6. ANIMALS AND RATIONAL AGENCY
In this chapter I shall consider how much truth there is in the simplifying assumption made throughout Chapter 5 - namely, that no animals are rational agents in the sense that would be necessary to ensure that they have moral standing within contractualism.
Clever Hans and the Sphex wasp
Plainly there is no problem of principle about animals being rational agents - whatever, precisely, the relevant sense of ‘rational agent’ might turn out to be (this will be investigated as we go along). For example, if the story I told in Chapter 3 of the university educated apes had turned out to be true - or, indeed, if almost any story from children’s literature, such as Richard Adams’s Watership Down, were to turn out to be true - then it is clear that the animals involved would be rational agents. Our question is one of fact, not of principle. We are to consider whether there is sufficiently good reason for thinking that any animals are rational agents. I shall begin with two (true) tales designed to illustrate the pitfalls involved in interpreting animal behaviour.
Clever Hans was a horse who lived in the late nineteenth century, who was widely believed to be able to count. If an array of objects was placed before him, and he was asked how many there were, Hans would stamp his hoof the appropriate number of times. There seemed to be no question of cheating. His trainer could not have been giving him covert signals, for example, since Hans would perform equally well whether or not his trainer was present. But then it was discovered that if the only people present were, themselves, unable to count, or if they were so positioned that they could not see the objects to be counted, then Hans did not know when to stop - he just went on stamping his hoof. What had been happening was that Hans had been responding to subtle behavioural changes in his audience, such as a slight intake of breath, when the audience knew that he had reached the right number. Without these changes, he was lost.
Now, the moral of this story is not that Hans’s behaviour was entirely unintelligent. On the contrary, it did display intelligence of a sort, only not the kind originally attributed to him. He had learned to recognize and respond to a variety of subtle behavioural changes, rather than to count. The real moral is that we need to be cautious in interpreting animal behaviour in experiments that require animals to interact with human beings. For it is hard to be sure that we have not been unwittingly encouraging the animals to do what we want, by conditioning them to respond to unconscious human signals. Many of the experiments that have claimed to be able to teach chimpanzees to use an articulate sign language, for example, are infected with this problem.
Now consider the story of the Sphex wasp. The female of the species lays her eggs in a burrow, leaving them to hatch on their own. Before she abandons her eggs, however, she captures and paralyses a cricket, and brings it to the burrow where she leaves it to provide fresh food for her young when they hatch. Before taking the cricket into the burrow, she places it on the ground outside and goes in alone, apparently to check for intruders. She then emerges to drag the cricket inside, leaving it close to her eggs. All this seems remarkably intelligent - indeed, an example of long-term planning and foresight. This appearance, however, is an illusion. If the cricket is moved by an experimenter very slightly while she is in the burrow, she will re-emerge, drag the cricket back to its original position outside the entrance, and then re-enter the burrow alone once more. This cycle of behaviour can be repeated dozens of times. What looked like intelligent behaviour turns out to have been rigid - presumably a pre-programmed action pattern.
The moral of this story is that it is not enough, if we are to show that an animal is intelligent, to show that the direction of its behaviour may be characterised as intelligent, being such as to fulfil the creature’s long-term needs. In order to count as exhibiting genuine intelligence, the behaviour must also be appropriately flexible. It must display sensitivity to changes in the environment in a way that suggests beliefs are being formed, up-dated, and acted upon.
The anthropomorphic tendency
The stories above bear emphasising, because we have a pervasive tendency towards anthropomorphic interpretations of animal behaviour. There are two sources of this, one of which is culture-specific, but one of which is, arguably, universal. I shall consider each of them briefly in turn.
Many cultures tell stories in which animals act out human roles. But in our Western culture the literature and entertainment directed at young children has been almost entirely monopolised by anthropomorphic treatments of animals. There is hardly a children’s story, now-a-days, that does not have an animal as its central character, engaging in human-like projects, and exhibiting patterns of thought and feeling that are distinctively human. It is surely inevitable that all this covert propaganda in childhood should have an effect in later life - if not actually encouraging a belief that animals entertain conscious thoughts just like ours (a belief that may seem surprisingly widespread), at least reinforcing a tendency to seek explanations for animal behaviour in terms of rational planning for long-term goals. This is not to say, of course, that all such explanations are false - only that we should be on our guard against attributions of intelligence that go beyond what the behavioural evidence would really warrant.
No doubt the anthropomorphic tendency derives partly from facts about our culture. But it also seems to me to have sources that go much deeper, grounded in the very structure of human cognition, as I shall now try to explain. One relevant claim here, is that our knowledge of the psychology of our own species is very probably innate, having been selected for in evolution because of its decisive advantages in ensuring our survival. A number of considerations support this claim. One is that our common-sense psychology is highly complex, but is acquired by young children within the space of a very few years. Yet children are never explicitly taught it, and it is hard to see how they could learn it for themselves, since most of the phenomena with which it deals - people’s thoughts, feelings, and decisions - are hidden from casual observation. (While young children may be supposed to have access to the occurrence of such phenomena in themselves, through introspection, this cannot be the source of their knowledge of the causal relationships between those states, which is what the bulk of common-sense psychology is concerned with.) Another consideration is that knowledge of common-sense psychology must be presupposed before co-operation and communication can take place. If you had no knowledge of beliefs and desires and the characteristic manner in which they interact, then, plainly, you could neither co-operate with others, nor understand what they said to you. It would hardly be surprising, therefore, if knowledge of common-sense psychology should turn out to be innate, given the crucial role of both co-operation and communication in human survival.
