The Act of Creation

by Arthur Koestler

  It is assumed, then, that muscle movement is due to the chemical action of ATP on contractile proteins; and moreover that essentially the same process is responsible for the streaming motion of amoeba, the rowing motion of cilia, and the tail-stroke motion of flagella. Amoeba, as I have mentioned before (note to p. 423, previous chapter) are capable of changing from amoeboid to flagellate form -- and back. Thus at the very bottom of the hierarchy we find the same basic, universal mechanism -- the archetype of organic motility as it were -- throughout the whole animal kingdom from amoeba to man; and within man, we again see it at work, serving such varied functions as the swimming of his spermatozoa, the bristling of his hair, and the flexing of his muscles in a tennis-stroke. It is a mechanism or apparatus with a high degree of autonomy -- and it reminds one of the equally autonomous functioning and universal occurrence of the power-plant-organelle, the mitochondria.

  However close we seem to get to rock-bottom in the organic hierarchy, we find complex, integrated sub-wholes leading a relatively autonomous existence. Viruses have been compared to 'nomadic' genes (though 'freelancing' genes might be a more appropriate description).

  Even a dead muscle cell (dissected and soaked in a cold glycerine solution for months, which makes it very dead indeed) will contract when exposed to the chemical trigger-action of ATP. The glycerine destroys cell-components serving its higher functions, but the essential structure of the fibrils -- the structural matrix of the basic motor unit -- remains intact; and it is still capable of functioning according to its built-in code. This function is of course a fixed reaction, as one would expect on the lowest level of the hierarchy; it is activated by the trigger-action of ATP on the acto-myosin. The environment of this matrix is represented by the temperature, oxygen supply, degree of fatigue in the cell, but the degrees of freedom of the fibre to adjust to these conditions boil down to an 'all-or-nothing' strategy: the alternative is to twitch or not to twitch.

  On the higher levels of the hierarchy, the autonomous function-patterns of muscles and muscle-complexes are even more in evidence; at the same time the degrees of freedom in the matrix allowing for adaptable performance increase with each level. Muscles dissected from the body and put into Ringer solution will contract normally for hours. Practically any part of an animal's heart, a muscular strip, and even a single muscle cell grown on a blood dot, will continue to go on beating in its own intrinsic rhythm. The heart of the chick embryo starts beating before any nerve cells have grown into it, and the hearts of frogs and tortoises will go on beating normally if the nerve supply is cut. Some smooth muscles equally show a rhythm of their own.

  But complementary to this Eigenfunktion, or functional autonomy, is control by a centre or system on the next higher level. The heart has its own pacemaker-system which is in itself a threefold hierarchy; under normal conditions, the sino-auricular node, the fastest part, acts as a pacemaker; but if it is prevented from doing so, the auro-ventricular node will take over, and as a last resort (in experiments carried out on frog and tortoise) yet a third centre may enter into action. The pacemaker-system is, in its turn, subject to regulatory control by sympathetic and parasympathetic nerves and by hormones, which will speed up or slow down the rate of beat by order of centres in the hypothalamus. Other organs -- kidney, intestine, stomach -- also have their self-regulating, intrinsic codes which assure their status as autonomous subwholes, while at the same time they function as parts in a multi-levelled hierarchy. Even the mid-brain centres which control temperature, metabolism, food and liquid intake, respiration, etc. -- even these homeostatic controls responsible for maintaining the equilibrium of the milieu intérieur, turn out to be subject to the control of still higher levels. They are 'biassed homeostats' which can be 'set', as a thermostat is 'set' by the tenant to keep a higher or lower room temperature. [8] And so the top of the hierarchy which controls the controls recedes into a cloud -- just as its base is embedded in the murky problem of what constitutes living matter -- and ultimately dissolves into genetic mutations with thresholds on the quantum level.

  The Goldfish and the Crab

  Getting back to earth, that is, to the medium levels of the hierarchy -- the levels 3, 4, and 5 in Weiss's schema -- we find, fortunately, more precise indications about its manner of working.

