These valuable essays by Peter Railton date from 1984 to 2000. Taken as a whole, they show that the fundamentals of his moral philosophy have remained constant over that time; they also show how conscientious he is in responding to, and wherever possible trying to accommodate, competing points of view. Railton's general ethical position is consequentialist, naturalistic and realist – in all cases in sweepingly radical ways. At the same time he has great insight into how people actually feel about questions of value, whether ethical or aesthetic, and he tries to accommodate those feelings. The resulting tensions make for fascinating attempts at reconciliation, more in the style of Mill than of Bentham. Like Mill, Railton tries to show that his argument is capable of surprising accommodations, and far from incompatible with what a serious and open-minded reader will find it natural to think.

The essays are divided into three groups. Part I sets out Peter Railton's meta-ethical views on moral realism and part II his views on moral and political theory; part III deals with general questions about normativity. Fine qualities are revealed throughout. But it is not possible to review Railton's insightful treatment of all these big topics, even in a critical notice. So I shall give particular attention to his theory of reason, the good and morality, and then, more briefly, to his influential essays on meta-ethics and ‘the problem of normativity’.

I

Some people think that reasons should be defined in terms of axiological and/or deontic concepts, such as that of a person's good, the general good, virtue, duty, obligation, etc. Some others think that the latter concepts should, on the contrary, be defined in terms of reasons. A notable feature of Railton's moral philosophy is its bipolarity. He belongs to neither camp. He takes it that reasons should be characterized in terms of an instrumental view of rationality. But he also introduces a further pair of notions, individual good and social rationality, and he gives an ‘objective’ or non-instrumental view of each. As for the distinctively deontic concepts, Railton thinks they should be defined in terms of the notion of social rationality. Thus deontic concepts aren't conceptually related to reasons any more than social rationality is; it's not, for example, a conceptual truth that if you have a duty to do something – that is, if the standpoint of social rationality requires you to do it – you have reason to do it.

To this bipolarity the instrumentalist conception of practical reasons – namely, that there's reason to do something if and only if it will promote one's ends – is basic:

Now these comments look as if they leave open the possibility that there are less clear, less minimal conceptions of practical reasons and rationality which are nonetheless correct. In fact, however, as far as I can see, Railton takes a stronger position. He endorses a fully instru-mental conception: he holds, that is, that one acts rationally if and only if one does what one has reason to believe will promote one's ends.

This conception could be adopted in the practical domain without prejudice to one's view of epistemic (theoretical) rationality and reasons. However Railton takes an instrumental view of epistemic as well as practical rationality:

Again these remarks could be read in a moderate or a strong way. The moderate view would start by distinguishing reasons to believe from reasons to bring it about that one believes. The former, it would say, are epistemic reasons, the latter are practical reasons. In a similar way, a moderate would distinguish reasons to feel from reasons to bring it about that one feels – where only the latter are practical reasons. For example, you might have very good reason to be angry with your boss, but also very good reason to bring it about that you're not angry – to suppress and even eradicate any tendencies to anger, in so far as you can. Only the latter is a practical reason, and so an instrumental conception of practical rationality would apply only to the latter.

‘Belief-formation’ may refer to the practical activity of searching out evidence, considering argument, etc. Clearly, whether to engage in such activity with respect to this or that topic is a practical question. Suppose, for example, that I couldn't care less which party will win the next election. In that case, at least on the instrumental theory of practical reasons, there may be no reason for me to engage in any belief-formation about that matter: no reason for me to bring it about that I have any beliefs on the subject. It does not follow, on the moderate view, that there is no reason for me to believe something on the subject. Suppose there are already plenty of facts available to me that, taken as a whole, constitute strong evidence for the conclusion that party A will win. All I have to do is to spend a little time attending to them. If that is so, then there is reason for me, and anyone else in my epistemic situation, to believe that party A will win: I just can't be bothered to work that out. Whether there's reason to believe something depends on the facts available to you. Whether there's reason to set about attending to those facts, or discovering more, is another matter.

