Campbell Purton's website
HOME Publications: BUDDHISM COUNSELLING FOCUSING PHILOSOPHY
Published in: Animal Behaviour 26 (1978) pp. 653-670
Abstract. The main purpose of this paper is to establish a bridgehead between contemporary ethology and some recent work in the philosophy of mind. Ethologists categorize behaviour according to its form, its function and its causation, but have not always been aware of the conceptual problems arising from the employment of several different conceptual schemes in one science. Philosophers of mind have recently discussed in detail the difference between thinking of human behaviour as action and thinking of it as movement, and I argue that this distinction is in some ways analogous to the distinction between behaviour categorized by function and by form. Wherever more than one conceptual scheme is employed in a science there is a pressure to run together concepts taken from the different schemes, and I show that such conceptual confusion in ethology can lead to obscurity, misrepresentation of experimental results, and the construction of unfruitful theories.
In this paper I shall try to relate an important ethological problem to the discussion of a similar sort of problem in current analytic philosophy. On the whole it is fair to say that ethologists are not familiar with contemporary philosophical discussions (though see Beer 1973a, p. 51; 1973b, pp. 47-49), and few philosophers have much familiarity with modern ethology, except through semi-popular works. This situation makes it worthwhile to point out that certain conceptual problems in modern ethology are closely linked with issues to which philosophers have recently devoted a great deal of thought. More specifically, ethology is devoted to the description and explanation of animal behaviour, but ethologists have in recent years been forced to some awareness of the ambiguities concealed in the term ‘behaviour’. Hinde (1966) and others have pointed out that the activities of animals can be classified in several different ways, so that what counts as ‘the same’ sort of behaviour according to one classificatory scheme may not count as ‘the same’ according to another classificatory scheme. For instance, courtship behaviour, defined in functional terms, may involve quite different sorts of causal mechanisms in different species, so that behaviour patterns that are functionally similar in that they exemplify courtship, may be of quite different types when categorized according to the sorts of physiological systems that give rise to them. In the same way behaviour patterns classified together functionally as appeasement displays, for instance, may involve, in different species, quite different sorts of movements or other physical changes.
To some extent the ethologist is bound to classify behaviour simply by its physical form (‘nodding’, ‘facing away’, ‘ventral roll’, etc.), but this sort of classification by form is distinct from both causal classification and functional classification. It would be a mistake to think that all appeasement displays involve, say, the hiding of conspicuous markings or the presentation of a minimum surface area to the opponent, since ‘appeasement’ is normally defined in functional terms. An appeasement display is usually taken to be a display whose function is to prevent attack, without eliciting flight, and this is quite compatible with the display of prominent markings. As examples, one might compare the ‘nodding’ behaviour of the sandwich tern which displays the black top of the bird’s head against a white background (van Iersel & Bol 1958) with the ‘greeting ceremony’ of the night heron in which a similar movement displays three white feathers protruding from a black back-ground (Lorenz 1938). These displays are formally similar, but the first is a threat display, the second an appeasement display.
Classical ethology had as one of its aims the preparation of ethograms for each species investigated. That is, the aim was to provide a detailed inventory of the ‘behaviour patterns’ of each species. The methodological principle involved was that one needs to know in detail what sorts of things animals do before one can begin any serious attempt to explain their behaviour. As a counter-balance to the experimental-theoretical emphasis of the comparative psychologists this approach was, and is, admirable. However, a difficulty is inherent in any programme that says simply ‘Describe what you see’. All observation is selective; indeed what one sees depends on the sort of questions one has in mind, and on one’s general background preconceptions. This is not to say that the request for ‘simple description’ is always misplaced, for it may be understood as a demand to get rid of as many theoretical preconceptions as possible; however, no programme of observation can be satisfactory if it eliminates all assumptions about what it is relevant to record (Popper 1959; Hampshire 1959). In point of fact, classical ethology was no freer from a special background ‘set’ or limitation of attention than was comparative psychology. The classical ethologists approached the problems of animal behaviour with specific questions in mind and specific expectations concerning the kind of explanation of behaviour which was to be sought. The main question raised by the comparative psychologists was that of causation, in the sense of that term which is associated with the discovery of regularities, laws, and underlying mechanisms accounting for the holding of the laws. (There was, of course, a division of opinion between those who were satisfied with the aim of establishing regularities in terms of stimulus and response, and those who insisted on making some reference to underlying mechanisms, which would account for the stimulus-response connections, but we may leave that issue aside.) The point is that comparative psychologists searched for the causes of behaviour, and if any question was raised about’ the meaning of the word ‘behaviour’ the answer would usually be given in terms of the movements animals make or the effects of these movements. (For example, the behaviour of a rat in a Skinner box might be defined in terms of the effect of a movement pattern, rather than in terms of its form. The behaviour pattern described as ‘bar pressing’ would be a good example.) By contrast with the comparative psychologists, the classical ethologists (e.g. Whitman, Craig, Heinroth, Lorenz, Tinbergen) were naturalists rather than experimentalists, and the question that guided their observations was not the question of causality so much as that of function. When confronted with an element of behaviour the question these naturalists were inclined to ask was ‘What is this for?’, and they expected an answer in terms of how the behaviour contributes to the life of the species.
For obvious reasons, questions about function can be asked and answered more satisfactorily in connection with behaviour that is ‘innate’ than in connection with behaviour that involves large components of learning, so it was natural that the classical ethologists should focus attention on those aspects of animal behaviour which can be referred to, vaguely, as ‘instinctive’. The result of this focus of attention was that the classical ethologists devoted their energy to the description and explanation of a range of behaviour quite distinct from that investigated by the comparative psychologists, and for many years the two schools continued their work in almost total isolation from one another. However, the publication of Tinbergen’s The Study of Instinct in 1951 and Lehrman’s (1953) critique of Lorenz’s theory did much to change the situation, and in the last 25 years the barriers between the two schools have been largely dismantled. As a result of the merging of the two disciplines much has been achieved, and several misleading dichotomies (such as that between the ‘innate’ and the learned’) have been overthrown; however, I want to suggest that in some respects the ‘blending’ of the two traditions of thought, which involves in part the blending of functional and causal conceptual frameworks, has not been entirely satisfactory, and this is one of the points I wish to discuss. In doing so I shall draw on a parallel discussion which has engaged philosophers of mind for some time.
It is well known that the psychological system of Hull was closely linked with a particular epistemological theory, namely, the logical empiricism which grew out of the earlier doctrines of logical positivism. As applied to animal behaviour the theory involved a sharp division between the behaviour (i.e. observed movements, etc.) of the animal and the factors (understood in terms of hypothetical constructs) which explained the behaviour. The behaviour was simply there to be observed; that part of the psychologist’s task was relatively unproblematical. The time and energy of the investigator were to be employed in constructing a hypothetico-deductive system which would account for the behaviour. Hullian theory, and much associated psychological theory, is so bound up with this picture of the task of psychology as to make any dissociation of the two absurd. Yet, as we have seen, in classical ethology the emphasis is on the descriptive aspect of the scientific task; if the description involving function, can be satisfactorily given there is no further, and separate, problem of explanation, since the more one is able to specify a behaviour pattern in terms of its function(s), the more understanding one has of ‘what it is for’. The descriptive part of classical ethology is not separable from the explanatory part in so far as the descriptions involve function. To describe a piece of behaviour as a. threat display, thereby saying something about its function, is in one sense to explain it, i.e. to render it comprehensible.
I have said that comparative psychology was closely linked with the epistemology of logical empiricism. I want now to suggest that there is a parallelism, though in this case no historical connection, between the methodology of classical ethology and the kind of epistemology deriving from Wittgenstein which has to a large extent replaced logical empiricism among contemporary philosophers. Logical empiricism begins with a ‘data base’ of either sensations (sense-data) or material objects and the description of these data is taken to be relatively unproblematic. The important thing is to ‘get behind’ the data via the ‘construction’ of hypothetical entities whose existence and nature will account for the sensations one has, or for the observable events in the physical world. What is to be explained can easily be described; the problem is with the explanation. In some of the later writings of Wittgenstein (1953, 1967) we find quite a different emphasis. Here the difficulty is seen to lie in giving an adequate description of the world. For instance, it is one thing to say that a man raised his arm, another to say that his arm went up, yet the observable physical events may be the same. The question arises, concerning any element of a person’s or animal’s behaviour: Is this an action, i.e. something the person did, or is it just a movement of his body, i.e. something which happened to him? (On this distinction between action and mere bodily movement see Hamlyn 1953, 1964; Peters 1958; Melden 1961; Charles Taylor 1964; Richard Taylor 1966; White 1968; Mischel 1969.) To wink at someone is an action, but to blink in response to a puff of air is not an action.
