle. But as we move upward, towards more specific universes of discourse, the codes, too, become more explicit.

  The simplest examples of explicit codes are the verbal commands in word-association tests, e.g.: 'name opposites!'. The experimenter then says 'dark', and the answer 'light' pops out promptly, as if produced by a slot-machine -- although in a free-association test the subject would probably associate 'dark' with 'night' rather than with 'light', and 'hot' with 'Italy' rather than with 'cold'. Thus the verbal command 'opposites' has acted as a master-switch, as it were, which changed the entire pattern of verbal organization. Even more striking is what happens to my verbal matrices si je continue de developer ma pensée en français -- if I continue to develop my argument, but use French words and French grammar to express it. My line of thinking has remained the same (or almost entirely so) but that single command-word 'French' has triggered off an instant reorganization affecting millions of neurons and their mode of inter-action. Not only has the vocabulary been changed, but also the method of converting ideas into syntactic language-units according to the sub-codes of French grammar.

  I do not want to go into neuro-physiology, but let us note that the model of the telephone-switchboard plugging into another localized 'language-area' simply will not do. Let us return for a moment to word-association tests. 'Light-dark', 'hot-cold' are primitive examples of matrices governed by codes with a single parameter in semantic space. Even so, are we to assign, by analogy with the 'language-areas', different cortical territories to operations controlled by the commands 'opposites', 'synonyms', 'super-ordinate class', etc.; not to mention 'inter-polation', 'extra-polation', and the mighty hierarchies of symbolic logic or of mathematical operations? Moreover, if the command is again 'opposites', and the stimulus-word 'Napoleon'; or if the command is 'supraordinate class' in the species-genus game, and the stimulus-word 'nail' or 'birth' -- what is the 'correct' response? Semantic space is multi-dimensional and cannot be represented by purely spatial connections activated by all-or-nothing signals in the three-dimensional rind of the brain; a model must include at least specific signals (e.g. frequency modulation pulses) and chemical (RNA) changes in the neurons to account for selectivity of response. [2]

  In all controlled associations and symbolic operations we again find a principle confirmed which we found operating on all levels -- namely, that a relatively simple 'releaser' signal from higher quarters ('opposites', 'speak French', 'find the square root of . . .') triggers off the operation of a complex code -- a whole universe of discourse in fact, with a hierarchy of implicit sub-codes, and a flexible strategy. (Even to find an 'opposite' to the word hairpin requires complex operations governed by individual 'strategies', perhaps involving visual images -- a process continuous with problem-solving). The trigger-releaser may be a verbal or visual, or even chemical command -- a tumblerful of gin or an amphetamin tablet: yet look what remarkably new rules of the game are triggered off by them.*

  Explicit Rules and Implicit Codes

  The implicit codes of grammar and syntax were, as we saw, acquired empirically 'as the gypsy learns to fiddle'. But the rules which govern more advanced symbolic skills -- mathematics or chemistry, or Law -- are learned in explicit verbal form. They may be acquired by rote-learning, or by guided learning, but at some stage they must be stated in explicit form. Take an example from elementary algebra: the average student learning the rule for solving quadratic equations: x[1,2] = -p/2 ± SQRT(p²/4-q). For the next few days or weeks, every time he has to use the formula, he must look it up in his textbook (or, if he is very brilliant, derive it afresh). At this stage of the learning process the formula is not yet an automatically functioning rule of the game; it is not yet a 'code' impressed on his nervous system. But after some practice, a single glance at an expression of the form x² + px + q = 0 will tell him that it is a member of the matrix of quadratic equations and trigger off the rule -- by now an automatized code -- for solving it.

  Thus rules which at first have to be looked up, or reconstructed by a conscious effort, become codified and automatized by routine. It is irrelevant in this particular context whether the student has by now forgotten the derivation of the formula and merely uses it as a mechanical gimmick; or whether he is aware of its binomial ancestry. This makes a vast difference in terms of understanding, but need not affect the process of automatization.