Another relevant claim to have emerged recently, is that common-sense psychology provides the source for one of our most basic explanatory strategies. One body of evidence supporting this claim derives from studies of primate behaviour, which suggest that the driving force behind the evolution of human intelligence was distinctively social intelligence, not technical intelligence as has often been believed. Another source of evidence comes from studies of child concept acquisition, which suggest that the basic conceptual repertoire of young children includes the concepts of common-sense psychology. These concepts are initially over-extended, being applied well beyond their proper domain, until more varied explanatory strategies are learned.
Put the above claims together, and what you get is the thesis that it is an innately determined aspect of human cognition that explanation in terms of beliefs, thoughts, and desires forms one of our most basic explanatory strategies. Other things being equal, we tend to try to explain a given phenomenon in terms of attributions of intelligence, having a natural tendency to offer such explanations until the evidence forces us to think otherwise. This is certainly consistent with the impulse towards animistic explanations of natural phenomena, such as storms and earthquakes, common amongst primitive peoples. So what follows, again, is that we need to tread carefully in interpreting animal behaviour, taking care that we are not tempted to attribute intelligence beyond what the evidence will allow.
Whatever else a rational agent may be, it is, plainly, a creature that has beliefs, and acts on them to satisfy its desires. Our common-sense view would certainly be that beliefs and desires may be attributed to most species of animal, including all mammals, as well as many birds, reptiles, and amphibians (though it would not be correct to attribute them to insects, if the points made in Chapter 3 were sound). We say of the dog who leaps up at the familiar sound of his owner’s car, for example, and begins scratching excitedly at the door, that he believes his owner to have returned home. And we say of the cat creeping carefully round the side of a bush, that she believes the bird to be behind it. Such explanations of animal behaviour (and also, to some extent, the corresponding predictions) seem remarkably successful. This creates a powerful presumption that many species of animal are, indeed, truly believers. This may be further reinforced by recalling the claim defended in Chapter 3 - that all mammals and birds, at least, should be counted as genuinely sentient - together with the facts on which that claim was based, namely the extensive similarities in behaviour, and of brain structure and function, between such animals and ourselves.
There have been a number of arguments to the contrary, however. Many of these are in fact weak, or make eminently deniable assumptions - taking for granted, for example, without further argument, that all beliefs and desires must be expressed in natural language. Perhaps the most powerful of these arguments, however, are those due to Donald Davidson, that have also been very influential. I propose to consider the two main ones. The first of these is, in outline, this: that in order to have beliefs at all you must possess, in addition, the concept of belief; but that in order to possess the concept of belief, in turn, you must have a language; in which case it will follow that non-linguistic creatures cannot have beliefs.
Davidson has defended the premisses of this argument rather differently in different publications. Consider first the claim that having beliefs requires you to possess the concept of belief. In ‘Thought and Talk’ he tries to defend this by claiming that having a belief requires understanding the possibility of being mistaken, which requires, in turn, a grasp of the contrast between true belief and false belief. Now, the latter claim is undeniable. But why should we grant the former? No reason for it is given, and it is hard to see what such a reason might be. For why should there not be simple, but genuine, kinds of belief where all beliefs are treated as certainties (without even the possibility of thought of a mistake) until they are eradicated by conflicting evidence that overwhelms them? In ‘Rational Animals’, on the other hand, Davidson argues that having a belief requires the possibility of being surprised, which involves, in turn, the belief that your original belief was false. Presumably the thought here is that it would be unintelligible that you should, while retaining a particular belief, just carry on as before when confronted with the fact that it is false (that is, with a contradictory belief). This may be true, but does not establish the point. For to say that any organism with beliefs must have cognitive mechanisms for identifying and resolving conflicts of belief, is not to imply that the organism need be capable of thinking about, or have any concept of, its own beliefs. And it is difficult to see why this latter claim should be true.
Since we have rejected the first premiss of Davidson’s argument, we have already done enough to avoid the conclusion. But let us consider, in any case, how he tries to defend the further claim that possession of the concept of belief requires language. In ‘Thought and Talk’ he argues that the concept of belief is only acquired in connection with the activity of interpreting the speech of others. But this just begs the question at issue, given that we also employ the concept of belief in explaining the non-verbal behaviour of both animals and ourselves. And it is unlikely, moreover, that the concept of belief is one that we have to acquire at all. Rather that concept is, plausibly, a component in a theory of the mind (common-sense psychology) that we know innately. In ‘Rational Animals’, on the other hand, Davidson argues that the concept of belief presupposes the concept of objective truth, which presupposes, in turn, the concept of inter-subjective, communicable, truth. But these claims are again undefended, yet are highly contentious. For example, why would not a thought, by description, of the way things are irrespective of what I may take to be the case, be sufficient for a concept of objective truth? Yet such a thought need not presuppose that I am a potential communicator or language user.
