A dog has been hit by a car, and lies unconscious on a busy highway in Chile. The dog’s canine companion, at enormous risk to his or her own life, weaves in and out of traffic, and eventually manages to drag the unconscious dog to the side of the road (see www.youtube.com/watch?v=-HJTG6RRN4E). A female elephant, Grace, tries to help the dying matriarch of another family of elephants, and appears distressed when she is unable to do so effectively (Douglas-Hamilton et al. 2006). A gorilla lifts the unconscious body of a small boy, who has fallen into her enclosure, and carries him to the gate where she hands him over to a keeper (Bekoff and Pierce 2009). A rhesus monkey refuses to take food, when doing so will subject another monkey to an electric shock. The monkey persists in this refusal for twelve days, nearly starving himself to death (Wechkin et al. 1964). What should we make of cases such as these? Here is one possibility: these cases form parts of a large and growing body of evidence for the claim that some nonhuman animals (henceforth ‘animals’) can exhibit moral behavior (see de Waal 2006 and Bekoff and Pierce 2009 for excellent surveys of the evidence). Most philosophers and scientists reject this possibility. I shall defend it.
Two issues should be distinguished. One is an empirical issue about the actual nature of an animal’s motivation in any given case. The near blanket rejection of the possibility of moral behavior in animals, however, is not driven by an empirical examination of particular cases but by conceptual considerations. There is, allegedly, something about the nature of moral behavior that renders animals incapable of engaging in it. And, skeptics aver, we can establish this claim without the need for empirical examination of individual cases (Dixon 2008). The basis of this denial is always the same: whatever else is true of animals, they are not responsible for what they do, and being responsible for what one does is a necessary condition of acting morally. I shall argue that, on the contrary, there are no insuperable conceptual obstacles to regarding animals as capable of moral behavior. I have developed this case in much more detail elsewhere (Rowlands 2011, 2012). Here, I can merely provide an outline of that case.
If something falls within moral space at all, it is generally thought to belong to one or both of two categories: it can be a moral patient or a moral agent or both, where:
I shall argue for the existence of a third category – an additional, curiously overlooked, region of moral space:
I doubt any animals are moral agents. I’m not sure humans are. But I think at least some animals are moral subjects.
The category of the moral subject has languished unrecognized because it is almost universally thought to collapse into that of moral agent. The reason for this stems from Kant’s dictum: ‘ought’ implies ‘can’. Imagine someone – for entirely obvious reasons, we can call him Sigmund – whose motivations are always hidden from him. The motivational component of his mind is akin to a black box: replete with states that successfully guide Sigmund’s behavior, but to which he has no first-person access. Sigmund, it seems, would be ‘at the mercy’ of his motivations: having no idea what motivates him and, therefore, having no control over those motivations. Sigmund just finds himself doing things on the basis of motivations he neither recognizes nor understands. Sigmund, therefore, cannot be a moral subject. Having no control over his motivations, he can neither embrace nor resist them. But if he can neither embrace nor resist his motivations, it makes no sense to say that he should embrace or resist them. Sigmund’s motivations, in this sense, make no normative claim on him. However, moral motivations are precisely things that make normative claims on their subjects. Good motivations should be embraced, and evil ones should be resisted. Therefore, it seems Sigmund’s motivations cannot be moral ones. Generalizing: without control, there can be no moral motivation. Therefore, the category of the moral subject collapses into that of the moral agent.
In the Kantian tradition, also, we find a specific conception of control: control over one’s motivations derives from the ability to critically scrutinize them. Thus, Korsgaard writes:
Once you are aware that you are being moved in a certain way, you have a certain reflective distance from the motive, and you are in a position to ask yourself, “but should I be moved in that way?’ Wanting that end inclines me to do that act, but does it really give me a reason to do that act? You are now in a position to raise a normative question about what you ought to do.
