The Act of Creation

by Arthur Koestler

  On the level of the zygote, the cell-matrix consists of biochemical gradients and organelles; it is a structural mosaic equipped with axial polarity which under normal conditions predetermines the head and tail region, the blastopores, etc., of the embryo; but it also has striking regulative properties revealed by experimental manipulation. With progressive differentiation the regulative properties of the cell diminish, and its degrees of 'genetic freedom' freeze up. On the level of the morphogenetic field we again find self-regulating properties -- half the field will still form a complete limb -- and a mosaic-matrix of cell-populations. The autonomous, self-assertive character of morphogenetic fields is manifested in grafting and centrifuging experiments; their dependent, or part character by the trivial fact that they are kept in their proper size and place in the normally growing organism. This will sound less of a truism when we turn to the phenomena of regeneration (Chapter III). We shall find regulative character and mosaic character, autonomy and subservience, the self-assertion of the part and its dependence on the whole to be complementary aspects on every level of the hierarchy in normal development and behaviour; but also, that under abnormal conditions this ceases to be the case, and that the part may then assert itself at the expense of the whole, with sometimes beneficial, mostly destructive, effects.*

  NOTES

  To p. 418. This is, of courre, not meant to belittle the enormous advantages of sexual over asexual reproduction.

  To p. 419. The three-letter 'dictionary', for instance, is partly a dictionary of synonyms: there are 4^3 = 64 triplets, but only 20 arnino acids, and many of the latter are represented by more than one code syllable.

  To p. 420. 'Respiration' is an approximate term. The process is in fact oxidative phosphorylation.

  To p. 422. In some, probably extreme cases, the nuclear changes are even more drastic. The nucleus of the fertilized egg of the gall-midge contains forty chromosomes, and in the course of the first few divisions these are faithfully duplicated. But in the fifth division, only eight sets of chromosomes in the soma cells duplicate in the orthodox manner; the other thirty-two fail to do so and gradually dissolve in the cytoplasm. The future germ calls, however, which have previously been segregated from the rest of the eggs, do not participate in the fateful fifth division and preserve their chromosome complement intact. Thus the nuclei of all specialized body cells have only eight chromosomes, whereas the germ cells have forty. Cf. Fischberg, M. and Blackler, A. W. (1961).

  To p. 423. In a paper read at the British Association Meeting in August 1962, L. Wolpert suggested that differentiation resulted from the single cell's tendency to stick on to that part of the gastrula wall best suited for it (the idea seems to have been originated by T. Gusthafson). At the same meeting E. N. Willmer showed that changes in the salt balance of the surrounding medium made amoebae change from amoeboid to flagellate form, reversibly -- the implication being that chemical gradients played an important part in the early stages of differentiation (New Scientist, No 303, 6.9.1962, p. 492).

  To p. 425. During maturation in the higher species, the two types of control overlap; and pre-set biological time-clocks seem to exercise some influence throughout adult life.

  To p. 428. In this necessarily simplified discussion of morphogenetic processes I have made no mention of cytoplasmic inheritance and other complicating factors, which do not affect the basic argument of this book.

  II

  THE UBIQUITOUS HIERARCHY

  Development of the Nervous System

  The pioneer work on the development of the nervous system in vertebrate embryos is G. E. Coghill's monumental study of ambystoma, a larval form of salamander). [1] Coghill published his results in a series of papers spread over a period of twenty-five years, 1914-39. Since they are surprisingly seldom quoted outside the technical literature -- presumably because they ran against the behaviourist Zeitgeist -- I must briefly summarize his conclusions.

  The traditional assumption about the development of the nervous system was that elementary, local reflexes arise first, and are chained together at a later stage. Thus the segmental reflex arcs of the earthworm would develop first, as independent units aligned in a series perpendicular to its axis, and only later on would they become connected, like rings hanging from a festoon string, by the spinal cord. Coghill's work showed that the opposite is true. In the salamander, development starts with the growth of the motor-tracts of the cord axially from head to tail; then this central bundle sends out collateral branches into the segmental muscles, coordinating their actions in primitive, unitary patterns; the sensory neurons become functional only at a later stage, and the local reflex-arcs come last, as segregations of 'partial patterns' out of the 'total pattern' which preceded them. The whole development is centrifugal: the stem precedes the branches, spontaneous undifferentiated movements involving the whole neuro-muscular apparatus precede differentiated movement, total responses precede specialized local responses. To give an example: when the limbs develop, their first movements are entirely dependent on and synchronized with the movements of the trunk. Only later on do the limbs begin to move independently; the same applies to the motions of head, mouth, etc. The growth of the nervous system from beginning to end is dominated by 'a totally integrated matrix, and not a progressive integration of primarily individuated units'. The organism is not a sum of its reflexes, but on the contrary 'the mechanism of the total pattern is an essential component of the performance of the part, i.e. the reflex'. The stimulus-response scheme cannot explain even embryonic behaviour, because movements appear long before the motor neurons of the reflex arc are connected with the sensory neurons. This centrifugal mode of development means that the individual acts on its environment before it reacts to its environment.

