|
STEPS TO AN ECOLOGY OF MIND: COLLECTED ESSAYS IN ANTHROPOLOGY, PSYCHIATRY, EVOLUTION AND EPISTEMOLOGY |
|
Part IV: Biology and Evolution On Empty-Headedness Among Biologists and State Boards of Education [1] My father, the geneticist William Bateson, used to read us passages of the Bible at breakfast—lest we grow up to be empty-headed atheists; and so I find it natural to wonder what broadening of the mind may come from the strange anti-evolutionary ruling of the State Board of Education in California. [2] Evolution has long been badly taught. In particular, students—and even professional biologists—acquire theories of evolution without any deep understanding of what problem these theories attempt to solve. They learn but little of the evolution of evolutionary theory. The extraordinary achievement of the writers of the first chapter of Genesis was their perception of the problem: Where does order come from? They observed that the land and the water were, in fact, separate and that species were separate; they saw that such separation and sorting in the universe presented a fundamental problem. In modern terms, we may say that this is the problem implicit in the Second Law of Thermodynamics: If random events lead to things getting mixed up, by what nonrandom events did things come to be sorted? And what is a “random” event? This problem has been central to biology and to many other sciences for the last 5000 years, and the problem is not trivial. With what Word should we designate the principle of order which seems to be immanent in the universe? The California ruling suggests that students be told of other attempts to solve this ancient problem. I myself collected one of these among the Stone Age headhunters of the Iatmul tribe in New Guinea. They, too, note that the land and the water are separate even in their swampy region. They say that in the beginning there was a vast crocodile, Kavwokmali, who paddled with his front legs and paddled with his back legs, and thereby kept the mud in suspension. The culture hero, Kevembuangga, speared the crocodile, who then ceased to paddle, causing the mud and the water to separate. The result was dry land upon which Kevembuangga stamped his foot in triumph. We might say he verified that “it was good.” Our students might have their minds broadened somewhat if they would look at other theories of evolution and consider how a man’s spirit must take a different shape if he believes that all sorting in the universe is due to an external agent, or if, like the Iatmul and modern scientists, he sees that the potentiality for order and pattern is immanent throughout this world. And then the student may be forced by the new system to look at the “Great Chain of Being,” with Supreme Mind at the top and the protozoa at the bottom. He will see how Mind was invoked as an explanatory principle all through the Middle Ages and how Mind later became the problem. Mind became that which needed explanation when Lamarck showed that the Great Chain of Being should be inverted to give an evolutionary sequence from the protozoa upward. The problem then was to explain Mind in terms of what could be known of this sequence. And when the student reaches the mid-nineteenth century, he might be given as a textbook Philip Henry Gosse’s Creation (Omphalos): An Attempt to Untie the Geological Knot. He will learn from this extraordinary book things about the structure of animals and plants which are today scarcely mentioned in many courses of biology; notably, that all animals and plants show a time structure, of which the rings of growth in trees are an elementary example and the cycles of life history, a more complex one. Every plant and animal is constructed upon the premise of its cyclic nature. After all, there can be no harm in Gosse, who was a devout fundamentalist—a Plymouth Brother—as well as a distinguished marine biologist. His book was published in 1857, two years before the Origin a f Species. He wrote it to show that the facts of the fossil record as well as those of biological homology could be made to fit with the principles of fundamentalism. It was to him inconceivable that God could have created a world in which Adam had no navel; the trees in the Garden of Eden, no rings of growth; and the rocks, no strata. Therefore, God must have created the world as though it had a past. It will do the student no harm to wrestle with the paradoxes of Gosse’s “Law of Prochronism”; if he listens care-fully to Gosse’s groping generalizations about the biological world, he will hear an early version of the “steady state” hypothesis. Of course, everybody knows that biological phenomena are cyclic-from egg, to hen, to egg, to hen, etc. But not all biologists have examined the implications of this cyclic characteristic for evolutionary and ecological theory. Gosse’s view of the biological world might broaden their minds. It is silly and vulgar to approach the rich spectrum of evolutionary thought with questions only about who was right and who was wrong. We might as well assert that the amphibia and reptiles were “wrong” and the mammals and birds “right” in their solutions to the problems of how to live. By fighting the fundamentalists, we are led into an empty-headedness analogous to theirs. The truth of the matter is that “Other men have laboured and ye are entered into their labours” (John 1:38), and this text is not only a reminder of the need for humility, it is also an epitome of the vast evolutionary process into which we organisms are willy-nilly entered. The Role of Somatic Change in Evolution [3] All theories of biological evolution depend upon at least three sorts of change: (a) change of genotype, either by mutation or by redistribution of genes; (b) somatic change under pressure of environment; and (c) changes in environmental conditions. The problem for the evolutionist is to build a theory combining these types of change into an ongoing process which, under natural selection, will account for the phenomena of adaptation and phylogeny. Certain conventional premises may be selected to govern such theory building: (a) The theory shall not depend upon Lamarckian inheritance. August Weismann’s argument for this premise still stands. There is no reason to believe that either somatic change or changes in environment can, in principle, call (by physiological communication) for appropriate genotypic change. Indeed, the little that we know about communication within the multicellular [4] individual indicates that such communication from soma to gene script is likely to be rare and unlikely to be adaptive in effect. However, it is appropriate to attempt to spell out in this essay what this premise implies: Whenever some characteristic of an organism is modifiable under measurable environmental impact or under measurable impact of internal physiology, it is possible to write an equation in which the value of the characteristic in question is expressed as some function of the value of the impacting circumstance. “Human skin color is some function of exposure to sunlight,” “respiration rate is some function of atmospheric pressure,” etc. Such equations are constructed to be true for a variety of particular observations, and necessarily contain subsidiary propositions which are stable (i.e., continue to be true) over a wide range of values of impacting circumstance and somatic characteristic. These subsidiary propositions are of different logical type from the original observations in the laboratory and are, in fact, descriptive not of the data but of our equations. They are statements about the form of the particular equation and about the values of the parameters mentioned within it. It would be simple, at this point, to draw the line between genotype and phenotype by saying that the forms and parameters of such equations are provided by genes, while the impacts of environment, etc. determine the actual event within this frame. This would amount to saying, e.g., that the ability to tan is genotypically determined, while the amount of tanning in a particular case depends upon exposure to sun-light. In terms of this oversimplified approach to the overlapping roles of genotype and environment, the proposition excluding Lamarckian inheritance would read somewhat as follows: In the attempt to explain evolutionary process, there shall be no assumption that the achievement of a particular value of some variable under particular circumstances will affect, in the gametes produced by that individual, the form or parameters of the functional equation governing the relationship between that variable and its environmental circumstances. Such a view is oversimplified, and parentheses must be added to deal with more complex and extreme cases. First, it is important to recognize that the organism, considered as a communicational system, may itself operate at multiple levels of logical typing; i.e., that there will be instances in which what were above called “parameters” are subject to change. The individual organism might as a result of “training” change its ability to develop a tan under sunlight. And this type of change is certainly of very great importance in the field of animal behavior, where “learning to learn” can never be ignored. Second, the oversimplified view must be elaborated to cover negative effects. An environmental circumstance may have such impact upon an organism unable to adapt to it, that the individual in question will in fact produce no gametes. Third, it is expectable that some of the parameters in one equation may be subject to change under impact from some environmental or physiologic circumstance other than the circumstance mentioned in that equation. Be all that as it may, both Weismann’s objection to Lamarckian theory and my own attempt to spell the matter out share a certain parsimony: an assumption that the principles which order phenomena shall not themselves be supposed changed by those phenomena which they order. William of Occam’s razor might be reformulated: in any explanation, logical types shall not be multiplied beyond necessity. (b) Somatic change is absolutely necessary for survival. Any change of environment which requires adaptive change in the species will be lethal unless, by somatic change, the organisms (or some of them) are able to weather out a period of unpredictable duration, until either appropriate genotypic change occurs (whether by mutation or by redistribution of genes already available in the population), or because the environment returns to the previous normal. The premise is truistical, regardless of the magnitude of the time span involved. (c) Somatic change is also necessary to cope with any changes of genotype which might aid the organism in its external struggle with the environment. The individual organism is a complex organization of interdependent parts. A mutational or other genotypic change in any one of these (however externally valuable in terms of survival) is certain to require change in many others—which changes will probably not be specified or implicit in the single mutational change of the genes. A hypothetical pregiraffe, which had the luck to carry a mutant gene “long neck,” would have to adjust to this change by complex modifications of the heart and circulatory system. These collateral adjustments would have to be achieved at the somatic level. Only those pregiraffes which are (genotypically) capable of these somatic modifications would survive. (d) In this essay, it is assumed that the corpus of genotypic messages is preponderantly digital in nature. In contrast, the soma is seen as a working system in which the genotypic recipes are tried out. Should it transpire that the genotypic corpus is also in some degree analogic—a working model of the soma—premise c (above) would be negated to that degree. It would then be conceivable.that the mutant gene “long neck” might modify the message of those genes which affect the development of the heart. It is, of course, known that genes may have pleiotropic effect, but these phenomena are relevant in the present connection only if it can be shown, e.g., that the effect of gene A upon the phenotype and its effect upon the phenotypic expression of gene B are mutually appropriate in the overall integration and adaptation of the organism. These considerations lead to a classifying of both genotypic and environmental changes in terms of the price which they exact of the flexibility of the somatic system. A lethal change in either environment or genotype is simply one which demands somatic modifications which the organism cannot achieve. But the somatic price of a given change