Version classiqueVersion mobile
OpenEdition Books

Behaviour, Development and Evolution

Patrick Bateson

3. Rules and Reciprocity1

Texte intégral

  • 1 Parts of this chapter were taken, with permission, from Bateson, P. (1976), Rules and reciprocity i (...)

1The study of development has attracted some of the most bitter and protracted controversies in the whole of psychology and ethology. The arguments reflected more general ideological battles about nature and nurture. Consequently, much research was concerned merely to establish that a particular kind of internal or external factor could be important, or that evidence could be obtained for a certain logical possibility. In recent years, such activity has abated with the growing acceptance that both internal and external factors can play important roles in the development of any one pattern of behaviour. Also, the air has been cleared by the realization that an interest in how behaviour has been adapted to its present uses is not the same as an interest in what makes one individual animal different from another one.

  • 2 A discussion of the concept of innateness is given in Mameli, M. & Bateson, P. (2006), Innateness a (...)

2Two radically different models have been proposed for what is happening when behaviour develops. On one view straightforward correspondence can be found between genes and innate behaviour on the one hand and between the environment and learned behaviour on the other. The word ‘innate’ has many different meanings attached to it: present at birth; a behavioural difference caused by a genetic difference; adapted over the course of evolution; unchanging throughout development; shared by all members of a species; and not learned. ‘Instinct’ is deployed in similar ways to innate. When the justification for using one of the meanings of innate or instinctive has been demonstrated it does not follow that another of the meanings will necessarily apply. Even if a behaviour pattern develops without obvious practice or example, it may subsequently be modified by learning. For instance, a blind baby may start to smile in the same way as a normal baby. But that does not mean that later on in their lives, sighted people will not modify their smiles to expressions that are characteristic of their own culture. A classification of behaviour into innate and not innate merely causes confusion.2

  • 3 Lehrman, D.S. (1970), Semantic and conceptual issues in the nature–nurture problem. In: Aronson, L. (...)

3The second view of development does not recognize the distinction between behaviour that is not learned and behaviour that is acquired. It proposes instead that as the animal develops, it is not merely affected by its genes and its environment. The animal’s state influences which genes are activated from time to time and the animal also alters the character of the environment as it develops.3 While such transactions of this kind seem reasonable, this second model is often perceived as being too complicated and too vague. Such objections start to fall away as the nature of developmental processes are unravelled. Gradually scientists have become aware that what is needed is an approach that will cope with the multiple and variegated nature of the factors that make individuals different from each other and the interactions that take place between those factors. This amounts to a systems approach.

Models of development

  • 4 Waddington, C. H. (1957), The Strategy of the Genes. Allen & Unwin: London.

4In a helpful visual aid to the biologist who has difficulty in grasping the abstractions of a mathematical model, Waddington4 represented the development of a particular part of a fertilised egg as a ball rolling down a tilted plane which is increasingly furrowed by valleys. He called the surface down which the ball rolls the ‘epigenetic landscape’. The essential point is that the mounting constraints on the way tissue can develop are represented by the increasing restriction on the sideways movement of the ball as it rolls towards the front lower edge of the landscape. The landscape represents, therefore, the mechanisms that regulate development.

5Waddington’s model is attractive to the visually minded because it provides a way of thinking about developmental pathways and about the astonishing capacity of the developing system to right itself after a perturbation and return to its former track. To take a specific example from post-embryonic development, if a juvenile rat is starved during its development, its weight falls while it is being deprived. When it is put back onto a normal diet, its weight curve rapidly picks up and rejoins the growth curve of the rat that has not been deprived. Similar examples of growth spurts after illness are well known in humans. For the moment the possibility that the individuals showing the catching-up phenomenon may differ in undetected ways from normal individuals can be ignored. The prime question is how weight gain is controlled and how two individuals with different dietary histories end up weighing the same.

  • 5 Capra, F. & Luisi, P.L. (2014), The Systems View of Life. Cambridge: Cambridge University Press, ht (...)

6The systems theorists have laid considerable emphasis on the self-correcting features of development, and have called the convergence of different routes on the same steady state ‘equifinality’.5 Waddington’s epigenetic landscape suggests a way of handling equifinality. If the ball rolling down the landscape encounters an obstacle in one of the valleys and is not stopped dead, it will ride up round the obstacle and fall back into the valley down which it had been rolling.

