1. Activity, Life and Behaviour

Can animals act? This question must be answered in the affirmative by those who accept non-intentional agency, including agency by inanimate objects. If

1. (1) The wind knocked over the vase

   and

2. (2) The bleach ruined the linen

   report agency in this encompassing sense, then so do

3. (3) The cat knocked over the vase

   and

4. (4) The dog ruined the linen.

In English, it is perfectly acceptable to say that in (1) and (2) inanimate agents act on something else. Similarly for French agir sur quelque chose and Spanish actuar sobre algo; note also phrases like ‘chemical agent’ or agent blanchissant. It is different for German handeln. But the latter is not a transitive verb, by contrast to auf etwas (ein-)wirken, ‘to act (up)on’. Furthermore, lexical diversity apart, Indo-European languages feature a grammatical distinction between active and passive voice. That distinction applies not just animate but also to inanimate objects and stuffs. Finally, in all the aforementioned languages one can apply many specific verbs of agency univocally to animate and inanimate objects, as in (1) – (4).

I shall capture this most pervasive kind of acting through the term activity. A paradigmatic instance is a body x either moving another y or moving itself. The capacity to discern things that are self-moving or self-propelled—notably animate objects--is a basic feature of human perception manifest in infants from the age of seven months onwards.Footnote ⁵ A more demanding notion is integrated activity, i.e. activity of complex agents resulting from a cooperation of functionally differentiated parts.Footnote ⁶ Such activity can be inanimate, e.g. in a machine. To pass from integrated activity in general to animate activity, simply add the ingredient of life! Alas, the concept of life is extremely plastic and contested. An analytic definition that is neither stipulative nor peculiar to specific empirical theories about the mechanisms and genesis of life is not in sight. Therefore, life is best treated as a family-resemblance concept. Among the parameters bandied about in contemporary debates – homeostasis, autopoesis, organization, adaptation, susceptibility to selection, etc. – two stand out, namely metabolism and reproduction. Both can come apart: viruses reproduce yet lack metabolism, mules have a complex metabolism yet are infertile. And they allow of border-line cases. This is to be expected, however, since life is subject to evolutionary development.

When it comes to living things, the idea of activity is supplemented by that of needs, which derive from their metabolism and reproduction. Plants fulfil their needs by growing roots, orienting their leaves, emitting chemical compounds to deter predators, flowering, etc. Plants are also subjected to influences that are neutral or detrimental to these needs. Conversely, passivity may also fulfil their needs, e.g. when they are drenched by rain or pollenised by insects.

Of course, animals have biological needs as well. According to Aristotle higher animals also have perception and locomotion. But there is no conceptual connection between these two capacities and being an animal by contemporary biological criteria, whether cladistic or phenotypical. The idea of an autotrophic organism that perceives and moves is perfectly coherent. So is the idea of a heterotrophic organism that is static and without sense-perception. For this reason, my portrait of the conceptual landscape would not be overturned if reports of perception, communication and learning in plants withstood empirical, methodological and conceptual scrutiny.

At the same time the Aristotelian taxonomy pinpoints a distinctive type of activity, which, for biological reasons, is the rule among animals and the exception among plants. It is activity in pursuit of needs through movement guided by perception. The terminus of such activity is its goal. And although biologists are happy to speak of plant ‘behaviour’, I shall reserve that label for animate activity that pursues a goal informed by perception.

What sets perception and behaviour apart from the information processing and activities of plants and micro-organisms? My answer appeals to the idea that our notions of behaviour and perception constitute a Gestalt. They intertwine behavioural with morphological-cum-anatomical criteria. Animal behaviour is animate activity along a line of orientation determined by the position and alignment of sense organs. The well-known alignment of sunflowers towards the sun, by contrast, is the direct effect of differential growth of cells. Those cells process Shannon information – concerning the direction of the sun; yet they are not specifically geared to informing the organism of that direction.

2. From behaviour to agency

There are other ways of distinguishing (animal) perception.Footnote ⁷ But for the purpose of comparing animal and human activity, the crucial point is: goal-directed activity, activity aimed at satisfying needs guided by perception constitutes behaviour. According to a prominent psychological definition, it also constitutes action.Footnote ⁸ Judged by that standard, some animals act; unsurprisingly, since the definition is meant to allow for precisely this fact. By contrast, rationalist philosophers make agency dependent on rationality. Contemporary versions of this stance have been developed within the Pittsburgh school. Thus, McDowell writes: ‘movements of limbs without concepts are mere happenings, not expressions of agency’.Footnote ⁹ Both extremes elide important differences.

Admittedly, we speak more readily of animal behaviour than of acts animals perform.Footnote ¹⁰ But once again, when we turn from generic terms like ‘action’ and ‘behaviour’ to specific action verbs, anthropocentric scruples vanish. Animals do many of the things that we do. They seek out food by foraging, hunting and killing prey; they eat; they court and mate; they protect their offspring; they bond; they defend their territory; etc. Some of them also play, prepare and hide food, make and use tools, socialize and act in concert, form alliances, fight and make up, etc. And there is no rationale for diagnosing any ambiguity between human and animal agency in these cases.