must depend, not absolutely upon the change in question, but upon the range of somatic flexibility available to the organism at the given time. This range, in turn, will depend upon how much of the organism’s somatic flexibility is already being used up in adjusting to other mutations or environmental changes. We face an economics of flexibility which, like any other economics, will become determinative for the course of evolution if and only if the organism is operating close to the limits set by this economics. However, this economics of somatic flexibility will differ in one important respect from the more familiar economics of money or available energy. In these latter, each new expenditure can simply be added to the preceding expenditures and the economics becomes coercive when the additive total approaches the limit of the budget. In contrast, the combined effect of multiple changes, each of which exacts a price in the soma, will be multiplicative. This point may be stated as follows: Let S be the finite set of all possible living states of the organism. Within S, let s1 be the smaller set of all states compatible with a given mutation (ml), and let s2 be the set of states compatible with a second mutation (m2). It follows that the two mutations in combination will limit the organism to the logical product of s1 and s2, i.e., to that usually smaller subset of states which is composed only of members common to both s1 and s2. In this way each successive mutation (or other genotypic change) will fractionate the possibilities for the somatic adjustment of the organism. And, should the one mutation require some somatic change, the exact opposite of a change required by the other, the possibilities for somatic adjustment may immediately be reduced to zero. The same argument must surely apply to multiple environmental changes which demand somatic adjustments; and this will be true even of those changes in environment which might seem to benefit the organism. An improvement in diet, for example, will exclude from the organism’s range of somatic adjustments those patterns of growth which we would call “stunted” and which might be required to meet some other exigency of the environment. From these considerations it follows that if evolution proceeded in accordance with conventional theory, its process would be blocked. The finite nature of somatic change indicates that no ongoing process of evolution can result only from successive externally adaptive genotypic changes since these must, in combination, become lethal, demanding combinations of internal somatic adjustments of which the soma is incapable. We turn therefore to a consideration of other classes of genotypic change. What is required to give a balanced “theory of evolution is the occurrence of genotypic changes which shall increase the available range of somatic flexibility. When the internal organization of the organisms of a species has been limited by environmental or mutational pressure to some narrow subset of the total range of living states, further evolutionary progress will require some sort of genotypic change which will compensate for this limitation. We note first that while the results of genotypic change are irreversible within the life of the individual organism, the opposite is usually true of changes which are achieved at the somatic level. When the latter are produced in response to special environmental conditions, a return of the environment to the previous norm is usually followed by a diminution or loss of the characteristic. (We may reasonably expect that the same would be true of those somatic adjustments which must accompany an externally adaptive mutation but, of course, it is impossible in this case to remove from the individual the impact of the mutational change.) A further point regarding these reversible somatic changes is of special interest. Among higher organisms it is not unusual to find that there is what we may call a “defense in depth” against environmental demands. If a man is moved from sea level to 10,000 feet, he may begin to pant and his heart may race. But these first changes are swiftly reversible: if he descends the same day, they will disappear immediately. If, however, he remains at the high altitude, a second line of defense appears. He will become slowly acclimated as a result of complex physiological changes. His heart will cease to race, and he will no longer pant unless he undertakes some special exertion. If now he returns to sea level, the characteristics of the second line of defense will disappear rather slowly and he may even experience some discomfort. From the point of view of an economics of somatic flexibility, the first effect of high altitude is to reduce the organism to a limited set of states (si) characterized by the racing of the heart and the panting. The man can still survive, but only as a comparatively inflexible creature. The later acclimation has precisely this value: it corrects for the loss of flexibility. After the man is acclimated he can use his panting mechanisms to adjust to other emergencies which might otherwise be lethal. A similar “defense in depth” is clearly recognizable in the field of behavior. When we encounter a new problem for the first time, we deal with it either by trial and error or possibly by insight. Later, and more or less gradually, we form the “habit” of acting in the way which earlier experience rewarded. To continue to use insight or trial and error upon this class of problem would be wasteful. These mechanisms can now be saved for other problems. [5] Both in acclimation and in habit formation the economy of flexibility is achieved by substituting a deeper and more enduring change for a more superficial and more reversible one. In the terms used above in discussing the anti-Lamarckian premise, a change has occurred in the parameters of the functional equation linking rate of respiration to external atmospheric pressure. Here it seems that the organism is behaving as we may expect any ultrastable system to behave. Ashby [6] has shown that it is a general formal characteristic of such systems that those circuits con-trolling the more rapidly fluctuating variables act as balancing mechanisms to protect the ongoing constancy of those variables in which change is normally slow and of small amplitude; and that any interference which fixes the values of the changeful variables must have a disturbing effect upon the constancy of the normally steady components of the system. For the man who must constantly pant at high altitudes, the respiration rate can no longer be used as a changeable quantity in the maintaining of physiological balance. Conversely, if the respiration rate is to become avail-able again as a rapidly fluctuating variable, some change must occur among the more stable components of the system. Such a change will, in the nature of the case, be achieved comparatively slowly and be comparatively irreversible. Even acclimation and habit formation are, however, still reversible within the life of the individual, and this very reversibility indicates a lack of communicational economy in these adaptive mechanisms. Reversibility implies that the changed value of some variable is achieved by means of homeostatic, error-activated circuits. There must be a means of detecting an undesirable or threatening change in some variable, and there must be a train of cause and effect whereby corrective action is initiated. Moreover, this entire circuit must, in some degree, be available for this purpose for the entire time during which the reversible change is maintained—a considerable using up of available message pathways. The matter of communicational economics becomes still more serious when we note that the homeostatic circuits of an organism are not separate but complexly interlocked, e.g., hormonal messengers which play a part in the homeostatic control of organ A will also affect the states of organs B, C, and D. Any special ongoing loading of the circuit controlling A will therefore diminish the organism’s freedom to control B, C, and D. In contrast, the changes brought about by mutation or other genotypic change are presumably of a totally different nature. Every cell contains a copy of the new genotypic corpus and therefore will (when appropriate) behave in the changed manner, without any change in the messages which it receives from surrounding tissues or organs. If the hypothetical pregiraffes carrying the mutant gene “long neck” could also get the gene “big heart,” their hearts would be enlarged without the necessity of using the homeostatic pathways of the body to achieve and maintain this enlargement. Such a mutation will have survival value not be-cause it enables the pregiraffe to supply its elevated head with sufficient blood, since this was already achieved by somatic change but because it increases the overall flexibility of the organism, enabling it to survive other demands which may be placed upon it either by environmental or, genotypic change. It appears, then, that the process of biological evolution could be continuous if there were a class of mutations or other genotypic changes which would simulate Lamarckian inheritance. The function of these changes would be to achieve by genotypic flat those characteristics which the organism at the given time is already achieving by the uneconomical method of somatic change. Such a hypothesis, I believe, conflicts in no way with conventional theories of genetics and natural selection. It does, however, somewhat alter the current conventional picture of evolution as a whole, though related ideas were put forward over sixty years ago. Baldwin [7] suggested that we consider not only the operation of the external environment in natural selection but also what he called “organic selection” in which the fate of a given variation would depend upon its physiologic viability. In the same article, Baldwin attributes to Lloyd Morgan the suggestion that there might exist “coincident variations” which would simulate Lamarckian inheritance (the so-called “Baldwin effect”). According to such a hypothesis, genotypic change in an organism becomes comparable to legislative change in a society. The wise legislator will only rarely initiate a new rule of behavior; more usually he will confine himself to affirming in law that which has already become the custom of the people. An innovative rule can be introduced only at the price of activating and perhaps overloading a large number of homeostatic circuits in the society. It is interesting to ask how a hypothetical process of evolution would work if Lamarckian inheritance were the rule, i.e., if characteristics achieved by somatic homeostasis were inherited. The answer is simple: it would not work, for the following reasons:
The principle involved here is general and by no means trivial. It obtains in all homeostatic systems in which a given effect can be brought about by means of a homeostatic circuit, which circuit can, in turn, be modified in its characteristics by some- higher system of control. In all such systems (ranging from the house thermostat to systems of government and administration) it is important that the higher system of control lag behind the event sequences in the peripheral homeostatic circuit. In evolution two control systems are present: the homeostases of the body which deal with tolerable internal stress, and the action of natural selection upon the (genetically) nonviable members of the population. From an engineering point of view, the problem is to limit communication from the lower, reversible somatic system to the higher irreversible genotypic system. Another aspect of the proposed hypothesis about which we can only speculate is the probable relative frequency of the two classes of genotypic change: those which initiate something new and those which affirm some homeostatically achieved characteristic. In the Metazoa and multicellular plants, we face complex networks of multiple interlocking homeostatic circuits, and any given mutation or gene recombination which initiates change will probably require very various and multiple somatic characteristics to be achieved by homeostasis. The hypothetical pregiraffe with the mutant gene “long neck” will need to modify not only its heart and circulatory system but also perhaps its semicircular canals, its intervertebral discs, its postural reflexes, the ratio of length and thickness of many muscles, its evasive tactics vis-a-vis predators, etc. This suggests that in such complex organisms, the merely affirmative genotypic changes must far outnumber those which initiate change, if the species is to avoid that cul-de-sac in which the flexibility of the soma approaches zero. Conversely, this picture suggests that most organisms, at any given time, are probably in such a state that there are multiple possibilities for affirmative genotypic change. If, as seems probable, both mutation and gene redistribution are in some sense random phenomena, at least the chances are considerable that one or other of these multiple possibilities will be met. Finally, it is appropriate to discuss what evidence is avail-able or might be sought to support or disprove such a hypothesis. It is clear at the outset that such a testing will be difficult. The affirmative mutations upon which the hypothesis depends will usually be invisible. From among the many members of a population which are achieving a given adjustment to environmental circumstances by somatic change, it will not be possible immediately to pick out those few in which the same adjustment is provided by the genotypic method. In such a case, the genotypically changed individuals will have to be identified by breeding and raising the offspring under more normal conditions. A still greater difficulty arises in cases where we would investigate those homeostatically acquired characteristics which are achieved in response to some innovative genotypic change. It will often be impossible, by mere inspection of the organism, to tell which of its characteristics are the primary results of genotypic change and which are secondary somatic adjustments to these. In the imaginary case of the pregiraffe with a somewhat elongated neck and an enlarged heart, it may be easy to guess that the modification of the neck is genotypic while that of the heart is somatic. But all such guesses will depend upon the very imperfect present knowledge of what an organism can achieve in way of somatic adjustment. It is a major tragedy that the Lamarckian controversy has deflected the attention of geneticists away from the phenomenon of somatic adaptability. After all, the mechanisms, thresholds, and maxima of individual phenotypic change under stress must surely be genotypically determined. Another difficulty, of rather similar nature, arises at the population level, where we encounter another “economics” of potential change, theoretically distinguishable from that which operates within the individual. The population of a wild species is today conventionally regarded as genotypically heterogeneous in spite of the high degree of superficial resemblance between the individual phenotypes. Such a population expectably functions as a storehouse of genotypic possibilities. The economic aspect of this storehouse of possibilities has, for example, been stressed by Simmonds. [8] He points out that farmers and breeders who demand 100 per cent phenotypic uniformity in a highly select crop are in fact throwing away most of the multiple genetic possibilities accumulated through hundreds of generations in the wild population. From this Simmonds argues that there is urgent need for institutions which shall “conserve” this storehouse of variability by maintaining unselected populations. Lerner [9] has argued that self-corrective or buffering mechanisms operate to hold constant the composition of these mixtures of wild genotypes and to resist the effects of artificial selection. There is therefore at least a presumption that this economics of variability within the population will turn out to be of the multiplicative kind. Now, the difficulty of discriminating between a characteristic achieved by somatic homeostasis and the same characteristic achieved (more economically) by a genotypic short cut is clearly going to be compounded when we come to consider populations instead of physiologic individuals. All actual experimentation in the field will inevitably work with populations, and, in this work, it will be necessary to discriminate the effects of that economics of flexibility which operates inside the individuals from the effects of the economics of variability which operates at the population level. These two orders of economics may be easy to separate in theory, but to separate them in experimentation will surely be difficult. Be all that as it may, let us consider what evidential sup-port may be available for some of the propositions which are crucial to the hypothesis:
A system will be additive insofar as the units of its currency are mutually independent. Here there would seem to be a difference between the economic system of the individual, whose budgetary problems are additive (or sub-tractive) and those of society at large, where the overall distribution or flow of wealth is governed by complex (and perhaps imperfect) homeostatic systems. Is there, perhaps, an economics of economic flexibility (a metaeconomics) which is multiplicative and so resembles the economics of physiological flexibility discussed above? Notice, however, that the units of this wider economics will be not dollars but patterns of distribution of wealth. Similarly, Lerner’s “genetic homeostasis,” insofar as it is truly homeostatic, will have multiplicative character. The matter is, however, not simple and we cannot expect that every system will be either totally multiplicative or totally additive. There will be intermediate cases which combine the two characteristics. Specifically, where several independent alternative homeostatic circuits control a single variable, it is clear that the system may show additive characteristics—and even that it may pay to incorporate such alternative pathways in the system provided they can be effectively insulated from each other. Such systems of multiple alternative controls may give survival advantage insofar as the mathematics of addition and subtraction will pay better than the mathematics of logical fractionation.
He calls it the “genetic assimilation of acquired characteristics.” Similar phenomena have also probably occurred in various experiments when the experimenters set out to prove the inheritance of acquired characteristics but did not achieve this proof through failure to control the conditions of selection. We have, however, no evidence at all as to the frequency of this phenomenon of genetic assimilation. It is worth noting, however, that, according to the arguments of this essay, it may be impossible, in principle, to exclude the factor of selection from experiments which would test “the inheritance of acquired characteristics.” It is precisely my thesis that the simulation of Lamarckian inheritance will have survival value under circumstance of undefined or multiple stress.
Under such variable circumstances, it might pay the organisms in survival terms to achieve the converse of the genetic assimilation of acquired characteristics. That is, they might profitably hand over to somatic homeostatic mechanisms the control of some characteristic which had previously been more rigidly controlled by the genotype. It is evident, however, that such experimentation would be very difficult. Merely to establish the genetic assimilation of such characteristics as bithorax requires selection on an astronomical scale, the final population in which the genetically determined bithorax individuals can be found being a selected sample from a potential population of something like 1050. or 1060 individuals. It is very doubtful whether, after this selective process, there would still. exist in the sample enough genetic heterogeneity to undergo a further converse selection favoring those individuals which still achieve their bithorax phenotype by somatic means. Nevertheless, though this converse corollary is possibly not demonstrable in the laboratory, something of the sort seems to operate in the broad picture of evolution. The matter may be presented in dramatic form by considering the dichotomy between “regulators” and “adjusters.” [11] Prosser proposes that where internal physiology contains some variable of the same dimensions as some external environmental variable, it is convenient to classify organisms according to the degree to which they hold the internal variable constant in spite of changes in the external variable. Thus, the homoiothermic animals are classified as “regulators” in regard to temperature while the poikilothermic are “adjusters.” The same dichotomy can be applied to aquatic animals according to how they handle internal and external osmotic pressure. We usually think of regulators as being in some broad evolutionary sense “higher” than adjusters. Let us now consider what this might mean. If there is a broad evolutionary trend in favor of regulators, is this trend consistent with what has been said above about the survival benefits which accrue when control is transferred to genotypic mechanisms? Clearly, not only the regulators but also the adjusters must rely upon homeostatic mechanisms. If life is to go on, a large number of essential physiological variables must be held within narrow limits. It the internal osmotic pressure, for example, is allowed to change, there must be mechanisms which will defend these essential variables. It follows that the difference between adjusters and regulators is a matter of where, in the complex network of physiologic causes and effects, homeostatic process operates. In the regulators, the homeostatic processes operate at or close to the input and output points of that network which is the individual organism. In the adjusters, the environmental variables are permitted to enter the body and the organism must then cope with their effects, using mechanisms which will involve deeper loops of the total network. In terms of this analysis, the polarity between adjusters and regulators can be extrapolated another: step to include what we may call “extraregulators” which achieve homeostatic controls outside the body by changing and controlling the environment—man being the most conspicuous example of this class. In the earlier part of this essay, it was argued that in adjusting to high altitude there is a benefit to be obtained, in terms of an economics of flexibility, by shifting from, e.g., panting to the more profound and less reversible changes of acclimation; that habit is more economical than trial and error; and that genotypic control may be more economical than acclimation. These are all centripetal changes in the location of control. In the broad picture of evolution, however, it seems that the trend is in the opposite direction: that natural selection, in the long run, favors regulators more than adjusters, and extraregulators more than regulators. This seems to indicate that there is a long time evolutionary advantage to be gained by centrifugal shifts in the locus of control. To speculate about problems so vast is perhaps romantic, but it is worth noting that this contrast between the overall evolutionary trend and the trend in a population faced with constant stress is what we might expect from the converse corollary here being considered. If constant stress favors centripetal shift in the locus of control, and variable stress favors centrifugal shift, then it should follow that in the vast spans of time and change which determine the broad evolutionary picture, centrifugal shift of control will be favored. In this essay the author uses a deductive approach. Starting from premises of conventional physiology and evolutionary theory and applying to these the arguments of cybernetics, he shows that there must be an economics of somatic flexibility and that this economics must, in the long run, be coercive upon the evolutionary process. External adaptation by mutation or genotypic reshuffling, as ordinarily thought of, will inevitably use up the available somatic flexibility. It follows—if evolution is to be continuous—that there must also be a class of genotypic changes which will confer a bonus of somatic flexibility. In general, the somatic achievement of change is uneconomical because the process depends upon homeostasis, i.e., upon whole circuits of interdependent variables. It follows that inheritance of acquired characteristics would be lethal to the evolutionary system because it would fix the values of these variables all around the circuits. The organism or species would, however, benefit (in survival terms) by genotypic change which would simulate Lamarckian inheritance, i.e., would bring about the adaptive component of somatic homeostasis without involving the whole homeostatic circuit. Such a genotypic change (erroneously called the “Bald-win effect”) would confer a bonus of somatic flexibility and would therefore have marked survival value. Finally, it is suggested that a contrary