  I have emphasized before that the term 'code' is used in this book not in the metaphorical way in which Freud, for instance, used the word 'censor', but to refer to concrete processes or patterns of organization in the central nervous system. However, it sounds somehow more plausible to attribute physiological reality to codes on lower levels of the organic hierarchy -- the genetic code, the codes of instinct-behaviour and sensory-motor skills -- than to claim that the rules of such esoteric games as non-Euclidean geometry or quantum mechanics are physiologically represented, by appropriate coding methods, in the nervous system, and can be triggered off by simple releasers (such as the command 'let's talk shop'). I must therefore underline once more that the term 'code' is meant to apply only to those rules of behaviour which govern established routines and function automatically without conscious effort. In the initial stages of learning a complex symbolic skill, the rules of the game (laws, theorems, mathematical or chemical formulae) must be constantly memorized, looked up, or consciously recalled; so long as this is the case they are not yet incorporated into the physico-mental organization and are not to be called 'codes'. Codes are only those fixed stable rules which, once switched on, automatically govern the thinking routine. The problem in problem-solving consists firstly in discovering which routine is appropriate to the problem -- what type of game is to be played; and secondly, how to play it -- i.e. which strategy to follow, which members of the flexible matrix are to b
e brought into play according to the lie of the land.

  We learn, or discover, with strenuous effort, a new method of thinking; after a while, with practice, the novelty changes into semi-automatized routine, based on an invariant code with an adaptable matrix, and is incorporated into our repertory of habits. It is astonishing how soon, once a new railroad is built across desert and mountains, the passenger-trains start running on schedule.

  But the process of habit-formation does not stop there: not only the rules of the newly learnt game become soon automatized to such an extent that it becomes increasingly difficult to go against them, but strategy, too, tends to become stereotyped and incorporated into the code. Take progress in chess, an example I have mentioned before. The beginner is uncertain about the rules; then the rules become automatic codes and it becomes almost impossible for him to move his men in impermissible ways; after protracted practice certain tactical principles, which are no longer 'rules' in the formal sense, also begin to operate automatically in his mind -- e.g. to avoid pins, to seek open rook-files. But this reification of tactical pseudo-rules into automatized sub-codes contains a mortal danger, because considerations of strategy on a higher level demand that each of these tactical rules should be broken if the occasion warrants it. Sacrifices in material, and moves which look cockeyed (that is: positionally unsound), are signs of combinative power, i.e. originality; the mediocre player always remains a slave of habit and cautious orthodoxy.

  At this point the argument merges into that of Book One (IX-XI), concerning the pitfalls of orthodoxy, over-specialization, and one-sided development in the history of science and philosophy. In biology or theoretical physics there are no clean-cut distinctions between canonical rules of the game and heuristic rules of strategy and tactics. We are inclined to believe, as popular books on science tell us, that the 'permissible moves' are laid down for ever by the laws of formal logic and the criteria for judging evidence; and that strategy is determined only by the lie of the land, that is, the data of observation. In fact, however, the rules turn out to be infiltrated with implicit assumptions and 'self-evident axioms' which as often as not are specious contraband; and the empirical strategies are often weighted by a stubborn adherence to methods of interpretation and biassed techniques, promoted to canonical status. Habit is heir to originality; without the hierarchies of organized habits life would be chaos; creativity means breaking up habits and joining the fragments into a new synthesis.

  Matrix Categories

  I have tried to outline the hierarchic organization of levels of understanding, levels of consciousness, and levels of habit and flexibility -- the last ranging from implicitly acquired codes, through the master-switches of controlled association, to the explicitly learned rules and pseudo-rules in the universes of discourse of science and philosophy. To avoid giving undue dominance to the abstractive hierarchies in the mental landscape, I must briefly mention some different types of language matrices -- without aspiring at anything like a complete catalogue.

  Phonetic matrices (of rhythm, meter, alliteration, assonance, rhyme, and euphony) do not properly belong to symbolic thought, though they interact with it not only in poetry and word-games but also in ordinary discourse, often more persistently than we are aware of.

  Chronological matrices, naively regarded, seem to be linear chains of events, but are of course nothing of the sort. They are multi-dimensional structures in semantic space, governed by a diversity of selective codes, whose criteria of relevance are often quite indifferent to temporal order. This applies to personal memories, which always unfold within specific frames of reference, but also to written History: historians organize their material according to highly idiosyncratic rules for sifting and interpreting evidence, and for constructing causative theories.

  Classificatory codes in taxinomy, indexing systems and in certain branches of mathematical logic are hierarchic par excellence but rigid; they resemble stone pyramids in the mental landscape.

  Dogmatic matrices could be described as closed systems with distorted feedback and impaired sub-skills of reasoning. They are ruled by a fixed code derived from an act of faith, a circular argument, or supposedly self-evident axioms. In other respects, however, they are remarkably adaptable, and their dialectical strategies are of great subtlety. Related to these are frames of value which determine ethical or aesthetic judgements or attitudes, and emotion-dominated matrices which need not be discussed as a separate category, since emotion enters in various guises and intensities into all form of thought.