(2006, p. 113)
We might call this the Scrutiny-Control-Normativity-Morality (S-C-N-M) nexus. The ability to critically scrutinize one’s motivations gives one control over them. This control permits these motivations to make a normative claim on their subject, and so makes them the sort of motivations that might be moral. Thus: animals cannot have moral motivations because they have no control over their motivations. And they have no control over their motivations because they lack the ability to critically scrutinize them.
This argument is only as strong as the claim that ‘ought’ implies ‘can’. And while it has the status of dogma, I think it is untenable. To see why, consider two arguments.
Three-year-old Jamie Bulger was abducted, tortured, and murdered by two boys, Robert Thompson and Jon Venables. At the time of the murder, both Venables and Thompson were ten years old, and became the youngest convicted murderers in English history. They kicked Bulger, threw bricks at him, and hit him with an iron bar. Bulger suffered ten skull fractures, and so many injuries that none could be definitively identified as the mortal one. Following the murder, Venables and Thompson placed Bulger’s body on railway tracks in the hope that a train would make his death appear an accident. Under questioning, they revealed that they had planned to abduct and murder a child that day (and also that their initial intention was to take him to a busy road and push him into oncoming traffic – that aspect of the plan later changed).
Suppose one decides – as many theories of moral development suggest – that Venables and Thompson were below the age at which they could be regarded as morally responsible. One deems them to lack the control over their motivations necessary for responsibility. Even if this were true, would we really want to deny that their motivations were morally bad ones? Even if we wished to rescind from attributions of moral responsibility to Venables and Thompson, would we really want to say that, therefore, their motivations carry no moral weight? That claim, if one is not in the grip of a peculiarly warped moral psychology, is as counterintuitive as a claim can get.
Rescinding from moral evaluation of individuals is one thing; rescinding from moral evaluation of their motivations is quite another. However, if ‘ought’ implies ‘can’, the former rescindment entails the latter. A subject’s lack of control over a given motivation would entail that the motivation makes no normative claim on the subject: it would not be the sort of thing he should resist because it is not the sort of thing he can resist. Therefore, the motivation could not be a moral one (it would be amoral). If they were not responsible, Venables and Thompson’s motivations would not be morally bad. To avoid this conclusion, we need to reject the idea that ‘ought’ implies ‘can’.
In determinism world – which may or may not be the actual world – hard determinism is true, and no one is, therefore, ever responsible for what they do. Would we really want to say that, in this world, there is no such thing as moral motivation? When Hitler (or the worldly equivalent thereof) starts a World War and attempts to exterminate various races, would we want to say that his motivations do not count as morally evil? We might, in such a world, justifiably rescind from evaluation of Hitler, the person: we might, that is, refuse to blame or hold him responsible for what he does. But refusing to classify his motivations as even falling into the category of the moral is highly counterintuitive. The principle that ‘ought’ implies ‘can’ commits us to this. If Hitler cannot control his motivations, no sense can be made of the idea that he should resist them. His motivations make no normative claim, and therefore cannot qualify as moral. To avoid this, we must reject the idea that ‘ought’ implies ‘can’.
The category of the moral subject is, therefore, a desirable one. But the desirability of a category is one thing; its actual existence is quite another. We can bolster these considerations by attacking the S-C-N-M nexus. Two links in the chain are particularly vulnerable: (i) scrutiny-control, and (ii) control-normativity.
Generally speaking, the idea of critical scrutiny of motivations comprises three distinguishable elements: recognition, interrogation, and judgment: a subject recognizes that she has certain motivations, and on the basis of this recognition can ask herself questions such as ‘Is this motivation one I should embrace or resist?’ Moreover, she also understands how she should attempt to answer this question by judging the motivation in the light of moral principles or propositions that she antecedently holds.
The idea that the ability to engage in critical scrutiny of her motivations gives a subject control over those motivations does, admittedly, have intuitive appeal. In the absence of this ability, she is, it seems, at the ‘mercy’ of her motivations. She has them, she acts on them, and that is pretty much that. The ability to critically scrutinize her motivations would, seemingly, transform her. She could, now, sit above the motivational fray: observing, judging, and evaluating her motivations, coolly deciding the extent to which she will allow them to determine her decisions and actions. This gives her a control over her motivations that she would otherwise lack.