The strong view, in contrast, does not see the distinction between reasons to believe and reasons to bring it about that one believes as important, because it takes epistemic rationality to be a species of practical rationality. Railton seems to take the strong view: ‘What it would be rational for an individual to believe on the basis of a given experience will vary not only with respect to his other beliefs, but also with respect to what he desires’ (p. 6). And in a note: ‘I am insisting that questions about what it would be rational to believe belong to practical rather than theoretical reason’ (p. 35).

So Railton thinks that there's reason for me to believe something if and only if it would promote my ends to believe it. But here again there is more to his view. He also distinguishes the ‘epistemic warrant’ that may exist for a belief from the rationality of holding it. Epistemic warrant, he notes, may be tied to an external criterion – as it is by causal or reliabilist theories of knowledge (p. 9). So if one thinks of good evidence that p, for example, as evidence that warrants the belief that p, then on Railton's view good evidence will not be definable in terms of reasons to believe, say as evidence that gives one reason to believe. And nor will reasons to believe be definable in terms of epistemic warrant. One may be epistemically warranted in believing something that it's irrational for one to believe, and it may be rational for one to believe something that one is not epistemically warranted in believing. How much reason there is for me to believe what I'm epistemically warranted in believing will depend on how strongly I aim to believe the true, or how strongly believing the true conduces to other aims I have.

I don't find this view of reasons to believe more plausible than the similar view of reasons to feel: how much reason there is for me to be irritated by someone's remark does not seem to me turn on how much being irritated would advance my aims. The moderate view strikes me as the common-sense view. To ask what reasons there are for us to think that global warming is occurring, for example, is one thing, to ask in what ways thinking that would conduce to our objectives is another. It may be highly contested what the evidence gives one reason to believe; nonetheless, given the same evidence the same answer holds for everyone in the world, irrespective of whether they want to drive petrol-heavy cars, say.

In a very interesting paper ‘On the Hypothetical and Non-hypothetical in Reasoning about Belief and Action’ (1997) Railton agrees that this is at least ‘the usual view’: ‘On the usual view of things, two agents in the same epistemic situation (same evidence, same background beliefs) would have the same reasons for believing any given proposition, regardless of possible differences in their personal goals’ (p. 293). He proceeds to ask whether this usual view of things, according to which one's epistemic reasons are not hypothetical on one's practical aims, can be given a principled basis, and wonders whether, if so, a similar non-hypothetical basis can be found in the practical realm. His method is to consider possible answers to an intelligent and awkward student, Gary, who ‘wants to know whether there are any considerations that require or favor following epistemic standards that don't depend at all on our personal goals’ (p. 294). Gary, as I understand it, is not a sceptic about epistemic warrant or epistemic standards. He is not raising the question of how we know they are truth-conducive. He just wants to know what reasons there are to engage in the activity of taking epistemic standards into account in one's belief-formation. If, following the usual view, we conceptually tie ‘epistemic reasons’ to epistemic warrant or epistemic standards, Gary will simply ask what reasons there are to take into account such epistemic reasons.

Railton considers ‘constitutive’ answers to Gary's question, according to which you do not count as an agent if you don't take account of epistemic standards in your thinking and acting, and he plausibly dismisses them. Gary can simply accept this definition of ‘agency’ and ask what reason he has to be an ‘agent’. It seems then – though Railton does not put it so bluntly – that his own answer to the question posed by Gary is, in the end, that there are no non-hypothetical considerations (considerations that don't depend at all on our personal goals) that ‘require or favor’ following epistemic standards. One could put this another way. On the usual or moderate view of reasons to believe, Gary's question is whether there is sufficient reason to bring it about that one believes that which there is, in fact, sufficient reason for one to believe. Why should one bother? The answer to this question obviously depends on your theory of practical reasons. Given an instrumental theory of practical reasons, there may be no reason; much will then hang, for example, on how deeply or widely Gary wants to believe the true. In contrast, on Railton's more radical, instrumentalist, view of reasons to believe, Gary's question is simply whether there is any reason to believe what epistemic standards require or favour that one believes. Railton's discussion with Gary does not resolve the question whether ‘reason to believe’ and ‘epistemic standard’ are conceptually tied. It does not show that the non-hypothetical view of reasons to believe, the view that conceptually connects them to epistemic standards, is wrong, unless one has already shown that a ‘principled basis’ for non-hypothetical epistemic reasons must somehow get them out of hypothetical practical reasons.