Davidson’s other main argument against animals having beliefs, is again a sophisticated defence of the claim that beliefs cannot really be possessed by any creatures that are not users of language. The argument is that, in the absence of language, we cannot draw the sorts of fine distinctions amongst beliefs necessary for them to have genuine intentionality. But all this will take some explaining.
First, the concept of intentionality in philosophy is a technical one, although the phenomenon it picks out is easily recognized. (Note that this technical use of ‘intentional’ applies primarily to beliefs and other representational mental states, including intentions. In contrast, in the everyday sense it is primarily overt actions that are intentional - meaning that they are caused in the usual way by beliefs, desires and intentions.) Intentional states are distinctive in that they contain representations of things that may or may not exist, and represent them in one way rather than another. Someone can believe, or hope, that Atlantis once supported a great civilization, although there is, in reality, no such place. In contrast, if there is no such place as Atlantis then it is impossible for anyone to go there. So belief, but not physical travel, can stand you in relation to a non-existent thing. Moreover, someone can believe that there is water in the jug without believing that there is H2O in the jug, even though water is H2O; and someone can believe that Mr Hyde is the murderer without believing that Dr Jekyll is, even though Jekyll is Hyde; and so on. In contrast, if water boils at 100 degrees, then so too must H2O, and if Hyde is 32 years old, then so too must Jekyll be. So the property of belief, but not the properties of boiling point or age, can apply differently to one and the same thing, depending on how that thing is represented in the description of it.
Now Davidson’s argument is that these fine-grained distinctions amongst beliefs can only be drawn on the basis of evidence that is linguistic. Only if a creature can do something like assert that Hyde is the murderer while denying that Jekyll is, can we have reason for distinguishing the one belief from the other. And similarly, only if the creature can do something like respond to a request to fetch some water by bringing the jug, while failing to respond in the same way when asked to fetch some H2O, can the belief that water is in the jug be distinguished from the belief that H2O is in the jug. In which case, those animals that lack an articulate language, as presumably almost all do, cannot be said to possess fine-grained beliefs. There is nothing that a dog can do, for example, that can make a difference between the statements ‘Attila believes that his master is home’, ‘Attila believes that Mr Smith is home’, and ‘Attila believes that the president of the bank is home’, provided that Attila’s master is in fact Mr Smith, who is president of the bank. It can make no difference which description we use, no matter what Attila may do.
One response to this argument would be to allow that animals cannot have fine-grained beliefs (or, at least, that those without language cannot), but to insist that they may, nevertheless, have coarse-grained ones. To take this line, would be to defend a notion of belief-content for animals according to which one and the same belief may be characterised indifferently as ‘the belief that my master is home’, ‘the belief that Mr Smith is home’, or as ‘the belief that the president of the bank is home’. But this would be the wrong move to make against Davidson, since it would, in effect, concede to him that animal beliefs lack intentionality. Yet to concede the intentionality of belief is to concede belief, since it is essential to the very notion of belief that beliefs should represent things in one way rather than another.
The correct response is not to claim that animal beliefs are indeterminate between fine-grained descriptions, but rather to insist that the sorts of descriptions canvassed above are falsely attributed to animals. Very likely Attila does not have any beliefs that can be characterised correctly using such terms as ‘master’, ‘Mr Smith’, or ‘the president of the bank’, since dogs lack the requisite concepts. On the contrary, taking animal beliefs seriously must involve trying to describe the way in which they represent things. It seems plausible, for example, that Attila will represent his master in terms of some schema of appearance - some complex set of properties of sight, smell, and voice. Equally, in place of our concept ‘home’ Attila may employ something like the concept ‘protectable territory’. Then it will be straightforwardly false to describe Attila as believing that his master is home. Rather, what Attila believes will be something like ‘The one who appears like this is on protectable territory’. Needless to say, the ways in which animals represent things will not be easily discoverable. But this is no argument for saying that such modes of representation do not exist.
Can this sort of approach to animal beliefs preserve for them all the features of intentionality? Surely yes. Animals, like us, can have beliefs in non-existent objects. Thus the dog who barks wildly in the night when a lamp is blown over by the wind might do so because of the belief that something is intruding on protectable territory. And it ought to be easy enough to find cases where a dog has contradictory beliefs about what is, in fact, one and the same thing, by virtue of representing that thing in two different ways. For example, suppose that Delia always appears to Attila in one of two different guises - now being recognized by smell (with her appearance disguised) and bringing food, now being recognized by appearance (with her smell disguised) and bringing rough treatment. Then Attila might easily manifest the beliefs that this person is a source of food while that person is not, although it is, of course, the same person in each case. There seems no essential difference between this, and the example that forms a paradigm of human intentionality, where someone believes that Hyde is the murderer while believing that Jekyll is not.