7Waddington’s model is, of course, informal and he would have been the first to point to its limitations. It is not difficult to simulate with greater rigour a system that compensates for short periods of food deprivation during development. If the amount of food an animal attempts to eat is determined by a comparison between a predetermined setting and the actual weight of the animal and if the value of the preferred weight is increased as the animal ages, a model similar to biological reality can be obtained. To make things more realistic the predetermined increments in the preferred value first increase and then decrease as the hypothetical animal gets older.

8By arranging for the preferred value of the closed feedback loop to be changed according to some predetermined plan, the system has been made interactive. In the simplest case the developmental process is essentially ballistic — its pathway is determined in advance and does not depend on a dynamic interaction between the system and other factors that might change during the course of development. Even rather simple explanations can account for different developmental routes leading to precisely the same steady state. The phenomena, which were so entrancing to an old-fashioned vitalist, do not pose inordinate conceptual problems.

9Children differ astonishingly in the age at which language development begins. Some may begin before the end of their first year and others may not utter a recognisable word until they are three or more. Furthermore, during acquisition, styles of forming word patterns may be markedly different. Despite these enormous differences, it is remarkably difficult to pick out the early developers when the children are older. Put cautiously, behaviour at one stage of development is an exceedingly poor predictor of behaviour at another. Put more boldly, a child which has been initially slow to develop can demonstrate the catch-up effect seen in tissue growth and reach the same steady state in one aspect of language ability as a more precocious child.

10This example from language acquisition can be matched by many others from child development, but it is sufficient to demonstrate not only the advantages but also the difficulties of employing the concept of equifinality in developmental studies of behaviour. Despite practical and philosophical difficulties, achieving equifinality in behavioural development does not pose insuperable problems of principle. Explanations for the control of weight can be readily adapted to behavioural examples. The preferred value against which the actual state is compared can be for, say, the proprioceptive feedback from a certain action or, at another level, the feedback provided by the behaviour of a parent. The justification for thinking in these terms is that it provides a different perspective from the more conventional interactional approach and suggests new ways of looking at the evidence.

  • 6 Hoyle, G. (1970), Cellular mechanisms underlying behavior-neuroethology. Adv. Insect Physiol. 7, 34 (...)

11Even a simple way of generating convergence could lead to marked differences in the pattern of development even though the final outcome was the same. Just as two rats with different food preferences can put on weight at the same rate, so different types of action can lead to the same behavioural end-point. A feature of a system dependent for its control on feedback is that it need not be fussy about how a match between the actual value and the preferred value is achieved. It is the consequences of an action that counts, not the precise form and patterning of that action. Admittedly, possible courses of action may be so constrained that the system is likely to do only one thing when a mismatch between the actual state and the preferred value is detected. The constraints need not be great. For example quite different combinations of muscles in a locust’s leg contract to produce the same overall movement of the leg — the explanation being that movement is controlled by means of sensory feedback.6

Alternative pathways

12Different routes to the same goal may be achieved even more dramatically than in the cases already considered if the young individual is equipped with two or more alternative systems controlling development of the same pattern of behaviour. Redundancy of this kind is common enough in man-made machines when lives are at stake, as in an airliner. Clearly, redundant developmental systems could be highly adaptive for an individual, particularly if the alternative control systems were matched to different environmental conditions to which they were appropriate—the provision of special horses for particular courses. The existence of other systems protects against failure, but from time to time individuals are faced with the situation where no amount of tactical manoeuvring will enable one of their developing systems to proceed along a particular route. Such an individual is a bit like a traveller who arrives at a station only to find that the trains have been cancelled. He or she can still reach their destination but only by choosing a different method of getting there.

13If contingency arrangements of this kind have been adapted during evolution, Waddington’s epigenetic landscape would have to be redrawn so that some valleys ran together again. It could be argued that a ball that had descended by one valley had had a different history from one that had descended by another so that even though the balls ended up in the same place, the concept of equifinality was valueless. This answer would depend on whether the different histories did indeed leave distinctive traces on the metaphorical ball. Even if they did, the objection might still not be serious since the resulting differences might be biologically trivial by comparison with the ultimate similarities.

14In other contexts, inputs that may be relatively non-specific are frequently required to facilitate the development of particular systems. The inputs may be provided by external environmental conditions or by feedback from the animal’s activities such as its own vocalisation. It may not yet be possible to give a clear instance where different developmental control mechanisms generate the same behavioural end-product.

15The biological function of some of the behavioural mechanisms found in many developing animals, particularly higher vertebrates, is the gathering of information. Their predispositions to learn the characteristics of certain things can be highly specific. Such proclivity can be extremely important in directing the course of development. A good example is provided by the active role of the young precocial bird in imprinting (discussed in Chapter 2). This example also illustrates the more general point about modifiability of control mechanisms.