I think that most contemporary philosophers of mind would agree that there is no single, simple answer to the question ‘What is the difference between an action and a bodily movement that is not an action?’ They would agree too that we are not dealing with a simple dichotomy here, for there are many half intentional, habitual, compulsive, inadvertent or accidental bits of behaviour which do not fall easily on either side of the fence (Austin 1956). One might say that the difference between action and mere move-ment is that actions are ‘intentional’ or ‘purposive’ but this simply shifts the problem to the difference between purposive and non-purposive behaviour, and no appeal to differences in neural activity, etc. can help here, since it is clearly not on that basis that we distinguish actions from mere bodily movements. Some writers have tried to distinguish purposive behaviour from non-purposive behaviour in terms of ‘directiveness’ or ‘convergence of activity on a goal state’ (Sommerhoff 1950;
Braithwaite 1953; Nagel 1961) but these attempts have not been very successful. (For criticisms see Scheffler 1959; C. Taylor 1964; Hamlyn 1967). It would be out of place to discuss the notion of purpose in detail here, although I shall say a little more about it later; our main concern for the moment is with ‘action’ and ‘movement’ and the relationship between these two. I suggest that the relationship is in many ways parallel to that between functionally defined behaviour categories such as threat, and the movement patterns in which they are embodied. To say that someone signed a cheque is to say fairly precisely what he did qua action, but it does not specify in detail what movements he made. (One can sign a cheque with the pen held in one’s mouth, etc.) The same action can be embodied in quite different movements, and correspondingly in ethology the same functionally defined behaviour pattern (e.g. threat) can be embodied in quite different physical displays in different species. It is not the form of the display that makes it a threat display. The converse also holds: different actions may be embodied in the same movements, depending on the context. (More pedantically, though more accurately, different action-types may be embodied in the same movement-type.) For instance, what is a wink in one context may be an attempt to remove a speck of dust from the eye in another context. Correspondingly in ethology we may encounter similar movement patterns in two species which embody quite different behaviour, where ‘behaviour’ is understood functionally (e.g. a nodding movement of the head may be a threat in one species, appeasement in another).
One conclusion to be drawn from the preceding discussion is that it makes no sense to enquire about the ‘behaviour’ of a person or animal, unless it has previously been agreed what this term is to mean. (ironically, the great defect of behaviourist psychology lies in its lack of attention to the task of clarifying and making less ambiguous the use of the term ‘behaviour’.) If one wants an explanation of the person’s (animal’s) movements then normally the explanation will have to be given in terms of neurophysiology, input to the nervous system, feedback, etc. If one wants an explanation of a person’s (or perhaps animal’s) actions then normally the explanation will make reference to his beliefs, his general view of the world, his wants, intentions, plans, etc. (This is not to say that a neurophysiological account cannot be given of the movements embodying the person’s actions, although that view has sometimes been held (Taylor 1964; Malcolm 1968; Taylor 1970); the point is rather that even if we had a full physiological account of movement we would still lack any psychological account of the action.) Analogously, if one wants greater understanding of a functionally defined behaviour pattern such as a threat display, the explanation needed may, depending on the interests of the questioner, be one which gives not the mechanisms underlying the movements embodying the threat pattern, but a more detailed account of how it is that this movement pattern serves the function of discouraging attack, and how it is that discouraging attack in this way has survival value. Such explanations might specify in detail what the functions of the display’s components are, and how they contribute summatively to the function of discouraging attack, or they might show the further effects of attack-discouragement in the life of the species. In brief, such explanations fit the observed behaviour into its wider biological context, and in doing so they explain it.
I have been suggesting that what counts as ‘the animal’s behaviour’ depends to a large extent on the context of explanation one has in mind, that is, on the questions one wants to answer. In a sense, of course, there is no limit to the types of questions which can be asked concerning an animal’s behaviour. One specialist may be interested in questions concerning oxygen consumption, another in questions concerning whether certain hormones are active in the animal’s body, and so on. However, there are certain types of questions which are on a much more general level than these; questions which define the most general fields of. scientific enquiry and the most general categories in terms of which one can think about behaviour.
Four such questions are: ‘What is the form of this behaviour?’; ‘What is its function?’; ‘What is its cause?’; and ‘What is its purpose (in the psychological sense)?’ If we take ‘behaviour’ as a completely general notion covering anything from the behaviour of electrons in magnetic fields to the behaviour of prime ministers in parliament, then it will often be the case that not all of the four questions can sensibly be raised. In the case of inorganic things generally it will be inappropriate to raise the question of function. It makes no sense to ask about the function of the electron in the hydrogen atom, for instance, since the notion of function only gains a hold where we are dealing with a biological species reproducing under conditions which allow us to speak of the survival value of the organs and behaviour patterns of its members. (‘Function’ can be understood in other ways, it is true, but as I shall argue below the relevant ethological use of the term is best linked with ‘survival value’.) Similarly, it may be agreed that it would be foolish to ask about the aims or intentions of amoebae or bees, although just why this is so is a matter that requires extended discussion. (See Bennett 1964 for the beginnings of an account of the conditions that would have to be satisfied before we could seriously begin to ask questions about bees’ purposes in performing their dances.) Now while not all of our four questions may have application to a particular instance of behaviour, more than one may have, and it is this fact that lies at the root of the confusion which I shall now discuss. One is inclined to say that a piece of behaviour is at the same time a nodding of the head, a display whose function is to cause an opponent to flee and an aggressively motivated display, in the sense in which ‘aggressively motivated behaviour’ is behaviour caused by mechanisms which are also causally responsible for attacking, fighting, etc. However, suppose one does say that there is just one ‘piece of behaviour’ here, and suppose one labels it ‘threat behaviour’. One is then committed to employing a concept which may well turn out to have a very limited range of application, for it may be that head-nodding, the function of eliciting flight, and aggressive motivation are not often found together. The temptation then is to assimilate the concept of threat to concepts of the sort which Putnam (1966) has called ‘cluster concepts’. A cluster concept is one such that it applies to a thing just if that thing has many or most (but not necessarily all) of a cluster of characteristics which tend to occur together.
Thus a concept of a metal, such that substances count as metals if they have many or most of the characteristics: shiny, dense, good conductor of heat and electricity, malleable, ductile, would be a cluster concept. Things to which such a concept applies do not necessarily have any one characteristic in common; instead there is what Wittgenstein (1953) called a ‘family resemblance’ between them. The criteria for the application of such concepts are obviously multiple: a material will count as a metal if it satisfies any one of a set of criteria, each of which involves having a set of characteristics. Now I do not think that anyone can reasonably deny that cluster con-cepts exist and are important in science; what I do deny is that ethological concepts such as threat satisfy the basic condition for the employment of a cluster concept. That basic condition is that the set of characteristics must be known (or at least believed) to be associated in a lawful way. Thus we can work with the cluster concept of a metal because there is a lawful connection between a material’s having the characteristics A, B, C, D, E, and its normally having F; between its having A, C, D, F, and its normally having B; and so on. As chemistry has developed we have learned more about why these ‘low-level’ laws hold: it is a matter of chemical structure.
But the crucial point is that if we had no good reason to believe that they did hold, the cluster concept of a metal would have no use. Now consider ethological concepts such as threat and appeasement. The question is whether we can reasonably think of such concepts as cluster concepts in which some of the elements are formal, some functional. Consider some typical displays: (1) The nod-gackering of the sandwich tern (van Iersel & Bol 1958) which includes the elements (a) head-nodding (b) display of prominent markings (c) function of eliciting escape; (2) the tail-waving of felids which includes (a) sideways movements of tail (b) display of markings (c) function of eliciting escape; (3) the ‘greeting ceremony’ of the night heron (Lorenz 1938) which involves (a) head-nodding (b) display of markings (c) function of eliminating attack; and (4) the display of the dark neck-band by juvenile kittiwakes (Cullen 1957) which involves (a) sideways movement of the head (b) display of markings (c) function of eliminating attack. Now one could invent a cluster concept Q such that behaviour would count as Q-behaviour if and only if it has at least two out of three of the characteristics: head-nodding, display of markings, function of eliciting escape. The behaviour of the sandwich tern would be a ‘paradigm case’ of Q-behaviour, since it involves all three characteristics. We might then identify Q-behaviour as ‘threat’ behaviour, and this would fit the fact that the tail movements of felids (which count as Q) are recognized by ethologists to be threat. However, such an identification would not fit the fact that the night heron’s Q-behaviour is classified by
ethologists as ‘appeasement’. Similarly, we could define a cluster concept R such that behaviour counts as R-behaviour if and only if it has at least two out of three of the characteristics: head-nodding, display of markings, function of eliminating attack. The behaviour of the night heron would then be a paradigm case of R-behaviour, and we might be tempted to identify R-behaviour with ‘appeasement’. However the identification would be unsatisfactory, since it would not work in the case of the R-behaviour of the sandwich tern, which ethologists identify as ‘threat’.