  Von Holst's study of the swimming motions of fish revealed a distinct three-step hierarchy: (a) the motions of the rays within a single fin, due to the alternative contractions of two antagonistic pairs of muscles; (b) the motions of the fin as a whole; and (c) the co-ordination of the motions of all the fins. In the anaesthetized goldfish, the swinging motions of each individual ray remain perfectly regular, but their co-ordination within the fin is disturbed: they flutter in disorder. The anaesthesia evidently does not affect the integrative centres on the lowest (a) level, but puts the higher nervous centres on the (b) level out of action. On the next higher, (c) level, the pectoral fin acts as a kind of pacemaker by imposing, or superimposing, its own rhythm on the caudal fins -- the so-called 'magnet effect'. This whole locomotor hierarchy is relatively independent of sensory stimuli, for fishes and tadpoles go on swimming in perfect co-ordination if they have been disafferentated, i.e. if all the main sensory connections have been severed. Von Hoist concludes that the stimulus-response schema does not apply to the autonomous locomotive hierarchy, and that 'the reflex is not the primary element of behaviour but a device for adapting the primary automatism to changing peripheral conditions'. [9]

  Higher up on the evolutionary ladder we find increasing flexibility of motor skills. In a series of famous experiments, von Buddenbrook and Bethe have shown that the removal of one or several legs from centipedes, spiders, and other insects does not lead to disorganization, but to a spontaneous rearrangement of the whole pattern of locomotion which is instantaneous and not preceded by trial-and-error learning. The normal progress of an insect or crab is the so-called 'cross-amble'. If 'L' and 'R' stand for left and right, and the index numbers stand for the order of legs from front to rear, the crab's locomotive code is as follows: R1, L2, R3, L4, R5, etc., are stepping simultaneously; then L1, R2, L3, R4, L5 -- are stepping simultaneously; and so on. If, now, the left front leg is removed, the pattern changes instantaneously to: R1, L3, R3, L5, etc.; followed by L2, R2, L4, R4, L6, etc. The crab's progress before and after loss of the left front leg:

  This transformation indicates that the front legs act as 'pacemakers'; this makes it impossible for the animal to adopt the simpler solution of preserving the original pattern minus L1: for in this case R2 would become the second pacemaker and both pacemakers would be on the right side of the animal. The crab's behaviour provides us with a rather elegant example how a motor skill can be adapted to changed conditions while preserving the basic pattern laid down in its code. I have mentioned other examples earlier on -- from the spider's net to the pianist who transposes a tune from one key to another. The experiments to be described presently illustrate the challenging nature of the problem.

  Shuffling the Salamander's Limbs

  Weiss's transplantation experiments date back to the 1920s and proved to be, as one author said, 'of immeasurable positive significance for the appraisal of centro-peripheral co-ordination in nervous function'. [10]

  Weiss grafted fully developed limbs of salamander as super-numeraries on to normal animals which thus had five limbs instead of four. The additional limb was always grafted next to a normal one, and in the process some of the nerve-fibres supplying the normal limb were severed. At first the transplant limb hung inertly from the body as a mere appendage -- the fifth wheel of the cart. However, after a few weeks, it began to give signs of movement, and within a short time it functioned in complete synchronicity with the adjacent normal limb, as its equal in vigour and co-ordination.

  The implications of this phenomenon were described by Weiss as follows (italics in the original):

  As could be incontrovertibly gathered from the microscopical (post-mortem) investigation and reco
nstruction of the course of the nerves in the original limb and in the transplanted limb, this is what took place. The severed nerve fibres had vigorously split up in the scar at the place of grafting. The branches had pressed forward, and some of them had eventually met the degenerated nerve paths of the transplanted limb. As fortuitously as they were located and distributed, they had penetrated into these and so had reached the muscles . . . in the most extraordinary and indiscriminate tangle. . . . Moreover, those few paths belonging to the normal extremity which had also been previously cut (in order to obtain severed nerve stumps capable of regeneration for the supply of the grafted limb) these too were filled with fresh nerves. In the end, therefore, the relatively small number of ganglion cells, which originally led to a small, limited section of the musculature of the normal extremity were now not only connected with this very section of muscle again, but in addition with the entire musculature of the grafted limb. . . . Thus not only have the ganglion cells involved to serve a terminal area several times as large as before; and not only have they to serve muscles altogether different from. the previous ones . . . but above all the previous rule, that one ganglion cell had connections with only one muscle, now becomes the exception. Instead the rule is now a boundless confusion of conduction paths. [11]