argument can be applied in those cases where a population must acclimate to variable stress. Here natural selection should favor an anti-Baldwin effect. Problems in Cetacean and Other Mammalian Communication [12] The Communication of Preverbal Mammals [12] Of the Cetacea I have had little experience. I once dissected in the Cambridge Zoological Laboratories a specimen of Phocoena bought from the local fishmonger, and did not really encounter cetaceans again until this year, when I had an opportunity to meet Dr. Lilly’s dolphins. I hope that my discussion of some of the questions that are in my mind as I approach these peculiar mammals will assist you in examining either these or related questions. My previous work in the fields of anthropology, animal ethology, and psychiatric theory provides a theoretical framework for the transactional analysis of behavior. The premises of this theoretical position may be briefly summarized: (1) that a relationship between two (or more) organisms is, in-fact, a sequence of S-R sequences (i.e.,. of contexts in which proto-learning occurs); (2) that deuterolearning (i.e., learning to learn) is, in fact, the acquiring of information about the contingency patterns of the contexts in which proto-learning occurs; and (3) that the “character” of the organism is the aggregate of its deutero-learning and therefore reflects the contextual patterns of past protolearning. [13] These premises are essentially a hierarchic structuring of learning theory along lines related to Russell’s Theory of Logical Types. [14] The premises, following the Theory of Types, are primarily appropriate for the analysis of digital communication. To what extent they may be applicable to analogic communication or to systems that combine the digital with the analogic is problematic. I hope that the study of dolphin communication will throw light on these fundamental problems. The point is not either to discover that dolphins have complex language or to teach them English, but to close gaps in our theoretical knowledge of communication by studying a system that, whether rudimentary or complex, is almost certainly of a totally unfamiliar kind. Let me start from the fact that the dolphin is a mammal. This fact has, of course, all sorts of implications for anatomy and physiology, but it is not with these that I am concerned. I am interested in his communication, in what is called his “behavior,” looked at as an aggregate of data perceptible and meaningful to other members of the same species. It is meaningful, first, in the sense that it affects a recipient animal’s behavior, and, second, in the sense that perceptible failure to achieve appropriate meaning in the first sense will affect the behavior of both animals. What I say to you may be totally ineffective, but my ineffectiveness, if perceptible, will affect both you and me. I stress this point because it must be remembered that in all relationships between man and some other animal, especially when that animal is a dolphin, a very large proportion of the behavior of both organisms is determined by this kind of ineffectiveness. When I view the behavior of dolphins as communication, the mammalian label implies, for me, something very definite. Let me illustrate what I have in mind by an example from Benson Ginsburg’s wolf pack in the Brookfield Zoo. Among the Canidae, weaning is performed by the mother. When the puppy asks for milk, she presses down with her open mouth on the back of his neck, crushing him down to the ground. She does this repeatedly until he stops asking. This method is used by coyotes, dingoes, and the domestic dog. Among wolves the system is different. The puppies graduate smoothly from the nipple to regurgitated food. The pack comes back to the den with their bellies full. All regurgitate what they have got and all eat together. At some point the adults start to wean the puppies from these meals, using the method employed by the other Canidae; the adult crushes the puppy down by pressing its open mouth on the back of the puppy’s neck. In the wolf this function is not confined to the mother, but is performed by adults of both sexes. The pack leader of the Chicago pack is a magnificent male animal who endlessly patrols the acre of land to which the pack is confined. He moves with a beautiful trot that appears tireless, while the other eight or nine members of the pack spend most of their time dozing. When the females come in heat they usually proposition the leader, bumping against him with their rear ends. Usually, however, he does not respond, though he does act to prevent other males from getting the females. Last year one of these males succeeded in establishing coitus with a female. As in the other Canidae, the male wolf is locked in the female, unable to withdraw his penis, and this animal was helpless. Up rushed the pack leader. What did he do to the helpless male who dared to infringe the leader’s prerogatives? Anthropomorphism would suggest that he would tear the helpless male to pieces. But no. The film shows that he pressed down the head of the offending male four times with his open jaws and then simply walked away. What are the implications for research from this illustration? What the pack leader does is not describable, or only insufficiently described, in S-R terms. He does not “negatively reinforce” the other male’s sexual activity. He asserts or affirms the nature of the relationship between himself and the other. If we were to translate the pack leader’s action into words, the words would not be “Don’t do that.” Rather, they would translate the metaphoric action: “I am your senior adult male, you puppy!” What I am trying to say about wolves in particular, and about preverbal mammals in general, is that their discourse is primarily about the rules and the contingencies of relationship. Let me offer a more familiar example to help bring home to you the generality of this view, which is by no means orthodox among ethologists. When your cat is trying to tell you to give her food, how does she do it? She has no word for food or for milk. What she does is to make movements and sounds that are characteristically those that a kitten makes to a mother cat. If we were to translate the cat’s message into words, it would not be correct to say that she is crying “Milk!” Rather, she is saying something like “Ma-ma!” Or, perhaps still more correctly, we should say that she is asserting “Dependency! Dependency!” The cat talks in terms of patterns and contingencies of relationship, and from this talk it is up to you to take a deductive step, guessing that it is milk that the cat wants. It is the necessity for this deductive step which marks the difference between preverbal mammalian communication and both the communication of bees and the languages of men. What was extraordinary—the great new thing—in the evolution of human language was not the discovery of abstraction or generalization, but the discovery of how to be specific about something other than relationship. Indeed, this discovery, though it has been achieved, has scarcely affected the behavior even of human beings. If A says to B, “The plane is scheduled to leave at 6.30,” B rarely accepts this remark as simply and solely a statement of fact about the plane. More often he devotes a few neurons to the question, “What does A’s telling me this indicate for my relationship to A?” Our mammalian ancestry is very near the surface, despite recently acquired linguistic tricks. Be that as it may, my first expectation in studying dolphin communication is that it will prove to have the general mammalian characteristic of being primarily about relationship. This premise is in itself perhaps sufficient to account for the sporadic development of large brains among mammals. We need not complain that, as elephants do not talk and whales invent no mousetraps, these creatures are not overtly intelligent. All that is needed is to suppose that large-brained creatures were, at some evolutionary stage, unwise enough to get into the game of relationship and that, once the species was caught in this game of interpreting its members’ behavior toward one another as relevant to this complex and vital subject, there was survival value for those individuals who could play the game with greater ingenuity or greater wisdom. We may, then, reasonably expect to find a high complexity of communication about relationship among the Cetacea. Because they are mammals, we may expect that their communication will be about, and primarily in terms of, patterns and contingencies of relationship. Be-cause they are social and large-brained, we may expect a high degree of complexity in their communication. The above hypothesis introduces very special difficulties into the problem of how to test what is called the “psychology” (e.g., intelligence, ingenuity, discrimination, etc.) of individual animals. A simple discrimination experiment, such as has been run in the Lilly laboratories, and no doubt elsewhere, involves a series of steps:
It must be our first focus for methodological reasons. Consider the arguments that are conventionally based upon experiments of this kind. We argue always from the later steps in the series to the earlier steps. We say, “If the animal was able to achieve step 2 in our experiment, then he must have been able to achieve step 1.” If he could learn to behave in the way that would bring him the reward, then he must have had the necessary sensory acuity to discriminate between X and Y, and so on. Precisely because we want to argue from observation of the animal’s success in the later steps to conclusions about the more elementary steps, it becomes of prime importance to know whether the organism with which we are dealing is capable of step 4. If it is capable, then all arguments about steps 1 through 3 will be invalidated unless appropriate methods of controlling step 4 are built into the experimental design. Curiously enough, though human beings are fully capable of step 4, psychologists working with human subjects have been able to study steps 1 through 3 without taking special care to exclude the confusions introduced by this fact. If the human subject is “cooperative and sane,” he usually responds to the testing situation by repressing most of his impulses to modify his behavior according to his personal view of his relationship to the experimenter. The words cooperative and sane imply a degree of consistency at the level of step 4. The psychologist operates by a sort of petitio principii: if the subject is cooperative and sane (i.e., if the relational rules are fairly constant), the psychologist need not worry about changes in those rules. The problem of method becomes entirely different when the subject is noncooperative, psychopathic, schizophrenic, a naughty child, or a dolphin. Perhaps the most fascinating characteristic of this animal is derived precisely from his ability to operate at this relatively high level, an ability that is still to be demonstrated. Let me now consider for a moment the art of the animal trainer. From conversations with these highly skilled people —trainers of both dolphins and guide dogs—my impression is that the first requirement of a trainer is that he must be able to prevent the animal from exerting choice at the level of step 4. It must continually be made clear to the animal that, when he knows what is the right thing to do in a given context, that is the only thing he can do, and no non-sense about it. In other words, it is a primary condition of circus success that the animal shall abrogate the use of certain higher levels of his intelligence. The art of the hypnotist is similar. There is a story told of Dr. Samuel Johnson. A silly lady made her dog perform tricks in his presence. The Doctor seemed unimpressed. The lady said, “But Dr. Johnson, you don’t know how difficult it is for the dog.” Dr. Johnson re-plied, “Difficult, madam? Would it were impossible!” What is amazing about circus tricks is that the animal can abrogate the use of so much of his intelligence and still have enough left to perform the trick. I regard the conscious intelligence as the greatest ornament of the human mind. But many authorities, from the Zen masters to Sigmund Freud, have stressed the ingenuity of the less conscious and perhaps more archaic level. Communication About Relationship As I said earlier, I expect dolphin communication to be of an almost totally unfamiliar kind. Let me expand on this point. As mammals, we are familiar with, though largely unconscious of, the habit of communicating about our relationships. Like other terrestrial mammals, we do most