Must we Weep for Sentimentalism?

This paper is published in James Dreier, ed. Recent Debates in Moral Theory. (Blackwell, 2006) All references should be made to, and from, the published version.

1. A misunderstanding

Hume said that the distinct boundaries and offices of reason and taste are easily ascertained, including under the heading of ‘taste’ the moral sentiments.[1] Alas, he proved over-optimistic. I doubt if any question in moral theory has proved more vexatious. The area is confounded by difficulties over the identification of attitudes and beliefs, over the distinction between senses of the word ‘reason’ that sentimentalists can admit from those they must deny, over the relation between properties and concepts, over the metaphysics of the categorical imperative, and over much else besides. In this brief essay I can therefore not attempt a full-scale defence of sentimentalism. I shall simply defend the theory against various recent assaults, one of which is mounted in Samuel Kerstein’s defence of rationalism in this debate. My impression is that Kerstein does not stand alone, but is a spokesman for a whole phalanx of people, perhaps calling themselves ‘Kantians’, who would sympathize with his assault, or at least fail to understand how a sentimentalist could withstand it.

It is fortunate, then, that the misunderstanding that permeates Kerstein’s treatment of sentimentalism is highly visible, and I shall concentrate on one particularly exposed passage. After giving an account of my own neo-Humean description of the emotions and attitudes that underlie our propensity to go in for ethics and morals, he considers the issue of justice to strangers or outsiders. He writes that on my view:

It is a person’s displeasing sentiments, ones such as unease or shame, that form the basis of her obligation to acquire the character trait of being just to strangers, or at least to act in a way that a person with this trait would act.

He continues:

This last point is crucial to the issue of whether sentimentalism coheres with the idea that there are categorical imperatives. On this account the basis for an agent’s obligation to do something is a displeasing sentiment she has when, after taking the “common point of view,” she contemplates her not doing it or, perhaps, her not possessing the character of someone who would do it…If an agent does not have this sentiment, then she has no obligation. Of course, if an agent has no obligation to perform a certain action, then a principle commanding that action does not count as a categorical imperative. For it belongs to the concept of a categorical imperative that everyone within its scope is obligated to do what it enjoins. So in order for sentimentalism to ground a particular categorical imperative, each and every person, after taking the common point of view and so forth, must have a displeasing sentiment towards not doing what the imperative commands.

So, although Kerstein also chides me for failing to answer the question of how moral obligations ‘stem from’ the processes I have described, he supposes that such an account, were it provided, would inevitably suppose that people without the sentiments are free of the obligation: the ‘basis’ of her obligation is a sentiment, so that ‘if an agent does not have this sentiment, then she has no obligation’. In a similar vein he imagines someone with no sympathy for members of some minority within his society, and says that ‘on Blackburn’s sentimentalist account, you have at this point no obligation to refrain from abusing the minority’.

And then, unsurprisingly, he can go on to point out that there are legions of unhappily bad Samaritans, and what I called foreign-office knaves, when I was stressing and lamenting the same sad fact about humanity. These people do not have the appropriate sentiments. Hence, Kerstein concludes, for the sentimentalist, there are people who lie under no obligation to universal justice. Hence, there are no categorical imperatives, for the categorical imperative embraces everyone.

Whatever else is to be said about it, we should notice that this argument is remarkable for its scope. It can be directed not only against sentimentalism, but against any theory that seeks to explain our moral capacities in terms of contingent and potentially variable facets of human nature: language, culture, upbringing, acquired ‘second nature’, and so on. Even reason, insofar as it is empirically variable, or leaves its possessors liable to partial and self-serving policies, will not be enough. Only a universal birthright – and one strong enough to deliver commands to the will – could withstand it. It is a pity then, that Kerstein himself is not confident of a Kantian story of this kind, since it seems to be the only hope for a theory of the requisite standing. Otherwise there seems to be a straightforward empirical problem. If there is an inner mechanism of reason strong enough to dragoon us all into the ranks of the caring and just, it seems odd that so few of us get affected by it.

Why does Kerstein suppose that on a sentimentalist story, the knaves and villains are exempt from obligations? I should have thought no moral philosopher, except perhaps Gilbert Harman, and certainly not Hume nor myself, could have been thought to suggest such a thing. In fact, it seems to me such a shocking thing to say, that I am at a loss to understand how Kerstein could have read Hume, or me, and perhaps others such as Allan Gibbard, as saying anything that implies it. For the record, I explicitly say the reverse, fairly often.[2]

The only explanation I can offer for the misreading is that it comes from conflating two different projects. One, the project of the anatomist, in Hume’s terms, is to give an accurate and complete account of the states of mind that gain expression in moral thinking. The other, a moralistic project appropriate to Hume’s painter, is to give an account of the ‘sources’ of our obligations. In my discussion of Christine Korsgaard’s account of ‘the normative question’ I voice some doubts about how to conduct this second project. Like other pluralists, I think obligations arise for different reasons, and I am not myself wedded to the idea that any one, clear, univocal concept, such as ‘utility’ or ‘self-legislation’ might have been thought to be, plays the same explanatory role when we try to describe why we lie under one or another obligation. But it is the anatomist’s project that occupies the bulk of my work, and that justifies calling me a ‘sentimentalist’.