II

In the epistemic sphere, then, Railton accepts that we have such notions as epistemic warrant and epistemic standard, but denies that these are conceptually linked to the notion of reason to believe. Likewise in the practical sphere: he thinks that we have notions of individual good, social rationality, and moral rightness, but denies that these are conceptually linked to the notion of reason to act.

His account of a person's good, or objective interests, takes the form of an idealized-desire account with refinements designed to cater for well-known difficulties. One of these difficulties is that, intuitively, satisfying the desires an agent would have if idealized, for example by having improved information, may not advance his good as he actually is, with his actual desires. Railton's way of dealing with this is to say that A's good consists in the satisfaction of those wants that an idealized version of himself, A+, would want him to have as he is (pp. 11–12, 49ff.). This strikes me as problematic. The problem is that we don't know whether A+ would have any desires about A's wants as he actually is. To stipulate that we're considering what A+ would want if he had A's good at heart would be plainly circular. (In general, this point about circularity seems a fatal objection to all ‘well-informed advice’ theories of a person's good.)

Another broad problem for accounts of a person's good that are couched in terms of idealizing counterfactuals arises from familiar difficulties about the determinateness of counterfactuals. To this Railton responds (pp. 57–63) by arguing that hypotheses about what desires an individual would have if idealized serve a heuristic or evidential rather than a constitutive function.

To me this seems a much more promising line of thought (and closer to our practice) than the idea of appealing to how an ideal version of me would advise me. A person's desires, actual and hypothetical, ex ante and ex post, together with their probable development over time, are evidential (or criterial) as to what his good comprises. To adapt Mill: the sole evidence it is possible to produce that anything is desirable for a person (notwithstanding that such evidence is always defeasible) is that that person desires it, would desire it if he knew this or that, would still desire it when he had it, and so on. What then is this sort of evidence evidence for? One overall view is this. What A desires, actually and hypothetically, is evidence for normative propositions about what is desirable for A – where that means, propositions about what there's reason for A to desire. Furthermore, the good for A is definable in terms of what is desirable for A. On this view there is an evidential link between A's desires and A's good, and a conceptual link between A's good and A's reasons; there is then a further, substantive, question about what in fact would contribute to A's good. Finally, granting that there is reason for A to pursue what there is reason for A to desire, there is reason for A to pursue his good, whatever it may be. (There may be plenty of reasons for A to do other things as well.) For Railton, in contrast, the conceptual link between A's good and A's reasons drops out. He too will hold, given the heuristic approach, that A's desire is evidence of A's good, but he will then propose a reduction of A's good to those properties of A that, on that evidence, A's good consists in. Whether there is reason for A to pursue A's good will be a further question, turning on what A's actual aims are.

I would argue that the first view is a better account of this complex of connections. But there is a further objection that applies to either view. It is that this sort of account of a person's good is too broad. If we consider all a person's desires, actual and hypothetical, as evidence, we'll make all sorts of things a part of A's good that intuitively are not.

This is plainly a considerable objection. A response to it can go in one or both of two ways. The first (which I would follow) is to observe pre-theoretical constraints on the notion of desire. We have a perfectly natural way of using the notion of desire on which it can be true to say that one is doing what one has no desire to do. And we think there can be good reason to do it even though there is no reason to desire, in this sense, to do it. This is a narrower notion of desire than that introduced by philosophers who postulate a ‘belief-desire’ psychology. And we can then say that A's good is what there is reason for A to desire in this narrower, substantial sense.

