Janus: A Summing Up

by Arthur Koestler

  Generally speaking, the performance of any purposeful action, whether instinctive, like the nest-building of birds, or acquired as most human skills are, follows the same pattern of spelling out a general intent by the stepwise activation or triggering of functional holons -- sub-routines -- on successively lower levels of the hierarchy. This rule is universally applicable to all types of 'output hierarchies', regardless whether the 'output' is a human baby, a sentence spoken in English, the playing of a piano sonata or the action of tying one's shoelaces. (For input hierarchies, as we shall see later, the reverse sequence holds.)

  7

  The next point to emphasize is that every level in a hierarchy of any type is governed by a set of fixed, invariant rules, which account for the coherence, stability, and the specific structure and function of its constituent holons. Thus in the language hierarchy we found on successive levels the rules which govern the activities of the vocal chords, the laws of grammar and above them the whole semantic hierarchy concerned with meaning. The codes which govern the behaviour of social holons, and lend them coherence, are written and unwritten laws, traditions, belief -- systems, fashions. The development of the embryo is governed by the 'genetic code'. Turning to instinctive activities, the web which the spider weaves, the nest which the blue tit builds, and the courting ceremony of the greylag goose all conform to fixed, species-specific patterns, produced according to certain 'rules of the game'. In symbolic operations, the holons are rule-governed cognitive structures variously called 'frames of reference', 'associative contexts', 'universes of discourse', 'algorithms', etc., each with its specific 'grammar' or canon. We thus arrive at a tentative definition: the term 'holon' may be applied to any structural or functional sub-system in a biological, social or cognitive hierarchy, which manifests rule-governed behaviour and/or structural Gestalt-constancy.* Thus organelles and homologous organs are evolutionary holons; morphogenetic fields are ontogenetic holons; the ethologist's 'fixed action-patterns' and the sub-routines of acquired skills are behavioural holons; phonemes, morphemes, words, phrases are linguistic holons; individuals, families, tribes, nations are social holons. **

  * The 'or' is necessary to include configurations in symbolic hierarchies -- which do not manifest 'behaviour' in the usual sense. ** Various authors have pointed to certain affinities between the concept of the holon and Ralph Gerard's 'org'. Thus D. Wilson in Hierarchical Structures: 'Koestler (1967) elects to designate these "Janus-faced" entities by the term holon . . . We note that Gerard uses the term org to designate the same concept (Gerard, 1957).' This of course amounts to a veiled hint at plagiarism. The two quotations from Gerard that follow indicate the similarities and differences between his org and the holon (my italics): 'Those material systems or entities which are individuals at a given level but are composed of subordinate units, lower level orgs'. [13] The limitation to 'material systems' is made more explicit in the second quotation, where he defines the org as 'that sub-class of systems composed of material systems, in which matter enters into the picture; this excludes formal systems, for example.' [14] Thus the term 'org' cannot be applied to behavioural or linguistic or cognitive hierarchies where the concept of the holon proved especially useful. Orgs, as defined by Gerard, represent a sub-category of holons confined to material systems.

  8

  The set of fixed rules which govern a holon's structure or function we shall call its code or canon. However, let us note at once that while the canon imposes constraints* and controls on the holon's activities, it does not exhaust its degrees of freedom, but leaves room for more or less flexible strategies, guided by the contingencies of the environment. This distinction between fixed (invariant) codes and flexible (variable) strategies may sound at first a little abstract, but it is fundamental to all purposeful behaviour; a few examples will illustrate what is meant.

  * 'Constraint' is a rather unhappy scientific term (reminiscent of the strait-jacket) which refers to the rules which govern organized activity.

  The common spider's web-making activities are controlled by a fixed inherited canon (which prescribes that the radial threads should always bisect the laterals at equal angles, thus forming a regular polygon); but the spider is free to suspend his web from three, four or more points of attachment -- to choose his strategy according to the lie of the land. Other instinctive activities -- birds building nests, bees constructing their hives, silkworms spinning their cocoons -- all have this dual characteristic of conforming to an invariant code or rule -- book which contains the blueprint of the finished product, but using amazingly varied strategies to achieve it.

  Passing from the instinctive activities of the humble spider to sophisticated human skills like playing chess, we again find a code of fixed rules which define the permissible moves, but the choice of the actual move is left to the player, whose strategy is guided by the environment -- the distribution of the chessmen on the board. Speech, as we saw, is governed by various canons on various levels, from semantics through grammar down to phonology, but on each of these levels the speaker has a vast variety of strategic choices: from the selection and ordering of the material to be conveyed, through the formulation of paragraphs and sentences, the choice of metaphors and adjectives, right down to enunciation -- the selective emphasis placed on individual vowels. Similar considerations apply to the pianist improvising variations on a theme. The fixed 'rule of the game' in this case is the given melodic pattern, but he has almost infinite scope for the strategic choices in phasing, rhythm, tempo or transposition into a different key.* A lawyer's activities are very different from a pianist's but the lawyer, too, operates within fixed rules laid down by statute and precedent, while he disposes of a vast range of strategies in interpreting and applying the law.