If you confuse these two projects, you might end up saying that moral obligations ‘stem from’ or ‘are based in’ psychological states, and thence infer that in the absence of the psychological states, the obligations disappear as well. The anatomical view is then  supposed to lead to bad morals or bad painting. But it is not I who says that. I would say, for instance, that your obligations as a parent stem from the dependency of your children, their needs, and the absence of other social resources provide a substitute if you fail to meet those needs. One of these needs is affection, so if you don’t care about your child, you are in breach of the obligation that the child’s need places on you.[3] The obligation does not come and go according to your affections, any more than your debt comes and goes depending on whether you care about it. And I think it shocking to suppose otherwise. The obligations you lie under, like the debts you owe, don’t decrease or disappear when you stop caring about them.

I think, then, that parents of young children lie under a complex obligation, O. According to the sentimentalist, I say this by way of expressing a complex of attitudes and feelings towards the relationship between parents and their young children—what I shall call ‘these sentiments’. Now let us say that someone who ignores or negligently or deliberately falls short in fulfilling an obligation, fails O. Finally, suppose we say that people who have no sentiments corresponding to feeling the weight of an obligation, laugh-off O. Then all I ask is that we recognize the distinction between:

If I (we) had not had these sentiments, I (we) would not have been condemning parents who fail O or even those who laugh-off O.

If I (we) had not had these sentiments, I (we) would have failed O or laughed-off O.

If parents X do not have these sentiments, then they are likely to fail O or laugh-off O

If parents X do not have these sentiments, then they are under no obligation O.

The first three of these are true and harmless. The last is false and deadly. But it is the last that is foisted upon the sentimentalist in the passages I quoted.

I think the transition from the harmless to the deadly is lubricated by careless use of phrases like ‘is based upon’, or ‘stems from’. If you ask me what moral thought itself stems from or is based upon, then, as an anatomist, I give the sentimentalist reply. If you ask me what a particular obligation or duty stems from or is based upon, then my painterly answer may vary, but will seldom cite the feelings of the agent. In this case it stems from the needs of the children, and the sociological structures that makes the parent the person responsible for meeting those needs. In the case of justice to outsiders, again it may stem from the needs of the outsiders, or our overall needs for accommodation with them, or perhaps it stems from fundamental rights to equal treatment. I am not sure: the relationship between justice and mutual advantage, and reciprocity, and equality, is obscure enough for me and many others to feel insecure about exactly how best to paint it. What I am sure about is that you cannot get rid of the obligations by not feeling them, or laughing them off.

2. Peacocke’s  Hi-Tec Version

A trivial misunderstanding becomes worrying when you find it shared by enough people. As I have said, I fear that Kerstein is not alone. Christopher Peacocke has recently suggested that sufficient attention to two-dimensional modal logic conjures up a dependency claim which the quasi-realist must accept, but which offends against some conviction that we hold.[4] So, contrary to what I have repeatedly claimed, there is a mind-dependency claim which causes trouble for any sentiment-oriented theory of value. The technology in Peacocke’s discussion will doubtless shock and awe enough readers for it to be worth some trouble to show that it is in fact a smokescreen. The issues can be put simply enough, and when they are, the objection disappears.

In a nutshell, the issue goes like this. Peacocke  recognizes the general strategy I have repeatedly used. It is an integral part of our ethical lives that we can evaluate scenarios that are described to us, whether past, present, or merely possible or fictional. So if you bring me a story about people and their doings, I can train my thoughts on it, and according to the attitudes it elicits, I will admire it or condemn it, or hold a whole variety of more or less nuanced responses. If you tell me a story in which people fail to meet their childrens’ needs, I react badly, and I express the conviction that what they are doing is cruel and wrong. I hope we all do. If asked why I condemn their behaviour, at least a prime part of my answer is about the needs that are not being met. Perhaps this simplifies a little, since the indifference of the perpetrators also matters, and that is a feature of the perpetrator rather than of the children. But for clarity, and as a harmless simplification, I shall say the verdict is child-dependent. If you told me a story about people causing pain to animals such as dogs, my verdict would be dog-dependent.

So far so good. Now suppose your story is more complicated. You tell me a story in which people not only ignore their childrens’ needs, but also fail to condemn such behaviour, or even admire it. They congratulate and esteem especially harsh or negligent parents. Their moral sensibilities are here the opposite from ours. What am I to say about this? It looks equally bad or somewhat worse to me. In the first story we could imagine some guilt attaching to the behaviour: perhaps it is mainly adolescents or criminals or failures who are bad with children, and their ordinary morality condemns it. But in the second story, there is no condemnation from the people who are described. They admire negligence or brutality. It is a horrible scenario, and I deplore it the more.