My discussion here has inevitably been schematic, and it is intended merely for purposes of illustration, but it is important to see that difficulties of the sort just mentioned are almost bound to arise if we construct cluster concepts from a combination of formal and functional elements. For, the formal element(s) in the cluster fix the type of movement pattern (or other physical change) which must occur, while the functional element(s) allow room for enormous variation in the movement patterns. But one cannot have a satisfactory concept which both fixes the form and allows great latitude in the form of the behaviour. Correspondingly, one cannot reasonably expect that there will be any general laws linking functional characteristics with formal characteristics; it is in the nature of a function that it can be fulfilled in many different ways. This is not to say that there is no relation between form and function, but only that the relation varies from species to species. Nor is it to say that there is no way of combining form and function in a single concept. Clearly one may form the concept ‘behaviour pattern whose function is to elicit flight through
the display of prominent markings’. But this is not a cluster concept; it is basically a functional concept that contains, in the language of the older logicians, formal differentia. It is the concept of a kind of threat. Similarly, there are different kinds of eyes in vertebrates and invertebrates, the distinction being made within the functional concept of an eve on formal grounds. I would emphasize that this sort of combining of form and function is quite different from what I shall refer to later as ‘category conflation’.
In what follows I shall discuss first the blurring of the distinction between functional and formal categories; then I shall go on to discuss the conflation of functional with causal categories. Before considering the observational and experimental details, however, it would be as well to discuss briefly certain general points concerning the functional classification of behaviour.
First, it may be suggested that contrary to my account above, ethologists often do classify displays by form, rather than by function. There is a sense in which this is true; i.e. particular named displays of particular species or related species, such as the ‘Upright’ of most gulls, certainly exemplify a formal classification. However, I am concerned not with particular named displays but with functional types of display such as threat, appeasement, courtship, which may be observed not just in one species (or in a few related species), but in widely separated genera. It may be objected that even display-types of this sort are formally rather than functionally defined, and quotations from ethological writings can be produced which at first sight seem to favour such a view. For instance Carthy (1965) writes:
‘As might be expected, where aggressive display is found so also is that of submission, arid, as often happens, this is almost the inverse /of aggressive display. A submissive fiddler crab, for example, holds its body close to the ground, not raised as an aggressive crab holds his.’
Here it may seem at first sight that Carthy includes ‘holding body close to ground’ as part of what ‘submission’ means, but a little reflection shows that this is not so. Carthy succeeds, after all, in giving us factual information about the fiddler crab’s submissive display and he could not do this without presupposing some other account of what is to count as a submissive display. I do not mean to suggest that a completely sharp line can always be drawn between the characteristics of a behaviour pattern which make it count as that sort of pattern and other characteristics which it is empirically found to have; nevertheless, the absence of a sharp line does not eliminate .the importance of the distinction. The underlying philosophical question here is whether a useful distinction can be made between ‘analytic’ and ‘synthetic’ statements, and although some philosophers have questioned the possibility of the distinction (Quine 1953), there are good reasons for retaining it (Grice & Strawson 1956). Whether the distinction should be construed as a dichotomy is a separate question.
Whatever the philosophical niceties of the matter, it seems clear that in practice ethologists do distinguish between the criterion for a display’s being a threat and the criterion for its being, e.g. an Upright. Further, the criterion for its being a threat is almost always taken to be a matter of function. I have found that some ethologists object to this claim, so it would be well to document it by reference to the literature on the subject. For example, Moynihan (1955a) states that:
‘All attempts to arrange or classify the various hostile displays must be somewhat arbitrary in one way or another; but a classification based on functional criteria would seem ‘to be the most nearly ‘natural’ one, and the most convenient for our purposes.’
Tinbergen (1960, pp. 24 and 27) writes concerning certain postures of gulls, that when they are ‘shown by a territory owner in his territory, they have an intimidating effect; it is this effect which has led us to call them “threat displays’” and ‘an aggressive bird, meeting a bird which faces away, calms down. For this reason we have suggested the term “appeasement display” for this category.’ Hinde (1966, p. 12) writes:
the adaptive (in an evolutionary sense) con-
sequences they serve. “Threat”, “courtship”,
“hunting”, refer to categories of this type. Such
terms correspond to words such as “legs” or
“eyes” in morphology. A structure is classed as
a leg because of its function, and there is no
implication that the legs of, for example,
arthropods and vertebrates are related either
embryologically, i.e. causally, or phylogenetically.
A similar structure which has a different
function is given a different name; for instance,
“maxilliped”.’ -
Bastock (1967, p. II) writes:
Two displays are common in agonistic situations. Both are best defined in terms of their function. Threat displays tend to cause withdrawal on the part of the adversary; appeasement or submissive displays tend to reduce attacks.
Taken together, the import of these quotations seems clear: threat and appeasement, as display types, are to be defined in functional terms; the form of a threat display may vary widely from one species to another, but this is to be expected, just as it is to be expected that the form of a wing will be different in different groups of animals. What is essential to a display’s counting as a threat display is not its form but its function
of eliciting flight.
What I have said so far inevitably involves certain oversimplifications. I have, for instance, passed over the fact that Tinbergen (in the passage quoted) speaks of the ‘effects’ of certain postures, whereas Hinde speaks of ‘adaptive consequences’ and Moynihan and Bastock of ‘functions’. Now I think it is clear that to speak of the effects of a posture is not the same as to speak of its adaptive consequences or its functions, even though in practice one often identifies functions through observation of effects. The observation of ‘effects’ is itself not quite so straightforward as it sounds. For one is concerned with the usual effect, and this must be determined by many observations, bearing in mind that the same formal display may elicit different responses in males and females or in differently motivated males, and so on. However, within a suitably restricted range of circumstances it is often the case that a particular display has a quite characteristic effect on nearby conspecifics (e.g. it results in their flight); and when this is so it is very plausible to suggest that the function of the display is to cause flight. To
establish that this is its function (and not merely an accidental effect) one would have to produce evidence that the effect of the display has some survival value, that it makes some contribution to the life of the species. There is, conceptually, a
difference between a behaviour pattern which happens (even regularly) to produce a certain effect, and a behaviour pattern whose function is to produce that effect. No doubt the gackering threat pattern of the sandwich tern regularly produces certain characteristic patterns of disturbance in the air which may slightly ruffle the feathers of the threatened bird, but such regular effects are not part of me function of the display. It may be objected that such effects are not prominent, but suppose that a regular and prominent effect of a behaviour pattern could be shown to contribute nothing to the life of the species. One might then suggest that it was a relic from a period in the evolution of the species when such effects had survival value, and if this was plausible one would say that the function of the pattern used to be such-and-such, whereas now the effect is no longer a function of the behaviour. Alternatively, one might discover that the effect E in question is, for instance, a purely mechanical effect of a movement pattern which is itself functional, and which is maintained in the population since E has little or no deleterious effect. However, seriously misleading cases of this sort are the exception rather than the rule, and given this it becomes highly plausible to identify, though not define, function on the basis of ‘usual effect’.
One further point may be made regarding the difference between identifying and defining a function. It has recently been suggested by Hinde (1976) that a behaviour pattern has a function in a strong sense of ‘function’ only if, in addition to being biologically advantageous, the consequences of the pattern are such that natural selection may act through those consequences. For this to be so the consequences must be achieved to a varying extent amongst members of the population; hence, according to Hinde, consequences not subject to adequate variation will not count as functional (in the ‘strong’ sense) however essential they may be for the continued existence of the species. Now as an attempt to secure a more precise and useful meaning for the term ‘function’ this seems unsatisfactory. It is true that natural variability of consequences may allow the investigator to discover more easily what the function of the behaviour pattern is (e.g. he will not have to introduce artificial variations), but from the fact that lack of variability makes it difficult to identify the function it hardly follows that in such cases there ‘is no function in a strong sense’. The question of which consequences are the ones through which natural selection is operating is obviously an important one, but it is unhelpful to regard it as the same as, or as one aspect of, the question of what the function of the pattern is. As Tinbergen (1963) has emphasized, questions about function are linked with questions about ‘survival value’ and should be separated from questions about evolution and natural selection.
