AUTONOMOUS HOLONS
The evolutionary stability of these sub-assemblies -- organelles, organs, organ-systems -- is reflected by their remarkable degree of autonomy or self-government. Each of them -- a piece of tissue or a whole heart -- is capable of functioning in vitro as a quasi-independent whole, even though isolated from the organism or transplanted into another organism. Each is a sub-whole which, towards its subordinated parts, behaves as a self-contained whole, and towards its superior controls as a dependent part. This relativity of the terms 'part' and 'whole' when applied to any of its sub-assemblies is a further general characteristic of hierarchies.
It is again the very obviousness of this feature which tends to make us overlook its implications. A part, as we generally use the word, means something fragmentary and incomplete, which by itself would have no legitimate existence. On the other hand, there is a tendency among holists to use the word 'whole' or 'Gestalt' as something complete in itself which needs no further explanation. But wholes and parts in this absolute sense do not exist anywhere, either in the domain of living organisms or of social organizations. What we find are intermediary structures on a series of levels in ascending order of complexity, each of which has two faces looking in opposite directions: the face turned toward the lower levels is that of an autonomous whole, the one turned upward that of a dependent part. I have elsewhere [9] proposed the word 'holon' for these Janus-faced sub-assemblies.
The concept of the holon is meant to supply the missing link between atomism and holism, and to supplant the dualistic way of thinking in terms of 'parts' and 'wholes', which is so deeply engrained in our mental habits, by a multi-level, stratified approach. A hierarchically-organized whole cannot be 'reduced' to its elementary parts; but it can be 'dissected' into its constituent branches of holons, represented by the nodes of the tree-diagram, while the lines connecting the holons stand for channels of communication, control or transportation, as the case may be.
FIXED RULES AND FLEXIBLE STRATEGIES
The term holon may be applied to any stable sub-whole in an organismic, cognitive, or social hierarchy which displays rule-governed behaviour and/or structural Gestalt constancy. Thus biological holons are self-regulating open systems' [10] governed by a set of fixed rules which account for the holon's coherence, stability and its specific pattern of structure and function. This set of rules we may call the canon of the holon.* The canon determines the fixed, invariant aspect of the open system in its steady state (Fliessgleichgewicht -- dynamic equilibrium); it defines its pattern and structure. In other types of hierarchies, the canon represents the codes of conduct of social holons (family, tribe, nation, etc.); it incorporates the 'rules of the game' of instinctive rituals or acquired skills (behavioural holons); the rules of enunciation, grammar and syntax in the language hierarchy; Piaget's 'schemes' in cognitive hierarchies, and so on. The canon represents the constraints imposed on any rule-governed process or behaviour. But these constraints do not exhaust the system's degrees of freedom; they leave room for more or less flexible strategies, guided by the contingencies in the holon's local environment.
* Cf. the 'organizing relations' or 'laws of organization' of earlier
writers on hierarchic organization (e.g., Woodger (1929), Needham
(1941), and the 'system-conditions' in general system theory.
It is essential at this point to make a sharp, categorical distinction between the fixed, invariant canon of the system and its flexible (plastic, variable) strategies. A few examples will illustrate the validity of this distinction. In ontogeny, the apex of the hierarchy is the zygote, and the holons at successive levels represent successive stages in the development of tissues. Each step in differentiation and specialization imposes further constraints on the genetic potential of the tissue, but at each step it retains sufficient developmental flexibility to follow this or that evolutionary pathway, within the range of its competence. guided by the contingencies of the cell's environment -- Waddington's [11] 'strategy of the genes'. Turning from embryonic development to the instinctive activities of the mature animal, we find that spiders spin webs, birds build nests according to invariant species-specific canons, but again using flexible strategies, guided by the lie of the land: the spider may suspend its web from three, four or more points of attachment, but the result will always be a regular polygon. In acquired skills like chess, the rules of the game define the permissible moves, but the strategic choice of the actual move depends on the environment -- the distribution of the chessmen on the board. In symbolic operations, the holons are rule-governed cognitive structures variously called 'frames of reference', 'universes of discourse', 'algorithms', etc., each with its specific 'grammar or canon; and the strategies increase in complexity on higher levels of each hierarchy. 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 responsible for symbolic thought.
TRIGGERS AND SCANNERS
Let me discuss briefly some specific characteristics of what one might loosely call output hierarchies, regardless whether the 'output' is a baby, or a sentence spoken in English. However much their products differ, all output hierarchies seem to have a classic mode of operation, based on the trigger-releaser principle, where an implicit coded signal which may be relatively simple, releases complex, pre-set mechanisms.