I have often stressed two further related points. The first is that someone could disagree with me about what I have just said. He could urge that the fact that they find it admirable makes all the difference, makes it admirable in fact. We have a moral disagreement, for I deny that. I hold that it is the sad life of a child that is so shocking, and in this imagined society, the parents’ self-congratulation at what they are doing takes none of that away, but actually adds to it, making it even more shocking. The second addition is that there might be examples—call them etiquette examples—where the structure looks similar, but my opponent would be right. For there are cases where the bad we do would not be bad at all were it not for the community’s unfavourable take on it. I can imagine communities (perhaps there are some) where it is very bad form indeed to give a gift in return for a gift received. In such a community it would be insulting and wrong to do something which amongst us would be a normal expression of gratitude or reciprocated friendship, and it would be right to do something—omitting to reciprocate— which amongst us would be a breach of manners, and even insulting and wrong. In such a case it is true that had we had these different attitudes, different actions would have been right or wrong.[5] Their value is due to the conventions of etiquette that people follow, and these might have been harmlessly different. In the child case that is not so; as I said, it is due to the unmet needs of the child.

How does Peacocke hope to embarrass this analysis? He says some curious things about it. He says at the beginning of the discussion  ‘it is very hard to see how it can be denied that, under (my) approach, the conditions under which someone is correct in asserting a moral proposition have something to do with expressed mental states’ [6] And the intention is to show that although, as he recognizes, I claim not to have a ‘mind-dependent’ treatment of morality, in fact I do. Unfortunately these wordings, like Kerstein’s above, and others to which we shall come, are ambiguous. Obviously an expressivist treatment of ethics is ‘mind-dependent’ in one sense—it starts from reflections on the kind of mental state that gets expressed when values are made public and exchanged. Obviously as well ‘the conditions under which someone is correct in asserting a moral proposition have something to do with expressed mental states’ in one sense. Were the expressed mental states different, the proposition would be different and would be correct under different circumstances. For example, if the sentence ‘kicking dogs is wrong’ standardly expressed approval of kicking dogs, anyone voicing it would be correct only in quite different circumstances, such as ones in which dogs have no conscious states. But there is nothing worrying to the expressivist (or anyone else) in thoughts such as those. The conditions under which someone is correct in asserting any proposition has something to do with expressed mental states, in this sense. It has to do with which beliefs are being expressed. Of course, in another sense it has nothing to do with mental states: unless a proposition is explicitly about the mind, its truth condition will be world-dependent rather than mind-dependent. But similarly the truth (for I say it is a truth) that you have an obligation to your children is child-dependent, and the truth that you should not kick friendly dogs for fun is dog-dependent.

Peacocke  pursues his attack via an indexing of propositions, corresponding to reference first to the ‘world’ from which an evaluation is made, and secondly to the ‘world’ that is being evaluated. To this end he introduces the double index P(w1, w2), explained as:

Proposition P, when evaluated from the standpoint of psychological states in w1, holds with respect to w2. [7]


P here is some moral proposition, such as ‘it is wrong to kick dogs for fun’ or ‘the infliction of avoidable pain is wrong’. The overall proposition P(w1, w2) is, however, not entirely clear, because of the curious and treacherous word ‘holds’ (with its shades of ‘based on’ and ‘stems from’). Suppose someone says that the proposition that the war in Iraq was justified holds from George Bush’s point of view. I can hear that as a contorted way of saying that George Bush believes that the war in Iraq was justified, and it is probably true. What I should not do is hear it as some kind of insinuation that the war in Iraq was justified. It’s a description of what George Bush thinks, not an endorsement of the way he thinks. Only a confused relativist of some sophomoric stamp would accept the transition from ‘the war in Iraq is justified from George Bush’s point of view’ to ‘the war in Iraq is justified’.

With that clear for the moment, we can turn to the ‘w2’ variable. For that to do any work, there has to be some space between the proposition P and the variety of worlds to which it applies, or in which it is evaluated. And this may be granted. ‘it is wrong to kick dogs for fun’ can be tested against this world, or, if we are imaginative enough, against slightly different worlds, for instance in which there are still dogs and people, but only dogs that are unconscious, or in which there are only people who can survive by nutrition from the pain of other animals. And then it may turn out that the moral proposition is only true contingently on aspects of our world, and would get a different truth value were these other things different. Or of course, it may not. We may suppose that however worlds vary, it is always wrong to cause unnecessary pain, although even that may wobble if we bring in, for instance, apparently possible people who like pain.

With these explanations we can agree with Peacocke when he says that nobody can object to the employment of this doubly indexed proposition. Nobody can object to it, for P(w1, w2) can be the form of good enough propositions, that can be regarded as true or as false in various cases, although they will often be indeterminate, when we have not given definite enough interpretations of the variables. Given what I have said, the evaluation of such propositions goes like this. We tell what we might call a treble story. First we introduce a moral proposition P. Second, we introduce some possible people with attitudes. And third we present a possible scenario, and we imagine the people we just introduced evaluating what goes on in the scenario, in accordance with the attitudes we gave them. If the people introduced evaluate the introduced scenario in the way that would properly gain expression by P, then P (w1, w2) should be accorded T, otherwise not. It corresponds to ‘the people we have introduced evaluate the scenario we imagined them to be contemplating, in a way that could be expressed by saying that P’. More concisely, we can say that the people we have introduced evaluate the scenario we imagine them to be contemplating, in the P-way.