It is worth noting that the distinction between the effect of a movement and its function is paralleled by a distinction between the effect of an action and its purpose. From a common-sense point of view one might say that a rat’s purpose in pressing a lever in a Skinner box is to obtain food. Comparative psychologists, on the other hand, wishing to eschew such notions as ‘purpose’, would define this sort of behaviour by its effect. Here it may seem that an operational definition has been given for ‘the action of obtaining food’. But this is a superficial view of the situation. Not all actions defined in terms of their purpose of obtaining food are successful. For example, not all instances of ‘hunting’ (defined in terms of the purpose of capturing prey) are instances of ‘hunting’ defined in terms of the effect of prey capture. Actions classified according to their purpose may fail (and may often fail) in achieving their purpose, whereas to classify behaviour in terms of (usual) effect leaves no room for (more than occasional) failure. The same applies even at the level of description where we say that what the rat is doing is pressing the bar. Perhaps that is its purpose in moving as it does, but note that the effects of its movements include all sorts of things that are not part of the animal’s purpose, e.g. making the lever click, throwing a particular sort of shadow on the floor of the box, speeding up the circulation of blood in its body, and so on (Anscombe 1957). An animal’s purposes in acting form at best only a small subset of the effects of its movements, and the purposes informing some segments of its behaviour may not coincide with the effects of that behaviour at all.
Returning to the topic of the functional classification of behaviour, let us consider
another complicating factor (Beer 1976). It is clear that a single formally defined display may, at the same time, serve several different functions; and also that it may serve different functions at different times. For example, a particular sort of posturing by a gull may serve the functions of (a) presenting prominent markings to an opponent (b) causing the opponent to flee and (c) spacing out the population over the breeding area. This sort of case is analogous to that in which we ask for a man’s purpose in raising his arm and are told that it is (a) to reveal his clenched fist (b) to frighten someone and (c) to keep people away in accordance with his employer’s instructions. His overall aim is (c); he does (b) in order to do (c), and he does (a) in order to bring about (b). The relation here is that of ends and means. Similarly, in the ethological case the three functions of the gull’s movements are hierarchically arranged. The gull’s behaviour has the function of spacing out the population; it contributes to this function by causing the opponent to flee, and that is achieved by presenting prominent markings. If we ask what the gull is ‘doing’ the answer is that it is doing all these things, but it is doing (c) by doing (b) and doing (b) by doing (a).
Consider now a second sort of case where there is multiple function. This is where a behaviour pattern serves two or more functions at the same time, but the functions are not hierarchically arranged in a manner analogous to means and ends. For instance, the Kitter calls of tyrannid flycatchers (Smith 1963) serve to attract unmated females and repel intruding males. The vocal display serves two functions at once, and we may therefore say that it embodies both invitation behaviour and threat behaviour. This sort of case is like that where a man acts from more than one motive. He jumps into a river in order (a) to save a child’s life and (b) to display his courage and manliness to his girlfriend. In such cases of ’overdetermination of motives’ it would often be inappropriate to ask for a decision on which action he was ‘really’.
performing: He was saving a life and he was showing off; both types of action were embodied in his movements. It is plausible to suppose that most human actions are multiply motivated in this way, and it is equally plausible to suppose that most instances of animal display serve multiple functions. They have, presumably, evolved through the action of a whole range of selection pressures, so that it is unlikely that any formally-defined display will be a pure threat-display, i.e. will serve only that function. However, this does not mean that it is pointless to speak of ‘the function of a display’; Often it seems that one or perhaps two functions are dominant, i.e. it is in virtue of them that the display has most of its survival value, and in such cases there will be no harm in speaking of, say, a ‘threat display’ even though the movements involved contribute marginally to other functions. On the other hand, where two or more functions are clearly important a problem arises which highlights a point I am specially concerned with in this paper, i.e. that ethological description and explanation involve more than one conceptual framework. In this particular context we are apparently faced with the choice between speaking of one display-type that has two functions, and speaking of two display-types which are embodied in a single sort of movement pattern. However, the two alternatives are not really incompatible; there appears to be an incompatibility only if we assume that the term ‘display-type’ is univocal.
‘Display’, I take it, is a functional notion; a movement pattern that serves no communicatory function is not a display. But this does not mean that the criteria we employ in differentiating displays must themselves be functional. As I pointed out above, it is common in biology for a concept to be generically functional while the
species it encompasses are distinguished from one another on a formal basis. Hence there can be no objection to classifying displays according to their formal characteristics as Uprights, Head-flaggings, etc., and such a classificatory
scheme will be the one which is most valuable in studies of homology and phylogeny. This sort of classification gives us one clear sense to the phrase ‘type of display’; but there is also the possibility of a thoroughgoing functional classification. Here the display types will be ‘threat’, ‘appeasement’, ‘courtship’, etc. and as we have
seen there is no guarantee that displays which are of the same type according to this functional classification will be of the same type according to the formal scheme. It does not follow that we must ‘choose’ between the schemes; both are essential to a satisfactory understanding of display behaviour. What does follow is that one cannot construct a satisfactory ethogram for a species by ‘listing its displays’. One needs at
least two lists: a thoroughgoing functional list, and a list which involves function only in so far as one must refer to the function of communication in picking out patterns of behaviour as displays in the first place. If it is asked which of the classificatory schemes has priority, the answer can only be that it depends on what sort
of priority one has in mind. On the whole, ethologists first classify displays according to their formal characteristics and then begin to enquire about function. In this sense the formal system has a certain priority. On the other hand once a reasonably detailed knowledge of a species’ displays has been acquired, the formal characteristics will in some contexts be of only secondary interest. For instance, if one is investigating the frequency with which threat behaviour on the part of an animal is followed by
the flight of that animal, in different species, the formal differences between the various threats will be, for the moment, irrelevant. The classificatory scheme which is prior in one’s thinking depends on the sort of problem one is investigating: there is no absolute priority of one scheme over the other.
Both the cases of multiple function considered above were cases where a movement pattern in a given context serves more than one function. But there is a third sort of case of multiple function to be considered, i.e. that where the same movement pattern serves different functions in different contexts. (Note that ‘context’ here may be taken in a wide or a narrow sense. The narrow sense is that in which the ‘context’ is the rest of the displaying animal’s behaviour, the other movements which accompany the movement in which we are interested. The wider sense brings in such things as whether the display is directed at an intruder or a potential mate, or even such things as the time of year or whether the displaying animal is within its own territory.) For example Smith (1963) reports that Kitter calls in the Tyrannidae occur not only in the context I mentioned above (i.e. sexual invitation/threat) but also as appeasement displays during mating. That is, in this latter context, the call seems to serve the function of preventing attack without eliciting escape. Now is this sort of Kitter call the same sort of vocal display as the one discussed previously? The question is unanswerable until it is made clear which system of display classification we are employing; for the displays are of the same sort formally, but are not of the same sort functionally. So this kind of example, too, shows that we must modify the programme of constructing display ethograms simply by listing display-types which the species have: display-types cannot be counted until it is determined whether we are counting formal or functional types.
I hope that I have said enough to show that the notion of a ‘functionally defined behaviour type’ has certain complexities and raises certain difficulties which need further exploration, but at the same time I would suggest that these complexities and difficulties do not mean that the notion is of no value or that studies of
function are of less value than studies of causation. It is perhaps true to say that in the period following the classical period of ethology, i.e. from the mid-fifties to the mid-sixties, there was a shift of interest from questions of function to
questions of causality, but contemporary concern with the ‘socio-ecology’ of behaviour (Crook 1970; Lack 1968) has once more focused attention on function. (Studies such as that of Simmons (1970) on the behaviour of fish-eating birds clearly belong to the tradition of Huxley and Lorenz.)
I turn now to the point that the existence of different classificatory schemes in one science may give rise to a rather special set of problems. In particular there is a constant pressure towards running together, or conflating, categories drawn from the different schemes, so that it is no longer clear to which category a term belongs. The result of such conflation is at best obscurity, but at worst it can lead to real misunderstanding, misinterpretation of experimental results and the attempt to construct theories in ways which are bound to be unfruitful. I shall discuss first the sort of confusion which arises when formal categories are conflated with functional categories, and show that such conflation sometimes leads to the misinterpretation of experimental results. Then I shall discuss the even more serious functional-causal conflation and indicate briefly its unfortunate effects on ethological theory. I begin by considering a series of papers concerning a certain ‘hypothesis’ said to have been first put forward by Tinbergen (1959). The relevant passage is:
It is interesting that the effect of the “facing away” movement (and probably of many other appeasement signals) is not identical with that of threat displays. It is true that threat displays frequently have the same effect of reducing attack; but they do so by stimulating flight (i.e. by intimidation) which in turn suppresses aggression. Facing away, on the other hand, does not induce flight in the aggressor; it appears to inhibit both flight and attack. It seems to do this by bringing about the disappearance of some of the stimuli that release aggression and withdrawal.