  * Incidentally, transposition of a musical theme into a different key on the piano, where the sequence of finger movements is totally different, amounts to a complete refutation of the behaviourists' chain-response theory.

  9

  In ontogenesis -- the development of the embryo -- the distinction between 'rules' and 'strategies' is at first sight less obvious, and requires a slightly longer explanation.

  The apex of the hierarchy in this case is the fertilized egg; the axis of the inverted tree represents time: and the holons on successive lower levels represent successive stages in the differentiation of tissues into organs. The growth of the embryo from a shapeless blob to a 'roughed in' form and through various stages of increasing articulation has been compared to the way in which a sculptor carves a figure out of a block of wood -- or, as already mentioned, to the 'spelling out' of an amorphous idea into articulate phonemes.

  The 'idea' to be spelt out in ontogeny is contained in the genetic code, housed in the double helix of nucleic acid strands in the chromosomes. It takes fifty-six generations of cells to produce a human being out of a single, fertilized egg-cell. The cells in the growing embryo are all of identical origin, and carry the same set of chromosomes, i.e., the same hereditary dispositions. In spite of this, they develop into such diverse products as muscle cells, kidney cells, brain cells, toe-nails. How can they do this if they are all governed by the same set of laws, by the same hereditary canon?

  This is a question which, as W. H. Thorpe recently wrote, 'we are not yet within sight of being able to answer'. [15] But at least we can approach it by a rough analogy. Let the chromosomes be represented by the keyboard of a grand piano -- a very grand piano with a few thousand million keys. Then each key will represent a gene or hereditary disposition. Every single cell in the body carries a complete keyboard in its nucleus. But each specialized cell is only permitted to sound one chord or play one tune, according to its speciality -- the rest of the keyboard having been sealed off by scotch-tape.*

  * This sealing-off process also proceeds step-wise, as the hierarchic tree branches out into more and more specialized tissues -- see The Ghost in the Machine, Ch. IX, and below, Part Three.

  But this analogy immediately
poses a further problem: quis custodiet ipsos custodes -- who or what agency decides which keys the cell should activate at what stage and which should be sealed off? It is at this point that the basic distinction between fixed codes and adaptable strategies comes in once again.

  The genetic code, defining the 'rules of the game' of ontogeny, is located in the nucleus of each cell. The nucleus is bounded by a permeable membrane, which separates it from the surrounding cell-body, consisting of a viscous fluid -- the cytoplasm -- and the varied tribes of organelles. The cell-body is enclosed in another permeable membrane, which is surrounded by body-fluids and by other cells, forming a tissue; this, in turn, is in contact with other tissues. In other words, the genetic code in the cell -- nucleus operates within a hierarchy of environments like a nest of Chinese boxes packed into each other.

  Different types of cells (brain cells, kidney cells, etc.) differ from each other in the structure and chemistry of their cell-bodies. These differences are due to the complex interactions between the genetic keyboard in the chromosomes, the cell-body itself, and its external environment. The latter contains physico-chemical factors of such extreme complexity that Waddington coined for it the expression 'epigenetic landscape'. In this landscape the evolving cell moves like an explorer in unknown territory. To quote another geneticist, James Bonner, each embryonic cell must be able to 'test' its neighbour-cells 'for strangeness or similarity, and in many other ways'. [16] The information thus gathered is then transferred -- 'fed back' -- via the cell-body to the chromosomes, and determines which chords on the keyboard should be sounded, and which should be temporarily or permanently sealed off; or, to put it differently, which rules of the game should be applied to obtain the best results. Hence the significant title of Waddington's important book on theoretical biology: The Strategy of the Genes. [17]

  Thus ultimately the cell's future depends on its position in the growing embryo, which determines the strategy of the cell's genes. This has been dramatically confirmed by experimental embryology: by tampering with the spatial structure of the embryo in its early stages of development, the destiny of a whole population of cells could be changed. If at this early stage the future tail of a newt embryo was grafted into a position where a leg should be, it grew not into a tail, but into a leg -- surely an extreme example of a flexible strategy within the rules laid down by the genetic code. At a later stage of differentiation the tissues which form the rudiments of future adult organs -- the 'organ-buds' or 'morphogenetic fields' -- behave like autonomous self-regulating holons in their own right. If at this stage half of the field's tissue is cut away, the remainder will form, not half an organ, but a complete organ. If the growing eye-cup is split into several parts, each fragment will form a smaller, but normal eye.

  There is a significant analogy between the behaviour of embryos at this advanced stage and that very early, blastular stage, when it still resembles a hollow ball of cells. When half the blastula of a frog is amputated, the remainder will develop not into half a frog but a smaller normal frog; and if a human blastula is split by accident, the result will be twins or even quadruplets. Thus the holons which at that earliest stage behave as parts of the potentially whole organism manifest the same self-regulating characteristics as the holons which at a lower (later) level of the developmental hierarchy are parts of a potential organ; in both cases (and throughout the intermediary stages) the holons obey the rules laid down in their genetic code but retain sufficient freedom to follow one or another developmental pathway, guided by the contingencies of their environment.