At the same time, our preference for ‘behaviour’ over ‘action’ for simple activities of both animals and humans indicates that we do not count every behaviour as bona fide agency. One difference that can be marked by this distinction runs as follows. Some behaviour is guided by perception in a way that is immediate and rigid. Other behaviour involves a decoupling of stimulus and response, of perceptual input and behavioural output. By a similar token, belief is at one remove from behaviour, among other things because of a deferment or delay between cognitive state and behavioural disposition.Footnote ¹¹ It enables the subject to adjust behaviour to phenomena outside its current perceptual range and to engage in problem-solving and learning.

This type of decoupling will be discussed in section 3. Before then, note another type of decoupling. It marks the difference not between stereotypical and intelligent behaviour, but between behaviour and agency. Such agency requires a differentiation of cognitive states like belief and conative states like desire or intention in the non-technical sense. There are animals which pursue goals guided by perception, yet without displaying this difference. According to a popular tale, leopard frogs snap at any object looking like an insect, irrespective of their state of satiation, i.e. independently of their motivational state. There is no differentiation into cognitive and conative, informational and motivational mental states. Millikan explains such behavior by ‘pushmi-pullyu representations’, which are in one breath both descriptive and directive.Footnote ¹²

The two ways of decoupling perceptual input and behavioural output do not coincide. Flexible behaviour does not per se evince a differentiation of cognitive (belief) and conative (desire) determinants. The flexibility can result instead from variation in just one of these two parameters: an animal could be lumbered with fixed goals, yet pursue the latter in a flexible and adaptive manner.

In any event, the complex behaviour of higher animals is a function of two independent vectors: what they know or believe on the one hand, what they desire or want on the other. Neither common sense nor science has a better way of explaining such behaviour than by reference to an interplay between cognitive states on the one hand, conative states on the other. To both humans and animals, we apply intentional verbs, and we explain behaviour by reference to the fact that A believes that p, desires X, intends to Φ, etc. Both cases, moreover, display the same basic pattern:

Behaviour that is part of this pattern does not just serve a purpose (as in plants) or pursue a fixed goal (as in leopard frogs); it is guided by a want, a goal that is variable rather than fixed. In such cases the active/passive distinction cuts across a distinction between voluntary and involuntary, from which it needs to be kept apart. A can remain inactive because it wants to; and A can be active against its will, e.g. when convulsed by pain. An animal can be thrown over a fence or trip accidentally (involuntary activity), or it can jump over a fence or lie down as part of a trick it has learnt (voluntary activity). Mutatis mutandis for not doing something. An animal can be prevented from doing something, but it can also refrain from doing something. For instance, a dog can be static because it is held in place by fetters (involuntary passivity), yet it can also hold still because it wants to be groomed (voluntary passivity). In cases of involuntary activity and passivity, wants and desires of the animal are often frustrated or thwarted, irrespective of whether the frustration militates against the satisfaction of objective needs. Such frustration of wants, just like their fulfilment, is manifest in behaviour. In the case of some animals – notably great apes and dogs – we can also recognize them on account of postures and facial expressions.

In short, some animals are capable of doing what they want to do, pursuing goals that are not simply dictated by and may downright conflict with their biological needs (see below). I shall call such flexible behaviour ‘action’, a terminological choice that is not dictated by yet perfectly compatible with established parlance. On the basis of that explanation, my answer to question (a) is that higher animals can act.

3. Intelligence

How do we get from action to rational action? The idea of intelligence affords a starting-point. Intelligence is often granted to animals. The term derives from classical Latin intellegere, i.e. choosing between. It was later confined to cognitively informed choices, as opposed to mindlessly plunging for a particular option. Unlike related concepts, e.g. intellect, the notion never gained prominence in philosophy. At the same time, it was pivotal to the emergence of psychology as a distinct academic discipline. In the wake of Spencer and Galton, much of 19^th and 20^th century psychology revolved around the attempt to make this conceptual construct at once more precise and quantifiable.

The eventual result was the idea of an ‘intelligence quotient’ (IQ). But what is the IQ? Edwin Boring's notorious response runs: What is measured by an IQ test!Footnote ¹³ By these lights, intelligence is a capacity to solve linguistic, computational and spatial (concerning in particular spatial configuration and imagination) problems. The IQ orthodoxy also postulates a g-factor, understood as the underlying cause of a unified phenomenon – ‘general intelligence’. But even if there is such an underlying factor, it is a common cause of three abilities – linguistic, computational and spatial – that remain conceptually distinct. In so far as IQ is defined operationally, through a tri-partite test, as a conceptual construct it divides into components.

This point matters, since the allegedly unified IQ concept of intelligence defies application to animals ab initio. The linguistic component of IQ tests simply does not apply. But since the appearance of an indissoluble package is deceptive, we can consider each IQ component separately. There is no reason to suppose that distinct mental capacities which as a matter of fact tend to coincide in humans also have to coincide in animals. Au contraire. Since the behavioural capacities of a species reflect its form of life and ecological niche, there is every reason to expect that these capacities will be found not just to different degrees, but also in a variety of configurations. Each species will display its own suite of mental powers.

A proper conception of intelligence must go beyond the Boring orthodoxy. The refusal of its adherents to furnish a substantial definition of intelligence is belied by the prose required to make sense of their own operational explanations. The non-technical notion of intelligence employed in everyday parlance and explicated or willy-nilly presupposed in psychology is a functional one. Intelligence is a highly general cognitive capacity; and it is displayed in a certain type of behaviour. It is often equated with the ability to solve problems. Not all problem-solving or adaptive behaviour counts as intelligent, however. We exclude problem-solving strategies which are predetermined either by the genetic outfit of the species, the genetic endowment of individual specimen or epigenetic factors.