Not surprisingly, we can vary the people or psychological states introduced, and we can vary the scenarios we conjure for them to be contemplating. So P(wi, wj) can vary in two dimensions: there are two variables to be given interpretations before we turn it into a definite claim. And filling in one does not determine how we fill in the other. Or, of course we could quantify. ("wi)P(wi, x) would mean that everyone from any possible story evaluates some given scenario x in the P-way, and ("wj)P(z, wj) would mean that the introduced persons with the psychological states z, evaluate every possible scenario in the P-way. ("wi) ("wj)P(wi, wj) would mean that everyone, whatever their other differences evaluates everything in the P-way.

This is the machinery, so what happens when it is set in motion? Alas, nothing at all. We get a variety of rather cumbersome descriptions of what different people think about different scenarios, and whether they would express themselves as agreeing with some moral or ethical proposition. We get things like ‘we, as we are, think that in the world, as it is, kicking dogs is wrong’ (true, I hope). Or, ‘we, as we would be were we to become coarse and callous, would think that in the world as it is, kicking dogs is wrong’ (false, no doubt). We can keep the people constant (‘us’) giving what Peacocke calls the ‘vertical’ reading, or we can vary people and scenarios together, giving what he calls a ‘diagonal’ reading, such as the true ‘we, as we would be were we to become coarse and callous, would think that in a possible world in which dogs feel pain slightly less than they do, kicking dogs is OK’. 

Peacocke claims that since there is this diagonal reading, there is a ‘mind dependency’ claim that the quasi-realist has not acknowledged. But that is just not true. Propositions such as this last one amount to descriptions  of how people of some particular attitude (which we may or may not share) react to different scenarios. And there is nothing in general in these descriptions to offend the quasi-realist (or anyone else). It does not amount to giving our own verdict on those same scenarios, although if make ourselves the topic, and describe ourselves rightly, there will be the coincidence that what we say about ourselves will be true just if we do assent to the verdict P.

The locutions that Peacocke uses reveal him to be in the same swamp as Kerstein:

on the quasi realist’s theory the acceptability of basic moral principles depends on some psychological attitudes. However this dependence is formulated, it must be possible in thought to consider which propositions are correct when we vary the standpoint of evaluation; that is, when we vary the first parameter…[8]


The first sentence is again ambiguous. On the quasi-realist’s theory the question of which basic moral principles are accepted by people indeed depends upon (is the same thing as) their psychological attitudes. Whether they are right to accept those principles is a different thing altogether, and we will only settle it by ourselves finding a verdict on their approvals and disapprovals. If people in outer modal space, or for that matter people in benighted corners of the earth, accept the principle that it is OK to cause unnecessary pain to sentient creatures for fun, then they are cruel and callous and it would be good if they would change. The second of Peacocke’s quoted sentences is therefore technically correct but highly misleading, for it implies that in general changing the first parameter, that is, iconsidering different evaluative standpoints, changes the correctness of a verdict. But it doesn’t. It only changes whether it is supposed to be correct, by whichever evaluators are introduced. Except in the cases that I called those of etiquette, it merely brings the evaluators into the embrace of our verdict, perhaps to their discredit, as in this case.

            Peacocke continues:

Take a specific moral principle identified by its content, say ‘Prima facie, the infliction of avoidable pain is wrong (w,w). It seems to me that the quasi-realist, like other mind-dependent theorists, must say this is false. It is false at those entries in the diagonal for worlds in which we have different attitudes to the infliction of avoidable pain


This is hard to follow, because in accordance with his own explanation of the notation, propositions of the form P(w1, w2), are not moral principles at all. First, they describe whether the evaluation from w1, of the scenario of w2, could gain expression by P: they are descriptions, not evaluations. And secondly, they are not propositions at all until the variables are bound or replaced by actual values, so neither the quasi-realist nor anyone else has any business saying that P(w,w) is true, or false.

            Perhaps Peacocke  is thinking of the double universal quantification ‘Everyone, from whatever evaluative standpoint, and considering any scenario whatever, would agree that inflicting avoidable pain is wrong’ I do indeed doubt whether this is true, but that doubt has nothing to do with quasi-realism. It has more to do with pessimism about varieties of the wicked human heart, and if we are in outer modal space, the even more wicked Martian heart. And after all, Peacocke shares the doubt, for he allows worlds in which inhabitants have different attitudes to the infliction of pain. That’s the point on the diagonal that he is inviting to the feast. Or perhaps it is not a double quantification but an anaphoric reference back to the world of the people with the evaluative standpoint: ‘everyone, from whatever evaluative standpoint, and considering the world they inhabit, would agree that inflicting avoidable pain is wrong’. Alas, the same pessimism is appropriate.

Peacocke’s  ambition is clearly to get the quasi-realist both to treat some proposition of the form P(w1, w2) as a genuine moral principle, and to evaluate it as false when we think of worlds in which the wicked hearts rule. But the machinery takes him not one inch nearer to that goal. Worlds in which the wicked hearts rule are still worlds in which, prima facie, the infliction of avoidable pain is wrong. The wicked hearts may not agree with this, but then that is just what’s wrong with them.