Categorisation and concepts
Attempts to argue that creatures lacking an articulate language cannot have beliefs, in advance of detailed consideration of the evidence, have been seen to fail. We therefore need to look at what animals can actually do, and how their abilities are best explained. And in fact, the evidence is overwhelming that almost all animals have cognitive abilities that go beyond mere connections of stimulus and response, of the sort beloved by behaviourists. Even goldfish can retain in short-term memory (for a period of about one minute) the location of previously discovered food. However, although these abilities may be, in some sense, genuinely cognitive, this does not mean that they must involve beliefs and desires - which is what we need if animals are to stand any chance of counting as rational agents. This point will come out most clearly in the contrast between the ability to categorise things into two or more classes, and possession of a genuine concept, as I shall now try to explain.
Pigeons, for example, are capable of learning remarkably sophisticated perceptual discriminations. They can learn to categorise slides depending on whether those slides contain a triangle or not, or depending on whether they contain a human being (in whatever pose) or not, and so on. They can soon come to peck at the slide to gain a reward only if it contains a triangle, or a human being. But do the pigeons thereby possess the concept of a triangle, or of a human appearance? Being able to sort things into categories, in a series of yes/no choices, is surely different from having a concept. For a machine can categorise potatoes by weight or size without, of course, possessing any concepts of weight or size. But what more is needed? Well plainly, if something is to possess a concept it must be capable of having beliefs or desires in which that concept figures. Now, this might not seem to take us very far, beyond explaining why it is that the potato-sorting machine possesses no concepts. For after all, what we started off wanting to know was whether pigeons may be said to possess beliefs. But in fact the answer is useful, in two respects.
First, it is essential to the very notions of belief and desire, that beliefs and desires are states that interact with one another to produce behaviour. In fact, genuine attributions of belief and desire go along with a certain standard for explaining behaviour, that I call the practical-reasoning-model. On this account, to explain an item of behaviour is to exhibit it as the consequence of a piece of practical reasoning, of the form ‘If I do X then I shall get Y, and I want to get Y, so I shall do X’. (It need not be assumed that the reasoning process in question is a conscious one, in animals any more than in ourselves. What is crucial for the application of the practical-reasoning-model is only that there should be states of belief and desire that interact together to produce an intention in the way that the structure of practical reasoning outlines.) It follows, then, that pigeons possess the concept of a triangle only if they exhibit patterns of behaviour that are best explained using the practical-reasoning-model, by attributing to them beliefs and desires within the content of some of which, at least, the concept of a triangle figures.
Second, it is essential to beliefs and desires that they should be structured out of elements that can be recombined with others. The concepts that fit together to make up the content of any given belief or desire must be capable of fitting together with other concepts to form yet other contents. Any creature capable of believing that grass is green, for example, must be capable of believing that grass is something else (edible, perhaps), and of believing that something else is green (emeralds, say).
For these reasons, it is doubtful whether a child who can sort bricks into red and green, but can do nothing else involving those colours, as yet possesses the concepts of red and green. When the child begins to form beliefs such as that green apples are sour, red ones sweet; that red things are often hot; that green lights mean go, red ones mean stop; and so on, then it will possess the concepts of red and green. In the same way, we should not attribute the concept of a triangle, or of a human appearance, to a pigeon, unless we are prepared to take seriously explanations of its behaviour on the practical-reasoning-model. (For example, ‘Pecking at triangles is a way of getting food. I want food. Here is a triangle. So I shall peck at it.’) And we should only take these sorts of explanations seriously, where the pigeon’s behaviour displays sufficient flexibility for us to be able to attribute to it a variety of different contents involving the concept of a triangle.
While these points may make it doubtful whether pigeons have beliefs (or, at least, beliefs about triangles), they do nothing to undermine the attribution of beliefs to most, if not all, species of mammal. For we surely do take seriously the use of the practical-reasoning-model to explain their behaviour. For example, we might explain the behaviour of a dog by attributing to it the sequence, ‘I want to get the ball. The ball is on the table. If I jump on to the chair I can reach the table. So I shall jump on to the chair.’ Moreover, a dog’s behaviour certainly exhibits a wide variety of ways in which it can interact with a ball - fetching, chewing, chasing, and catching - suggesting that the concept of a ball does form a component in a number of different canine beliefs and desires.
I propose to grant that all mammals have beliefs and desires. They form beliefs about their immediate environment on the basis of their perceptions, and are able to act in the light of those beliefs to satisfy their immediate desires. Still, this is by no means enough for these animals to count as rational agents. For recall that rational agents, in the context of contractualism, are required to agree with one another on the rules to govern all of their future interactions. They must therefore be capable of representing in thought a variety of long-term futures, and of making rational choices between those futures. So to count as a rational agent, an animal must not only be capable of acting to satisfy its immediate desires, but also of constructing and following a long-term plan. For our purposes, rational agents are planners.