  Lastly, 'style codes' represent a person's idiosyncrasies, mannerisms, etc., which, in their ensemble, constitute his individuality. Gait, gesture, voice, hand-writing are all governed by stable automatic codes. If a person, deprived by accident of the use of his right hand, learns to write with the left, his signature, before long, recovers its true character. 'Even the suspicious bank clerk will cash his cheque because the old form of signature returns,' wrote Penfield [3], who had several such patients. 'The pattern of the signature and of the writing is in the brain, not in the hand.'

  This applies, to a considerable extent, even to the style of writing. Hemingway or Proust can be identified -- and parodied -- after reading a few lines -- as you identify the timbre of a drum or violin after a few transients. Even the language of common mortals whose style is undistinguished and seemingly indistinguishable, appears to have fixed characteristic ratios -- e.g. between the number of adjectives to verbs.

  The total matrix, which comprises all these frames of behaviour, constitutes the personality structure. But even here, the code can apparently be triggered on and off by some super-master-switch -- as the spectacular cases of multiple personalities indicate. Once more the hierarchy fades into a receding series.

  NOTES

  To p. 634. Cyberneticists have discussed at length models which are supposed to be capable of this feat. But they have no bearing on the question of awareness.

  To p. 637. A summary of earlier work on controlled association tests, and the controversies around it (e.g. complex theory versus constellation theory) can be found e.g. in Woodworth (1939), pp. 790-800.

  XVII

  ASSOCIATION

  'Multiple Attunements'

  Associationism is dead, but association remains one of the fundamental facts of mental life. So far I have considered mental organization chiefly in its 'vertical' aspect -- hierarchic structures formed by abstractive processes in ascending series. But each verbal concept, apart from being a member of a 'vertical' hierarchy, is also a member of several connotative matrices, each of which could be represented by an inclined plane. The concept's place in the vertical, abstractive hierarchy provides the dictionary definition -- as far as that goes -- of its meaning. But the concept as a psychological reality, its aura of connotations, and its individual significance to the person who actually uses it, is determined by the multitude of matrices which intersect in it. Each of them provides an associative context governed by a selective code; and the more there are of these inclined planes in semantic space the richer and more multi-dimensional the concept.

  If concepts are to be regarded as atoms of thought, they are certainly not the hard lumps of classical physics. In the first place, they are unstable and subject to change -- to change both in definition and in connotation. My concept of a 'gene' or a 'seductress', or of 'President Eisenhower' is certainly not the same as it was ten years ago, though the verbal label attached to each of these concepts has remained the same. It is strange to reflect that a major part of our scientific and philosophical vocabulary consists of old Greek bottles filled and refilled with new wine; that "electron" once meant a piece of amber, and Homer's "cosmos" a flat disc covered by a vault. It is even stranger that the same Sanskrit root "matr" split, by mitosis, as it were, into "maya" -- the Oriental's web of illusions, and "metron", metre, the Occidental's yardstick to measure the world.

  A concept has as many dimensions in semantic space as there are matrices of wh
ich it is a member. Let me return for a moment to the example of the parlour game mentioned before, 'towns starting with M'. In playing that game, I write down on my list 'Madrid' -- which proves that the concept Madrid is a member of the phonetic matrix governed by the code 'initial M'. Since I am bored with the game, I permit my thoughts to wander, and at once an image arises: the crowd at the Puerta del Sol stampeding in the panic of an air bombardment -- and off we go along the emotion-charged matrix of my memories of the Spanish Civil War. At this moment Brenda's little girl -- who was watching the game, equally bored -- asks, 'What is Madrid?'; and I oblige with the information: 'Madrid is the capital of Spain and of the Province of Madrid, situated on the left bank of the river Manzanares, which falls into the river Jarama' -- whereby I have produced a definition of sorts of the concept 'Madrid'. A moment later I remember the Prado, with its Goyas, Velasquezes, and El Grecos -- which are items in a mental catalogue indexed under the code 'Painters', sub-code 'Painters, Spanish' (but also under 'Spain', sub-code 'Spain, painters of'). These connotations presented themselves more or less automatically, but now my repertory of associations is nearly exhausted, and my mind a momentary blank. Add to the repertory the printed and auditory-vocal images of the word, plus the location of the town on a mental map, and you get about half a dozen matrices which will be activated by, and which will activate, the concept 'Madrid' without effort. The associative contexts of a concept that are firmly established in a person's repertory of thought-habits, are less numerous than we are inclined to believe -- as free association tests demonstrate.