The second response is to distinguish a philosophical notion of a person's good from a more restrictive notion of a person's welfare. Thus Railton remarks in a note on page 67 that his notion of a person's good

Presumably, then, in the relevant sense of ‘welfare’ the welfare or indeed good of others is not a part of this individual's welfare, but can still be a part of his good. I'm on Railton's side in this view of a person's good, because it seems to me that even on a suitably constrained, substantial concept of desire it is a fact about people that they often really desire, in that substantial sense, the well-being of others. Moreover, substantial desires of that kind pass any test of what counts as a reasonable desire. So for these people, with these reasonable desires, the good of others really is a part of their good. As Railton eloquently notes:

In making these remarks he follows important idealist traditions in ethics: traditions that highlight a deep truth. We should note however that this deep truth raises an important ethical question. Suppose we are concerned to take all individuals into account in an impartial way. When we take up this impartial standpoint, which concept should we be working with – the expansive concept of a person's good, or the restrictive concept of a person's welfare? Consider a mother in a fire, for whom it matters much more that her children should be saved than that she should be, and suppose that we can save either her children or her, but not both. Her welfare is diminished by not saving her, her good is increased. Which of these, her welfare or her good, should we be considering from the impartial point of view?

III

I am one of those, like Railton, who hold that impartiality is the appropriate starting point of ethical reflection or theory, though not of concrete moral reasoning or practical decision-making. So put, this is a vague but nonetheless substantive thesis that would already be disagreed with, for example, by thinkers who hold that agent-relativity goes all the way down. But it does not yet imply utilitarianism, in any customary sense.

We take a step towards utilitarianism if we add that the way in which individuals are all equally important is that the good, the objective interest, of each individual matters ethically neither more not less than that of any other. Then we have ‘the guiding utilitarian idea, namely, that the final ground of moral assessment – including assessment of character – must lie in its effects on people's well-being’ (p. 226). Railton calls this core thesis ‘valoric utilitarianism’. The ultimate standard of morality is the general good, understood as some impartial positive function of the good of all individuals. To this core utilitarians have added other theses, such as hedonism about individual good, aggregative criteria of impartiality, and not least, definite interpretations of what is meant by ‘ultimate standard of morality’.

Railton also describes this impartial ethical standpoint as the standpoint of social rationality: it determines ‘what is rational from a social point of view with regard to the realization of intrinsic non-moral goodness’. And he assumes, in earlier essays, that the socially rational action is the morally right action: this, he says, ‘seems to me to be a recognizable and intuitively plausible – if hardly uncontroversial – criterion of moral rightness. Relative moral rightness is a matter of relative degree of approximation to this criterion’ (p. 22). In later essays, however, this robust stance gives way to increasing awareness of the implausibilities that objective act-utilitarianism seems to incur. A subtilizing process takes place in several thoughtful papers, including ‘Alienation, Consequentialism, and the Demands of Morality’ (1984), ‘How Thinking about Character and Utilitarianism Might Lead to Rethinking the Character of Utilitarianism’ (1988) (see n. 19, p. 248 for some change of view) and ‘Pluralism, Dilemma, and the Expression of Moral Conflict’ (1992, 2001). Thus:

The more familiar sense of moral obligation is captured better, he notes (p. 265), by Mill's account, which distinguishes between obligatory and supererogatory acts and links moral obligation to blameworthiness for non-compliance. (One might note further, as also relevant, the important idea that my moral obligations are something I can tell for myself I have, so that moral obligation, unlike optimality of consequences, must be information-relative, and also the striking agent-relativity of many moral obligations.)

This is part of a general shift of emphasis. At the outset Railton approaches morality with the optimistic, business-like air of someone who has a sound scientific plan for reconstructing and improving it in his pocket. On closer inspection of the article in question, however, he becomes increasingly hesitant:

At best, he concedes, reconstruction will take a long time; he also considers the possibility of an error theory of morality (e.g. p. 283), that is, of denying that there is such a thing as moral obligation. In the last essay of the volume, however, ‘Morality, Ideology, and Reflection; or, The Duck Sits Yet’ (2000) – which originated as a paper to a conference on the work of Bernard Williams – he rejects that option:

Neurath's duck – or better, let's hurriedly revert to his original boat – is indeed an appropriate metaphor to reflect on at this point. Among (valoric) utilitarians there are two temperaments and traditions: one might say the New Constructors and the Conservative Holists. New Constructors reject the inherited boat of morality. They want to construct a new boat, which might or might not be called ‘morality’. Conservative Holists think that without the boat of morality we will drown, and that such repairs as it needs – which might be quite extensive – must be done while standing in it. Railton often talks like a New Constructor but in the end comes across as a Conservative Holist.