Intuitively appealing or not, this account has familiar problems (Friedman 1986, Thalberg 1989, Noggle 2005, Rowlands 2012). There is a recalcitrant property – the property of being under the control of the subject – that first-order states (motivations) lack. We, therefore, introduce higher-order states – recognitions, interrogations, and judgments of those motivations – to supply this control. But the same issue of control will, logically, arise at this higher order too. Do we have control over our higher-order recognitions, interrogations, and judgments of our motivations? If so, then we have not explained the notion of control, but simply assumed it. But if not, then it is difficult to see how these higher-order interrogations and assessments could supply us with control over our motivations. The ability to critically scrutinize our first-order motivations was supposed to lift us above the motivational fray, allowing us to dispassionately pass judgment on our motivations, and thus providing us with control over them. But it, in fact, does no such thing: either because the higher-order states and processes are themselves part of that fray or because the appeal presupposes what it is supposed to explain – a subject’s control over her mental life.
This problem is reasonably familiar, and there is an equally familiar response: additional specification of the conditions under which this critical scrutiny is to take place. For example, it must take place under conditions free of distorting factors, or must reflect an adequate causal history, and so on (Christman 1991, Mele 1995). For our purposes, this kind of response is irrelevant: whether or not the scrutiny takes place under conditions free of distorting factors or reflects an adequate causal history is not something that is, generally, under the subject’s control. The response seeks – in effect, by stipulation – to divorce the idea of control from what is often called autonomy. The response attempts to preserve a viable notion of the latter in the absence of the former. Our question, however, is whether we can explain control over motivations by appeal to higher-order states such as recognitions, interrogations, and judgments – for it is the lack of these that is what is thought to deny animals the possibility of controlling their motivations. It is, therefore, simply irrelevant to point out that there is another conception of autonomy that does not require control over motivations.
The control-normativity link is also vulnerable: it is clear that normativity, in general, does not require control. After all, other cases of ‘ought’ do not seem to imply ‘can’. Consider, for example, the prudential ought. An alcoholic ought – prudentially speaking – to give up drinking. Whether he is capable of doing so is irrelevant to the truth of this claim. He ought to give up drinking – to resist this motivation to drink – even if he, in fact, can’t. Why must the moral ‘ought’ require ‘can’ when at least some other versions of ‘ought’ do not?
There is, in fact, a conception of the moral ‘ought’ under which it does not imply ‘can’. This is an externalist, consequentialist one. The type of externalism in question is evaluational and the type of consequentialism is objective (Driver 2000). Such an account will assume a reasonably robust sense of ethical objectivity: situations contain features that make them good or bad independently of the subjective states of the agent. The evaluation of a motivation will then be a function of whether it systematically (as opposed to accidentally) promotes good- or bad-making features of situations. The normative status of a motivation is, therefore, explained in terms of relations it bears to certain external factors, rather than the subject’s control over it.
With this general model in mind, consider the case of the dog described earlier. His companion lies unconscious on a busy road. This is, let us suppose, a bad-making feature of the situation. The first requirement is that the would-be rescuer is sensitive to this bad-making feature. Such sensitivity does not require that the dog is able to think thoughts such as ‘This is bad!’ The appropriate sensitivity can, in fact, be purchased by other means. Most obviously, suppose the dog possesses a mirror neuron system. When he sees another animal in distress, he is caused to feel distress too. The other’s distress is not merely a cause of the dog’s distress. Rather, the other’s distress is an intentional object of the dog’s distress: he is distressed that the other is distressed. (Or, if one has qualms about attributing mindreading abilities to the dog, one can reframe his distress in terms of behavior reading: the dog is distressed at the other’s distress behavior. See Monso, 2015.) The distress experienced by the dog will cause him to attempt to ameliorate the distress of the other.