  These self-regulating properties of holons within the growing embryo ensure that whatever accidental hazards arise during development, the end-product will be according to norm. In view of the millions and millions of cells which divide, differentiate, and move about, it must be assumed that no two embryos, not even identical twins, are formed in exactly the same way. The self-regulating mechanisms which correct deviations from the norm and guarantee, so to speak, the end-result, have been compared to the homeostatic feedback devices in the adult organism -- so biologists speak of 'developmental homeostasis'. The future individual is potentially predetermined in the chromosomes of the fertilized egg; but to translate this blueprint into the finished product, billions of specialized cells have to be fabricated and moulded into an integrated structure. It would be absurd to assume that the genes of that one fertilized egg should contain built-in provisions for each and every particular contingency which every single one of its fifty-six generations of daughter-cells might encounter in the process. However, the problem becomes a little less baffling if we replace the concept of the 'genetic blueprint', which implies a plan to be rigidly copied, by the concept of a genetic canon of rules which are fixed, but leave room for alternative choices, i.e., adaptive strategies guided by feedbacks and pointers from the environment.

  Needham once coined a phrase about 'the striving of the blastula to grow into a chicken'. One might call the strategies by which it succeeds the organism's 'pre-natal skills'. After all, the development of the embryo and the subsequent maturation of the newborn into an adult are continuous processes; and we must expect that pre-natal and post-natal skills have certain basic principles in common with each other and with other types of hierarchic processes.

  The foregoing section was not intended to describe embryonic development, only one aspect of it: the combination of fixed rules and variable strategies, which we also found in instinctive skills (such as nest-building, etc.) and learnt behaviour (such as language, etc.). It seems that life in all its manifestations, from morphogenesis to symbolic thought, is governed by rules of the game which lend it order and stability but also allow for flexibility; and that these rules, whether innate or acquired, are represented in coded form on various levels of the hierarchy, from the genetic code to the structures in the nervous system associated with symbolic thought.

  10

  Ontogeny and phylogeny, the development of the individual and the evolution of species, are the two grand hierarchies of becoming. Phylogeny will be discussed in Part Three, but an anticipatory remark is required in our present context of 'rules and strategies'.

  Motor-car manufacturers take it for granted that it would make no sense to design a new model from scratch; they make use of already existing sub-assemblies -- engines, batteries, steering systems, etc. -- each of which has been developed from long previous experience, and then proceed by small modifications of some of these. Evolution follows a similar strategy. Compare the front wheels of the latest model with those of an old vintage car or horse-cart -- they are based on the same principles. Compare the anatomy of the fore-limbs of reptiles, birds, whales and man -- they show the same structural design of bones, muscles, nerves and blood-vessels and are accordingly called 'homologous' organs.

  The functions of legs, wings, flippers and arms are so different that one would expect them to have quite different designs. Yet they are merely modifications, strategic adaptations of an already existing structure -- the forelimb of the common reptilian ancestor. Once Nature has taken out a patent on a vital component or process, she sticks to it with amazing tenacity: the organ or device has become a stable evolutionary holon. It is as if she felt compelled to provide unity in variety. Geoffroy de St Hilaire, one of the pioneers of modern biology, wrote in 1818: 'Vertebrates are built upon one uniform plan -- e.g., the forelimb may be modified for running, climbing, swimming, or flying, yet the arrangement of the bones remains the same.' [18] That basic arrangement is part of the invariant evolutionary canon. Its utilization for swimming or flying is a matter of evolutionary strategy.

  This principle holds all along the line, through all the levels of the evolutionary hierarchy down to the organelles inside the cell, and the DNA chains in the chromosomes. The same standard models of organelles function in the cells of mice and men; the same ratchet-device using a contractile protein serves the motion of amoeba and of the concert-pianist's fingers; the same four chemical mo
lecules constitute the basic alphabet in which heredity is encoded throughout the animal and plant kingdoms -- only the words and phrases formed by them are different for each creature.

  If evolution could only create novelties by starting each time afresh from the 'primeval soup', the four thousand million years of the earth's history would not have been long enough to produce even an amoeba. In a much quoted paper on hierarchic structures, H. G. Simon concluded: 'Complex systems will evolve from simple systems much more rapidly if there are stable intermediate forms than if there are not. The resulting complex forms in the former case will be hierarchic. We have only to turn the argument around to explain the observed predominance of hierarchies among the complex systems Nature presents to us. Among possible complex forms, hierarchies are the ones that have the time to evolve.' [19]

  We do not know what forms of life exist on other planets, but we can safely assume that wherever there is life, it is hierarchically organized.

  11

  Neglect of the hierarchic concept, and the failure to make a categorical distinction between rules and strategies of behaviour, has caused much confusion in academic psychology.* Since its primary concern for the last fifty years was the study of rats in confined spaces ('Skinner boxes'), this failure is hardly surprising. Yet to any spectator of a game of football or chess it is at once obvious that each player obeys rules which determine what he can do, and uses his strategic skills to decide what he will do. In other words, the code defines the rules of the game, strategy decides the course of the game. The examples cited in the previous section indicate that this categorical distinction between rules and strategies is universally applicable to innate and acquired skills, to the hierarchies which make for social coherence, as well as to the hierarchies of becoming.