This passage can be understood in more than one way, but some investigators at least (e.g. Dunham et al. 1968) have understood Tinbergen to be proposing the hypothesis that threat displays tend to result in the flight of the opponent, while appeasement displays inhibit attack. On the other hand, if Tinbergen’s own functional definitions of threat and appeasement are accepted (Tinbergen 1960, quoted above) it would follow that the ‘hypothesis’ is really an analytic truth which is not (logically) open to experimental investigation.’
Dunham et al. (1968) refer to Moynihan as another source of the hypothesis they wish to test. Moynihan (1955b, p. 70) writes: ‘Threat displays, of course, may also prevent attack. They seem to achieve this, however, by increasing the escape drive of the opponent.’ This certainly sounds like an empirical hypothesis, and it appears to be the same hypothesis as that attributed to Tinbergen, i.e. that threat tends to produce flight. On the other hand, if we look at Moynihan’s own explanation of the meaning of the term ‘threat’ (Moynihan 1955a, quoted above) we see that the function of causing flight is written into the meaning of ‘threat’ (Moynihan’s definition in his (1955b) paper is different but I shall ignore this complication for the moment). I think that anyone who studies the papers just mentioned, by Tinbergen, Moynihan, and Dunham et al., will agree that there is present quite needless obscurity, and further that the source of this obscurity lies in the fact that it simply is not clear whether function is written into the meanings of ‘threat’ and ‘appeasement’ or not. I shall now try to show that the situation is rather worse than this. In a paper by van Iersel & Bol (1958) the statement that threat displays tend to result in flight whereas appeasement displays merely prevent attack is clearly presented as an empirical hypothesis. These author write that ‘The effect of a threat movement... is apparently an increase in the escape drive which inhibits the aggression . . . whereas the effect of an appeasement movement, though virtually the same (stopping aggression) does not result from inhibition of aggression via the increased escape drive, but from the non-activation of aggression’ (van Iersel & Bol 1958, pp. 48-49). In their work with terns they claim to have demonstrated the first effect quantitatively and suggest that further research should be conducted in order to test the second part of the hypothesis (p. 49). I shall first consider van Iersel and Bol’s claim, and then consider further work on testing both parts of the ‘hypothesis’.
The first part of the hypothesis is that threat behaviour operates by increasing the escape drive of the opponent, which inhibits his aggression. The behaviour investigated is ‘gackering’ and ‘nod-gackering’ in the sandwich tern, and it must first be established that these behaviour patterns are threat displays. The authors do not collect their evidence for this in one place, but from their discussion the following arguments can be picked out:
(a) In the case of nod-gackering the tern’s black cap is presented to the opponent, and it is known that the display of such a prominent mark is often a threat display. The difficulty with this argument is that there are, as I mentioned previously, displays involving the presentation of prominent markings which, far from being threat, are in fact appeasement. Threat displays cannot be identified conclusively on formal grounds; the criterion for threat is a functional one. Van Iersel and Bol themselves do not seem to come to any clear decision about whether ‘threat’ is a functional or a formal concept; they write that ‘By the “nod” the bill is turned away from the opponent and points to the ground; the movement might therefore have appeasement function, as the weapon is more or less with-drawn. At the same time, however, the black cap is turned towards the opponent. The contrast of the black cap with the white plumage could very well indicate a threat structure’. This seems, at first reading, to involve a formal understanding of ‘threat’ and ‘appeasement’. Van Iersel and Bol are saying that withdrawal of the bill is likely to be appeasement whereas display of prominent markings is likely to be threat. But the ‘could very well’ is a revealing qualification. What would show that the display is threat, or what would show that it is, after all, appeasement ? Here I think the answer can only be ‘Its function’.
(b) When one tern gackers or nod-gackers at another the second bird frequently shakes its head. Van Iersel and Bol argue (how plausibly I shall not discuss) that this head-shaking is an indication of an activated escape drive. They then say: It is hard to avoid the conclusion that the bending movement [nod] causes a relative increase of the escape drive, and thereby the chance of displacement in the opponent. The effect produced is further evidence that. the bending is a threat movement (as was argued above)’ (p. 48). Here it seems clear that the bending or nod is to be counted as threat because it tends to arouse escape.
(c) Perhaps a third argument in favour of gackering and nod-gackering being threat displays could be read into van Iersel and Bol’s comments ‘Gackering is a relatively aggressive activity of the Sandwich Tern.. . it is an ambivalent activity, not being a direct attack, though it may merge in to bill fighting’ (p. 42). It might be suggested that a display is a threat display in so far as it is activated by the same causal system which activates attack, i.e. in so far as it is an ‘aggressively motivated’ display.
Yet this is not really consistent with van Iersel and Bol’s account: they hold that gackering is ambivalently motivated by aggression and
escape. Perhaps, then, threat behaviour should be understood to be ‘display behaviour motivated by both attack and escape’? This was Moynihan’s (1955b, p. 20) suggestion as an alternative to a functional definition. But is any sort of behaviour thus doubly motivated going to count as threat ? Irrespective of its effect on other animals ? This does not seem to be a very plausible suggestion. What may well be true is that threat behaviour is often doubly motivated, just as it often does involve the display of prominent markings.
I conclude from the above discussion that van Iersel and Bol do not have any satisfactory criterion for threat behaviour apart from its function. Neither argument (a) nor argument (c) above is satisfactory, and argument (b) explicitly gives function as the defining characteristic of threat. Now let us consider the quantitative demonstration, which concerns the correlation of gacker threats with the head-shaking movements of the opponent. Van Iersel and Bol give a table of results which shows conclusively that as the gacker intensity increases, so does the frequency of occurrence of headshaking. Hence they conclude that the hypothesis that threat operates by increasing the escape drive of the opponent is confirmed. However, it is clear that this interpretation of the results table is impossible if the reason for counting gackering as threat consists in the fact that gackering tends to elicit escape. The question is, ‘What is the criterion for a display being a threat display?’, and as we have seen van Iersel and Bol do not provide any satisfactory criterion other than the standard one of having-the-function-of-producing-flight. Yet given the standard criterion their results-table does not show what they say it shows. What the results-table does show is that gackering and nod-gackering are in fact threat displays in this species of tern. The nod movement could well have been appeasement, but the results show that it is in fact threat.
At this point one may confidently predict that any other experimental tests of the Tinbergen-Moynihan-van Iersel ‘hypothesis’ will turn out not to be tests of the hypothesis at all, for there is no such hypothesis to be tested. Such a prediction is borne out by a consideration of the investigation by Dunham et al. referred to
above. These authors state at the beginning of their paper on the fish Barbus stoliczkanus (pp. 15-16) that their work is designed as a test of the hypothesis in question, i.e. that threat tends to cause escape, while appeasement tends merely to prevent attack. The experimental work they report is alleged to confirm the
hypothesis and the work is summarized by the authors as follows:
Study of agonistic encounters between male
Barbus stoliczkanus in aquaria shows that both
threat and appeasement lower the probability
of a fish being attacked. Correlation between action
and reaction indicates that threat inhibits attack in the
reactor by increasing his tendency to flee. Reaction to
appeasement indicates a “loss of interest” by the reactor, and
shows no evidence of inhibited attack or increased readiness to flee.
If we look at the authors own explanations of the terms ‘appeasement’ and ‘threat’ we see that the first paragraph of the Summary is simply
False. For these explanations run as follows:
Attack is much commoner as a reaction to
(fast) fleeing than as a reaction to (fast) fleeing
+ Roll, and therefore Ventral Roll lowers the
probability of attack by the reactor. Since
Ventral Roll appears to do this by presenting a
small surface area to the opponent, it can be
considered as an appeasement display. Spread
per se lowers the probability that the displaying
animal will be attacked. Since it is also a display
of conspicuous markings and presents a maxi-
mum area to the opponent, it can be considered
a threat display.