            I said that the word ‘holds’, as it occurs in the clarification of his notation, is treacherous, and at this point we are compelled to think that Peacocke has actually been betrayed by it. It seems he really does want to index the question of whether a moral principle is true to the various worlds whose inhabitants either agree or disagree with it. I think that is preposterous. It would be like saying that the proposition that the Iraq war was a good thing holds—really holds—in Republican circles in America, and really does not hold in most of the UK. And if that is what it means the quasi-realist simply refuses to adopt the notation. It differs from the legitimate meaning we have so far allowed it, aiming at something more like this: ‘the people we have introduced evaluate the scenario we imagined them to be contemplating, in a way that could be expressed by saying that P and as a result P is true’ But the quasi-realist has no use for this dog’s breakfast of an assertion (it will be false except in etiquette cases). The Iraq war was a bad thing whatever other people think about it. It is not true in London but false in Texas. Nor is it a matter of etiquette, so that enough thinking it a good thing might make it one. That way lies sophomoric relativism, not sentimentalism. The criminality of the Iraq War is dead-innocent-Iraqui-dependent, not Republican-sentiment-dependent.

Far from taking him into the sunny uplands of rationalism, then, Peacocke’s machinery grinds to a halt in the swamp of a relativism of his own devising. He finishes the discussion by considering the neighbouring case of colour, and the possibility that creatures with different perceptual systems might see physically different things and surfaces as red. He says that it is widely agreed that things would not stop being red if humans lost their colour vision and saw only in shades of grey. That may be so, although it ought also to be widely agreed that there is much more indeterminacy here than in the case of values. Jonathan Bennett’s example of phenol-thio-urea, which tastes bitter to some people and bland to others, led many people to think that if the former group breeds into a huge majority, the world becomes one in which the stuff is bitter, while if the latter group does, the world becomes one in which it is bland. In other words, the ‘response-dependency’ of secondary properties is a much better candidate for providing a genuine truth condition for ascriptions of them, than any similar attempt to provide a ‘truth-condition’ for ascriptions of value.

            However, Peacocke is also correct that two-dimensionalism allows different formulations of the idea that colours are mind-dependent. Where Q is some underlying physical power, such as a disposition to reflect light of a certain wavelength more than other light, and we imagine varying perceptual systems, we could say that:

For any world, whatever perceptual systems its inhabitants have, Q objects are red, as they would be judged by us, as we actually are.

We would also want to say:

In some worlds, Q objects are not red, as judged by the inhabitants of those worlds.

And given Bennett’s case, we might remain ambivalent about whether

For any world, whatever perceptual systems its inhabitants have, Q objects are red.

Since we would be ambivalent about, as it were, sticking with our own judgements, or entering into the world-view of the people with the other perceptual system.

            The reason this ambivalence is harmless is that once we bring other perceptual systems into view, the provided they are equally discriminatory, we lose any very robust attachment to the idea that ours is right and theirs is wrong. Similarly we do not maintain sceptical fears that perhaps our sense of smell, or sense of colour, may in general be letting us down, so that perhaps things really smell differently from the way we smell them, or have different hues from those we see them as having. People who taste phenol-thio-urea the other way are not wrong, just different. But there is no reason to suppose that this ambivalence extends similarly to the case of value. [9] People who are coarse and brutal are not ‘just’ different. They are also depraved, and as a result they are rotten judges of value. If we are invited to ‘see the world as they see it’ we can, perhaps, manage it, but we ourselves can attach no weight to the verdicts we would imagine giving as we do so.

            Before leaving this part of the discussion, it may be useful to reflect upon a difference between sentimentalism, as a theory of the origin of the moral sentiments, and a partly parallel exercise of quasi-realism, attempting to see our verdicts of modal necessity as the upshot of various features of the shape of our minds that determine what we can or cannot imagine.[10] Here there is a legitimate pressure to see a contingent source of imaginative limitation as an undermining or debunking account of logical or metaphysical necessity. If ‘we cannot think otherwise’ is sourced in contingent facts about us, an inference to ‘things could not be otherwise’ is compromised rather than explained.

            Someone might be tempted to use the modal case as a Trojan horse, bringing the same worry into the theory of morals. But if so they would be wrong. The asymmetry lies in what we say about the states of mind in question and how it relates to the kinds of verdict we are making. In the modal case, if we find that the modes of thought, or the absence of alternatives, are only contingent, their source as an explanation of real necessity is compromised. But in the moral case, it would be not finding that they are metaphysically contingent that would give them a parallel debunking power. The parallel would be finding that they are morally indifferent. If it were morally alright to have the other sentiments—say, approving of cruelty to dogs or neglect of children—then it would be hard to believe that the ones we actually have could source a robust confidence in an obligation to refrain from cruelty and neglect. At least in general, if it is OK to think that some action is OK, then the action is OK.

But the sentimentalist is not saying that it is OK to have the contrary sentiments. As I have already said, the sentiments of those who would think otherwise fall within the scope of proper disapproval. We do not just disapprove of neglect of children, but perhaps even more so, and certainly just as much, we disapprove of those who approve of it or even tolerate it.