At a first approximation, intelligence in the general pre-theoretical sense is the ability to solve problems – notably hitherto unencountered problems – in a flexible way.Footnote ¹⁴ But this explanation needs to be refined in several respects.

First, intelligence is a mental ability. It can be displayed by a subject that is barred from solving a problem by material circumstances, such as physical limitations. This option exists for non-linguistic subjects as well. An ageing chimpanzee that knows how to open a container with a heavy tool can be prevented from solving that problem if the tool is too distant from the container for her to carry there. She can display her intelligence by trying to carry it, or by casting around for other means of getting it to its destination. Secondly, intelligence is a cognitive capacity, rather than a conative or affective one such as resolve or strength of will. At first sight, established use pulls us in opposite directions here. On the one hand, strength of will or the capacity for delayed gratification count as virtues and akrasia and impatience as vices of the will. On the other hand, we characterize akratic behaviour as ‘stupid’ or ‘irrational’. Nevertheless, in pre-theoretical contexts, we use ‘intelligent’ ultimately as a cognitive term, by contrast, perhaps, to ‘clever’ and ‘irrational’, as well as their cognates and antonyms. For it is part and parcel of established use that, e.g., an addict smoker can be intelligent, provided that he knows what he should do, even if he is incapable of acting on this insight. We allow for a contrast between the intelligence or stupidity of a subject on the one hand, the rationality or irrationality of its behaviour on the other, not just in the case of individual actions, but also in the case of habits such as smoking.

Thirdly, intelligence is cogitative, unlike perceptual powers. One can be eagle-eyed, yet thick as two short planks; and one can be as sharp as a razor yet as short-sighted as a mole. What counts is not how much information a subject garners through perception, but how proficient she is in operating with that information, and ultimately the level of understanding she can achieve of the situation as it presents itself to her senses. Once again, this is not confined to linguistic creatures. The aforementioned chimpanzee could be too short-sighted to spot a tool she is looking for, while a conspecific might see the tool while being clueless about what to do with it.

Fourthly, the practical result of this type of understanding is not just the ability to solve problems in the sense of overcoming obstacles confronting a course of action, but also the pro-active ability to exploit ‘affordances’, i.e. positive opportunities that the situation offers.Footnote ¹⁵

Fifthly, as an ability underlying flexible adaptation, intelligence presupposes a capacity for learning. But one needs to distinguish types of learning. At the bottom of the scale, we find learning through strict conditioning. This presupposes intelligence of some kind. On the other hand, stimulus-response patterns established in this fashion are almost as mechanical and rigid as those which are genomically programmed; and like the latter, they do not involve a genuine contribution by the subject. For the opportunities for trying, erring and succeeding are imposed on subjects. This contrasts with a more advanced learning through trial and error, which emerges from the subject behaving spontaneously, notably if it pursues a strategy of manipulating objects motivated by curiosity. Finally, there is insight into the underlying processes of foresight and planning. In all these cases we can speak of intelligence, a capacity for behaviour that is plastic, responsive to altering circumstances, and therefore enables the subject to tackle problems it has never encountered before. But the differences make for levels of intelligence, rather than mere degrees.

Summing up: intelligence is a measure of a subject's insight into or understanding of a potentially novel situation or task, with a view to solving problems which that situation or task poses or exploiting opportunities it affords in a flexible manner. Different levels of intelligence are manifest in distinct types of learning.

4. Rationality

Intelligence as just explained is clearly a feature of some higher animals. It extends from problem-solving through trial and error, as with the Capuchin monkeys in the trap tube task, to the kind of insight and foresight displayed by chimpanzees or Caledonian crows in the fashioning of tools.Footnote ¹⁶ Cognitive ethology has uncovered peaks of animal intelligence in species from various taxa: in addition to primates (great apes), these include other mammals, notably marine mammals, and elephants, avians like crows and parrots, and even non-vertebrates like some cephalopods. Note that all four of the aforementioned refinements of the idea that intelligence is the ability to solve problems apply to non-linguistic creatures as well; and not just in principle, but also to actual cases. In reverse order: i) intelligent animals are not merely able to overcome obstacles, they can also exploit affordances opportunistically; ii) we can distinguish between the acuity of their senses and their capacity to use sensory information; iii) an animal can display a conative virtue like perseverance in pursuit of its goal as opposed to the cognitive virtue of understanding of how to attain it, and vice versa; iv) there is as much of a difference between cognitive insight on the one hand, capacities like physical prowess and dexterity on the other, in the case of animals as there is in humans.

In short, the answer to question (b) is a resounding Yes! Some animals are intelligent. But in order to answer question (c), we need to establish whether intelligent animals are ipso facto rational. How does intelligence relate to rationality?

Within contemporary academic debates one can distinguish four general conceptions of rationality. According to the first, it is the capacity to maximize satisfaction of one's interests or goals; according to the second, it is responsiveness to reasons; according to the third, it is the ability to reason – draw theoretical and/or practical inferences and to avoid inconsistencies; according to the fourth, it is the ability to justify one's actions and beliefs to others. For present purposes, we can set aside the last one in so far as it requires advanced linguistic capacities and hence excludes animals ab initio. But the second and third conception go in the same direction, without begging the question in favour of lingualism.