In order to count as a rational agent, of course, you do not have to act rationally on every occasion. To say that normal adult humans are rational agents is not to imply that they never make mistakes, or construct thoroughly muddled plans. It implies only that they are capable of representing different possible futures, of working out which one they want, and of constructing some sort of plan to achieve the future that they desire. It is enough that they are capable of engaging in these activities at all, not that they perform them successfully, let alone superlatively. While rational agents are planners, they do not have to be very good planners.
It might be said, then, that plenty of animals should be counted as rational agents. Think of squirrels who store nuts in the autumn, birds who migrate south for the winter or build elaborate nests for the protection of their young, and of dogs who bury bones for later retrieval and consumption. Surely these are all cases of long-term planning? But in fact, to say that an animal engages in behaviour adapted to meet a predictable future eventuality is not to say that the animal has itself predicted that future, or arrived at its behaviour as a result of a plan. (Remember the Sphex wasp.) For it is left open that the behaviour in question may be merely an acquired habit, or that it may be innately determined. For example, it might be written into the genes of certain species of birds that they are to fly at a particular orientation relative to the stars when the sun reaches a certain position in the sky. This would not be planning, but reacting. (Indeed, the nut-burying behaviour of the European red squirrel, at least, is an innately determined action pattern, that the squirrel will continue to perform in captivity, on a solid floor with no earth to dig.)
There are at least two general reasons for thinking that none of the sorts of animal activities mentioned above manifest genuine planning. The first is that the skills involved in planning are transferable. These include the abilities to represent and predict future states of affairs, and to work out ways of bringing about or preventing those states of affairs. So if any animals were planners, it would be remarkable that they do not do more of it. If a dog were really capable of predicting that unless its food is hidden it may be stolen by others, and of working out that burying the food would keep it out of sight and smell until needed, then it is strange that it should not make use of these abilities in other areas of its life. Why, for example, do dogs never lay out food as bait for an unwary cat? Another way to put the point is this. It is distinctive of human beings, and a mark of our rational agency, that we can adapt to almost any circumstance or habitat. No other single animal species even begins to approximate to this adaptability. To the extent that this is so, it suggests that no other species of animal approximates to the status of a rational agent.
The second reason for doubting whether the activities of squirrels, birds, and dogs manifest genuine planning, is that it would then be remarkable that members of the same species should not come up with alternative plans. It seems essential to the activity of planning, as we understand it, that there will always be a number of possible ways of trying to achieve a given objective, even if not all of those ways would be equally successful. It would then be strange, if squirrels were engaging in genuine planning when they gather nuts, that some individuals should not hit upon the alternative plan of observing where other squirrels have hidden their nuts, and later stealing them. And if birds were really planning for the future of their offspring in building a nest, it would be strange that members of the same species should not hit upon alternative modes of construction, or that some individuals should not avoid the labour of building altogether by laying their eggs in the nests of others, as does the cuckoo. (The cuckoo, presumably, does this innately.)
Long-term planning implies more than mere possession of beliefs about the remote future, of course, or the ability to predict future states of affairs. It must also involve possession of long-term desires, which serve to set the ultimate goals for any prolonged projects undertaken. Since many of the putative examples of animal planning are concerned with individual survival, this may be the point at which we should consider the question deferred from Chapter 4 - namely, whether any animals may be said to have desires for their own future existence.
Possessing a desire implies possession of its constituent concepts. So a desire for one’s own future existence must involve concepts of oneself, of the future, and of existence. Moreover, possession of any given concept must involve, in addition, possession of its contrasting concepts. To possess the concept of existence you must also possess the concept of non-existence. So if any animal were really capable of conceptualising, and desiring, its own future existence, it would also have to be capable of conceptualising non-existence. But there is no evidence that any animals are capable of this. True enough, if a dog returns to the spot where it had buried its bone to find the bone gone, it may express surprise. But there is nothing in this to manifest the thought that the bone has ceased to exist, rather than that it has been moved. Indeed, since a dog in such circumstances will generally hunt around the surrounding area before finally losing interest, it would seem to be the latter idea that is entertained, rather than the former. Of course, a human being in such a situation may behave similarly, at least initially. If I return to my desk to find my diary has disappeared from its usual position, I may begin by searching in the drawers and on the floor. But, in contrast to the dog, I can also manifest the belief that the diary has ceased to exist - for example, by accusing my secretary of having dropped it in the shredding machine by mistake.
There is a sense in which all animals will struggle for survival, of course, in that they will respond to perceived threats with aggression or fear. But this does not mean that they have desires for their own future existence, as opposed to non-existence. It only shows, at most, that they have desires to avoid damage or danger, which are conceptually simpler. All animals can make some distinction between things that are safe and things that are not, or between things that may damage them and things that will not. Some animals may, in addition, show sufficient variety in their behaviour for us to attribute to them possession of the corresponding concepts. But none of this shows that animals can have desires for their own future existence. Indeed, I assume that such desires should be denied of them.