Let me again run through a few examples. In phylogeny, Waddington [12] and others have convincingly shown that a single favourable gene-mutation can act as a trigger to release a kind of chain-reaction which affects a whole organ in a harmonious way. In ontogeny, the prick of a fine platinum needle on the unfertilized egg of a frog or sheep triggers off parthenogenesis. The genes act as chemical triggers, catalysing reactions. The implicit four-letter alphabet of the DNA chain is spelled out into the explicit, twenty-letter alphabet of amino-acids; the inducers or evocators, including Spemann's 'general organizer', again turn out to be relatively simple chemicals which need not even be species-specific to activate the genetic potentials of the tissue. In instinct behaviour, we have releasers of a very simple kind -- the red belly of the stickleback, the spot under the herring-gull's beak, which trigger off the appropriate behaviour. [13] In the performance of acquired skills you have the same process of step-wise filling in the details of implicit commands issued from the apex of the hierarchy, such as 'strike a match and light this cigarette' or 'sign your name', or 'use your phrase-generating machine' to transform an unverbalized image into innervations of the vocal chords.
The point to emphasize is that this spelling-out process, from intent to execution, cannot be described in terms of a linear chain of S-R units, only as a series of discrete steps from one open sesame, activated by a combination lock, to the next. The activated holon, whether it is a government department or a living kidney, has its own canon which determines the pattern of its activity. Thus the signal from higher quarters does not have to specify what the holon is expected to do; the signal merely has to trigger the holon into action by a coded message. Once thrown into action, the holon will spell out the implicit command in explicit form by activating its sub-units in the appropriate strategic order, guided by feedbacks and feed-forwards from its environment. Generally speaking, the holon is a system of relations which is represented on the next higher level as a unit, that is, a relatum.
If we turn now to the input hierarchies of perception, the operations proceed, of course, in the reverse direction, from the peripheral twigs of the tree towards its apex; and instead of trigger-releasers we have the opposite type of mechanisms: a series of filters, scanners or classifiers through which the input traffic must pass in its ascent from periphery to cortex. First you have lateral inhibition, habituation and presumably some efferent control of recep
tors. On the higher levels are the mechanisms responsible for the visual and acoustic constancy phenomena, the scanning and filtering devices which account for the recognition of patterns in space and time, and enable us to abstract universals and discard particulars. The colloquial complaint: 'I have a memory like a sieve' may be derived from an intuitive grasp of these filtering devices that operate first all along the input channels, then along the storage channels.
How do we pick out a single instrument in a symphony? The whole medley of sounds arriving at the ear-drum is scrambled into a linear pressure-wave with a single variable. To reconstruct the timbre of an instrument, to identify harmonies and melodies, to appreciate phrasing, style and mood, we have to abstract patterns in time as we abstract visual patterns in space. But how does the nervous system do it? I will play you the opening bars of the Archduke Trio; watch your reactions, because no text-book on psychology that I know of will give you the faintest clue [opening bars of Beethoven's Archduke Trio played]. If one looks at the record with a magnifying glass, one is tempted to ask the naive question why the nervous system does not produce engrams by this simple method of coding, instead of being so damned complicated. The answer is, of course, that a linear engram of this kind would be completely useless for the purpose of analysing, matching and recognizing input patterns. The chain is a hopeless model; we cannot do without the tree.
«serial vs random access»
In motor hierarchies, an implicit intention or generalized command is particularized, spelled out, step by step, in its descent to the periphery. In perceptual hierarchies, we have the opposite process. The peripheral input is more and more de-particularized, stripped of irrelevancies during its ascent to the centre. The output hierarchy concretizes, the input hierarchy abstracts. The former operates by means of triggering devices, the latter by means of filtering or scanning devices. When I intend to write the letter R, a trigger activates a functional holon, an automatic pattern of muscle contractions, which produces the letter R in my own particular hand-writing. When I read, a scanning device in my visual cortex identifies the letter R regardless of the particular hand that wrote it. Triggers release complex outputs by means of a simple coded signal. Scanners function the opposite way: they convert complex inputs into simple coded signals.
'ABSTRACT' AND 'SPOTLIGHT'
Let me briefly turn to the phenomena of memory and ask whether the hierarchic approach is capable of shedding some additional light on them. You watch a television play. The exact words of each actor are forgotten by the time he speaks his next line, and only their meaning remains; the next morning you can only remember the sequence of scenes which constituted the story; after a month, all you remember is that it was about a gangster on the run or about two men and a woman on a desert island. The same happens generally with the content of novels one has read, and episodes one has lived. The original experience has been stripped of detail, reduced to a schematic outline. Now this skeletonization of the input before it is put into storage, and the gradual decay of the stored material, would mean a terrible impoverishment of memory, if this were the whole story -- memory would be a collection of dusty abstracts, the dehydrated sediments in the wine-glass, with all flavour gone. But there are compensating mechanisms. I can recognize a melody, regardless of the instrument on which it is played, and I can recognize the timbre of an instrument, regardless of the melody played on it. There are several interlocking hierarchies at work, each with its own criteria of relevance. One abstracts melody and treats everything else as noise, the other abstracts timbre and treats melody as noise. Thus not all the information discarded as irrelevant by one filtering system is irretrievably lost, because it may have been stored by another filtering hierarchy with different canons of relevance. Recall would then be made possible by the cooperation of several interlocking hierarchies, which may pertain to different sense modalities -- sight and smell, for instance; or, what is less obvious, there may also be several distinct hierarchies with different criteria or relevance operating within the same sense modality. Recall could then be compared to the process of multi-coloured printing by the superimposition of several colour-blocks. This, of course, is speculative, but some modest evidence for the hypothesis can be found in a series of experiments by J. J. Jenkins and myself; * [14] and more tests on these lines can be designed without much difficulty.