According to all four conceptions, rationality shares one feature with intelligence. It presupposes a capacity for transferring and adapting insights and solutions from one context to another, a capacity that normally manifests itself in flexible and plastic behaviour. Rationality is sometimes held to go beyond mere intelligence through three features: greater abstraction from the here and now (‘decentring’); greater generality; and being subject to normative constraints.Footnote ¹⁷

Although it pinpoints important parameters of advanced cognition, this way of separating intelligence and rationality fails to capture the pre-theoretical contrast. As regards normative constraints, both rationality and intelligence presuppose a distinction between a general capacity and its successful or unsuccessful exercise. We contrast a subject A's being rational with her being irrational. Both epithets presuppose, however, that A possesses the faculty of reason in the first place, rather than being non-rational, a-rational or pre-rational. Similarly, only intelligent creatures, creatures able to behave intelligently, can strictly speaking behave stupidly, namely by either failing to exercise their intelligence in a pertinent context, or by exercising it in an inadequate manner.

At the same time, there is an important difference. Unlike ‘intelligence’, ‘rationality’ suggests a capacity for reasoning – the idea encapsulated in the third conception mentioned above. This difference is sometimes glossed as one between two kinds of rationality: ‘behavioural rationality’ is a feature of behavioural patterns and capacities that lie open to observation, whereas ‘process rationality’ is a special kind of behaviour-generating process.Footnote ¹⁸ But this way of capturing the difference is deficient in two respects. First, while intelligence must be seen in the context of behavioural capacities, it consists in the ability to understand a task or situation, irrespective of whether the subject in fact behaves in an optimal fashion. Secondly, even the rational performances of humans need not be accompanied (and a fortiori caused) by conscious processes of reasoning.Footnote ¹⁹ What counts is not what does or does not cross the mind of an agent A, but what A is capable of doing, e.g. by way of justifying her beliefs and actions post-hoc.

This capacity is the topic of question (e) and will not be further discussed here. Instead I want to pass from question (c) to question (d), by pointing out an underappreciated distinction. There is a difference between externalist and internalist conceptions of rationality. Are A’s performances assessed exclusively by reference to external standards of which A may be entirely unaware? Or are they also assessed by reference to A's own perspective, by reference to A’s beliefs, desires, intentions and preferences?

5. Externalist and internalist conceptions of rationality

Externalist approaches have gained favour in the wake of Darwinism. A’s behaviour is rational in this Darwinist sense in so far as it maximizes A’s biological fitness. But an externalist approach is not the prerogative of evolutionary biology. It is also suggested by neo-Aristotelian teleology: A’s behaviour is rational in so far as it maximizes A’s objective well-being, irrespective of A’s own preferences or reproductive success.Footnote ²⁰ Both positions adopt an outside perspective: behaviour is judged as rational not by reference to A’s subjective predilections, but by reference to objective goals. For instance, an agent that employs intricate ‘foraging’ techniques in order to acquire a constant supply of sweets may maximize the satisfaction of its own desires or cravings, yet doing so may be deleterious to its biological fitness or ultimate flourishing.

These phenomena are biologically important. Nevertheless, they should not occupy centre stage in a debate about animal rationality, even if we confine ourselves to the satisfaction of interests. Animals are indisputably capable of Darwinian rationality. They are also capable of teleological rationality, since they have needs and pursue these in a more or less efficient manner. But Darwinian rationality is miles away from our everyday notion of rationality. There is nothing irrational, for instance, about reducing one's biological fitness by using contraception. Teleological rationality caters for other interests, but like Darwinian rationality, it does not require higher mental faculties, or even intelligence. It belongs to the first aforementioned general conception of rationality, leaving out the parameters stressed by the other three.

Internalist conceptions revolve around the idea that agents adopt goals of their own. A is rational in this sense if she can perform an action in order to attain a good, subjectively conceived, something she herself wants or prefers over the alternatives. The agent needs to be capable of acting in pursuit of her own goals.

By contrast to plants, animals possess not just needs of the kind invoked in externalist conceptions, but also wants; they pursue goals and avoid things. In addition, intelligent animals pursue goals of their own, namely by adopting novel goals, in line with what they have learnt. Such animals engage in a wide range of activities. Some of these activities are directed not at meeting immediate needs, but rather at things that these animals enjoy or have learnt to enjoy. This holds true in particular of play behaviour, which can be astonishingly varied, plastic and sophisticated. Furthermore, in pursuit of their adopted goals, such animals can persist in the face of obstacles, and they can explore various routes for overcoming them.Footnote ²¹ Finally, these goals may conflict with their objective needs. A can be ‘instrumentally rational’ by adopting feasible or optimal ways of achieving its goals or satisfying its desires, independently of whether doing so is conducive to A’s biological fitness or objective well-being.

At this point we need to address a worry. Although rational action by humans need not be accompanied by inner speech, agents can justify it post-hoc by outer speech. Doesn't this exclude the possibility of rational animals?Footnote ²² It must be granted that A can only act rationally in Φ-ing if A can be justified in Φ-ing. Nonetheless, the objection goes through only if this in turn requires that S herself must capable of justifying her Φ-ing vis-à-vis others. It is far from obvious that this is so. A non-linguistic subject can be justified in pursuing a goal she has adopted in a certain way, e.g. if she has explored her environment through proficient use of her sense organs and has chosen intelligently between different options on that basis. In consequence, the answer to question (c) is that intelligent animals with goals of their own can act rationally.