In conclusion, many species of animal engage in short-term planning, if we take seriously the attribution of beliefs and desires to them. Consider the cat stalking a bird, or the dog jumping up on a chair to get a ball from the table. But this is not enough for the animals in question to be counted as a rational agents, in the sense that matters for contractualism. This also requires long-term planning. But so far as I can see, no suggested animal behaviours are at all convincing as examples of long-term planning. Moreover, this is, in any case, by no means the only obstacle to counting any animals as rational agents. To be so counted, animals would also have to be capable of planning for the results of implementing social rules, as we shall see in the next section.
In order to have the kind of intelligence necessary to be a rational contractor, it is not enough to have beliefs and desires, and to be able to construct long-term plans in the light of those beliefs and desires. You must also have an idea of what it is to act under a general rule, and of what it might be like if all were to act under the same rule. This will require that you have a conception of the beliefs and desires of others, and that you are able to work out what might be expected of those others in particular cases if the rule in question were implemented. So rational agency requires, not just beliefs and desires, but beliefs about beliefs and desires - second-order beliefs, in fact. Is there any evidence that animals are capable of entertaining second-order beliefs?
The clearest way in which an animal can manifest second-order beliefs is by deceit. For to act deceitfully is to act in such a way as to intentionally induce a false belief in another. But if such action can be intentional, it must presuppose a conception of the beliefs of the other. So, do any animals engage in intentional deceit? There is anecdotal evidence that they do. For example, Donna’s dog Dean likes to walk, and likes to sleep in Donna’s armchair. One day when Donna is sitting comfortably in her chair, the dog lying awkwardly on the floor, Dean gets up and fetches his lead. But when Donna gets out of the chair to take the dog for a walk, Dean jumps up into the chair she has vacated. Did the dog not act with the intention of inducing in Donna the false belief that he wanted to go for a walk?
The trouble with this sort of anecdotal evidence is that it is always amenable to more neutral description, precisely because it is merely anecdotal. For example, we may re-describe the case above by saying that Dean wanted to walk and also wanted to lie in Donna’s armchair. He set about trying to satisfy the first desire, but when, as an unintended consequence, the opportunity to satisfy the second arose, he set about doing that instead. All putative examples of deceptive behaviour in animals are, in principle, vulnerable to this sort of re-description.
In reply it might be said that there is very good reason why the evidence of deception in animals should be merely anecdotal. It is that deception, by its very nature, can only succeed if it is infrequent. Since there is always a real risk that a deception may be discovered and exposed, those who attempt to deceive others too often will soon find themselves without the opportunity to deceive at all, because no one will trust them. But this reply is only partially adequate. It can explain why the evidence of deception by any given individual animal should be merely anecdotal, but not why it should be so for the species as a whole. To make out a real case for intentional deception in animals, we should need frequent examples of apparent deception practiced by different individual members of the same species. Such evidence (which would, as it were, be systematically anecdotal) is entirely lacking in connection with almost, but not quite, all species of animal. The exceptions are the great apes, particularly chimpanzees.
Studies of chimpanzee behaviour, both in captivity and the wild, are rife with examples such as the following. One ape, who is female, knows where a store of food is buried. But she also knows from past experience that if she goes directly to that store, then one particular larger male will follow her, and take it from her. So she sets off in the opposite direction, and begins to dig. When the male pushes her aside and takes over the digging for himself, she rushes back to the actual location to retrieve and consume the food. True enough, these tales are anecdotal, and cannot be reliably repeated. But taken together they constitute an impressive body of evidence.
I propose to allow that chimpanzees, at least, have second-order beliefs about the beliefs and desires of others. But this is only a necessary condition of rational agency. It is by no means sufficient. In order to count as a rational agent, in the sense that matters for contractualism, an animal would also have to be capable of long-term planning, as we saw in the previous section. It would need, moreover, a conception of social rules, and of what it might be for all to act under the same social rules. Evidence of these aspects of rational agency would seem to be conspicuously absent, even in chimpanzees.
Bold claims have been made in recent years that chimpanzees, at least, count as rational agents in virtue of their capacity to use, and to understand, language. Many animals use systems of signs of one sort or another, of course. Bees do a figure-of-eight dance to show the direction of nectar, dogs bark in warning and growl in threat, and birds sing to attract mates or to defend territory. But plainly this sort of thing is too far removed from human language to be of any interest in the current debate. For the behaviours in question are, very likely, innately determined action sequences, as well as lacking the structural complexity of human natural language. The claim made, however, is that chimpanzees can be taught to use signs in ways much more closely resembling our own.