  Assuming D0 and E0 to be a pair of antagonistic muscles -- how can they properly function if both now depend on the same nerve supply? And what about At, Bt, Ct? Yet this 'boundless confusion of conduction paths' nevertheless produces perfectly coordinated movement. Weiss concludes that it is not the topographical layout of the pathways which matters, but the specific properties of the excitation transmitted by them; in other words, that although each muscle of the added limb will receive a chaotic medley of excitations, it will respond selectively to such excitations only which are appropriate to it:

  The means by which the central nervous system maintains concord with each muscle individually, does not consist in separate conduction paths. . . . If one and the same nerve cell has to supply excitation to several organs simultaneously, but if under these circumstances only one single route common to all these end-organs is at its disposal . . . then it is logical to assume that the periphery is so constituted that a control of its functioning in a coordinated manner inheres in itself. . . . We require . . . a mechanism of positive selectivity in the end organ, which must explain us why, when two muscles in the same state are given, one of them enters into function and the other does not, although both, being connected with the same nerve cell, receive excitation equally. . . . The nature of every muscle is such that it does not react to every excitation from the centre, but only to excitation of a quite definite form which is characteristic for it. [12]

  To account for the specific selectivity of muscle response, Weiss uses the analogy of selective resonance in a broad sense. The acoustic analysers of the ear each respond to one particular pitch and to one only, thus analysing a complex clang into in harmonic elements. Mutatis mutandis, Weiss assumes that:

  . . . the total impulse flowing towards a particular peripheral region from the central nervous system can, metaphorically speaking, forthwith be designated as an "excitation clang". The "excitation clang" is composed of "excitation tones" for the varying muscles which are to be activated at a given moment, and hence is constantly fluctuating in its composition. . . . The process now is as follows: at the very same time, the same "excitation clang" flows through all the motor root fibres (at least all those supplying a given functional area of considerable extent) towards the periphery. It flows equally through all the fibres as if it had been indiscriminately poured into a canal system and were flooding all the channels. Thus it arrives at all the muscles which are in any way whatever connected with the centre. But when it gets to this point it is analysed. Every muscle, in accordance with its constitution, selects the components appropriate to it from those eventually arriving, and acts as if these components alone had arrived. And thus, although the very same impulse streams to all the muscles and across every available route, only that combination of muscles comes into action -- as is now intelligible -- which the central nervous system has provided for. [13]

  He then proceeds to show that the theory of selective response is not contradicted by the indiscriminate responses of muscles to electro-galvanic stimulation. The latter is an artificial, gross stimulation which compares to natural stimulation like a violent a-periodic blast to a specific clang. 'Just as, both with the clang and with the blast, the substratum carrying the movement is always the same, i.e. the air, so obviously the medium in which both the organized and the unorganized nerve impulse run their course is always the same, i.e. the conductive substance of the nerve fibre. But just as the clang sets a definite selection of resonators vibrating, whereas a noise or blast causes them to resound all at once and without an exception; so also only the organized impulse, built up of specific impulse-tones, is capable of bringing the coordinated selection of muscles into activity, while 'the artificially induced, unorganized impulse, by contrast, forces every muscle whatsoever which it reaches into function.' [14]

  Let me translate the picture that emerges from the experimental evidence into the terms of the present theory. The locomotor matrix on level 4 of the hierarchy (p. 435) is represented by the muscular structure of the limb, plus the 'canal-system' of nervous pathways leading into it, and includes the apparatus -- whatever its nature -- which accounts for the selectivity of the response by enabling muscles to analyse incoming impulses. The code is the sequence of excitation-clangs which calls forth one complete motion -- say, one step of the limb. Members of the matrix are the several joints on the next-lower level No. 3, which are triggered off in a pre-set order by their sub-codes, i.e. by the appropriate components of the excitation-clang. We may remember by way of analogy, how part-sequences of the genetic code are triggered into action in a pro-set order.

