The Act of Creation

by Arthur Koestler

  We have seen how laughter is sparked off by the collision of matrices; discovery, by their integration; aesthetic experience by their juxtaposition. Snobbery follows neither of these patterns; it is a hotchpotch of matrices, the application of the rules of one game to another game. It uses a clock to measure weight, and a thermometer to measure distance. The creative mind perceives things in a new light, the snob in a borrowed light; his pursuits are sterile, and his satisfactions of a vicarious nature. He does not aim at power; he merely wants to rub shoulders with those who wield power, and bask in their reflected glory. He would rather be a tolerated hanger-on of an envied set than a popular member of one to which by nature he belongs. What he admires in public would bore him when alone, but he is unaware of it. When he reads Kirkegaard, he is not moved by what he reads, he is moved by himself reading Kirkegaard -- but he is blissfully unaware of it. His motions do not derive from the object, but from extraneous sources associated with it; his satisfactions are pseudo-satisfactions, his triumphs self-delusions. He has never travelled in the belly of the whale; he has opted for the comforts of sterility against the pangs of creativity.

  One cannot discuss the act of creation without devoting at least a few pages to the act of desecration. Snobbery is a poor word with too specifically modern connotations for that benightedness, due to the confusion of values, which is one of the leitmotifs of the history of man; he always seems to be groping his way through a labyrinthine world, armed with a compass which always points in the wrong direction. The symbol of creativity is the magic wand which Moses used to make water come out of the rock; its reverse is the faulty yardstick which turns everything it touches into dust.

  BOOK TWO

  HABIT AND ORIGINALITY

  INTRODUCTION

  So far I have discussed creativity in science and art, that is to say, the highest forms of mental activity, with only occasional references to the humbler routines of existence. I started at the roof, as it were -- what remains to be done is to build up the walls which support it.

  The main purpose of this somewhat perverse procedure was to deal first with those subjects which are of primary interest to the general reader and to establish a wider theoretical framework afterwards. But there exist additional considerations to justify this reversal of order. At the Hixon Symposium in 1948 K. S. Lashley quoted with approval a French author writing in 1887. 'The study of comparative grammar,' Lashley said, 'is not the most direct approach to the physiology of the cerebral cortex, yet Fournié has written, "Speech is the only window through which the physiologist can view the cerebral life." ' [1] The word 'only' is, of course, an unwarranted exaggeration, but perhaps no more unwarranted than the opposite claim, that the 'only' legitimate window is that through which we watch the workings of the salivary reflex in dogs or the behaviour of rats in mazes. To repeat an argument from the Preface to Book One: in the history of most sciences we find alternations between the downward approach from roof to basement, from the complex to the elementary, and the upward approach from the elementary to the complex; until the two finally merge. It was the study of complex electro-magnetic phenomena which provided the due to sub-atomic structures. Torn out of the larger context in which the 'elementary part' functions, it ceases to be a true elementary part -- whether we speak of electrons, tissue-cells, or 'elements of behaviour'. Genetics started with morphological classifications and comparisons of whole organisms long before anything was known about chromosomes, genes, and nucleic adds. The use of undefined, 'dirty' concepts as black boxes in theory-making has led into many cul-de-sacs in the history of science, but was nevertheless indispensable for its progress.

  In experimental psychology the pendulum oscillated in the course of the last century between the 'upward' and the 'downward' approach: from the atomism of the English associationists to the introspectionism of the Würzburg school; from the chain-reflex theory of Watsonian behaviourism to the Holism of Gestalt, and back again to the more sophisticated behaviourism of the forties and fifties. But there are signs which indicate that the controversies between 'cognitive' versus 'stimulus-response' theories of behaviour have become sterile, and that a new synthesis is in the making.

  So much by way of justification for the order of the two parts of this book. The methodological approach of the second is meant to be the reverse of the first. It starts 'from the bottom' with some elementary considerations which are non-controversial. Yet gradually, I hope, a structure will emerge which makes contact with and lends support to the controversial theories of creative activity outlined in Book One.