The issue is important, because it does seem clear that full competence in the use of a human natural language (or something closely resembling it) would be a sufficient condition for a creature to count as a rational agent. Anything capable of using a system of signs with the expressive power of a human natural language must be able to use those signs with the intention of inducing beliefs in other users, and must therefore have second-order beliefs about those others’ beliefs. Since such a creature must also be able to represent alternative futures, and the states of affairs on which those futures are contingent, it will be capable of long-term planning. Moreover, to have the expressive power of a human language, a system of signs must contain ways of representing various possible systems of rules, and the consequences of universal compliance with those rules. So a full language-user would be, without qualification, a rational agent, in the sense of the phrase that concerns us. In the light of the points made in earlier sections, indeed, it would seem that there can be no real prospect of showing any animals to be rational agents except by showing that they are capable of using an appropriately developed language. For there is little evidence of other sorts that even chimpanzees are capable of long-term planning, let alone that they are capable of conceptualising alternative systems of social rules. If we are to show that they are rational agents, our only remaining prospect is to show that chimpanzees are, at least potentially, language-users.
Many important criticisms have been made of the systems of signs taught to chimpanzees, even supposing that we set aside worries about the clever Hans phenomenon. Amongst these are that the sign-languages they have mastered show no significant syntax. In some cases there is no question, even, of the signs expressing articulate propositions, since only one sign is used at a time. But even in those cases where something like sentences are employed, it is in fact mere groupings of signs that are significant. Also, and relatedly, it has been objected that the systems taught to chimps are not genuinely productive, in the way that human natural languages are. We are capable, by virtue of our grasp of grammatical structure, of continually using old words in new ways, never before encountered, whereas the chimps are not. (You will, almost certainly, never before have come across the sentence ‘A green dragon sleeps beneath my word-processor’, for example. But now that you have confronted it, you will have no difficulty in grasping its meaning.) Finally, and perhaps most importantly, it has been pointed out that there is no evidence of the chimpanzees ever using their signs in thought, for solving problems or reasoning about what to do. They treat them merely as practical tools for enabling them to fulfil their immediate desires.
More important than any of the above criticisms for our purposes, however, is the point that all the various systems of signs so far taught to chimpanzees have been concerned only with immediately perceptible aspects of the chimps’ environment. Crucially, no chimpanzee has mastered the phenomenon of tense, or any way of representing particular future times. Nor have they mastered the concepts necessary for representing causality, conditionality (‘if...then...’), or general rules. But these concepts would be absolutely necessary if the chimps’ mastery of language were to establish that chimpanzees should be counted as rational agents. For as we have seen, the capacities for long-term planning, and for considering the consequences of adopting certain general rules, are crucial to rational agency.
It is hardly surprising that attempts to teach languages to animals have met with such limited success. For as Noam Chomsky and others have forcefully argued, the human capacity for language is very likely an innately determined aspect of our cognition. In Chomsky’s view, we ourselves are only able to learn languages because much information about the grammars of natural languages, as well as many linguistic concepts, are already contained in the inherited structure of our language faculty. Other animals, in lacking such a faculty, will find learning a full natural language impossible.
As we saw earlier, the capacity to speak a full natural language would be a sufficient condition for a creature to count as a rational agent. It might additionally be wondered whether such a capacity is also a necessary condition of rational agency. This is not for the reason considered - and rejected - in a previous section, namely that possession of natural language is a necessary condition for having beliefs. It is rather because a rational agent, for our purposes, must be a possible rational contractor. Yet to enter into an explicit contract surely requires prior communication on the terms of the contract, and it seems clear that a languageless creature could not communicate anything so abstract as a proposed system of rules. This argument is, however, too swift. For consider the following example. Suppose that we arrive on Mars to discover creatures that seem at least as intelligent as ourselves. They have a highly developed technology, and engage in activities that seem plainly to require long-term planning and knowledge of the beliefs and desires of others. But they lack any articulate system of communication. Perhaps the Martian creatures are extremely long-lived, and by nature solitary, in an environment that is by no means harsh. So they only need to meet one another to mate, and perhaps to exchange items of technology that each has developed independently.
In these circumstances it might be perfectly clear to us that the Martian creatures are rational agents. Could we, as contractualists, refuse to acknowledge that they have the same basic rights as ourselves, merely on the grounds that by being unable to communicate, they are incapable of entering into an explicit contract? I think not. As Scanlon remarks, the basic criterion, under contractualism, for whether or not a creature has moral standing is whether the idea of justification of a policy of action to that creature makes sense. We do not actually have to be able to justify our system of rules to a creature, or some action under the rules, in order for it to have the same basic rights as ourselves. It is enough that the creature should have all the mental qualities and capacities necessary to appreciate such a justification, if it could somehow be transmitted. In fact, we ought to regard the Martian creatures’ inability to communicate with others as a mere contingency, that could conceivably be overcome without altering anything fundamental in their mode of cognition.
The above point generalises to cover any other qualities that might be necessary for a creature to be capable of entering into an explicit contract with us. Any creature that is incapable of making and keeping promises, for example, would not, in any meaningful sense, be able to enter into a contract. But this need not prevent it from having the status of a rational agent, if it were otherwise capable of long-term planning, and of working out the consequences of implementing alternative sets of social rules. For recall that the contract after which contractualism is named is hypothetical, not actual. We are not proposing to grant moral standing to creatures only after they have entered into a definite agreement with us. Rather, our moral rules will extend to those creatures provided that we might intelligibly attempt to justify our actions to them, in terms that none could reasonably reject who shared the aim of reaching free and unforced general agreement. The basis of contractualism lies in just this conception of reasonableness, not in any tit-for-tat contract.