  'In so far as the correlation of nervous structure and function in the development of the individual has been carried, structural provision has been found for the perpetuation of spontaneity, autonomy, or initiative as a factor in its behaviour. Any theory of motivation, therefore, that attributes this function wholly to the environment, is grossly inadequate.' The idea that instincts are chained reflexes must be abandoned; instincts represent 'total action patterns in response to relatively general situations'. Comparing the embryonic development of ambystoma with that of the human foetus, Coghill sums up:

  In conclusion I am convinced by a study of all available records of movement in human foetuses during the first six months, that behaviour develops in man as it does in ambystoma by the expansion of a total pattern that is integrated as a whole from the beginning, and by individuation of partial patterns (reflexes) within the unitary whole.*

  We thus find in the development of the nervous system the same principles at work which we have discussed before. The neural plate starts as a primordium with multipotential cell-populations which differentiate in a series of steps into the brain, the spinal cord, and its substructures. The 'wiring diagram' of the organism has a standardized pattern -- an invariant code; but transplant experiments again show the great flexibility of the 'neurogenetic skill' which realizes that pattern. If a limb-bud from a salamander embryo is transplanted to another embryo's flank, outgrowing nerve-fibres locate the bud and establish a normal nerve pattern. The bulb-shaped tips of the outgrowing nerves are apparently guided by submicroscopic structures in the cell-matrix of the growing bud -- at least according to the current 'contact guidance' theory.

  I have called differentiation of structure and integration of function complementary aspects of a unitary process. But the 'functions' of the growing embryo are different from the 'functions' of the adult. It has been shown that the limb-buds and wing buds of chick embryos develop into almost normal legs and wings if nerves are prevented from entering them. This does not mean, of course, that differentiation of structure comes first, and integration of function later on, as a separate act. For the function of the leg-bud is to grow -- not to walk. Growth is a function controlled by the genetic code; when growth is completed and t
he time has come to walk, the nervous system takes over control; and if it fails to do so, the muscle tissues will degenerate, as denervated adult muscles do.*

  Locomotor Hierarchies

  'Whatever the nature of organizing relations may be,' J. Needham wrote in 1932, 'they form the central problem of biology, and biology will be fruitful in the future only if this is recognized. The hierarchy of relations, from the molecular structure of carbon compounds to the equilibrium of species and ecological wholes, will perhaps be the leading idea of the future.' [2]

  This prophecy has not come true. The Gestalt school's over-emphasis on 'wholeness', and the behaviourists' over-emphasis on 'simple elementary processes' -- the so-called S.-R. (stimulus-response) scheme -- created a controversy based on a fallacious alternative, and prevented a true appreciation of the multi-layered hierarchic order to be found in all manifestations of life.

  Yet the idea is of course by no means new; hierarchies in nervous function were proposed by Herbert Spencer in the 1870s, and elaborated by Hughlings Jackson, Sherrington, and others. [3] The hierarchical character of skills was demonstrated in great detail by Bryan and Harter in their study of telegraphy and in Book's study of touch-typing (see below, pp. 544 ff.) at the turn of the century, but neither S-R psychologists nor Gestaltists paid attention to them. Woodger (1929) attempted a formalization, by means of symbolic logic, of certain types of hierarchies ('divisional hierarchies', 'spatial hierarchies', 'genetic hierarchies', etc.) which are of somewhat abstract interest. Heidenhain (1923) [4] proposed a hierarchy of 'histo-systems' which are 'encapsulated' into one another (e.g. neuro-fibriles, neurons, nerve fibres). Bertalanffy (1952) tried to make a distinction between 'hierarchies of parts', 'hierarchies of processes', 'hierarchies of centralization', etc. Tinbergen defined instinct as a hierarchically organized nervous mechanism -- but his mechanism is fixed and rigid (see below p. 478). A stimulating discussion of the subject can be found in Miller, Galanter, and Pribram's remarkable essay on 'Plans and the Structure of Behaviour' (1960).

  The word 'hierarchy' can be used to mean simply rank-order. Hull's famous 'habit family hierarchy', for instance, means just that (the ordering of a group of interchangeable responses according to their strength), and is not a hierarchy at all in the sense in which the term is used in this book. I have summarized what I meant by it in the chapter 'Partness and Wholeness' (Book One, Chapter XIII). A hierarchy, in this sense, is not like a row of organ pipes; it is like a tree, arborizing downward. The structural or functional entities on each level are autonomous sub-wholes of complex pattern, but are represented on the next higher level as units. In every organic hierarchy, to paraphrase Gertrude Stein's statement about the rose, 'a part is a whole is a part is a whole'.