  I

  PRENATAL SKILLS

  Organic life, in all its manifestations, from embryonic development to symbolic thought, is governed by 'rules of the game' which lend it coherence, stability, and an appearance of purpose (or 'goal-directedness' if you prefer that term). These rules or codes, whether phylogenetically or ontogenetically acquired, function on all levels of the hierarchy, from the chromosomes to the neuron-circuits responsible for verbal thinking. Each code represents the fixed, invariant aspect of an adaptable skill or matrix of behaviour. I shall take the stylistic licenee of using the word 'skill' in a broad sense, as a synonym for 'matrix', and shall speak of the morphogenetic skills which enable the egg to grow into a hen, of the vegetative skills of maintaining homeostasis, of perceptual, locomotive, and verbal skills.

  We shall find as a fundamental characteristic of codes on all levels that they function on the trigger-release principle, so that a relatively simple signal releases pre-set, complex action patterns. The signal may be mechanical, as in artificial parthenogenesis induced by a pin-prick; chemical (e.g. inducers and evocator substances); or neural (Tinbergen and Lorenz's Internal Releasive Mechanisms). But the pre-set action pattern activated by the code is generally an elastic pattern, not a rigid automatism (such as suggested, for instance, in Tinbergen's schema). Skills have varying degrees of flexibility. The restraints imposed by the code do not exhaust the degrees of freedom possessed by the matrix; there are usually various alternative choices left to provide for a flexible strategy according to the 'lie of the land' -- i.e. guided by feedback from the environment. Matrices thus function under the dual control of an invariant code and a variable environment. These two factors jointly determine which members of the matrix should enter into action and in what order.

  The members of the matrix are sub-matrices governed by their sub-codes. Facing 'upward' and 'inward' in the organismic hierarchy, they act as dependent parts; facing 'downward' and 'outward', they act as relatively autonomous sub-wholes. On every level of structure and function in the living organism we find the same phenomenon of Janus-faced entities which behave partly as wholes, yet wholly as parts, as it were. The 'irreducible, elementary particle' has turned out to be a will-o'-the-wisp, even in inorganic chemistry; in the organism, the firing of a single nerve-cell turns out to be not an event, but a complex, autonomous pattern of events.

  Structure and Function

  I shall take as my starting point two complementary aspects of the evolutionary process: differentiation of structure and integration of function. Regardless of what causal theory one adopts to account for the evolution of higher forms of life, the means by which progress is achieved is the development of more complex and specialized bodily structures and of more elaborate ways of coordinating their function. These are two aspects of a unitary process, not two processes. In fact both terms, 'structure' and 'function', are abstractions derived from imaginary cross-sections along the spatial or temporal axis of indivisible spatio-temporal events.

  Structure is a static concept of a process frozen in the specious present. 'Anatomy abstracts from time. It studies the organism conceived "at an instant".' [1] A kidney, or a single kidney cell, which has stopped functioning no longer has the same spatial structure as before; and while it is still functioning its structure is changing at every moment. All living matter continuously alters its shape and composition; it can be said to preserve i
ts identity only if both structure and function are taken into consideration. Vice versa, function detached from structure is the grin detached from the cat. Physiological processes, unlike mechanical automata, are 'open systems'; Ostwald compared them to the flame of a gas-burner. The flame is a relatively stable phenomenon; but how is one to distinguish between its structure and function, between combustible and combustion?

  Most of the time it is both convenient and necessary for the biologist and physician to forget that when he is speaking of 'structure' and 'function' he is using 'dirty concepts' -- otherwise his work would come to a standstill. However, we shall soon come across problems where the tacit implications of both terms will become relevant.

  A second catch is the semantic ambiguity of the word 'function'. It can either mean the role or part which my liver ought to play to keep me in good health, or it can refer with heartless matter-of-factness to what is in fact happening in my liver, i.e. not to its function in the first sense but to its actual 'functioning'. In the first case, the organ is considered as part of the organism; in the second case it is treated, prima facie, as an independent whole. Whenever a distinction is indicated, I shall use the word 'function' in the first sense, and 'functioning' in the second.