16Another kind of modification dependent on environmental conditions is suggested by the stunting of growth if animals or humans are starved for long enough during development. The simple model used for the control of weight can be readily adapted to cope with such evidence by making the extent of the increments in preferred weight dependent on the size of the increments between the preferred weight and the actual weight. If the discrepancy is large the increment in preferred weight is modified so that it is less than if the discrepancy is small. This simple rule, which could be specified in advance, would greatly enhance the dynamic interaction with the environment. It would have one interesting consequence that would be particularly striking if the normal growth curve were sigmoid with the period of maximum growth occurring mid-way through development. The stunting effects of starvation would be particularly marked at times of rapid growth. This would give rise to periods in development when the animal was especially vulnerable to environmental disturbance.

Rules for changing the rules

17The biological advantage of a rule that allows for a change in the preferred value is that the animal does not endlessly attempt to reach a state that may never be achievable in the particular conditions in which it is developing. That the conditions for the development of one system are less than optimal does not imply that conditions are bad all round; normal development of the individual’s other systems may still be possible. Although it may be handicapped, its chances of surviving and leaving offspring may not be reduced to nothing.

18One behavioural example of settling for less than the best is the nest-site choice of the blue tit. In the spring the tits visit a large variety of crannies many of which are obviously unsuitable. One way of interpreting their behaviour would be that, if the actual site did not match up to the characteristics of an optimal nest-site, they kept searching — to begin with at least. If optimal sites were unavailable or already occupied, the birds would ultimately nest in places they had previously rejected. It would make good sense if they were equipped with a rule for gradually relaxing the conditions under which searching for a nest-site was brought to an end and nest building began. Once the bird has selected a sub-optimal site it will, for that breeding season at least, prefer it even if an optimal site should subsequently become available.

19The modification of preferred values can be examined in the context of emerging social relationships. Suppose that it is important for the maintenance of a relationship between two individuals that they both have the same general pattern of behavior — the same activity rhythm, for example. In the early stages of a relationship differences in pattern might well exist but these might reflect nothing more than the relatively unimportant peculiarities of personal history.

20It might be possible for one or both of the partners to change their preferred patterns without cost. If a pattern of behaviour is achieved by comparison with a preferred standard, that same standard could also be used for judging a companion’s behaviour. Individual A’s standard could be changed by individual B’s and vice versa. Any mismatch would lead to the individuals breaking-off contact with each other. It would therefore be necessary to provide for a mechanism that would, in the early stages of a developing relationship, over-ride or inhibit the consequence of a mismatch. For example, two individuals might be drawn together by the physical appearance of each other. During the ‘honeymoon’ period the relatively subtle differences of behaviour would be ignored. It would only be later, when the effects of physical appearances had started to wane that a mismatch of behaviour would become important and lead to a disruption of the relationship. In the intervening period one or both of the individuals could have changed its pattern of behaviour so as to correspond to that of the other. The flexibility of an individual might be constrained by some social roles and facilitated by others.

  • 7 Hinde, R.A. & Simpson, M.J.A. (1975), Qualities of mother-infant relationships in monkeys. Ciba Fou (...)

21An example of such behavioural meshing comes from observational studies of the relationship that develops between mother and infant rhesus monkeys.7 Independent measurements were made when the mother left her infant and when the infant left its mother. In some pairs the probability that the mother would leave the infant at any particular moment after they had come together was closely related to the probability that the infant would leave the mother. Such meshing could, of course, be obtained in a variety of ways. For example, the two individuals might become highly sensitive to the immediate cues provided by their partner. If other things are equal, and if apparent plasticity of preference is not merely elasticity, then the pattern of behaviour should be maintained for some time in the absence of the particular partner with which the pattern developed.

Mother rhesus monkey with her offspring. Photo by Laszlo Ilyes (2007), Flickr,​photos/​laszlo-photo/​495498455, CC BY 2.0.

22So long as a rule for changing a rule develops reliably in the individual, the outcome of Darwinian evolution is indifferent to how that came about. The outcome may have been arrived when certain environmental conditions were invariant and reaching adulthood successfully may depend on the maintenance of those conditions.