These explanations make it clear enough that lowering the probability of attack is written into the meanings of ‘threat’ and ‘appeasement’, but if this is so then the first paragraph of the Summary does not report any experimental
result at all. (And it will not do to regard this as simply a slip of the pen: the authors state quite explicitly that they are following up van Iersel and Bol’s recommendation to test just that empirical hypothesis.) Given the authors’ explanations the second paragraph of the Summary does follow, but it would not be acceptable as an experimental result to anyone who accepted the more usual functional definitions of
‘threat’ and ‘appeasement’. Dunham et al. make the point that threat and appeasement have similar functions (in that both lower the probability of attack) and then go on to distinguish between them on formal grounds. However, the formal distinction is unsatisfactory for reasons that are by now familiar, and in fact there is no great difficulty in making the distinction on functional grounds: the function of threat is to produce escape in the opponent, whereas the function of appeasement is simply the elimination of the opponent’s attack tendency, without activating any other tendency in him. These functional definitions are, as Moynihan states elsewhere (1955a), the most natural in ethology, but if we accept them we find that the second paragraph of Dunham et al.’s Summary becomes, like the first paragraph, true by definition. Hence what Dunham et al. take to be the results of their investigation cannot be regarded as the real results.
What then should we conclude from this investigation of B. stoliczkanus? I think the
following: First, the initial conclusion (not given in the Summary) which the authors draw from their results table remains valid. Namely, that since the figures show that Spread and Ventral Roll reduce the probability of attack, we can identify them as either threat or appeasement displays. Secondly, the table of results shows that it is Spread which is the threat display and Ventral Roll which is the appeasement display, since the first tends to cause flight and the second produces no special reaction. Thirdly, we find that in this species the threat display involves the showing of markings, while the appeasement display involves the concealment of markings. Fourthly, the investigation can only be said to confirm the Tinbergen-Moynihan-van Iersel ‘hypothesis’ if threat and appeasement are defined in a way that is ethologically unnatural and which is not clearly attributable to any of the cited originators of the ‘hypothesis’. If we change the hypothesis under test from ‘threat tends to result in increased probability of escape’ to ‘displays involving the presentation of prominent markings tend to result in increased probability of escape’ the investigation confirms the new hypothesis, but this hypothesis is not a consequence of the original hypothesis (given the standard ethological understanding of the terms involved). So we end with the dilemma that either (1) the reported investigation of B. stoliczkanus is irrelevant to the hypothesis it was supposed to test, or (2) it is relevant, but is conceptually muddled because the hypothesis is conceptually muddled.
Now it may be objected that in pursuing this line of argument I have taken too narrow a view of ethological terms such as ‘threat’ and ‘appeasement’. It may be said that part of the point of calling a display a ‘threat display’ is to indicate that it is aggressively motivated, rather than being motivated by sex or fear, for example. I referred above to van Iersel and Bol’s arguments for regarding gackering and nod-gackering in the sandwich tern as threat, one of which was that these behaviour patterns sometimes merge into bill-fighting. The suggestion here is that gackering and nod-gackering result from the operation of the same causal mechanisms which underlie bill-fighting, and that this common aggressive motivation is relevant in classifying the gackering behaviour as threat.
This suggestion leads us to a discussion of a second sort of category conflation in ethology, for what is being invoked here is a third principle of classification, distinct from the formal and functional principles. This third principle of classification is that behaviour patterns should be grouped together if they share common underlying mechanisms; it is the principle, or rather, as we shall see, one of the principles that
is involved in classifying behaviour according to the various ‘drives’ or ‘instincts’ that are involved. Of course ‘drives’, as underlying causal systems, are not entities which are in ordinary circumstances open to our view; the reason for postulating them must rest largely on observed correlations between the occurrences of various
elements of an animal’s behaviour. For instance, in so far as scraping, sideways building and collection of nest material tend to occur together in a species of gull we may infer that there is a single causal system in the bird (a ‘nest building’ system) which is responsible for the fact that the activities tend to occur in association with one another. Similarly, quivering, settling and ruffling may occur together and thus be regarded as manifesting the activity of an underlying ‘incubation system’ (Moynihan 1953,1955b). Now as Hinde has insisted, this ‘causal’ classification of behaviour is logically distinct from any functional classification, yet in practice the two are often run together. In practice ethologists have only rarely, and only fairly recently (Beer 1963a), calculated correlation coefficients in order to assess whether a group of behaviour elements ‘goes together’; instead, they have noticed that certain groups of
elements come together in serving a particular function. Thus scraping, sideways building and collecting appear to serve the common function of nest construction; quivering, sitting and ruffling serve the common function of incubation. It is then assumed that the functional classification corresponds to a causal classification, i.e. that there will be a common causal mechanism for the movements manifesting the nest-building ‘instinct’, and a common causal mechanism underlying the movements manifesting the incubation ‘instinct’. However this does not follow, and in some cases there is clear evidence to the contrary. For example Beer (1963a) found that the presence of an egg-model on the nest ‘increased the frequency and perhapsthe average intensity . . . of sideways building movements but did not increase the numbers of collecting trips or their average intensity’; that ‘significantly more collecting trips were preceded by performance of food-begging than was the case for settling and sideways building’; and so on. Such lack of correlation between the behaviour elements speaks against any theory which proposes a single underlying ‘nest building’ drive instinct or motivational system. The fact that the behaviour elements form a unity from a functional point of view is undeniable, but it does not follow that there is a corresponding causal unity. The causal story, it seems, must be more complex, and Beer gives some interesting hints about the kind of detailed analysis that may be required.
In the light of papers by Hinde (1956, 1960) and others many ethologists would now agree that the picture of behaviour as energized by ‘drives’ is seriously misleading. However, one still hears a good deal of talk about ‘motivation’. The question arises of what remains to be referred to by this term once the drive picture has been abandoned. One response to this question is to identify ‘motives’ with the causal systems whose existence is revealed by correlational (or physiological) studies. If it is true that certain
patterns of movement, etc. ‘go together’ then there is presumably a common causal mechanism. This hypothetical mechanism can be referred to as a ‘motivational system’ and the details of its physiological embodiment may one day be worked out. However, it should be emphasized that to use the term ‘motivation’ in this way is a far cry from the classical sense in which the ‘motives’ of an animal are constituted by such categories as nest-building, courtship, aggression, and so on. The classical motives were based largely on a functional classification which was assumed to correspond neatly with a causal classification. But now, as we have seen, there is good reason to think that the functional classification cuts across the causal classification, so that while we can speak of ‘motivational systems’ in a causal sense, these systems may not
be such that they can usefully be named ‘nest building system’, ‘aggressive system’, ‘courtship system’, etc. In the case of Beer’s (1963a) black-headed gulls, for example, the evidence suggests that there simply is no motivational system (causally defined) which can properly be described as a nest building system.