* See Appendix II.
I am aware that the hypothesis is in apparent contradiction to Penfield's [15] experiments eliciting what looks like total recall of past experiences by electrical stimulation of points on the patient's temporal lobe. But the contradiction may be resolved if we include in the criteria of relevance also criteria of emotional relevance which decide whether an input is worth storing. A detail might be emotionally relevant (on a conscious or unconscious level), and retained with almost photographic or cinematographic clarity. One might call this the spotlight type of memory which is stamped in, as distinct from abstractive memory which schematizes. Spotlight memories may be related to eidetic images; and they might even, unlike abstractive memories, originate in the limbic system. [16]
ARBORIZATION AND RETICULATION
I have used the term 'interlocking' or 'interlacing' hierarchies. Of course hierarchies do not operate in a vacuum. This truism regarding the interdependence of processes in an organism is probably the main cause of confusion which obscured from view its hierarchic structure. It is as if the sight of the foliage of the entwined branches in a forest made us forget that the branches originate in separate trees. The trees are vertical structures. The meeting points of branches from neighbouring trees form horizontal networks at several levels. Without the trees there could be no entwining, and no network. Without the networks, each tree would be isolated, and there would be no integration of functions. Arborization and reticulation seem to be complementary principles in the architecture of organisms. In symbolic universes of discourse arborization is reflected in the 'vertical' denotation (definition) of concepts, reticulation in their 'horizontal' connotations in associative networks. This calls to mind Hyden's proposal that the same neuron, or population of neurons, may be a member of several functional 'clubs'.
HIERARCHIC ORDER AND FEEDBACK CONTROL
The most obvious example of interlocking hierarchies is the sensory-motor system. The sensory hierarchy processes information and transmits it in a steady upward flow, some of which reaches the conscious ego at the apex; the ego makes decisions which are spelt out by the downward stream of impulses in the motor hierarchy. But the apex is not the only point of contact between the two systems; they are connected by entwining networks on various lower levels. The network on the lowest level consists of reflexes like the patellary. They are short-cuts between the ascending and descending flow, like loops connecting opposite traffic streams on a motor highway. On the next higher level are the networks of sensory-motor skills and habits, such as touch-typing or driving a car, which do not require the attention of the highest centres -- unless some disturbance throws them out of gear. But let a little dog amble across the icy road in front of the driver, and he will have to make a 'top level' decision whether to slam on the brake, risking the safety of his passengers, or run over the dog. It is at this level, when the pros and cons are precariously balanced that the subjective experience of free choice and moral responsibility arises.
But the ordinary routines of existence do not require such moral decisions, and not even much conscious attention. They operate by means of feedback loops, and loops-within-loops, which form the multilevelled, reticulate networks between the input and output hierarchies. So long as all goes well and no dog crosses the road, the strategy of riding a bicycle or driving a car can be left to the automatic pilot in the nervous system -- the cybernetic helmsman. But one must beware of using the principle of feedback control as a magic formula. The concept of feedback without the concept of hierarchic order is like the grin without the cat. All skilled routines follow a pre-set pattern according to cer
tain rules of the game. These are fixed, but permit continual adjustments to variable environmental conditions. Feedback can only operate within the limits set by the rules -- by the canon of the skill. The part which feedback plays is to report back on every step in the progress of the operation, whether it is over-shooting or falling short of the mark, how to keep it on an even keel, when to intensify the pace and when to stop. But it cannot alter the intrinsic pattern of the skill. To quote Paul Weiss [17] at the Hixon Symposium:
The structure of the input does not produce the structure of the
output, but merely modifies intrinsic nervous activities, which have
a structural organization of their own.
One of the vital differences between the S-R and SOHO concepts is that according to the former, the environment determines behaviour, whereas according to the latter, feedback from the environment merely guides or corrects or stabilizes pre-existing patterns of behaviour.
Moreover, the cross-traffic between the sensory and motor hierarchies works both ways. The input guides the output and keeps it on an even keel; but motor activity in its turn guides perception. The eye must scan; its motions, large and small -- drift, flicker, tremor -- are indispensable to vision; an image stabilized on the retina disintegrates into darkness. [18] Similarly with audition: if you try to recall a tune, what do you do? You hum it. Stimuli and responses have been swallowed up by feedback loops within loops, along which impulses run in circles like kittens chasing their tails.
Janus Page 34