  The complementary aspects of structural differentiation and functional integration are reflected at every step in the process of morphogenesis.

  The giant molecules of DNA (dioxyribonucleic acid) in the chromosomes of the fertilized egg are said to represent the blue-print of the future organism. Let us consider for a moment this first code which we meet at the base of the hierarchy. It is a 'code' in the twofold meaning in which I have used that term: a canon of rules for making an organism, represented in a 'coded', compressed language. The way in which the coding is done deserves some attention. Each DNA molecule is supposed to be capable of storing something of the order of 10^10 bits of information. [2] According to the Watson-Crick model, the molecule has the shape of a ladder twisted into a double helix. The two sides of the ladder consist of uniform, linear chains of sugar groups alternating with phosphate links. The rungs of the ladder are pairs of nitrogenous bases attached on each side to the sugar groups in the double helix. There are four of these bases: adenine, guanine, cytosine, and thymine, or A, G, C, T for short. A is always paired with T, and G always with C, so that the two sides of the ladder are complementary. The sequence of bases as read down along one side of the ladder is the genetic code, written in the four-letter alphabet, 'A', 'G', 'C', 'T'. Here, then, floating in the nuclear sap, is the code which governs the skill of creating a six-foot drum major with a slight squint and dimpled cheeks, out of an egg with a diameter of a few microns.

  The code's remarkably complex activities are set off by a relatively simple signal: the impact of the fertilizing agent. The principle involved must indeed be relatively simple, because in virtually all main groups of animals, including rabbits and sheep, normal development of the egg can be induced by stimulation with non-specific, physical, or chemical agents: heat or cold, salinity, ultra-violet rays, galvanic currents, or puncture with a fine needle. Though it may be hurtful to the male's pride, his seed seems to be a dispensable commodity. Its primary function is to trigger off the egg; but -- with frogs at least -- a platinum needle will do just as well.*

  The main task of the growing cell is the manufacture of proteins which constitute the basic stuffs of the body, and of protein-enzymes, which control all its biochemical activities. The growing cell is primarily a protein-producing factory, or rather a group of specialized plants, each engaged in synthesizing one variety at a time among the thousands of different types of protein-molecules. The order and quantity in which each substance is produced is regulated by the genetic code. The key-substances are the enzymes, which can be regarded as structural embodiments of the genetic instructions coded in the DNA chains. Each discrete set of instructions -- a code-word composed of several hundred code-letters -- contains the recipe, as it were, for a specific enzyme. It is usual to call these sets of instructions 'genes' (although the meaning of this traditional term has become somewhat blurred and some geneticists prefer more non-committal words). Each code-word is composed of sub-wholes: three-letter syllables like CTA, AAG, etc. Each of these sub-units in the code-sequence is capable of 'recognizing' and transporting one particular kind of sub-unit which goes into the making of proteins. These sub-units or 'building-blocks' are the amino adds; there are twenty common varieties of them. A protein usually contains all or most of the twenty varieties of amino-acid units, strung together into a chain of two hundred or more in a specific sequence. The number of permutations of twenty units in a sequence of several hundred links is of course enormous, and accounts for the impressive variety of proteins and protein-enzymes which go into the making of our all-too-solid flesh. Thus the main process in morphogenesis seems to consist in transcribing the implicit, 'functional' four-letter alphabet of the code into the explicit 'structural' twenty-letter alphabet of proteins.

  The details of the process are only beginning to be known at the time of writing; and progress has become so accelerated in this particular field that by the time these pages appear in print much in them will be outdated. Broadly speaking, we know at present of two kinds of instruction-carriers between the genes and the protein-plants of the cell. The first is 'messenger RNA'. The chemical structure of RNA -- ribonuclcic acid -- is closely related to that of DNA. The DNA chain apparently manufactures 'messenger RNA' chains by the same process of base-paring by which it replicates itself. Each messenger is supposed to contain the 'recipe' of one specific enzyme, and to carry it to the protein-plant, the ribosome. A second type of messenger is 'transfer RNA'. Each variety of it is a much shorter chain, supposed to represent a triplet of code-letters corresponding to a particular variety of amino adds. When a transfer RNA unit meets an amino unit of the proper type in the proper condition, it will attach itself to the unit like a sort of locomotive, drag it to the ribosome plant, and shunt it into its proper place in the nascent protein chain, according to the instructions of its bigger colleague, the 'messenger RNA'.