of our communicating on this subject by means of kinesic and paralinguistic signals, such as bodily movements, involuntary tensions of voluntary muscles, changes of facial expression, hesitations, shifts in tempo of speech or movement, overtones of the voice, and irregularities of respiration. If you want to know what the bark of a dog “means,” you look at his lips, the hair on the back of his neck, his tail, and so on. These “expressive” parts of his body tell you at what object of the environment he is barking, and what patterns of relationship to that object he is likely to follow in the next few seconds. Above all, you look at his sense organs: his eyes, his ears, and his nose. In all mammals, the organs of sense become also organs for the transmission of messages about relationship. A blind man makes us uncomfortable, not because he cannot see that is his problem and we are only dimly aware of it—but because he does not transmit to us through the movement of his eyes the messages we expect and need so that we may know and be sure of the state of our relationship to him. We shall not know much about dolphin communication until we know what onedolphin can read in another’s use, direction, volume, and pitch of echolocation. Perhaps it is this lack in us which makes the communication of dolphins seem mysterious and opaque, but I suspect a more profound explanation. Adaptation to life in the ocean has stripped the whales of facial expression. They have no external ears to flap and few if any erectile hairs. Even the cervical vertebrae are fused into a solid block in many species, and evolution has streamlined the body, sacrificing the expressiveness of separate parts to the locomotion of the whole. Moreover, conditions of life in the sea are such that even if a dolphin had a mobile face, the details of his expression would be visible to other dolphins only at rather short range, even in the clearest waters. It is reasonable, then, to suppose that in these animals vocalization has taken over the communicative functions that most animals perform by facial expression, wagging tails, clenched fists, supinated hands, flaring nostrils, and the like. We might say that the whale is the communicational opposite of the giraffe; it has no neck, but has a voice. This speculation alone would make the communication of dolphins a subject of great theoretical interest. It would be fascinating, for example, to know whether or not, in an evolutionary shift from kinesics to vocalization, the same general structure of categories is retained. My own impression—and it is only an impression unsupported by testing—is that the hypothesis that dolphins have substituted paralinguistics for kinesics does not quite fit in with my experience when I listen to their sounds. We terrestrial mammals are familiar with paralinguistic communication; we use it ourselves in grunts and groans, laughter and sobbing, modulations of breath while speaking, and so on. Therefore we do not find the paralinguistic sounds of other mammals totally opaque. We learn rather easily to recognize in them certain kinds of greeting, pathos, rage, persuasion, and territoriality, though our guesses may often be wrong. But when we hear the sounds of dolphins we cannot even guess at their significance. I do not quite trust the hunch that would explain the sounds of dolphins as merely an elaboration of the paralinguistics of other mammals. (To argue thus from our inability is, however, weaker than to argue from what we can do.) I personally do not believe that the dolphins have anything that a human linguist would call a “language.” I do not think that any animal without hands would be stupid enough to arrive at so outlandish a mode of communication. To use a syntax and category system appropriate for the discussion of things that can be handled, while really discussing the patterns and contingencies of relationship, is fantastic. But that, I submit, is what is happening in this room. I stand here and talk while you listen and watch. I try to convince you, try to get you to see things my way, try to earn your respect, try to indicate my respect for you, challenge you, and so on. What is really taking place is a discussion of the patterns of our relationship, all according to the rules of a scientific conference about whales. So it is to be human. I simply do not believe that dolphins have language in this sense. But I do believe that, like ourselves and other mammals, they are preoccupied with the patterns of their relationships. Let us call this discussion of patterns of relationship the t function of the message. After all, it was the cat who showed us the great importance of this function by her mewing. Preverbal mammals communicate about things, when they must, by using what are primarily μ-function signals. In contrast, human beings use language, which is primarily oriented toward things, to discuss relationships. The cat asks for milk by saying “Dependency,” and I ask for your attention and perhaps respect by talking about whales. But we do not know that dolphins, in their communication, resemble either me or the cat. They may have a quite different system. Analogic versus Digital Communication There is another side of the problem. How does it happen that the paralinguistics and kinesics of men from strange cultures, and even the paralinguistics of other terrestrial mammals, are at least partly intelligible to us, whereas the verbal languages of men from strange cultures seem to be totally opaque? In this respect it would seem that the vocalizations of the dolphin resemble human language rather than the kinesics or paralinguistics of terrestrial mammals. We know, of course, why gestures and tones of voice are partly intelligible while foreign languages are unintelligible. It is because language is digital and kinesics and paralinguistics are analogic. [15] The essence of the matter is that in digital communication a number of purely conventional signs -1, 2, 3, X, Y, and so on—are pushed around according to rules called algorithms. The signs themselves have no simple connection (e.g., correspondence of magnitude) with what they stand for. The numeral “5” is not bigger than the numeral “3.” It is true that if we remove the crossbar from “7” we obtain the numeral “1”; but the crossbar does not, in any sense, stand for “6.” A name usually has only a purely conventional or arbitrary connection with the class named. The numeral “5” is only the name of a magnitude. It is nonsense to ask if my telephone number is larger than yours, because the telephone exchange is a purely digital computer. It is not fed with magnitudes, but only with names of positions on a matrix. In analogic communication, however, real magnitudes are used, and they correspond to real magnitudes in the subject of discourse. The linked range finder of a camera is a familiar example of an analogue computer. This device is fed with an angle that has real magnitude and is, in fact, the angle that the base of the range finder subtends at some point on the object to be photographed. This angle controls a cam that in turn moves the lens of the camera forward or back. The secret of the device lies in the shape of the cam, which is an analogic representation (i.e., a picture, a Cartesian graph) of the functional relationship between distance of object and distance of image. Verbal language is almost (but not quite) purely digital. The word “big” is not bigger than the word “little”; and in general there is nothing in the pattern (i.e., the system of interrelated magnitudes) in the word “table” which would correspond to the system of interrelated magnitudes in the object denoted. On the other hand, in kinesic and paralinguistic communication, the magnitude of the gesture, the loudness of the voice, the length of the pause, the tension of the muscle, and so forth—these magnitudes commonly correspond (directly or inversely) to magnitudes in the relationship that is the subject of discourse. The pattern of action in the communication of the wolf pack leader is immediately intelligible when we have data about the weaning practices of the animal, for the weaning practices are themselves analogic kinesic signals. It is logical, then, to consider the hypothesis that the vocalization of dolphins may be a digital expression of μ functions. It is this possibility that I especially have in mind in saying that this communication may be of an almost totally unfamiliar kind. Man, it is true, has a few words for μ functions, words like “love,” “respect,” “dependency,” and so on. But these words function poorly in the actual discussion of relationship between participants in the relationship. If you say to a girl, “I love you,” she is likely to pay more attention to the accompanying kinesics and paralinguistics than to the words themselves. We humans become very uncomfortable when somebody starts to interpret our postures and gestures by translating them into words about relationship. We much prefer that our messages on this subject remain analogic, unconscious, and involuntary. We tend to distrust the man who can simulate messages about relationship. We therefore have no idea what it is like to be a species with even a very simple and rudimentary digital system whose primary subject matter would be μ functions. This system is something we terrestrial mammals cannot imagine and for which we have no empathy. The most speculative part of my paper is the discussion of plans for the testing and amplification of such a body of hypotheses. I shall be guided by the following heuristic assumptions:
The assumptions regarding the hierarchic structure of the learning process— upon which this whole paper is based —provide the basis for various kinds of experimentation. The contexts of proto-learning may be variously constructed with a view to observing in what types of contexts certain types of learning most readily occur. We shall pay special attention to those contexts that involve either relationships between two or more animals and one person, or relationships between two or more people and one animal. Such contexts are miniature models of social organization within which the animal may be expected to show characteristic behaviors and to make characteristic attempts to modify the context (i.e., to manipulate the humans). Mr. Wood: In the course of twelve years in Marine Studios in Florida, I spent a great deal of time watching what was perhaps the most natural assemblage of Tursiops in captivity, including animals of various ages, usually two or more of them in the process of growing up, and I saw remarkably little of what you are going to look for in a much more restricted group of animals in the Virgin Islands. One time I saw something very interesting. Early one morning about six or six-thirty, over a period of at least half an hour, the adult male assumed a position next to one of the females in the tank who was hanging motionless in the cur-rent. He would go up occasionally and move away and then come back and assume a position beside her, and he would stroke her side with his right flipper repeatedly. There was no indication that this had sexual significance. There was no erection on the part of the male, and no observable response on the part of the female. But it was as clear-cut a nonvocal signal as I ever observed in the tank. Mr. Bateson: I would like to say that the amount of signaling that goes on is much greater than is evident at first sight. There are, of course, the rather specific kinds of signals which are very important. I am not denying that. I mean the touching, and so on. But the shy individual, the traumatized female, staying almost stationary three feet be-low the surface while two other individuals fool around, is getting a great deal of attention just by sitting there and staying. She may not be actively transmitting, but in this business of bodily communication, you don’t have to be actively transmitting in order to have your signals picked up by other people. You can just be, and just by being she attracts an enormous amount of attention from these other two individuals who come over, pass by, pause a little as they pass, and so on. She is, we would say, “withdrawn,” but she is actually about as withdrawn as a schizophrenic who by being withdrawn becomes the center of gravity of the family. All other members of the group move around the fact of her withdrawal, which she never lets them for-get. Dr. Ray: I tend to agree with Mr. Bateson. We are working at the New York Aquarium with the beluga whale, and I believe these animals are much more expressive than we like to suspect. I