6. Acting intentionally and acting for a reason

As argued above, behaviour directed at achieving a goal of one's own qualifies as action. Such behaviour is also a case of acting intentionally in one important sense: the action is performed with the intention of achieving the goal. There is also a connection between intentional action and responsiveness to reasons. Intentional actions are commonly equated with actions that the agent performs for a reason. Admittedly, there is a case for holding that A can ϕ ‘intentionally’, in the sense of ϕ-ing in a controlled manner, without ϕ-ing ‘for a reason’, in pursuit of a goal.Footnote ²³ It is also true that some unintentional acts are done for a reason.Footnote ²⁴ Nevertheless, doing something with a specific intention is sufficient for doing it for a reason. In a perfectly clear and established sense, a goal or intention with which an action is performed counts as a reason for which the action is performed.

• (GR) A Φs in order to Ψ / bring it about that p A’s reason for Φ-ing is to Ψ / bring it about that p.

Accordingly, acting in pursuit of a goal of one's choosing qualifies as acting for a reason. By these lights, some animals are capable of acting for a reason. For instance, ‘Finding the bone’ is a legitimate answer to the question ‘What is A’s reason for digging?’, irrespective of whether A is a human or an intelligent animal.

Nevertheless, we shall see in section 9 that there remains an issue concerning what kinds of reason one can ascribe to animals by appeal to (GR). We must also note a different gloss on the ideas of acting for a reason and of intentional action. Following Davidson, it is customary to conceive of both as actions subject to ‘intentional’ or ‘rational explanations’, explanations that refer to the agent's reasons—her beliefs, desires and intentions.Footnote ²⁵

Now, intentional explanations revolve around a differentiation of cognitive states like belief and conative states like desire or intention in the non-technical sense. As noted above, higher animals display this differentiation. The answer to question (d₁) is affirmative, therefore. Some animals act intentionally; they behave in pursuit of goals of their own, and in a manner to be explained by reference both to what they want or intend and to what they believe or know. Moreover, in so far as acting for a reason means acting intentionally, this verdict carries over to (d): such animals can act for reasons.

7. Subjectivism vs. Objectivism

The idea of acting for a reason has also been construed in more ambitious ways. Before considering these, we need to look at the nature of reasons for actions. There is a widely accepted trichotomy between justifying, motivating and explanatory reasons.Footnote ²⁶ Justifying reasons are reasons for an agent to perform an action, i.e. considerations that actually favour her acting in this way. Motivating reasons are reasons for which an agent performs an action, considerations which make performing the action attractive from her point of view. Explanatory reasons are reasons why the agent acts a certain way, that is reasons which explain her acting. These types of reasons can come apart. There may be several justifying reasons for me to skip dessert, yet only some or perhaps even none of them are my actual (motivating) reasons for skipping dessert. And while all motivating reasons can contribute to explaining an action, explanatory reasons do not always motivate. If John punches Peter because he has learned that Peter betrayed him, the fact that he knows this explains his punching; yet, it is a fact about Peter, that he betrayed John, which motivates John to punch him.

Current debate revolves around a conflict between two opposing conceptions of motivating reasons. According to a subjectivist orthodoxy, the reasons for which A acts are mental states of A – states of desiring or wanting something and states of believing something.

1. (S) The reasons for which A acts are mental states of A (beliefs and ‘pro-attitudes’ like desires and intentions).

By contrast, according to a revisionist position currently gaining momentum, the reasons for which agents act are not mental states of the agent herself. Instead, they are objective facts or states of affairs; save for special cases, they concern A’s environment rather than her own mind. In so far as the reasons for which an agent acts can be said to be beliefs and desires at all, they are not subjective states of believing or desiring, but what is believed or what is desired.Footnote ²⁷ A acts for a reason if A acts on account not of her own mental states, but on account of facts, on account of how things are (from A’s perspective). Accordingly,

1. (O) The reasons for which A acts are facts (as A sees them) or states of affairs.

An objectivist conception of reasons does not imply an externalist conception of rationality. Although A’s reasons for Φ-ing are facts, they are facts which make Φ-ing good or attractive in A’s eyes (either instrumentally or intrinsically), that is, from the agent's own perspective. If A employs intricate ‘foraging’ techniques in order to acquire a constant supply of M&Ms, this is irrational by Darwinian lights, in that these exploits are deleterious to A’s biological fitness. Nevertheless, A has a reason for seeking M&Ms, namely that they are sweet, something which makes them attractive to A.

There are important differences among objectivists. But they agree on two points.

First, A’s reason for Φ-ing is something that makes Φ-ing good or valuable in A’s eyes; secondly, what makes Φ-ing good or valuable in A’s eyes is (by-and-large) not A’s believing something to be the case, but what is believed to be the case. Barring special cases like self-deception, my reason for Φ-ing is what I would or could specify in a sincere response to the question ‘Why are you Φ-ing?’. This answer typically takes the form ‘Because p’ rather than ‘Because I believe that p’. My reason for taking an umbrella is that it is raining, not that I believe that it is raining; for it is the weather rather than my own mental state that makes taking an umbrella good in my eyes. Now, animals can neither raise nor answer the question ‘Why are you Φ-ing?’ But the way we answer them shows: when we explain their behaviour by using the same intentional idiom as in explaining human behaviour, it is by reference not to their believing or desiring, but to what they believe or desire.