  This, in broad outlines, is the picture to date -- in the year after the Crick-Watson-Wilkins team received the Nobel prize for 'breaking the genetic code'. It is to a large extent derived from work on the genetics of micro-organisms such as the colon bacillus, and there are reasons to believe that something essential is still missing from the picture.* But we need not be concerned with these niceties. What matters is that we find, literally ab ovo, fixed, invariant codes in control of the variable cell matrix. The order in which the various sub-codes -- the genes -- become active and engage in producing their specific varieties of messenger RNA is supposed to depend on the pre-set activities of 'operators' and 'repressors' in the chromosomes, and perhaps on some additional biological clocks. But it must also depend to some extent on feedback signals from the cytoplasm, including signals from neighbouring tissues which reach the cell through its sensitive membranes. Finally, recent experiments by Hyden [2a] have shown that in the nerve-cells of the mature organism experiences retained in learning lead to more or less lasting alterations in the chemical composition of the cell's RNA content: a fact of great significance for the problem of memory.

  The Cell-Matrix

  The growing cell illustrates the relativity, or complementary nature, of such terms as part and whole, or structure and function. 'What are called structures are slow processes of long duration, functions are quick processes of short duration. If we say that a function such as the contraction of a muscle is performed by a structure, it means that a quick and short process-wave is superimposed on a long-lasting and slowly running wave' (Bertalanffy). [3]

  The same considerations apply to the relations between the genetic code and the cell-matrix. The latter is of course a very complex structure, equipped with an animal and a vegetal pole, with chemical and morphogenetic gradients, and with a variety of organelles -- these, in their ensemble, are the me
mbers of the matrix. Each member is a sub-whole governed by its sub-code which can be turned on and off by orders 'from above', but once triggered into action follows its autonomous functional pattern. The centrosomes and kinetosomes, for instance, are organelles of apparently similar structure but vastly different functions. Both possess the privilege of self-replication when the cell divides; both have a cylindrical shape and are made up, it seems, of eleven fibres -- two in the centre, nine outside. But these deceptively simple structures revealed by the electron-microscope are complete biochemical machines in miniature. The kinetosomes are attached to the cilia or flagella of motile cells and trigger off their activities. The centrosomes provide the poles of the spindle apparatus, which effects cell division; they establish connections with the chromosomes in the distant nucleus, tear the pairs of duplicated chromosomes apart as it were, and direct them towards their new locations in the centres of the two nascent daughter cells.

  To take another example, let us cast a quick glance at the most glorious of organelles, the mitochondria. There are fifty to five thousand of them according to the type of cell and (together with the chloroplasts of green-leafed plant-cells) they are 'the power plants of all life on earth'. As a physicist has said, waxing lyrical on a biological peccadillo, they 'feed on negative entropy and drink orderliness from the environment'. [4] The chloroplasts do it by photosynthesis, the mitochondria by extracting a very special form of energy from the chemical bonds in glucose, fat, and proteins. The end product is ATP -- adenosine-triphosphate -- the universal carrier used by the organism to supply the energy needed for the contraction of muscles, the transmission of nerve impulses, the manufacture of proteins, and so on. The synthesizing of ATP is carried out in three main stages (glycolysis, Krebs cycle oxidation, and 'respiration'*) involving about fifty different successive reactions, each of them catalyzed by a specific enzyme system. A single mitochondrion may possess up to forty thousand of such enzyme systems; their assemblies are arranged in the membrane of the organelle in patterns which have been described as 'wiring diagrams'. They represent the sub-codes which govern the operations of the energy plant. It also has a complex feedback system: in the presence of an excess of ATP, for instance, the organelle will contract.