think one of the reasons they don’t do very much in captivity is that they are bored to tears most of the time. There is nothing much of interest in their tank environment, and I would like to suggest that we have to manipulate their captivity much more cleverly than we do. I don’t mean handling the whales. They don’t like that. But the introduction of different types of animals, or clever little things that we might do would get them to respond more. Captive cetaceans are like monkeys in a cage. They are highly intelligent and highly developed, and they are bored. Another factor is our skill in observation, and in the beluga whale, at least, we have been able to notice visually the sounds they are making by watching the change in the shape of the melon, which is extremely marked in this animal. It can swell on one side or the other, or take several different shapes correlated with sound production. So, by very careful observation and/or skilled manipulation, I think a great deal can be done with these animals rather simply. Mr. Bateson: I had meant to point out that all sense organs among mammals, and even among ants, become major organs for the transmission of messages, such as, “Where are the other fellow’s eyes focused?” and, “Are his pinnae focused in one direction or another?” In this way sense organs become transmitting organs for signals. One of the things we must absolutely acquire if we are going to understand dolphins is a knowledge of what one animal knows and can read from another animals’ use of sonar. I suspect the presence of all sorts of courtesy rules in this business; it probably isn’t polite to sonar scan your friends too much, just as among human beings it is not polite, really, to look at another’s feet in detail. We have many taboos on observing one anothers’ kinesics, because too much information can be got in that way. Dr. Purves: It seems to me that the dolphin or the cetacean must suffer from an even greater disadvantage than man has in the past, because—I have forgotten the authority—it has been said that the origin of human speech is an analogue language. In other words, if you use the word “down,” you lower the hand and lower the lower jaw at the same time. If you say “up,” you raise the hand and raise the lower jaw. And if you use the word “table,” and, better still, pronounce it in French, your mouth widens out and you make a horizontal gesture. However complicated the human language is, it has its origin in an analogue language. The poor porpoise has nothing like this to start from. So he must have been highly intelligent to have developed a communication system completely de novo. Mr. Bateson: What has happened to this creature is that the information we get visually and the other terrestrial animals get visually must have been pushed into voice. I still maintain that it is appropriate for us to start by investigating what is left of the visual material. A Re-examination of “Bateson’s Rule” [17] Nearly eighty years ago, my father, William Bateson, be-came fascinated by the phenomena of symmetry and metameric regularity as exhibited in the morphology of animals and plants. It is difficult today to define precisely what he was after, but, broadly, it is clear that he believed that an entirely new concept of the nature of living things would develop from the study of such phenomena. He held, no doubt correctly, that natural selection could not be the only determinant of the direction of evolutionary change and that the genesis of variation could not be a random matter. He therefore set out to demonstrate regularity and “lawfulness” among the phenomena of variability. In his attempt to demonstrate a sort of order which the biologists of his day had largely ignored, he was guided by the notion, never clearly formulated, that the place to look for regularity in variation would be precisely where variation had its impact upon what was already regular and repetitive. The phenomena of symmetry and metamerism, themselves strikingly regular, must surely have been brought about by regularities or “laws” within the evolutionary process and, therefore, the variations of symmetry and metamerism should precisely exemplify these laws at work. In the language of today, we might say that he was groping for those orderly characteristics of living things which illustrate the fact that organisms evolve and develop with-in cybernetic, organizational, and other communicational limitations. It was for this study that he coined the word “genetics.” [18] He set out to examine the material in the world’s museums, private collections, and journals bearing upon the teratology of animal symmetry and metamerism. The de-tails of this survey were published in a large book [19] which is still of considerable interest. To demonstrate regularity within the field of teratological variation, he attempted a classification of the various sorts of modification that he encountered. With this classification I am not here concerned, except that in the survey he happened upon a generalization which can be called a “discovery.” This discovery came to be called “Bateson’s Rule” and remains one of the unexplained mysteries of biology. The purpose of the present note is to place Bateson’s Rule in a new theoretical perspective determined by cybernetics, information theory, and the like. Briefly, Bateson’s Rule asserts in its simplest form that when an asymmetrical lateral appendage (e.g., a right hand) is reduplicated, the resulting reduplicated limb will be bilaterally symmetrical, consisting of two parts each a mirror image of the other and so placed that a plane of symmetry could be imagined between them. He himself was, however, very doubtful whether such simple reduplication ever occurs. He believed and accumulated evidence to show that, in a very large proportion of such cases, one component of the reduplicated system was it-self double. He asserted that in such systems the three components are normally in one plane; that the two components of the doublet are mirror images of each other; and that that component of the doublet which is the nearer to the primary appendage is a mirror image of the primary. This generalization was shown by my father to hold for a very large number of examples of reduplication in the vertebrates and in arthropods, and for a few cases in other phyla where the museum material was, of course, more scarce. Ross Harrison [20] believed that Bateson underestimated the importance of simple reduplication. Whether or not simple reduplication is a real and common phenomenon, I shall begin this essay with a discussion of the logical problems which it would present. In 1894, it appeared that the problem centered around the question: What causes the development of bilateral symmetry in a context where it does not belong? But modern theory has turned all such questions upside down. Information, in the technical sense, is that which excludes certain alternatives. The machine with a governor does not elect the steady state; it prevents itself from staying in any alternative state; and in all such cybernetic systems, corrective action is brought about by difference. In the jargon of the engineers, the system is “error activated.” The difference between some present state and some “preferred” state activates the corrective response. The technical term. “information” may be succinctly de-fined as any difference which makes a difference in some later event. This definition is fundamental for all analysis of cybernetic systems and organization. The definition links such analysis to the rest of science, where the causes of events are commonly not differences but forces, impacts, and the like. The link is classically exemplified by the heat engine, where available energy (i.e., negative entropy) is a function of a difference between two temperatures. In this classical instance, “information” and “negative entropy” overlap. Moreover, the energy relations of such cybernetic systems are commonly inverted. Because organisms are able to store energy, it is usual that the energy expenditure is, for limited periods of time, an inverse function of energy in-put. The amoeba is more active when it lacks food, and the stem of a green plant grows faster on that side which is turned away from the light. Let us therefore invert the question about the symmetry of the total reduplicated appendage: Why is this double appendage not asymmetrical like the corresponding appendages of normal organisms? To this question a formal and general (but not particular) answer can be constructed on the following lines:
Converse cases can also be cited. Plants of many families bear bilaterally symmetrical flowers. Such flowers are all clearly derived from triadic radial symmetry (as in orchids) or from pentadic symmetry (as in Labiatae, Leguminosae, etc.) ; and the bilateral symmetry is achieved by the differentiation of one axis (e.g., the “standard” of the familiar sweet pea) of this radial symmetry. We again ask how it is possible to select one of the similar three (or five) axes. And again we find that each flower receives information from the outside. Such bilaterally symmetrical flowers can only be produced on branch stems, and the differentiation of the flower is always oriented to the manner in which the flower-bearing branch stem comes off from the main stem. Very occasionally a plant which normally bears bilaterally symmetrical flowers will form a flower at the terminus of a main stem. Such a flower is necessarily only radial in its symmetry—a cup-shaped monstrosity. (The problem of bilaterally asymmetrical flowers, e.g., in the Catasetum group of orchids, is interesting. Presumably these must be borne, like the lateral appendages of animals, upon branches from main stems which are themselves already bilaterally symmetrical, e.g., dorso-ventrally flattened.)
The problem of the bilateral symmetry of reduplicated limbs thus becomes simply a problem of the loss of a piece of information. This follows from the general logical rule that every reduction in symmetry (from radial to bilateral or from bilateral to asymmetrical) requires additional in-formation. It is not claimed that the above argument is an explanation of all the phenomena which illustrate Bateson’s Rule. Indeed, the argument is offered only to show that there are simple ways of thinking about these phenomena which have scarcely been explored. What is proposed is a family of hypotheses rather than a single one. A critical examination of what has been said above as if it were a single hypothesis will, how-ever, provide a further illustration of the method. In any given case of reduplication, it will be necessary to decide what particular piece of information has been lost, and the argument so far given should make this decision easy. A natural first guess would be that the developing appendage needs three sorts of orienting information to en-able it to achieve asymmetry: proximodistal information; dorso-ventral information; and antero-posterior information. The simplest hypothesis suggests that these might be separately received and therefore that one of these sorts of information will be lost or absent in any given case of reduplication. It should then be easy to classify cases of reduplication ac-cording to which piece of orienting information is missing. There should be at most three such types of reduplication, and these should be clearly distinct. Supernumerary Double Legs in Coleoptera But in the only set of cases where this deduction can be tested, facts clearly do not fit the hypothesis. The cases are those of supernumerary pairs of appendages in beetles. About a hundred such cases were known in 1894, and of these Bateson [22] describes about half and figures thirteen. The formal relations are remarkably uniform and leave no doubt that a single type of explanation should apply to the symmetry in all cases.
Fig. 1 Carabus scheidleri, No. 736. The normal right fore leg, R, bearing an extra pair of legs, SL and SR', arising from the ventral surface of the coxa, C. Seen from in front. (The property of Dr. Kraatz.) From Bateson, W., Materials for the Study of Variation, London:
Fig. 2 Pterostichus muhlfeldii, No. 742. Semidiagrammatic representation of the left middle tibia bearing the extra tarsi upon the antero-ventral border of the apex. L, the normal tarsus; R, the extra right; L' the extra left tarsus. ( The property of Dr. Kraatz. ) From Bateson, W., Materials for the Study of Variation, London: Macmillan, 1894, p 485.
Fig. 3 Symmetry of a doublet occurring in the dorsal region. Fig. 4 Symmetry of a doublet occurring in the dorso-anterior region.