But surely (to torture the metaphor a little further) valoric utilitarianism – that teleology – itself provides the dry dock for reconstruction, or new construction, that Neurath's original boat, which comprised the total system of beliefs, famously lacked?

This would be a good response by the New Constructors if valoric utilitarianism was rationally self-evident, while distinctively moral convictions about moral obligation had no degree of self-evidence at all. And that, I believe, is what some utilitarians have thought. It is also possible (and plausible) to hold that both valoric utilitarianism and particular moral convictions have their own intrinsic rational warrant. The question then arises, what should one think if they come into conflict: is valoric utilitarianism itself defeasible, or is it always the last court of appeal? I suppose the most minimal kind of utilitarianism says only that it is, indeed, the last court of appeal. Concrete moral convictions don't have to be derived from it, but do they to have to be rejected if shown to be inconsistent with it.

No doubt there are other possible views about what reason-conferring (or rejecting) authority valoric utilitarianism has over our practices. But none of them can be Railton's view. We must remember his overall instrumentalism about reason, theoretical as well as practical. Railton must think that valoric utilitarianism has no more rational self-evidence as a reason-giving guide to action than particular moral convictions – both having none. There are possible systems of ethical norms as there are possible systems of epistemic norms, some of them providing better models of our epistemic and ethical practice than others. What it is rational to believe and act on, however, is another matter, dependent on one's aims.

Within this picture it is hard to explain Railton's attachment to valoric utilitarianism. It is not obvious – as he effectively accepts – that it models our moral convictions and practice terribly well, and certainly it in no way follows from instrumental rationality. So what critical power can its impartial standpoint have?

In general, Railton's view of ethical standards must surely be like Gary's view of epistemic standards. Neither his or her own good, nor the social good, nor the morally right, is something that anyone necessarily has reason to pursue. It depends whether pursuing any of these satisfies the aims they actually have (as against those they would have under some idealized conditions). So why pick out these standards, as against those of the anti-social point of view, the Nietzschean perfectionist's point of view, etc.? Why do these standards in particular ‘support a critical evaluative practice’ (p. xiv)?

Railton's general answer must be that they link to characteristic human needs and desires in a way that does in fact motivate. But then so do plenty of other things. What is it about these standards that makes them, in particular, normative for our practice? The question leads us to his essays on meta-ethics and on normativity.

IV

Railton's meta-ethics combine naturalism and realism: realism says that true ethical propositions must have truth-makers, naturalism says that these must be natural facts. The essays in this volume that outline this view have been both influential and controversial. And rightly so. Railton has made us see how difficult it is to make ‘open question’ objections to it stick. And yet, realist naturalism still seems to many people somehow to close off questions that really are open. The worry remains: if Railton is right about them, how could ethical propositions have normative force?

What is this worry? One way to take it is as a worry about what reasons for action ethical truths give us, on Railton's interpretation of them. On this understanding of the word ‘normative’, a consideration is normative for action or belief if one cannot consistently accept it and yet deny that it affords a reason for action or belief. But if this is how we understand normativity then Railton simply concedes that ethical propositions about a person's good, the general good, duty and so on are not normative. For whether or not they give you reasons depends on your aims. Thus there is no inconsistency, for example, in accepting that you have a duty to do something, or that an action would advance your own good, while denying that you have a reason to do it.

However, this point does not get to the heart of the worries about normativity and realism, because it arises not from Railton's realism, but his externalism, about ethical propositions. So let us consider propositions about reasons themselves. On the understanding of ‘normative’ that we are working with, these are trivially normative: I cannot consistently accept that some fact gives me a reason for action or belief and then deny that it does. But given Railton's naturalist and realism, and given his instrumentalism about reasons, it seems that the truth-maker for the proposition that I have reason to X is the fact that X-ing would advance my aims. So we now ask whether I can consistently accept that X-ing advances my aims and deny that that affords any reason for me to act. The ‘open question’ response seems very powerful: even if it's true that X-ing advances your aims it remains an open question whether you have a reason to act (and vice versa).