Subjectivists often appeal to the difference between ‘motivating’ and ‘justifying reasons’: whereas the latter are facts, the former are held to be mental states after all. But if the ‘motivating’ reasons for which people act and the ‘justifying’ reasons they have for acting belonged to distinct ontological categories, agents could not possibly act for a reason there is for them to act, which is absurd.Footnote ²⁸

Admittedly, sometimes we state our reasons in terms of what we believe, e.g.

1. (5) I am taking an umbrella because I think / fear it is raining.

Yet these are precisely cases in which the agent is not confident of knowing the pertinent facts. They can be plausibly construed not as referring to a subjective mental state, but as entering a caveat concerning the claim that specifies their reason. Accordingly, (5) can be paraphrased through a parenthetical construction like

1. (5’) I am taking an umbrella because it is raining (I think / fear).

Even in cases in which answers to the question ‘Why are you Φ-ing?’ take this form, the agent does not and would not reason on account of her subjective mental states. To be sure, I might say

1. (6) I am going to the fridge because I hope/believe that there is beer in there

   in order to dampen excessive confidence on the part of my hearers. But I would not normally reason

1. (7) I hope/believe that there is beer in the fridge, so let me go there.

Nor would I say

1. (8) My reason for going to the fridge is that I hope/believe that there is beer in it.

At most, my reasoning would take account of the need to qualify my epistemic status with respect to what I take to be the case. But this is once more captured by parenthetical constructions that avoid reference to a subjective state of believing.

1. (7’) There is beer in the fridge (I hope/believe), so let me go there.

2. (8’) My reason for going to the fridge is that there is beer in it (I hope/believe).

Qualifying my epistemic status in divulging my reasons can also take the form

1. (9) There is probably beer in the fridge, so let me go there.

(9) provides even less succour to subjectivism, since it alludes to a fact involving beer and the fridge, albeit a probabilistic one.

Another defence of subjectivism appeals to error cases, cases in which the relevant beliefs of the agent are mistaken. If A Φs in the mistaken belief that p, A’s reason cannot be the fact that p, there being no such fact. Instead, subjectivists insist, these reasons must be A’s mental state of (falsely) believing that p. From this they infer, moreover, that in the veridical case these reasons must also be mental states of (truly) believing. Yet this automatic inference from the error to the veridical case is just as problematic concerning intentional action as the analogous argument from illusion concerning perception. Hence there is room for disjunctive accounts of reasons for action.Footnote ²⁹ Furthermore, one could maintain that in error cases there are no bona fide reasons for which the agent acts, but only apparent reasons.Footnote ³⁰ Finally, my favourite option: even in error cases the reasons for acting are what is believed rather than states of believing, namely possible states of affairs in the sense of Wittgenstein's Tractatus, i.e. potential facts. If A falsely believes that p, there is nonetheless a non-obtaining state of affairs that p, which makes Φ-ing attractive in A’s eyes, since A regards it as a fact. Such a state of affairs is no more a subjective mental state than an obtaining state of affairs or fact; it is something believed to be the case rather than a believing. A’s motivating reason is a state of affairs presumed or perceived by A to be a fact.

8. Acting in the Light of Reasons

This is not the place to develop a watertight version of objectivism. What matters are two points: first, there are noteworthy reasons for adopting such an account; secondly, it removes an important obstacle to crediting animals with the capacity to act for reasons. Intentional explanations make an action that has occurred intelligible by showing that it made sense from the agent A’s cognitive and conative perspective.Footnote ³¹ This suggests a gloss on the idea of acting for a reason which is potentially more demanding than that of pursuing an adopted goal or being explicable in terms of beliefs and desires. The agent can act ‘in the light of reasons’.Footnote ³² The action makes sense to the agent (as the thing to do), given those reasons. And this implies an epistemic relation between A and A’s reasons. A must be aware or conscious of those reasons.

At this point, the contrast between subjectivism and objectivism makes a crucial difference. For subjectivism the reasons in the light of which A acts are subjective states causally antecedent to A’s action. In Davidson, for instance, A’s reason for Φ-ing is roughly speaking A’s mental state of believing that p will lead to q, in combination with her mental state of desiring that q.

Now, if reasons for action were subjective states of agents, acting in their light would require A to entertain thoughts about A's own thoughts. More specifically, it would require A to think about or represent its own states of believing and desiring. To put it more grandly, rational agents would have to possess not just consciousness of their environment, but self-consciousness, awareness of their own mental phenomena (states, processes, abilities, etc.). The same would hold if the reasons in the light of which A acts were psychological facts about A, e.g. the fact that A believes that p, desires X or wants to Φ.

Such self-consciousness may be beyond the ken of animals. That, at any rate, is the view among many otherwise divided commentators.Footnote ³³ However, the apparent obstacle does not even arise on the objectivist account of reasons for action. According to objectivism, reasons for action are things that are believed or known rather than subjective states of believing. This immediately removes the threat posed to the idea of animals acting in the light of reasons. Animals can adjust their actions with a view to how things are or appear to them to be. Consequently, they are capable of acting in the light of facts or states of affairs, in the light of how things are in their environment or of how they believe them to be. This is just a corollary of their cognitive capacities. Animals can perceive and hence believe and know that things are thus-and-so.Footnote ³⁴

The upshot concerning (d₂): animals are capable of acting in the light of (motivating) reasons, provided that these are construed as they should be: as objective facts, states of affairs or conditions of which the agent is aware.