Fig. 5 A mechanical device for showing the relations that extra legs in Secondary Symmetry bear to each other and to the normal leg from which they arise. The model R represents a normal right leg. SL and SR represent respectively the extra right and extra left legs of the supernumerary pair. A and P, the anterior and posterior spurs of the tibia. In each leg the morphologically anterior surface is shaded, the posterior being white. R is seen from the ventral aspect and SL and SR are in Position VP. From Bateson, W., Materials for the Study of Variation, London: Macmillan, 1894, p. 480. Typically [23] one leg (rarely more than one) of a beetle is abnormal in bearing a branch at some point in its length. This branch is regularly a doublet, consisting of two parts which may be fused at the point of branching off from the primary leg but which are commonly separate at their distal ends. Distally from the point of branching there are thus three components—a primary leg and two supernumerary legs. These three lie in one plane and have the following symmetry: the two components of the supernumerary doublet are a complementary pair—one being a left and the other a right—as Bateson’s Rule would suggest. Of these two, the leg nearest to the primary leg is complementary to it. These relations are represented in Figure 3. (See page 387.) Each component is shown in diagrammatic cross section, and their dorsal, ventral, anterior, and posterior faces are indicated by the letters D, V, A, and P, respectively. What is surprising about these abnormalities—in that it conflicts with the hypothesis offered above—is that there is no clear discontinuity by which the cases can be classified according to which sort of orienting information has been lost. The supernumerary doublet may be borne on any part of the circumference of the primary leg. Figure 3 illustrates the symmetry of a doublet occurring in the dorsal region. Figure 4 (page 387) illustrates the symmetry of a doublet in the dorso-anterior region. It appears, then, that the planes of symmetry are parallel to a tangent of the circumference of the primary leg at the point of branching but, since the points of branching may be anywhere on the circumference, a continuous series of possible bilateral symmetries is generated. Figure 5 (page 388) is a machine invented by W. Bateson to demonstrate this continuous series of possible bilateral symmetries. If the bilateral symmetry of the doublet is due to a loss of orienting information, we should expect the plane of that bilateral symmetry to be at right angles to the direction of the lost information; i.e., if dorso-ventral information were lost, the resulting limbs or doublet should contain a plane of symmetry which would be at right angles to the dorso-ventral line. (The argument for this expectation may be spelled out as follows: a gradient in a lineal sequence creates a difference between the two ends of the sequence. If this gradient is not present, then the ends of the sequence will be similar, i.e., the sequence will be symmetrical about a plane of symmetry transverse to itself. Or, consider the case of the frog’s egg. The two poles and the point of entry of the spermatozoon determine a plane of bilateral symmetry. To achieve asymmetry, the egg requires information at right angles to this plane, i.e., something which will make the right half different from the left. If this something is lost, then the egg will revert to the original bilateral symmetry, with the original plane of symmetry transverse to the direction of the lost information.) As noted above, the supernumerary doublets may originate from any face of the primary leg, and therefore all intermediates occur between the expectedly discontinuous types of loss of information. It follows that if bilateral symmetry in these doublets is due to loss of information, then the information lost cannot be classified as antero-posterior, dorso-ventral, or proximo-distal. The hypothesis must therefore be corrected. Let us retain the general notion of lost information, and the corollary of this that the plane of bilateral symmetry must be at right angles to the direction of the information that was lost. The next simplest hypothesis suggests that the lost information must have been centro-peripheral. (I here retain this bipolar term rather than use the simpler “radial.”) Let us imagine, then, some centro-peripheral difference —possibly a chemical or electrical gradient within the cross section of the primary leg; and suppose that the loss or blurring of this difference at some point along the length of the primary leg determines that any branch limb produced at this point shall fail to achieve asymmetry. It will follow, naturally, that such a branch limb (if produced) will be bilaterally symmetrical and that its plane of bilateral symmetry will be at right angles to the direction of the lost gradient or difference. But, clearly, a centro-peripheral difference or gradient is not a primary component of that information system which determined the asymmetry of the primary leg. Such a gradient might, however, inhibit branching, so that its loss or blurring would result in production of a supernumerary branch at the point of loss. The matter becomes superficially paradoxical: the loss of a gradient which might inhibit branching results in branch formation, such that the branch cannot achieve asymmetry. It appears, then, that the hypothetical Centro-peripheral gradient or difference may have two sorts of command functions: (a) to inhibit branching; and (b) to determine an asymmetry in that branch which can only come into existence at all if the Centro-peripheral gradient is absent. If these two sorts of message functions can be shown to overlap or be in some sense synonymous, we shall have generated an economical hypothetical description of the phenomena. We therefore address ourselves to the question: Is there an a priori case for expecting that the absence of a gradient which would prohibit branching in the primary leg will permit the formation of a branch which will lack the information necessary to determine asymmetry across a plane at right angles to the missing gradient? The question must be inverted to fit the upside-downness of all cybernetic explanation. The concept “information necessary to determine asymmetry” then becomes “information necessary to prohibit bilateral symmetry.” But anything which “prohibits bilateral symmetry” will also “prohibit branching,” since the two components of a branching structure constitute a symmetrical pair (even though the components may be radially symmetrical). It therefore becomes reasonable to expect that loss or blurring of a Centroperipheral gradient which prohibits branch formation will permit the formation of a branch which will, however, itself be bilaterally symmetrical about a plane parallel to the circumference of the primary limb. Meanwhile, within the primary limb, it is possible that a Centro-peripheral gradient, by preventing branch formation, could have a function in preserving a previously determined asymmetry. The above hypotheses provide a possible framework of explanation of the formation of the supernumerary doublet and the bilateral symmetry within it. It remains to consider the orientation of the components of that doublet. According to Bateson’s Rule, the component nearest to the primary leg is in bilateral symmetry with it. In other words, that face of the supernumerary which is toward the primary is the morphological counterpart of that face of the periphery of the primary from which the branch sprang. The simplest, and perhaps obvious, explanation of this regularity is that in the process of branching there was a sharing of morphologically differentiated structures between branch and primary and that these shared structures are, in fact, the carriers of the necessary information. However, since information carried this way will clearly have proper-ties very different from those of information carried by gradients, it is appropriate to spell the matter out in some detail. Consider a radially symmetrical cone with circular base. Such a figure is differentiated in the axial dimension, as between apex and base. All that is necessary to make the cone fully asymmetrical is to differentiate on the circumference of the base two points which shall be different from each other and shall not be in diametrically opposite positions, i.e., the base must contain such differentiation that to name its parts in clockwise order gives a result different from the result of naming the parts in anticlockwise order. Assume now that the supernumerary branch, by its very origin as a unit growing out from a matrix, has proximo-distal differentiation, and that this differentiation is analogous to the differentiation in the axial dimension of the cone. To achieve complete asymmetry, it is then only necessary that the developing limb receive directional information in some arc of its circumference. Such information is clearly immediately available from the circumstance that, at the point of branching, the secondary limb must share some circumference with the primary. But the shared points which are in clockwise order on the periphery of the primary will be in anticlockwise order on the periphery of the branch. The information from the shared arc will therefore be such as to determine both that the resulting limb will be a mirror image of the primary and that the branch will face appropriately toward the primary. It is now possible to construct a hypothetical sequence of events for the reduplications in the legs of beetles:
The above speculations are intended to illustrate how the explanatory principle of loss of information might be applied to some of the regularities subsumed under Bateson’s Rule. But it will be noted that the data on symmetry in the legs of beetles have, in fact, been overexplained. Two distinct but not mutually exclusive—types of explanation have been invoked: (a) the loss of information which should have been derived from a centroperipheral gradient, and (b) information derived from shared peripheral morphology. Neither of these types of explanation is sufficient by itself to explain the phenomena, but when combined the two principles overlap so that some details of the total picture can be referred simultaneously to both principles. Such redundancy is, no doubt, the rule rather than the exception in biological systems, as it is in all other systems of organization, differentiation, and communication. In all such systems, redundancy is a major and necessary source of stability, predictability, and integration. Redundancy within the system will inevitably appear as overlapping between our explanations of the system. Indeed, without overlapping, our explanations will commonly be insufficient, failing to explain the facts of biological integration. We know little about how the pathways of evolutionary change are influenced by such morphogenetic and physiological redundancies. But certainly such internal redundancies must impose nonrandom characteristics upon the phenomena of variation. [24] Reduplicated Limbs in Amphibia At this point it is interesting to turn from analysis of reduplication in beetles’ legs to another body of data in which reduplication commonly occurs and has been referred to Bateson’s Rule. [25] These are the data on reduplication in the experimentally transplanted limbs of larval newts.
Such preparations, where the product is binary and the parts equal, certainly look like what would be expected from a simple loss of one dimension of orienting information. (It was Dr. Hibbard’s specimen that suggested to me that the hypothesis of lost information might be applicable to the amphibian material.)