Is this because the two propositions at stake are necessarily but not a priori equivalent: The fact that p makes it the case that x-ing would advance y's aims and The fact that p is a reason for y to x? The trouble with this line of thought, it seems to me, is that basic claims about what constitutes a reason for one to act must in the end be open to reflection that is ‘a priori’, or at least not a posteriori. Only by this kind of reflection can the truth or otherwise of instrumentalism about reasons be settled. So should we say instead that instrumentalism is true a priori? If it was, the objection we've just considered to realist naturalism about reasons would be removed; but the thesis itself would remain unsettled. The question then would be how any naturalistic reduction of propositions about reasons could be a priori. Natural facts give one reasons to act – that can be a priori – but is it itself a natural fact? If it is knowable a priori can it be a natural fact?

Leaving these large issues I also want to note that the apparent self-evidence of instrumentalism can be questioned. Consider a person who believes that an action would advance his aims but sees no reason at all to do it. That person, let's grant, is open to rational criticism. But does this show that the conjunction, ‘X-ing would advance my aims and there is no reason for me to X’ is inconsistent – could not be true? No. In adopting an aim I'm committed to holding that I have reason to do so, just as in forming a belief I'm committed to holding that I have reason to do so. Thus just as I'm open to criticism if I believe something that I think there's no reason to believe, so too I'm open to criticism if I aim for something I think there's no reason to aim for. Hence the person we're considering is open to rational criticism in the following way: either he has adopted an aim which he himself does not think there's reason to adopt, or he is saying that there is no reason to do something which would advance an aim that he thinks it reasonable to adopt. The criticizability of this overall position rests not on instrumentalism about reasons but on the weaker (arguably analytic) principle that if there's reason to advance an aim there's reason to do whatever would advance that aim. This is an uncontroversial minimum, but instrumentalism, pace Railton, is not.

Railton's own treatment of the ‘problem of normativity’ does not focus on the issue of reasons and their naturalizability. He is interested, rather, in the phenomenological question of how we experience normative force, be it in the epistemic case or in aesthetic and ethical value judgement. In an excellent paper of 1999, ‘Normative Force and Normative Freedom: Hume and Kant, but not Hume versus Kant’, he provides an insightful account of how. The experience is as of something that distinctively marries force and freedom. A response (aesthetic, epistemic or moral) carries normative weight when it is one we cannot help but feel, if we are serious with ourselves, yet also one that feels spontaneously or ‘freely’ ours – and that we expect to find, as an equally free or spontaneous response, in others. Sources for this picture can be found, as Railton notes, in both Hume and Kant.

I entirely agree with him that focusing on these phenomenological questions, of what in our experience and our discussion with others actually prompts us to basic normative judgements, is the most helpful way of demystifying philosophical worries about how normativity is even possible. What I can't see is how it supports the realism, as against the naturalism, that Railton favours. To my mind it points away from it: to a dialogical and hermeneutic model of the normative, in which basic normative judgement can be innocuously truth-apt while being founded, in Kant's terminology, not on receptivity but on pure (and shared) spontaneity. It shows us, in other words, how we can perfectly well be naturalists and cognitivists about normativity without being realists about it.

Talk about normative properties and facts if you want to – just don't assume that in doing so you have or need any notion of property or fact available to you other than the notion of the sense of a normative predicate or proposition. I have the impression that Railton thinks we do need a stronger notion to fight against subjectivists, nihilists and relativists – that getting toughly ontological with these miscreants somehow helps. In my view, on the contrary, it plays into their hands.

V

There is productive tension in Railton's philosophy. On the one hand there is a penetrating and flexible humanism, highly sensitive to the way human responses and discriminations actually go, able to grasp them from within, and reluctant to box them up in neat packages. On the other, there is the driving power provided by a battery of far-reaching philosophical and ethical theses. The theses are themselves in dynamic tension. Can the attractions of utilitarianism and instrumentalism be reconciled? Come to that, can the attractions of naturalism and realism? And how well can one reconcile these far-reaching philosophical theses with a humanistic, ideographic sense of the particular?

It seems that really ambitious moral philosophy must always attempt this task of reconciling abstract and particular. Something in the subject throws down this challenge. Peter Railton brings to it a broad, incisive and imaginative mind. His essays have theoretical sweep and imaginative, subtle detail. They are one of the most impressive attempts at synthesis in contemporary moral philosophy.