One conclusion that seems to follow from this is that if the notion of drive is to be phased out of ethology then so should the notion of motivation, unless this latter term is reinterpreted so as to refer to purely causal categories of behaviour. Yet such a suggestion seems almost to border on the absurd; surely, it may be said, theclassical account was not totally wrong in picturing motivation in the way it did. Some comments by Kennedy (1954) are relevant here.Kennedy suggests, I think rightly, that the general conceptual framework for thinking about animal behaviour developed by Craig and followed by Lorenz and Tinbergen does not derive simply and solely from observation of animal movements and their functions. In addition it has roots in our ordinary way of thinking about people in terms of their purposes, desires or motives. Kennedy regards such notions as ‘subjective’ and consequently regards classical ethology as permeated by subjectivity. Many contemporary philosophers, on the other hand, while accepting that our ordinary way of thinking about human behaviour involves essentially such notions as purpose and motive, reject the traditional view that such notions are, in the traditional sense,’ subjective’. (Wittgenstein’s (1953) arguments against the possibility of a ‘private language’ are crucial here; for a discussion of this see Malcolm 1954.) Philosophers in the Wittgensteinian tradition would reject the idea, commonly accepted by ethologists (Lorenz 1950, pp. 263-264; Tinbergen1951, pp. 4-5), that having purposes, in the psychological sense, is a matter of ‘having subjective experiences’. They would argue that if purposes were purely subjective, i.e. ‘private’, ‘open only to introspection’, then one could logically have no reason to think that anyone but oneself had them. But this is an absurd view; if one knows anything one knows that other people have purposes, intentions, beliefs and so on. How we know, what the criteria are for the application of these concepts, is a matter
for detailed investigation, but that we know cannot coherently be doubted. (On this aspect of the matter see Strawson 1959.) The view many contemporary philosophers is, then, something like this: we understand other people’s action in terms of their purposes, beliefs, etc. and there must be criteria for the application of these concepts to particular individuals in particular situations. However, what the criteria are is often context dependent and not obvious, for we learn to apply our non-technical concepts in contexts which we do not normally learn to describe. (Consider, for example, the action-qualifying terms ‘by accident’ and ‘by mistake’ (Austin 1956); normally speakers of English use these terms in ways which show clearly that there are two quite distinct concepts involved, but few people can say without careful thought what the conditions (criteria) are which make it appropriate to use one rather than the other.) Determining just what are the criteria in accordance with which we apply our psychological concepts is an important task of the philosopher of mind,
although another part of his task may be to recommend changes in our conceptual scheme in the light of new empirical discoveries. No doubt our conceptions of human behaviour will change as psychology develops, but it also seems undeniable that at present we do have a reasonably coherent, and indeed extremely subtle, everyday conceptual framework in terms of which we describe and interpret human activities, which is distinct from the causal and functional frameworks I have been
considering up to now. Given that we naturally think in terms of this framework of action, belief, purpose, etc. it is inevitable that when we observe the behaviour of an animal in a ‘naturalistic’ way, without bringing to our observations any special background ‘set’, we will tend to see the animal as acting, as having purposes, beliefs etc. The question then arises, in any particular case, of whether we are being reasonable in so thinking about the animal’s behaviour. Is it really the sort of creature to which these psychological concepts apply? In some circumstances there will be every reason to refrain from thinking of an animal’s behaviour in psychological terms. For instance, if the behaviour turns out to be rigidly stereotyped, we will have to stop seeing it in terms of purpose, since purposive behaviour is essentially variable. Similarly, although the details of this story are much more complex, we cannot think of an
animal as making judgments (and hence as having beliefs) unless it is the sort of creature which can have some appreciation of the right and wrong ways of doing things, and it is hard to see how any non-social animal could have such an appreciation. Or if we want to know whether bees really communicate with each other we
need to look closely at the criteria which must be fulfilled if a creature is to be said to have a language (Bennett 1964). In psychology and ethology one wants to avoid thinking of creatures in terms of concepts that are not really applicable to them, but this is not to say that all use of concepts such as belief and purpose in ethology is ‘anthropomorphic’. If one is interested in the behaviour of human children, for instance, the idea of ‘anthropomorphizing’ is terminologically absurd, and if, as seems not implausible, some of the higher mammals do satisfy the criteria for the application of our concepts of action, belief, purpose, etc. then it is not ‘anthropomorphic’ to describe their behaviour in such terms. What should be resisted is the temptation
to try to blend our ordinary notion of action with the notion of movement, or our ordinary notion of purpose or motive with some causal or functional notion that is at home in a quite different conceptual framework. Hence I am not suggesting that in describing the behaviour of, say, primates, we should totally abandon our ordinary concept of motivation; I am saying that if we do employ this concept we ought to be aware of the fact that it belongs to a conceptual framework that is distinct from those that centre around questions of function or physical cause. I suggest that it was a lack of such awareness which was, in part, responsible for the development of the classical ethological picture of behaviour motivated by ‘drives’: For the natural way of seeing animal behaviour is in terms of purpose, etc.; the observer is then bothered by the fact that ‘purpose’ seems to be too ‘subjective’ a notion to have any place in
scientific ethology, so while he continues to speak of ‘motivation’ he in practice insists on classifying behaviour by its function, and assumes that corresponding to the functional categories of behaviour (nest building, aggression, etc.) there are causal categories which can be called ‘motivational categories’. The confusions here are multiple, and they interlock with one another in a way that almost defies
analysis; however, I am convinced that something like the picture I have sketched lies at the root of the ethological controversy over the validity of concepts such as ‘drive’ and ‘motivation’.
Before concluding I should like to draw attention to one more specific case where it is
clear that category conflation has caused serious difficulties in the construction of ethological theories, namely, the case of ‘displacement
activity’. Beer (1963b) and Kruijt (1964) have already drawn attention to this matter, and their discussions are very relevant to my general thesis. The notion of displacement activity logically involves the notion of an activity that is ‘irrelevant’ or ‘out of context’. Thus the ‘digging’ movements of a male stickleback
defending its territory (Tinbergen & van Iersel 1947) seem ‘out of place’, and the classical account of such behaviour was that it was due to the mutual frustration of two conflicting drives (attack and flight), neither of which could gain the ascendency. According to the original displacement theory (Armstrong 1950) the energy dammed up by the conflict ‘sparks over’ into the displacement activity; according to a
later theory (van Iersel & Bol 1958) the tendency to perform the displacement activity is always present, but it is normally overridden by the presence of other, stronger tendencies. When these other tendencies come into conflict, however, they neutralize one another, thus allowing the weaker ‘displacement’ tendency to
manifest itself. Now in the case of either of these accounts it is obvious that we must, at the start, be able to identify ‘out-of-context’ behaviour when we come across it. In practice this is done by noting that certain behaviour is functionally irrelevant, i.e. it does not appear to fulfill the function which such behaviour normally has.
The digging of the stickleback is functionally relevant when it occurs in the context of nest building; it is functionally out of context when it occurs in a situation of confrontation with an intruding male stickleback. If we now assume that functional categories correspond with formal categories we will be led to think that the
displacement activity is causally out of context, and that it therefore requires a special causal explanation, such as that provided by the ‘sparking-over’ account or by the ‘neutralizing of other drives’ account. However, once it is remembered that causal categories may not correspond with functional categories such hypotheses may well become redundant. The functionally ‘out of place’ displacement activity may not be causally ‘out of place’ at all. Suppose that behaviour elements A, B, C and D ‘go together’ functionally as (functionally defined) P-behaviour; and suppose that elements E, F, G and H ‘go together’ as (functionally defined) Q-behaviour. Correlational studies may then show that the occurrence of element B correlates more highly with the occurrence of E and F than with the occurrence of C and D. This would be evidence that B belongs to the same causal system as E and F, and would make it quite
understandable that B should fairly frequently occur in the midst of Q-behaviour, where
it would be functionally irrelevant. However its causal background would be quite normal; no special causal explanation would be required for it. (Of course, functionally irrelevant behaviour will tend to be non-adaptive, and for
that reason one might expect it to be the exception rather than the rule; on the other hand through ‘ritualisation’ such behaviour may come to be adaptive after all, thus acquiring a function of its own.) I have discussed the notion of displacement activity in an abstract, schematic way for reasons of brevity; however, reference
to the papers by Beer and Kruijt mentioned above will confirm that in fact functional and causal categories do not always coincide, and that behaviour which is functionally out of place may be just what is to be expected causally. As a consequence, the classical concept of a ‘displacement activity’ will, in such cases, have no application.
One further point should be made here. Ethologists have noted that there are cases of human behaviour which seem to fall under the heading of ‘displacement activity’. In conditions of conflict people often perform seemingly irrelevant actions, and it has often been held that these cases are more than just analogous to the cases of displacement behaviour first observed in birds. However, while the discovery of ‘out of context’ behaviour in birds and other animals has helped to draw our attention to ‘out of context’ behaviour in human beings, there is no guarantee that the two phenomena are more than superficially similar. The context referred to in the case of animals is (at least ostensibly) a functional context, whereas in the human case it is a purposive context; the action is ‘out of place’ in that it does not fit into the pattern of
the agent’s purposes at the time. For instance, someone wants to object to a remark that has just been made, and he wants to avoid giving offence, the result being that he rubs his nose or drums his fingers on the desk. Such behaviour is ‘out of context’ from the point of view of his purposes, and from that point of view certainly
calls for an explanation. But only category conflation could make one suppose that the
explanation must be of the same sort as that which accounts for the displacement digging of the three-spined stickleback.
Finally, to summarize what has been a rather long and complicated discussion: it seems to me that ethology is a science which employs, and must employ, several fundamentally different conceptual schemes. One of these schemes is developed around questions of form, another around questions of function, and a third around
questions of causality. In addition, a fourth scheme plays an important background role,
i.e. our ordinary scheme of talk about human actions which centres around questions concerning motive or purpose. These four frameworks are conceptually distinct; they organize the data in different ways so that what counts as ‘the same’ sort of behaviour may be different in each of the schemes. On the other hand there are links between the schemes, some of which are logical, some empirical, and some theoretical. Logically, an animal cannot perform an action without making some sort of movement, or emitting some sort of noise, etc. Empirically, we may find that threat displays always involve the display of prominent markings. From the point of view of the theory of evolution one would expect that, on the whole, functionally defined categories should roughly correspond with causally defined categories of behaviour, since
one would expect this to be a more ‘economical’ arrangement (Hinde 1966, p. 13). That such connections exist is obvious; the important thing is that the existence of the connections should not lure us into conflating concepts drawn from the different schemes; that is, we should not let ourselves get into the position where we use a term such as ‘threat’ or ‘displacement’ sometimes with a functional meaning, sometimes with a causal meaning and sometimes with a motivational meaning, shifting
from one to the other while pretending all the time that we have not moved at all.