The limitations of a hypothesis are, however, as important as its applications, and I shall therefore summarize here the very complex data on orthotopic transplants. One schematic paradigm will suffice: if the right anterior limb bud is excised, turned through 180° and replaced in the wound, it will grow to be a left limb. But this primary limb may subsequently form secondary limb buds at its base, usually either immediately anterior or posterior to the point of insertion. The secondary will be a mirror image of the primary, and may even later develop a tertiary which will typically be formed outside the secondary, i.e., on that side of the secondary which is farthest from the primary. The formation of the left primary on the right side of the body is explained [26] by assuming that antero-posterior orientation is received by the limb bud earlier than dorso-ventral information, and that, once received, this antero-posterior information is irreversible. It is supposed that the graft is already antero-posteriorly determined at the time of grafting but later receives dorso-ventral information from the tissues with which it is now in contact. The result is a limb whose dorso-ventral orientation is correct for its new setting but whose antero-posterior orientation is reversed. It is tacitly assumed that the proximo-distal orientation of the bud is undisturbed. The result is a limb which is reversed in regard to one of its three sorts of asymmetry. Such a limb must logically be a left. This explanation I accept and proceed to consider the reduplications. These differ in four important respects from the reduplications in beetles’ legs discussed above:
In the beetles it is clear that the two supernumerary components form together a single unit. In many cases there is actual compounding of the two components (as in Figure 1). In no case [27] is that component of the doublet which is nearer to the primary compounded with it rather than with the other supernumerary. In the amphibian preparations, on the other hand, it is not clear that secondary and tertiary form a subunit. The relation between tertiary and secondary seems no closer than between secondary and primary. Above all, the relation is asymmetric in the time dimension. These profound formal differences between the two bodies of data indicate that the explanations for the amphibian data must be of a different order. It would seem that the processes are located not in the shaft of the limb but in its base and the tissues surrounding the base. Tentatively we may guess that the primary in some way proposes the later formation of a secondary by a reversal of gradient information, and that the secondary similarly proposes a reversed tertiary. Models for such systems are available in cybernetic theory in those circuit structures which propose Russellian paradoxes. [28] To attempt to construct any such model at the present time would be premature. This essay on the symmetry of reduplicated lateral appendages starts from an explanatory principle, viz., that any step of ontogenetic differentiation which reduces the symmetry of an organ (e.g., from radial to bilateral symmetry, or from bilateral symmetry to asymmetry) requires additional orienting information. From this principle it is argued that a normally asymmetrical lateral appendage, lacking some necessary piece of orienting information, will only be able to achieve bilateral symmetry, i.e., instead of a normal asymmetrical appendage, the result will be a bilaterally symmetrical doublet. To examine this explanatory principle, the writer has at-tempted to construct a hypothesis to explain Bateson’s Rule as this regularity is exemplified in the rare supernumerary double legs of Coleoptera. In the construction of this hypothesis, it was assumed that morphogenetic orienting information may undergo transformation from one type of coding to another, and that each transform or code is subject to characteristic limitations:
The data on reduplication in the experimentally trans-planted limb buds of amphibia are also examined. It is argued that these data are not to be explained by simple loss of orienting information. Simple loss, it is suggested, will expectably result in equal and synchronous bilateral symmetry. The amphibian reduplicates are, in general, unequal and successive. In a few cases, synchronous and equal reduplication occurs in the amphibian experiments, especially in heterotopic implants. Such cases could perhaps be regarded as due to simple loss of orienting information. Compare the bilateral symmetry in the supernumerary doublet of the beetle’s leg with the bilateral symmetry in the sweet pea or orchid flower. Both in the plant and in the animal, the bilaterally symmetrical unit comes off from a point of branching. In the plant, the morphology of the fork provides information enabling the flower to be not radially but bilaterally symmetrical, i.e., information which will differentiate the “dorsal” standard from the ventral lip of the flower. In the doublet on the beetle’s leg, the plane of bilateral symmetry is orthogonal to that in the flower. We might say that the information which the beetle’s leg has lost is precisely that information which the plant creates by the act of branching. The papers placed together in this part are diverse in that while each paper is a branch from the main stem of the argument of the book, these branches come off from very different locations. “The Role of Somatic Change in Evolution” is an expansion of the thought behind “Minimal Requirements for a Theory of Schizophrenia,” while “Problems in Cetacean and Other Mammalian Communication” is an application of “The Logical Categories of Learning and Communication” to a particular type of animal. “A Re-examination of Bateson’s Rule” may seem to break new ground, but is related to the remainder of the book in that it ex-tends the notion of informational control to include the field of morphogenesis and, by discussing what happens in absence of needed information, brings out the importance of the context into which information is received. Samuel Butler, with uncanny insight, once commented upon the analogy between dreams and parthenogenesis. We may say that the monstrous double legs of the beetles share in this analogy: they are the projection of the receptive context deprived of information which should have come from an external source. Message material, or information, comes out of a context into a context, and in other parts of the book the focus has been on the context out of which information came. Here the focus is rather upon the internal state of the organism as a context into which the information must be received. Of course, neither focus is sufficient by itself for our under-standing of either animals or men. But it is perhaps not an accident that in these papers dealing with non-human organisms the “context” which is discussed is the obverse or complement of the “context” upon which I have focussed attention in other parts of the book. Consider the case of the unfertilized frog’s egg for which the entry point of the spermatozoon defines the plane of bilateral symmetry of the future embryo. The prick of a hair from a camel’s hair brush can be substituted and still carry the same message. From this it seems that the external context out of which the message comes is relatively undefined. From the entry point alone, the egg learns but little about the external world. But the internal context into which the message comes must be exceedingly complex. The unfertilized egg, then, embodies an immanent question to which the entry point of spermatozoon provides an answer; and this way of stating the matter is the contrary or obverse of the conventional view, which would see the external context of learning as a “question” to which the “right” behavior of the organism is an answer. We can even begin to list some of the components of the immanent question. First there are the already existing poles of the egg and, necessarily, some polarization of the intervening protoplasm towards these poles. Without some such structural conditions for the receipt of the prick of the spermatozoon, this message could have no meaning. The message must come into an appropriate structure. But structure alone is not enough. It seems probable that any meridian of the frog’s egg can potentially become the plane of bilateral symmetry and that, in this, all meridians are alike. It follows that there is, to this extent, no structural difference between them. But every meridian must be ready for the activating message, its “readiness” being given direction but otherwise unrestricted by structure. Readiness, in fact, is precisely not-structure. If and when the spermatozoon delivers its message, new structure is generated. In terms of the economics of flexibility, discussed in “The Role of Somatic Change in Evolution” and later in “Ecology and Flexibility in Urban Civilization” (Part VI), this “readiness” is uncommitted potentiality for change, and we note here that this uncommitted potentiality is not only always finite in quantity but must be appropriately located in a structural matrix, which also must be quantitatively finite at any given time. These considerations lead naturally into Part V, which I have titled “Epistemology and Ecology.” Perhaps “epistemology” is only another word for the study of the ecology of mind. _______________ Notes: 1. This item in BioScience, Vol. 20, 1970, is reproduced by permission from that journal. 2. See “California's Anti-Evolution Ruling,” BioScience, March 1, 1970. 3. This essay appeared in the journal Evolution, Vol 17, 1963, and is reprinted with the editor's permission. 4. The problems of bacterial genetics are here deliberately excluded. 5. Bateson, “Minimal Requirements for a Theory of Schizophrenia,” A.M.A. Archives of General Psychiatry, 1960, 2: 447. 6. W. R. Ashby, “The Effect of Controls on Stability,” Nature, 1945, 155: 242; also Ashby, Design for a Brain, New York, John Wiley & Co., 1952. 7. J. M. Baldwin, “Organic Selection,” Science, 1897, 5: 634. 8. N. W. Simmonds, “Variability in Crop Plants, Its Use and Conservation,” Biol. Review, 1962, 37: 422-62. 9. I. M. Lerner, Genetic Homeostasis, Edinburgh, Oliver and Boyd, 1954. 10. C. H. Waddington, “Genetic Assimilation of an Acquired Character,” Evolution, 1953, 7: 118; also Waddington, The Strategy of Genes, London, Allen and Unwin, 1957. 11. C. L. Prosser, “Physiological Variation in Animals,” Biol. Review, 1955, 30: 22-262. 12. This article appeared as Chapter 25, pp. 569-799, in Whales, Dolphins and Porpoises, edited by Kenneth S. Norris, University of California Press, 1966. Reprinted by permission of The Regents of the University of California. 13. J. Ruesch and G. Bateson, Communication: The Social Matrix of Psychiatry, New York, Norton, 1951. 14. A. N. Whitehead and B. Russell, Principia Mathematica, London, Cambridge University Press, 1910. 15. The difference between digital and analogic modes of communication may perhaps be made clear by thinking of an English-speaking mathematician confronted with a paper by a Japanese colleague. He gazes uncomprehendingly at the Japanese ideographs, but he is able partly to understand the Cartesian graphs in the Japanese publication. The ideographs, though they may originally have been analogic pictures, are now purely digital; the Cartesian graphs are analogic. 16. Whitehead and Russell, op. cit. 17. This essay has been accepted for publication in the Journal of Genetics, and is here reproduced with the permission of that journal. 18. W. Bateson, “The Progress of Genetic Research,” In-augural Address, Royal Horticultural Society Report, 1906. 19. W. Bateson, Materials for the Study of Variation, London, Macmillan and Co., 1894. 20. R. G. Harrison, “On Relations of Symmetry in Transplanted Limbs,” Journal of Experimental Zoology, 1921, 32: 1-118. 21. In this connection, scales and feathers and hairs are of special interest. A feather would seem to have a very clear bilateral symmetry in which the plane of symmetry is related to the antero-posterior differentiation of the bird. Superposed on this is an asymmetry like that of the individual bilateral limbs. As in the case of lateral limbs, corresponding feathers on opposite sides of the body are mirror images of each other. Every feather is, as it were, a flag whose shape and coloring denote the values of determining variables at the point and time of its growth. 22. W. Bateson, Materials . . . , op. cit., pp. 477-503. 23. See Figures 1 and 2, pages 385 and 386. 24. G. Bateson, “The Role of Somatic Change in Evolution,” Evolution, 1962, 17: 529-39. 25. Harrison, op. cit.; also F. H. Swett, “On The Production of Double Limbs in Amphibians,” Journal of Experimental Zoology, 1926, 44: 419-72. 26. Swett, op. cit.; also Harrison, op. cit. 27. Bateson (Materials …, op. cit., p. 507) describes and figures one doubtful exception to this statement. This is a reduplication in the left hind tarsus of platycerus caraboides. 28. G. Bateson, “Minimal Requirements for a Theory of Schizophrenia,” A.M.A. Archives of General Psychiatry, 1960, 2: 477-91.
|