Coordination in development

23If internal mechanisms have developed, by some means or another, to control later stages of behavioural ontogeny, a considerable degree of coordination is likely to exist between different mechanisms. For example, the rates of development of two patterns of behaviour may be independently influenced by interactions with the environment; it may be important that the development of one does not outrun the development of the other. Alternatively, the order in which behaviour patterns develop may be important; for example, exploration of the environment may be disastrous if it occurs before a young animal has established some standards of what is familiar. In such cases acquisition of information must precede performance. Once a motor pattern producing the appropriate feedback has been established, dependence on feedback can be reduced or even eliminated and the animal can accelerate the output rate. This is a bit like a musicians learning a new part. While they are able to monitor the individual sounds they are making to ensure their accuracy, they must allow enough time between notes. In the final performance when such control is no longer needed, the gaps between notes can be reduced.

24The processes involved in plasticity can operate at many different levels, ranging from the molecular to the behavioural, some involving adaptability to what may be novel challenges and some responding conditionally to local circumstances. The results of development variation can be triggered in a variety of ways, some mediated through the parent’s characteristics. Sometimes variation arises because the environment triggers a developmental response that is appropriate to those ecological conditions. Sometimes the organism makes the best of a bad job in suboptimal conditions. Sometimes the buffering processes of development may not cope with what has been thrown at the organism, and a bizarre set of characteristics is generated. Whatever the adaptedness of the characteristics, each of these effects demonstrate how a given genotype will express itself differently in different environmental conditions.

25The contrasting properties of resistance to change and changeability — of elasticity and plasticity — are often found within the same material object. Stretch a metal spring a little and it will return to its former shape. Stretch it too far, however, and it will permanently take on a new shape. Adult humans, too, exhibit plasticity as well as elasticity in their values and personalities; they remain recognisably the same individual in a variety of situations, yet retain the capacity to change (see Chapter 4). Compare the robustness of most people in response to life’s buffetings with the way that some individuals profoundly modify their behaviour and attitudes. Continuity and change are not incompatible. The brains that generate behaviour do not consist of springs, of course, but the general property of getting back on track coexists with an ability to alter direction.

26The implication of examples such as these is that when certain conditions have been satisfied, new mechanisms of control can be brought into operation. In self-modifying systems, for instance, the conditions necessary for progressing to the next stage of development could be the levelling-off of modification — in other words, the achievement of a steady state. This type of explanation would side-step an unprofitable debate about the precise chronology of developmental stages. It would focus attention on the environmental conditions and on the state of the individual associated with a transition from one stage of development to the next rather than on age as such.


27The two points of view alluded to at the beginning of this chapter that have sometimes seemed incompatible can be made compatible. Far from being irreconcilable, the approaches of theorists interested in interactions and those interested in control mechanisms usefully complement each other. In brief, the development of behaviour often requires internal rules for its guidance, but reciprocity between the organism and its environment is also needed in order to give those rules greater flexibility and definition.

28Plasticity in response to different environmental conditions may often usefully reside in those mechanisms that determine action by matching actual input values with preferred values. The main point is that if individuals have rules by which their behaviour is controlled, functional reciprocity between the developing individual and its environment can be usefully achieved by equipping the animal with rules for changing the rules. This feature of development has all the appearance of being well designed.


1 Parts of this chapter were taken, with permission, from Bateson, P. (1976), Rules and reciprocity in behavioural development. In: P. Bateson & R.A. Hinde (eds.), Growing Points in Ethology. Cambridge: Cambridge University Press, pp. 401–421.

2 A discussion of the concept of innateness is given in Mameli, M. & Bateson, P. (2006), Innateness and the Sciences. Biol. Philos. 21.2, 155–188,

3 Lehrman, D.S. (1970), Semantic and conceptual issues in the nature–nurture problem. In: Aronson, L. Tobach, E. & Rosenblatt, J.S. Development and Evolution of Behavior. San Francisco: Freeman, pp. 17–52.

4 Waddington, C. H. (1957), The Strategy of the Genes. Allen & Unwin: London.

5 Capra, F. & Luisi, P.L. (2014), The Systems View of Life. Cambridge: Cambridge University Press,

6 Hoyle, G. (1970), Cellular mechanisms underlying behavior-neuroethology. Adv. Insect Physiol. 7, 349–444,

7 Hinde, R.A. & Simpson, M.J.A. (1975), Qualities of mother-infant relationships in monkeys. Ciba Foundation Symp. 33, 39–67,

Table des illustrations

Légende Mother rhesus monkey with her offspring. Photo by Laszlo Ilyes (2007), Flickr,​photos/​laszlo-photo/​495498455, CC BY 2.0.
Fichier image/jpeg, 18k