  Perhaps the most satisfactory theoretical treatment of the concept of hierarchic order was given by Paul Weiss -- whose experimental work was a major contribution towards providing the concept with a firm empirical basis. The quotation which follows is from the celebrated Hixon Symposium; its vivickss is enhanced by the fact that it is taken from an ex tempore contribution by Weiss to the discussion of Lashley's paper on 'The Problem of Serial Order in Behaviour'* (my italics):

  While the physiologist and psychologist deal with the ready-made machine of the nervous system and can add to it as many properties as he thinks necessary, the embryologist must explain just how such an immensely intricate, yet orderly, thing can develop. These studies are still in their infancy, but a few things have already come out . . . for instance, the relative autonomy of structured patterns of activity, and the hierarchical principle of their organization . . . . The nervous system is not one big monotonic pool whose elements can be freely recombined in any number of groupings, thereby giving an infinite variety of nervous responses. This used to be the old idea of the associationists, and it is utterly incompatible with what we have learned about the development of the nervous system and its function in animals. The working of the central nervous system is a hierarchic affair in which functions at the higher levels do not deal directly with the ultimate structural units, such as neurons or motor units, but operate by activating lower patterns that have their own relatively autonomous structural unity. The same is true for the sensory input which . . . operates by affecting, distorting, and somehow modifying the pre-existing, preformed patterns of central co-ordination. . . . The final output is then the outcome of this hierarchical passing down of distortions and modifications of intrinsically preformed patterns of excitation, which are in no way replicas of the input. The structure of the input does not produce the structure of the output, but merely modifies intrinsic nervous activities that have a structural organization of their own. This has been proved by observation and experiment. Coghill has shown that the motor patterns of the animal develop prior to the development of sensory innervation. I have shown, as others have, that the removal of the sensory innervation does not abolish the co-ordination of motor activities. Moreover, coordinated motor functions of limbs and other parts develop even if these parts have been experimentally prevented from ever becoming innervated by sensory fibres. Therefore, the sensory pathway can have nothing to do with the structure of the motor response. There are still some authors who try to save the old associationist idea that actually the input shapes the structure of the output. I think that they are fighting a losing fight, and I think that today's discussion ought to have given them the coup de grâce. Intrinsic automatic rhythms have been shown, for instance, by Adrian in the brain stem of the goldfish and in insect ganglia, by Prosser in other arthropods, by Bremer and by von Holst in the spinal cord, and by Bethe in jellyfish. I have shown experimentally that any group of bulbar or spinal nerve cells taken from vertebrates, if deprived of their structural bonds of restraining influences and allowed to undergo a certain degree of degradation, will display permanent automatic, rhythmic, synchronized activity of remarkable regularity. Rhythmic activity, therefore, seems a basic property of pools of nervous elements . . . . The rhythm is not something generated through an input rhythm; but is itself a primary rhythm which may be released and even speeded up or retarded by the input, but is not derived from the input. So we have experimental evidence that rhythmic automatism, autonomy of pattern, and hierarchical organization are primary attributes of even the simplest nervous systems , and I think that this unifies our view of the nervous system. [5]

  Let me enlarge on some of these points and add a few facts which have emerged since.

  In the first place it has been found that intrinsic, rhythmic activity of an autonomous character is not confined to motor nerves, but that 'receptors also are spontaneously active even in the absence of stimulation from environment.' [6] This spontaneous receptor activity, while modified by environmental events, is under efferent control from the central nervous system. The central control (both of the spontaneous receptor activity and of the input) is, as we shall see, primarily of a restraining, inhibitory nature. But for the time being let us confine ourselves to motor organization.

  In an earlier paper (1941 a, p. 23) Weiss distinguished the following levels of the hierarchy:

  1. The level of the individual motor unit. 2. All the motor units belonging to one muscle. 3. Co-ordinated functions of muscular complexes relating to a single joint. 4. Coordinated movements of a limb as a whole. 5. Coordinated movements of a number of locomotor organs resulting in locomotion. 6. "The highest level common to all animals", the movements of "the animal as a whole".

  This is as far as the schema proposed by Weiss goes. Now let me extend it one step further downward in the hierarchy. Even the lowest among the six levels is a very complex affair. The individual muscle cell of a striped muscle -- usually called a muscle fibre -- is a long, cylindrical structure surrounded by a membrane. Its principal functions are: (a) to serve as a receptor for nerve impulses which reach it at the synapse through a chemical transmitter; (b) to re-code this message into an electro-chemical excitation spreading along its surface; (c) to relay it
to the actual contractile structure, and to provide the energy for the contraction in the cell's internal energy currency -- ATP; (d) to contract.* This involves at least three distinct processes (acetylcholine transmission; sodium-potassium action potentials; activation by ATP) entering successively into action between the synaptic and the filament levels, with mechanical contraction as the end product. Thus the ultimate 'motor unit' at the base of the hierarchy is not the cell itself, but the apparatus within it which provides the contraction.

  That apparatus is, broadly speaking, a kind of cylindrical cable, the fibre, composed of a bundle of fibrils, each in turn composed of bundles of filaments. The filaments are of two varieties, a thick and a thin one, the former supposed to be consisting of molecular threads of actin, the second of myosin. The combination of these two proteins -- acto-myosin -- is a substance which contracts when activated by the energy carrier of the cell, ATP. The mechanism of the contraction is presumed to be a telescoping into each other of the thick and fine bundles of filaments. [7*]