  The 'motor unit' at the bottom of the hierarchy responds according to the all-or-nothing rule, but the musculature of a joint is capable of graded responses, and the motions of the whole limb follow a flexible strategy -- shorter or longer step, swift or groping -- dependent on the input from the environment. Weiss accounted for these variations by proposing that the excitation-clang 'is constantly fluctuating in its composition', and thus determines which single muscle should be activated at any given moment. But this conception does not seem to agree well with the basic principle that centres on high levels do not deal directly with units on low levels of the hierarchy. A way out of this difficulty is to be found in suggestions by Ruch (1951) and Miller et al. (1960), according to which pre-set patterns of skilled movements are triggered off as units by the brain; but the signal -- i.e. the 'excitation clang' -- would merely 'rough in' the sequence of movements 'and thus reduce the troublesome transients involved in the correction of movement by output-informed feedbacks'. [15] Since feedback circuits must be assumed to operate on every level, down to the single cell, the adjustment of the details of the 'roughed-in' movement could be handed over to lower levels. Miller et al. have made the further suggestion that this handing-down procedure may be the equivalent of converting an order coded in a 'digital' language, into a graded, 'analogue' output. The excitation-clang could thus consist in a series of 'on', 'off' signals like the dots and dashes of the Morse code; but each sub-unit could respond to its specific 'on' signal by a 'more' or 'less' intense activity, dependent on local conditions (see also below, pp. 569 ff.). These are speculations, and offered by their authors as such; but there are various alternative possibilities to account for the 'filling in' of details which were left open in the generalized excitation pattern, by feedback devices on successively lower levels.

  Limits of Control

  In the experiments previously discussed, a super-numerary limb was grafted next to a normal one, facing in the same direction. In another series of experiments Weiss exchanged and reversed the position of the limbs of newts. The result is again best described in his own words:
>
  The essential independence of the structure of motor activity is dramatically demonstrated when one exchanges and reverses the limbs of animals and then finds them crawling backwards whenever they aim to crawl forwards and vice versa. . . . This has been done in the developed animal, but the same operations have been done in embryos, and these animals have then functioned in reverse from the very beginning. What more spectacular expression can there be of the intrinsic primacy of the motor patterns of behaviour for which the external input acts only as a selective trigger? [16]

  In other transplant experiments, only the position of the two fore-limbs was interchanged and reversed: The grafted limbs moved just as they would have done had they been left in their original position, causing backward motion when the rest of the animal was trying to move forward, and forward motion when the rest of the animal was trying, for example, to avoid a noxious stimulus presented in front of it. A year's experience did not change this reversed movement of the grafted legs. [17]

  The experiment beautifully illustrates the autonomy of the limb-matrix. As in the fifth-limb experiment, the nerves growing out of the stump reached the muscles of the grafted limb in a random manner. Once more, the graft-limb achieved perfectly coordinated motion, thanks (we assume) to the analyser devices in the muscles which respond to one component in the clang signal only. But since the limb was grafted in reverse, it has to step in reverse. No doubt the poor creature senses that something is wrong if its fore-limbs move backward and the back limbs move forward. But owing to the principles Of hierarchic order, the centre which co-ordinates the movements of the animal as a whole -- level 5 in Weiss's schema -- cannot interfere with the functioning of the analyser devices on level 2, to reverse their responses. It apparently cannot even prevent excitation-clangs being triggered off automatically by level 4 to the useless limbs. And when the excitation reaches the latter, they respond as they must. We have here a first, artificially produced example of 'faulty integrations' which will occupy our attention later on.