And rationalists had better not find the metaphysical contingency of modes of moral thought unsettling. If rationalist moral conviction is to falter whenever it comes upon people who do not share it or do not feel its force, then it is a fragile thing indeed. For knavery exists. Indeed, it often rules, and this is why a robust conviction of its baseness is so important.

Justice and Gentle Usage

Kerstein is not the first to worry about the scope of justice on Hume’s theory. According to Manfred Kuehn’s biography, Kant himself was led to reject Hutcheson’s sentimentalism for a very similar reason.[11] Reading Rousseau apparently convinced Kant that while the sentimentalist allows that we have duties of charity to the dispossessed of the world, this is not enough. The poor or excluded have a right to more than charity. It is not charity they want or need, but justice. If sentimentalism cannot deliver that, then it delivers an inadequate account of the actual nature of our moral thought.

            Hume makes himself a target for this kind of outraged reaction:

Were there a species of creatures intermingled with men, which, though rational, were possessed of such inferior strength, both of body and mind, that they were incapable of all resistance, and could never, upon the highest provocation, make us feel the effects of their resentment; the necessary consequence, I think, is that we should be bound by the laws of humanity to give gentle usage to these creatures, but should not, properly speaking, lie under any restraint of justice with regard to them, nor could they possess any right or property, exclusive of such arbitrary lords.  Our intercourse with them could not be called society, which supposes a degree of equality; but absolute command on the one side, and servile obedience on the other.  Whatever we covet, they must instantly resign:  Our permission is the only tenure, by which they hold their possessions:  Our compassion and kindness the only check, by which they curb our lawless will:  And as no inconvenience ever results from the exercise of power, so firmly established in nature, the restraints of justice and property, being totally useless, would never have place in so unequal a confederacy.  (190-191)


There are many things to say about this passage, and quite how Hume thought it related to the human cases he goes on to discuss, which are firstly European relationships to indigenous American people, and secondly mens’ relationships with women. The clear implication is that the model applies in neither case, but only at best to our relationship with animals, or perhaps imagined animals.

Nevertheless we might want to modify the account to make justice clearly applicable, even in the circumstances of the thought experiment. I shall consider how that might be done in a moment. Meanwhile, the important point is that it is not Hume’s sentimentalism that leads him here, but his strict delineation of the circumstances of justice and its source in mutual advantage. Hume does not deny that we have obligations to the creatures he presents. He says that we are ‘bound by the laws of humanity’ to give them gentle usage. The only issue is the way we are to understand this obligation. Remembering that for Hume the virtue of justice is both ‘cautious and jealous’ and above all artificial, it may not be so bad for these creatures if the source of that obligation lies elsewhere. But it is important to see that if we insist on the word ‘justice’, the sentimentalist can give it to us.

Hume mentions the resentment these creatures have, but which, because of their inferior strength and power, they can do nothing to visit upon us. This opens up a new sentimentalist vista, much more thoroughly explored by Adam Smith.[12] As Raphael explains, for Smith the sympathy that lies at the bottom of our capacity for morals has a slightly different shape than it does in Hume.[13] In Hume, we sympathize with the pleasure or pain that an action gives to a person. In Smith, we sympathize with different states of mind, including the motives of an agent, and more relevantly to the current case, with the gratitude or resentment of those affected by the action. Indeed our sense of justice, for Smith, is dependent on reactions of resentment or gratitude to actions, which need not vary as the actual quantity of harm or benefit they bring about.

The ‘sympathy’ that is so prominent in each of Hume and Smith is translated, by the one into respect for the general point of view, and by the other into the voice of the impartial spectator, the ‘man within the breast’ who represents the reactions of those without. There is, of course, much to be said about the ways in which each writer identifies and handles the mechanism, and the relation between them.[14] There is also much to be said about whether either mechanism implies some concession to rationalism, bringing in as they do some notion of ‘corrected’ sentiments. I cannot rehearse all that needs saying about that here, but shall have to take it as given that neither writer betrays sentimentalism by their construction of the more complex sensitivity.[15] So suppose we bring in the lynchpin of Smith’s sentimentalism, the ‘real, revered, and impartial spectator’ whose function is to bring home to us the resentment of those affected by our delinquencies. When the voice of this spectator is heard as it should be, we may recoil from our own contemplated or actual conduct. In Hume’s terms, we can no longer bear our own survey. Recognizing this resentment of our conduct, and feeling no defence against it, we admit the injustice.

Suppose, then, that we have been minded to take from one of Hume’s creatures something which they evidently cherish. They cannot visit their resentment upon us, but somehow we know that they feel it, and we know that we would feel it in their shoes. The man within the breast voices this resentment on their behalf, and we find we cannot dismiss it (we cannot resent their resentment, as we sometimes can). This unpleasant impact is the same as guilty awareness of the injustice of our conduct. What more could the sternest moralist ask from us?