Acknowledgments
I am very grateful to Dr C. (J. Beer for his comments on an earlier version of this paper.
Acknowledgment is also due to Professor D. W. Hamlyn who supervised my London M. Phil. thesis which was a previous attempt to deal with the problems discussed here. My thanks are due also to Professor Konrad Lorenz for the opportunity to visit the Max-Planck-Institut für Verhaltensphysiologie at Seewiessen, to Professor H. R. Post who first drew my attention to the philosophical problems of ethology, and to my wife Valerie for her help in translating curious writings on seagulls which had not at
the time appeared in English.
Anscombe, G. E. M. 1957. Intention. Oxford: Basil Blackwell.
Armstrong, E. A. 1950. The nature and function of displacement activities. Symp. Soc. exp. Biol. 4, 361-384.
Austin, J. L. 1956. A plea for excuses. Proc. Arist. Soc. 57, 1-30. Reprinted in White (1968).
Bastock, M. 1967. Courtship. A Zoological Study. London: Heineman.
Beer, C. G. 1963a. Incubation and nest building behaviour in Black-headed gulls. III: the pre-laying period. Behaviour, 21,13-77.
Beer, C. G. 1963b. Incubation and nest building behaviour in Black-headed gulls. IV: nest building in the laying and incubation periods. Behaviour, 21, 155-176.
Beer, C. G. 1973a. A view of birds. In: Minn. Symp. Child Psychol., 7 ,47-86.
Beer, C. G.1973b. Species typical behaviour and ethology. In: Comparative Psychology. A Modern Survey. (Ed. by D. A. Dewsbury & D. A. Rethlingshafer pp. 21-77. New York: McGraw-Hill.
Beer, C. G. 1976. Multiple functions and gull displays. In: Function and Evolution of Behaviour (Ed. By G. Baerends, C. G. Beer & A. Manning), pp. 16-54. Oxford: Clarendon Press.
Bennett, J. 1964. Rationality. London: Routledge & Kegan Paul.
Carthy, J. D. 1965. The Behaviour of Arthropods. London: Oliver & Boyd.
Crook, J. H. (Ed.) 1970. Social Behaviour in Birds and Mammals. London: Academic Press.
Cullen, E. 1957. Adaptations in the kittiwake to cliff nesting. Ibis, 99, 275-302.
Dunham, D. W., Kortmulder, K. & van Iersel, J. J. 1968. Threat and appeasement in Barbus stoliczkanus. Behaviour, 30, 15-26.
Grice, H. P. & Strawson, P. 1956. In defense of a dogma. Philosoph. Rev., 65, 141-158.
Hamlyn, D. W. 1953. Behaviour. Philosophy, 28, 132-145
Hamlyn, D. W. 1964. Causality and human behavior. Arist. Soc. Suppl. Vol., 38, 125-142.
Hamlyn, D. W. 1967. Critical notice of C. Taylor (1964) Mind, 76, 127-136.
Hampshire, S. 1959. Thought and Action. London: Chatto & Windus.
Hinde, R. A. 1956. Ethological models and the concept of drive. Br. J.Phil. Sci., 6, 321-331
Hinde, R. A. 1960. Energy models of motivation. Symp. Soc. exp. Biol., 14, 199-213.
Hinde, R. A. 1966. Animal Behaviour. New York: McGraw-Hill.
Hinde, R.A. 1976. The concept of function. In: Function and Evolution of Behaviour (Ed. by G. Baerends, C. G. Beer & A. Manning), pp. 3-15. Oxford: Clarendon Press.
van Iersel, J. J. A. & Bol, A. C. 1958. Preening in two tern species. A study in displacement behaviour. Behaviour, 13, 1-88.
Kennedy T 1954. Is modern ethology objective? Br. J. Anim. Behav., 2, 12-19
Kruijt, J. P. 1964. Ontogeny of social behaviour in Burmese Red Jungle Fowl. Behaviour Suppl. Vol. No. 12.
Lack, D. 1968. Ecological Adaptations for Breeding in Birds. London: Methuen.
Lehrman, D. S. 1953. A critique of Konrad Lorenz’s theory of instinctive behaviour. Q. Rev. Biol. 28, 337-363.
Lorenz, K. Z. 1938. A contribution to the comparative sociology of colonial-nesting birds. Proc. 8th Int. Orn. Congr. Oxford: Oxford University Press
Lorenz, K.. Z. 1950. The comparative method of studying innate behaviour patterns. Symp. Soc. exp. Biol. 4, 221-268
Malcolm, N. 1968. The conceivability of mechanism. Philosoph. Rev., 77, 45-72.
Melden, A. I. 1961. Free Action. London: Routledge & Kegan Paul.
Mischel, T. (Ed.) 1969. Human Action: Conceptual and Empirical Issues. New York: Academic Press
Moynihan, M. 1953. Some displacement activities of black-headed gulls. Behaviour, 5, 58-80.
Moynihan, M. 1955a. Types of hostile display. Auk 72, 247-259.
Moynihan, M. 1955b. Some aspects of reproductive behavior in the black-headed gull (Larus ridibundus L.) and related species. Behaviour Suppl. Vol. 4
Nagel, E. 1961. The Structure of Science. New York: Harcourt, Brace & World.
Peters, R. S. 1958. The Concept of Motivation. London: Routledge & Kegan Paul.
Popper, K. R. 1959. The Logic of Scientific Discovery. London: Hutchison.
Purton, A. C. 1970. Philosophical aspects of explanation in ethology. M. Phil. thesis. University of London.
Putnam, H. 1966. The analytic and the synthetic. In: Minnesota Studies in the Philosophy of Science, Vol. 3 (Ed. by H. Feigl & G. Maxwell), pp. 358-397. Minneapolis: University of Minnesota Press.
Quine, W. 1953. Two dogmas of empiricism. In: From a Logical Point of View (Ed. Quine, W., pp. 20-46. Cambridge, Mass.: Harvard University Press.
Scheffler, I. 1959. Thoughts on teleology. Br. J. Phil. Sci., 9, 265-284.
Simmons, K. E. L. 1970. Ecological determinants of breeding adaptations and social behaviour in two fish-eating birds. In: Social Behaviour in Birds and Mammals, (Ed. by J. H. Crook), pp. 37-77. London: Academic Press.
Smith, W. J. 1963. Vocal communication of information in birds. Am. Nat., 97, 117-125.
Sommerhoff, G. 1950. Analytical Biology. London: Oxford University Press.
Strawson, P. F. 1959. Individuals. London: Methuen
Taylor, C. 1964. The Explanation of Behaviour. London: Routledge & Kegan Paul.
Taylor, C. 1970. The explanation of purposive behaviour. In: Explanation in the Behavioural Sciences (Ed. By R. Borger & F. Cioffi), pp. 49-95. Cambridge: Cambridge University Press.
Taylor, R. 1966. Action and Purpose. Englewood Cliffs, N.J.: Prentice-Hall
Tinbergen, N. 1951. The Study of Instinct. Oxford: Clarendon Press.
Tinbergen, N. 1959. Einige Gedanken über ‘Beschwichtigungsgebärden’. Z. Tierpsychol., 16, 651-665. English translation in N. Tinbergen, The Animal in its World. Vol. 2 (1973). London: Allen & Unwin.
Tinbergen, N. 1960. Comparative studies of the behaviour of gulls (Laridae): A progress report. Behaviour, 15, 1-70.
Tinbergen, N. 1963. On aims and methods of ethology. Z. Tierpsychol., 20, 410-33.
Tinbergen, N. & van Iersel, J. J. A. 1947. ‘Displacement reactions in the three-spined stickleback’. Behaviour, 1, 56-63
White, A. R. (Ed.) 1968. The Philosophy of Action. Oxford: Oxford University Press.
Wittgenstein, L. 1953. Philosophical Investigations. Oxford: Basil Blackwell.
Wittgenstein, L. 1967. Zettel. Oxford: Basil Blackwell.
(Received 23 October 1973; revised 4 June 1974; second revision 7 May 1976; third revision 29 October 1976; MS. number A1499)