9. Reasons and Having Reasons

There are lingualists who deny that animals can act in the light of reasons without appealing to subjectivism. Thus, Hacker suggests that there are reasons why animals act, but that they do not act for reasons in the sense of having reasons for acting.Footnote ³⁵ This implies that there are explanatory reasons for the actions of intelligent animals, yet no motivating reasons. Consider

1. (10) The elephant licks the rockface because it is salty

According to the proposal under consideration, the propositional clause following ‘because’ does not signify something objective that makes licking the rockface attractive to the elephant. In that case, however, it does not signify a reason, properly speaking. That the rockface is salty is merely a causal factor contributing to the licking. And as a causal factor it must be mediated by internal causes, mental states or occurrences of the elephant. So, this compromise seems to lead back to subjectivism-cum-causalism.

A more accommodating alternative runs as follows. The ‘because’-clause indicates neither a reason for which the elephant drinks nor merely an inner cause; instead, it (obliquely) indicates the elephant's goal in licking the rockface, namely to quench its craving for salt. Animals, the story goes, are capable of acting in pursuit of a goal and hence of acting purposively, but not of acting for a reason.Footnote ³⁶ This proposal is objectivist. In so far as animals can be cognisant of the goals they are pursuing, it does not per se militate against holding that animals act in the light of reasons, namely with a view to attaining their goals. Kenny, for one, accepts that animals can be ‘conscious of their goals in the quite literal sense [sic] they may see or smell what they are after’.Footnote ³⁷ Furthermore, we can move to the conclusion that animals can act in the light of reasons by exploiting the aforementioned

• (GR) A Φs in order to Ψ/bring about that p A’s reason for Φ-ing is to Ψ/bring about that p.

But this move faces another objectivist objection. Alvarez maintains that the only motivating reasons are facts.Footnote ³⁸ She grants that we can answer the question ‘Why are you Φ-ing?’ by citing a goal. She nonetheless resists the suggestion that goals can be motivating reasons, on the following grounds:

[b] acknowledges that the dog digs in order to achieve a goal. But against GR, [a] takes for granted that reasons should not be construed as something for which animals can act. Do [c] or [d] vindicate that assumption? It is perfectly unexceptional to say

1. (11) The reason why the dog digs is that it is looking for a bone.

It is similarly unproblematic to say that

1. (12) The dog's reason for digging is that there is a bone hidden there.

Now, (11) states an explanatory and (12) a normative reason. To ascribe a motivating reason one could avail oneself of

1. (13) The reason for which the dog is digging is that there is a bone buried there.

(13) sounds stilted, yet this holds equally for parallel statements about human agents. For both cases, moreover, a more idiomatic version is readily available.

1. (14) The dog is digging because there is a bone hidden there.

But even if (13) or (14) were problematic, they are required for ascribing motivating reasons to the dog only on the assumption that a subject cannot be motivated by a goal, and that ascribing motivating reasons must therefore be transposed into the format ‘The reason for which A Φs is that p’. There is absolutely nothing amiss with locutions like

1. (15) The dog has a reason for digging: it wants to find a bone

2. (16) The dog is digging because it knows (believes) that there is a bone hidden there.

Restricting animals to goals while denying them motivating reasons also requires that the apparently unproblematic and indispensable explanations of animal behaviour in terms of ‘because p’—as in (10) and (14)—are reformulated uniformly in terms of goals. It is far from clear how this could be done. We apply intentional explanations to animals because they display the same fundamental nexus between cognitive and conative capacities, between knowledge, belief, desire and action that we find in humans. We explain a well-known behaviour among some chimpanzees by saying that they batter nuts with hard stones in order to, or for the purpose of cracking them. But imputing this goal only makes sense on the assumption that they can respond to reasons objectively understood. They pursue their goal of cracking nuts by battering them on account of the fact that nuts crack when treated in this way. In short, intelligent animals pursue their goals in certain ways rather than others because of what they know or believe to be the case in their environment.

10. Reflecting on and evaluating reasons

Let us turn to a final lingualist objection that does not presuppose subjectivism. Some authors deny that animals can act in the light of reasons, on the grounds that they cannot reflect on their reasons.Footnote ⁴⁰ It is not enough for A to have beliefs and desires that rationalise the action, i.e. render it intelligible to us; in addition, A herself must be aware of the rationalising relation. Put differently, it does not suffice if A is aware of facts that, as we realize, render the action desirable for A. A must also recognize that the action is desirable because it fits the reasons. And that presupposes that the agent is capable of thinking about or reflecting on her reasons, and of thinking of them as reasons.

Admittedly, deliberation does presuppose the ability to reflect on reasons. But it does not follow that it presupposes the capacity to justify one's action verbally. Furthermore, while deliberation requires a capacity to consider reasons – i.e. facts pertinent to one's desires and goals – it is at any rate not obvious that this in turn presupposes the capacity to consider reasons as reasons. There is no general principle according to which being aware of or considering an F entails being aware of it as an F, and hence possession of the concept of an F. And even if there is a rationale favouring such a demand in this specific case, what precisely does it amount to? If it restricts the consideration of reasons to creatures in possession of the entire conceptual apparatus of reasons and practical reasoning, it is implausible. One need not possess the full concept of a reason in order to be aware of reasons, just as one need not possess the concept of affordances in order to be aware of affordances. A good thing too! Judging by the controversies surrounding it, the concept of a reason is both vexing and contested. It is therefore unclear what precise concept human mortals would have to master simply in order to act for reasons.