Smith’s modification of Hume may still leave us falling short of full-blown Kantian rationalism. But it is at least telling that the most fervent contemporary Kantians find it hard to do better. Korsgaard, for instance gives us the crucial moment in the genesis of obligation to others like this:

How does this obligation come about?  Just the way that Nagel says that it does.  I invite you to consider how you would like it if someone did that to you.  You realize that you would not merely dislike it, you would resent it.  You would think that the other has a reason to stop - more, an obligation to stop.  And that obligation would spring from your own objection to what he does to you.[16]

Korsgaard goes on to employ a cognitive or rationalistic vocabulary, but it is hard not to feel that the central process is exactly the same as in Smith. The potential victim forces you to recognize his resentment, and to ‘put yourself in his shoes’. His fundamental question is: how would you like it if someone did that to you’ — and once you find that you would not, then, other things being equal, his not liking or resenting it translates into your own discomfort at your own behaviour. Of course, as Smith sensibly recognizes, things are not always equal. We have an abundance of defence mechanisms against this incipient discomfort, including ignoring the impartial spectator, or more often convincing ourselves the he would be on our side or ‘of our faction’.[17]

            At this point the sentimentalist will certainly face another familiar challenge. The account finds the source of feelings of obligation and injustice in a certain emotional identification: in this case a contingent (of course) capacity to internalize the resentments of others. But might not this very notion of resentment itself import, and depend upon, an unacknowledged cognitivism? Resentment, as Korsgaard says, is more than mere dislike. Perhaps it is more like bitterness, but bitterness at the dispensation of some agent. Anyone suffering a third summer holiday in succession blighted by continuous rain might feel bitter, but only a theist can resent it. This quickly suggests that resentment is more like bitterness at the injustice of the behaviour, in which case a perception of injustice cannot be explained in terms of resentment and sympathy with it, but must be identified in some pre-existent cognition. Similar objections may be made to sentimentalist uses of notions like guilt, or even anger: if anger is the attitude or emotion of those who perceive themselves to have been wronged, and guilt is the attitude or emotion of those who recognize themselves to have done wrong, then we cannot understand the judgments by citing the emotions.

            There is a simple lacuna in this popular line of thought. The equations in question are things like ‘anger is perception of wrong’ or ‘resentment is recognition of injustice to oneself’. The objection implicitly supposes that these equations need to be read from right to left, so that the apparent cognition explains the emotion. The sentimentalist tradition, by contrast, reads them from left to right, so that the emotion or attitude explains the thought of wrong or injustice in terms of which it gets expressed. This is not the place to rehearse all the moves in this debate. You just have to try out the different directions of explanation, and you have to ask which is psychologically or metaphysically the more economical. But at least a preliminary remark is that as they stand, of course, the equations are absurdly simple. Anger is not perception of wrong, nor resentment recognition of injustice to oneself. Each is both more in one respect and less in another. More, because the pure cognition leaves out the upheaval and the motivational force, so that in fact perception of wrong may not lead to anger, and recognition of injustice to oneself may not lead to resentment. Less, because each has a primitive identity in which ethical thought is not yet present. After all, we should not forget that Darwin called his great work The Expression of Emotion in Man and Animals. The guard dog does a fair job of being angry at the intruder, and the pet which throws its food around the house on being left behind, does a fair job of resenting being neglected. [18]

A different strand in Smith is the idea that unlike obligations of benevolence, obligations of justice can be exacted from us. They bring in the potential force of the community or the civil power: ‘the person himself who meditates an injustice is sensible of this, and feels that force may, with the utmost propriety, be made use of both by the person whom he is about to injure, and by others, either to obstruct the execution of his crime, or to punish him when he has executed it.’[19] Applied to Hume’s example, this suggests that the question of whether there is an obligation of justice may hinge on whether we think a spectator, contemplating a breach of ‘gentle usage’,  should use pre-emptive or retaliatory force on the perpetrator. I am not sure whether we do think this in general. If we can take the case of animals as indicative, our actual animal welfare legislation suggests we think that if the breach is severe enough then the criminal law has a say, but at least in our jurisprudence, if not in our studies, we seem prepared to let a fair amount of not so gentle usage go on unprevented and unpunished.

            However we solve this issue, if we stand back for a moment it should be obvious that this particular issue about justice is not a promising basis from which to attack sentimentalism. The structure of the case disqualifies it from that task. The idea is to arouse our sense of what is due to these creatures, and to encourage shock and outrage at the base behaviour to which Hume’s agents might be led, or to excite us to lament the outrages which they might get away with, and to wring our hands over the sad plight of the poor defenceless creatures.  All this is excellent. It shows us sympathizing with the downtrodden and their resentment, perhaps desiring or wishing for a civil order in which the powerful would be punished, feeling that things are out of joint unless they are brought to account for their crimes, and so on. In other words, it shows that our sense of outrage and injustice is mobilized, not merely our benevolence. But it cannot show more than that. It cannot show that what is mobilized lies outside the sentiments altogether. Hume’s example may make us hot under the collar about the indignities the powerful may visit on the weak, but it does not afford any evidence that getting hot under the collar is anything else than feeling an attitude and emotion, directed upon a particular social structure and the abuses it looks set to allow.

Where This Leaves us.