No doubt rational agency, as such, admits of degrees. For the gradual development of a human infant, through childhood towards adulthood, is a process in which a fully-fledged rational agent slowly emerges, as we noted in Chapter 5. Yet I have been arguing, in effect, that no animals count as rational agents to any degree, since they lack even rudimentary versions of those qualities that are distinctive of rational agency. These are, namely, the capacities for long-term planning, for representing alternative sets of social rules, and for working out the likely consequences of implementing those rules. It therefore invites some comment that human beings should be unique in this respect. For we accepted in Chapter 3, after all, that human beings are continuous with the rest of the natural world, having evolved, like any other species of animal, through a process of natural selection. What follows are some highly speculative suggestions.
It may be our distinctive possession of an innately structured language faculty that underlies our uniqueness as rational agents. In the beginning, we may suppose, human beings came equipped with a working model of one another’s psychology, somewhat as, perhaps, chimpanzees do today. Our ancestors’ common-sense psychology might possibly have been more sophisticated than that possessed by chimpanzees, but would not have been different from it in kind. Given such a model, human beings would have been able to predict one another’s behaviour, to a limited extent, and to engage in rudimentary forms of co-operative activity. The next - and crucial - development may have been the evolution of an innately structured language faculty. This would immediately have conferred decisive advantages in survival. It would have made it possible for human beings to co-ordinate their behaviour, and to frame and execute joint plans of action, to their mutual advantage. It would also have made it possible for early humans to begin exchanging information, and to pass on the accumulated wisdom of a society from generation to generation. But most importantly, for our purposes, it is possible that the evolution of such a language faculty facilitated a wider range for human thought. For, as Chomsky argues, there are essentially the same reasons for thinking that a wide range of human concepts are innate, as there are for thinking that knowledge of universal grammar is innate. With the coming of language, and its associated grammatical forms, human beings could then frame thoughts about particular future times, about the long-term consequences of patterns of human behaviour, and agree rules with one another for the conduct of their affairs.
Whatever may be true of the hypothetical Martians considered earlier, it may thus be our unique status as natural language users that underlies our uniqueness, amongst creatures on Earth, as rational agents. In which case it will also be our possession of natural language that accounts for the fact that human beings are alone in their moral standing, and in having direct rights, if the contractualist approach to morality is correct. Note, moreover, that this story has been told in terms that, so far from denying our continuousness with the rest of the natural order, presupposes it. There is substantial evidence that we do possess innate knowledge of common-sense psychology, and an innate language faculty, and it is easy to see why these faculties might have emerged through natural selection. But if natural language is implicated in our capacity to represent future times, causes and conditionals, and general rules, then it will be our unique (but naturally explicable) possession of natural language that underlies our uniqueness as rational agents.
Many animals may be said to have beliefs and desires. Some animals (particularly apes) may be said to have second-order beliefs and desires. But no animals possess the other qualities necessary for rational agency. Specifically, no animals appear capable of long-term planning, or of representing to themselves different possible futures. And no animals appear capable of conceptualising (let alone acting under) general socially agreed rules. I therefore conclude that the simplifying assumption made in Chapter 5 is correct. No animals count as rational agents, in the sense necessary to secure them direct rights under contractualism.
 Penguin, 1972.
 See Walker, Animal Thought, pp. 372-4.
 For detailed defence of this idea, see my Human Knowledge and Human Nature, ch. 8.
 See Henry Wellman, The Child's Theory of the Mind (MIT Press, 1990).
 See R.Byrne and A.Whiten eds., Machiavellian Intelligence (Oxford University Press, 1988).
 See Susan Carey, Conceptual Change in Childhood (MIT Press, 1985).
 Many of these are detailed in R.G. Frey, Interests and Rights (Oxford University Press, 1980), chs.7-9.
 See ‘Thought and Talk' in his Inquiries into Truth and Interpretation (Oxford University Press, 1984), and ‘Rational Animals' in E. LePore and B. McLaughlin eds., Actions and Events (Blackwell, 1985).
 See my Human Knowledge and Human Nature, ch. 8.
 See Walker, Animal Thought, ch. 6.
 See Walker, Animal Thought, ch. 6.
 See Fred Dretske, Explaining Behaviour (MIT Press, 1988), p. 4.
 See the papers collected in Byrne and Whiten eds., Machiavellian Intelligence.
 See Singer, Practical Ethics, pp. 93-5.
 For detailed development of these points, see Walker, Animal Thought, ch. 9.
 See Noam Chomsky, Language and Problems of Knowledge (MIT Press, 1988). See also my Human Knowledge and Human Nature, chs. 6-8.
 Adapted from Robert Stalnaker, Inquiry (MIT Press, 1984).
 See ‘Contractualism and Utilitarianism', p. 113.
 See Language and Problems of Knowledge, ch. 1.
 See my Human Knowledge and Human Nature, chs. 6 & 8.