Nowhere in this general picture is there any suggestion that the dog has control over his sentiments, still less that he is able to critically scrutinize them. Rather, the dog has an emotional sensitivity to at least some of the good- and bad-making feature of the environment. This sensitivity is fragmentary. There is no suggestion that the dog is sensitive to all the good- and bad-making features of situations. Rather, the claim is simply that he can recognize some of them and, moreover, when he does so, he can be correct (or incorrect) in this recognition. This possibility is a consequence of the robust sense of ethical objectivity advertised above. If it is indeed true that the condition of his companion is a bad-making feature of the environment, then in his experienced distress and resulting attempt to help, the dog is responding to this bad-making feature in a morally appropriate way.
Is the dog morally motivated – motivated to act by way of a moral motivation? One way to address this question is by way of considerations of parity. Suppose there were a human who exhibited the same kind of profile as the dog. He is sensitive to some (but not all) of the (morally) good- and bad-making features of situations. This sensitivity is reliable and takes an emotional form: with respect to the good- and bad-making features to which this human is sensitive, he reliably experiences distress at the bad-making features and joy at the good-making features.
Is such a human a moral subject? A dyed-in-the-wool Kantian would deny that he is. I have raised some objections to the Kantian model, and attacked some of its crucial elements – the connection between scrutiny and control, and between control and normativity. But experience suggests that if someone really wants to be a Kantian, then that is precisely what he/she is going to be. But other accounts of morality are far more hospitable to the idea that our imagined human is a moral subject. A sentimentalist account will have to take seriously the idea that this imagined human is a moral subject. And if the human is a moral subject, then the sentimentalist will have to take seriously the idea that some animals are too. The blanket dismissal of the possibility of moral behavior in animals has no justification. Can animals be moral? On at least some, prominent and respectable, accounts of morality, yes they can.
Bekoff, Marc, and Pierce, Jessica (2009) The Moral Lives of Animals (Chicago: University of Chicago Press).
Christman, John (1991) ‘Autonomy and personal history’, Canadian Journal of Philosophy 21, 1–24.
de Waal, Frans (2006) Primates and Philosophers: How Morality Evolved (Cambridge: Harvard University Press).
Dixon, Beth (2008) Animals, Emotions, and Morality (New York: Prometheus Books).
Douglas-Hamilton, Iain, Bhalla, Shivani, Wittemyer, George, and Vollrath, Fritz (2006) ‘Behavioural reactions of elephants towards a dying and deceased matriarch,’ Applied Animal Behaviour Science 100, 67–102.
Driver, Julia (2000) Uneasy Virtue (New York: Cambridge University Press).
Friedman, Marilyn (1986) ‘Autonomy and the split-level self’, Southern Journal of Philosophy 24, 19–35.
Korsgaard, Christine (2006) ‘Morality and the distinctiveness of human action’, in Frans de Waal, Primates and Philosophers: How Morality Evolved (Princeton, NJ: Princeton University Press), pp. 98–119.
Mele, Alfred (1995) Autonomous Agents: From Self Control to Autonomy (New York: Oxford University Press).
Monso, Susana (2015) ‘Empathy and morality in behavior readers’, Biology and Philosophy 30(5), 671–690.
Noggle, Robert (2005) ‘Autonomy and the paradox of self-creation: Infinite regresses, finite selves, and the limits of authenticity’, in J. Taylor ed., Personal Autonomy (Cambridge: Cambridge University Press), pp. 87–108.
Rowlands, Mark (2011) ‘Animals that act for moral reasons’, in T. Beauchamp and R. Frey eds., Oxford Handbook of Animal Ethics (New York: Oxford University Press), pp. 519–546.
Rowlands, Mark (2012) Can Animals Be Moral? (New York: Oxford University Press).
Thalberg, Irving (1989) ‘Hierarchical analyses of unfree action’, in John Christman ed., The Inner Citadel (New York: Oxford University Press), pp. 123–136.
Wechkin, Stanley, Masserman, Jules H., and Terris, William (1964) ‘Shock to a conspecific as an aversive stimulus’, Psychonomic Science 1, 17–18.