I have not, in this paper, been exclusively defending expressivism. Other views which stress the place of sentiments, or imagination and culture, in the genesis of our ethical thought were equal possible targets of Kerstein and Peacocke’s attacks. Some of these others, perhaps less deft with the notion of ‘mind dependence’ might even fall to such attacks, for instance by giving the moral judgement a truth-condition that is not child-dependent or dog-dependent, but genuinely mind-dependent. Others may avoid them only by inappropriate reliance on ‘actually’ operators and other pieces of doubtful machinery. If so, I am glad to part company with them.

            The popularity of rationalism, and the general feeling that there ‘must be something to’ the kinds of argument I have been discussing, are very deep-rooted. Partly, they represent a noble dream. They answer a wish that the knaves of the world can be not only confined and counfounded, but refuted – refuted as well by standards that they have to acknowledge. Ideally, the will be shown to be in a state akin to self-contradiction. Kerstein acknowledges that Kant and neo-Kantians have not achieved anything like this result. But it is still, tantalizingly there as a goal or ideal, the Holy Grail of moral philosophy, and many suppose that all right-thinking people must join the pilgrimage to find it.

            We sentimentalists do not like our good behaviour to be hostage to such a search. We  don’t altogether approve of Holy Grails. We do not see the need for them. We are not quite on all fours with those who do. And we do not quite see why, even if by some secret alchemy a philosopher managed to glimpse one, it should ameliorate his behaviour, let alone that of other people. We think instead that human beings are ruled by passions, and the best we can do it to educate them so that the best ones are the most forceful ones.

We say of rationalistic moral philosophy what Hume says of abstract reasonings in general, that when we leave our closet, and engage in the common affairs of life, its conclusions seem to vanish, like the phantoms of the night on the appearance of the morning.[20]






[1] David Hume, An Enquiry Concerning the Principles of Morals, ed. Tom Beauchamp, Oxford: Oxford University Press, 1998, Appendix I, p. 163.

[2] Ruling Passions, (Oxford: Oxford University Press, 1998) p. 210, p. 230, p. 265, and elsewhere. And why does Kerstein think I call the foreign-office knave a knave?

[3] If you (wrongly) think that we cannot lie under an obligation to feel various ways, perhaps because ‘ought’ implies ‘can’, read it as ‘behaving as though you care for the child’.

[4] Christopher Peacocke, The Realm of Reason, Oxford: Oxford University Press, 2004.

[5] One could play with complexities introduced by the thought that in some sense ‘not giving a gift back’ in the described community is performing the same action as ‘giving a gift back’ in ours. They do not affect the point.

[6] Christopher Peacocke, The Realm of Reason, p. 208.

[7] p. 210.

[8] p. 214

[9] In her illuminating paper ‘History of Philosophy in Philosophy Today: and the Case of Secondary Qualities’ (Philosophical Review, 1992, pp. 191 – 226) Margaret Wilson comments on a constant tendency in ‘the modern philosophy’ to vacillate over whether colours are in the mind, are categorical or primary grounds in the atomic constitution of things and their surfaces, or are powers to excite human perceptual systems. The vacillation is more excusable if we reflect that in the case of colour we never have to decide on ‘what to say’ about the case when the same physical properties have different powers because of varying perceptual systems.

[10] Simon Blackburn ‘Morals and Modals’ in Essays in Quasi-Realism, New York: Oxford University Press, 1993.

[11] Manfred Kuehn, Kant, Cambridge: Cambridge University Press, 2001.

[12] In the paragraphs that follow, on Smith, I am much indebted to work by Michael Ridge.

[13] Adam Smith, The Theory of Moral Sentiments, ed. D.D. Raphael and A.M. Macfie, Oxford: Oxford University Press, 1976, editor’s introduction, p. 13.

[14] See, for instance, Rachel Cohon, ‘The Common Point of View in Hume’s Ethics’, Philosophy and Phenomenological Research, vol 57, 1997, Christine Korsgaard, ‘The General Point of View: Love and Moral Approval  in Hume’s Ethics’, Hume Studies, vol 25, 1999, Elizabeth Radcliffe, ‘Hume on Motivating Sentiments, the General Point of View, and the Inculcation of Morality’ Hume Studies, vol 20, 1994, Geoff Sayre-McCord.  ‘On Why Hume's 'General Point of View' isn't Ideal - and Shouldn't Be’ Social Philosophy and Policy vol 11, 1994.

[15] In Ruling Passions I argue that the idea of the general point of view involves no such concession, but only introduces what I call a ‘Hume-friendly’ notion of reason. See Chapters 7 and 8.

[16]Korsgaard, The Sources of Normativity, Cambridge: Cambridge University Press, 1996, p. 98.

[17] Smith: ‘The propriety of our moral sentiments is never so apt to be corrupted, as when the indulgent and partial spectator is at hand, while the indifferent and impartial one is at a great distance’. p. 154.

[18] For more on these themes, see my ‘How Emotional is the Virtuous Person?’ in Peter Goldie, ed. Understanding Emotions: Mind and Morals, Aldershot: Ashgate Publishing, 2002.

[19] Smith, pp. 79-80.

[20] David Hume, A Treatise of Human Nature, Book III, part 1, section 1.