Even if one makes reflecting on reasons dependent only on a partial grasp of the concept of a reason, a difficulty persists. Do young children capable of asking questions like ‘Why do you …?’ have a partial understanding of the concept of a reason? On the one hand, they are no more capable of explaining what reasons are than ‘dumb’ animals. On the other hand, they can recognize what is a reason for what, notably by answering ‘why? questions by ‘because …’. To that extent language does make a difference, though it is more gradual than lingualists suppose: reflecting on reasons as reasons comes in degrees, depending on the linguistic repertoire. My answer to question (d₃) – Can animals reflect on reasons? – is therefore hedged: in so far as reflecting on reasons requires reflecting on them as reasons and therefore mastery of the concept of a reason, it depends on some mastery of the idiom of intentional explanations.

However important, this concession does not reinstate a negative answer to (d): Can animals act for a reason? For the requirement that a subject acting for reasons should be able to reflect on her reasons is contestable. Indeed, there is a positive argument for keeping having reasons separate from reflecting on reasons: A couldn't develop a capacity to reflect on A’s reasons if A didn't have reasons to begin with.Footnote ⁴¹

Lingualists could retort that having reasons of a normative and explanatory kind is less demanding than acting for a reason. In one respect this is true: there are reasons for me to Φ that never motivate me to Φ. But a subject cannot have such reasons without being able to act for them. Lingualists would therefore have to insist that a subject cannot act for reasons without being able to reflect on them. Acting for reasons and being able to reflect on them come as a package.

But why accept the holistic package? It is a conceptual truism that while one cannot reflect on one's reasons without having such reasons, the reverse does not hold. Consequently, the burden of proof is on the holist to show why this should not also hold for acting for a reason. It will not do to insist that there is a concept of acting for a reason that requires an additional capacity compared to acting and having reasons. For that more demanding concept is not the one embodied in the established use of ‘having a reason’, ‘acting for a reason’ and ‘reflecting on one's reasons’. That use, as exemplified in (11) – (16), allows for explaining animal action by reference to reasons animals have for acting; it thereby implies that they can act for those reasons. At the same time, it leaves open the question whether they can think thoughts like ‘My reason for Φ-ing is that p’, or ‘Let me Φ because p’.

Holism faces another obstacle. How can it account for the intermediate stages that lie between a one-year old pre-linguistic human infant, who allegedly can neither act for reasons nor reflect on them, and the adult, who can do both? Let us grant for the sake of argument that merely having reasons is a stepping-stone and that acting for reasons is a more advanced phenomenon that develops in tandem with reflecting on reasons. Even then it would remain to be shown that an increase in the ability to act for a reason is ipso facto an increase in the ability to reflect on reasons, and vice versa.

A parallel response can be given to another worry. In crediting animals with the capacity to pursue their own goals I committed myself to the claim that animals are capable of pursuing a good subjectively conceived, something they themselves want. But, the worry runs, how can animals conceive of something as good? Turning this into a lingualist objection: animals cannot act in the light of reasons, since they cannot evaluate reasons. Once again, however, there is a riposte: one cannot evaluate one's reasons without having such reasons, so subjects must be capable of having reasons independently of evaluating them. Furthermore, on an objectivist account evaluating reasons is to evaluate facts, situations or outcomes with a view to attaining one's goals. Some animals are capable of doing so. The way they choose between different options or courses of action both in the wild and in the laboratory establishes one thing beyond reasonable doubt. They can recognize or perceive a phenomenon X as having certain affective, conative or emotional dimension – as dangerous, edible, interesting, appealing, etc. They can also recognize its instrumental significance: X can be recognized as a way of attaining Y. Indeed, great apes and cetaceans, at least, can also learn to distinguish means from ends. They display awareness of the fact that one and the same end can be achieved by different means and that one and the same means – for instance a tool – can serve different ends.Footnote ⁴² Finally, X can acquire significance, for instance when an animal adopts a neutral feature of the physical environment as a landmark, e.g. as marking the border of its territory. In short, animals cannot just perceive facts, they can also evaluate them – or, in a more realistic ethological idiom, features of their environment – in the sense of responding to them as good or bad (e.g. attractive or threatening), and they can act accordingly. Reasons objectively conceived are not just cognitively accessible to them, they are also conatively assessable by them.

As a last resort, the lingualist might deny that evaluating reasons is the same as evaluating facts. Evaluating facts is getting clear on which course of action to pursue. By contrast, the story goes, evaluating reasons is weighing one's reasons, and this requires a basic understanding of them as reasons. However, on the objectivist assumption that reasons are facts, this objection can only mean: there is an evaluation of facts as reasons, i.e. as things to be taken into consideration in one's deliberations. This capacity does presuppose a facility for reasoning. Indeed, it requires the additional capacity to think about one's reasoning or deliberation. The latter presupposes concepts like that of a reason, and thereby language (a point conceded above). But that is a presupposition for a subject explicitly reflecting on the role that facts qua reasons play in her self-conscious deliberations. It is not a presupposition for a subject choosing between different options on account of facts. Animals can act in the light of reasons, as required by (d₂), even though they cannot reflect on them as reasons, as required by a certain understanding of (d₃).Footnote ⁴³