Derek Parfit's aims in his three-volume work On What Matters are famously ambitious. They are also famously driven by his discomfort with the fact that those whom he regarded as his epistemic peers held normative and metaethical views that seemed to conflict, both with one another and with his own views. What Parfit calls the master argument of Volumes I and II is his response to this discomfort: a defense of what he calls his triple theory of normative ethics, which attempts to reconcile rule consequentalism, Kantianism, and contractualism with one another. This conciliatory project briefly recedes from view in the final part of Volume II, where Parfit presents a metaethical account but makes no attempt to reconcile it with its extant alternatives. With Volume III, however, the conciliatory project again takes center stage. In the first part of Volume III, Parfit now sets out to show that his preferred metaethical view can be reconciled with its extant alternatives after all. Furthermore, and in a return to the normative questions that drove Volumes I and II, Parfit argues that act consequentialism and common-sense morality can also be reconciled, both with each other and with his triple theory of normative ethics.

Thus this third volume of On What Matters really comprises two quite different, even if not completely independent, projects. The first part of the book pursues his metaethical project, and is framed as a response to twelve of the thirteen papers published in a companion volume entitled Does Anything Matter: Essays on Parfit on Objectivity (ed. Peter Singer, 2017). The second part of the book, in which Parfit returns to normative questions, contains a brief discussion of one of those articles, but is otherwise best read as a continuation of the normative arguments of Volumes I and II. From the scope and nature of this book, two things follow. First, while Parfit asserts in his preface that the entire book can be read on its own, it must be said that this is at best wildly optimistic but more realistically quite misleading. It is most profitably read not only with Volumes I and II in the background, but also with a copy of the companion volume at hand. Second, it is impossible to do justice to On What Matters: Volume III in a single book review. Given this, I will focus in what follows almost entirely on the second, normative project that Parfit pursues in Volume III, for two reasons: First, it seems appropriate for the focus of this journal, Utilitas. Second, many reviews of On What Matters: Volume III have already appeared in print, and virtually all of them focus on the first, metaethical project of the book, almost entirely to the exclusion of the second, normative, one. This is no criticism of these reviews; the scope of the book requires such a narrowing of focus. But it makes it appropriate to focus otherwise here.

Let me begin, however, by saying something very briefly about the first, metaethical, part of the book. Here, Parfit responds to each of the papers from Does Anything Matter: Essays on Parfit on Objectivity in turn and, while this impulse is philosophically generous, it does not lend itself to one continuous line of argument that can be easily summarized. Nevertheless, it is clear that Parfit's main goal here is twofold. First, Parfit aims to articulate more clearly the metaethical view he first put forward in Volume II of On What Matters. Parfit labels this view Nonrealist Cognitivism, which he defines as the view that our moral beliefs can indeed be true, but they are not made true by corresponding to how things are in reality. Second, Parfit argues that such a view is not only true, but compatible with both Naturalism and Expressivism. These are bold and surprising claims. The particulars of Parfit's arguments for them, and whether they succeed, are not the focus of this review. Interested readers are directed to the fine reviews by Mark van Roojen in Notre Dame Philosophical Reviews (2017), and Nicholas Laskowski in Ethics (2018), which focus on these questions.

In the second, normative, part of the book, Parfit tries to reconcile, to the degree possible, disagreements between act consequentialism and common-sense morality (hereafter AC and CSM). (He also very briefly argues that these views can both be reconciled with his triple theory of normative ethics. For reasons of space, I won't have more to say on this claim in what follows.) He focuses particularly on three key points where AC and CSM seem to conflict. First, he considers a possible conflict in the picture of practical reason underlying AC and CSM; here, Parfit argues that the conflict can be significantly narrowed, even if not eliminated. Second and third, Parfit focuses on the existence of agent-centered restrictions, on the one hand, and agent-centered prerogatives, on the other. AC denies the existence of both; CSM affirms the existence of both. Parfit, however, argues that the reasoning behind CSM itself can be brought to reject act-centered restrictions and the reasoning behind AC itself gives us reason to accept act-centered prerogatives. If Parfit is right about these three claims, then the distance between AC and CSM will indeed be significantly narrower than is generally thought. So let me say something about Parfit's arguments for each of them.

According to traditional forms of AC, all practical reasons are person-neutral and impartial; we all have the same reason to try to promote whatever outcome would be best for all. But according to CSM, only some reasons are person-neutral and impartial. While we may all have some reason to want all suffering to end, for example, we also each have personal reasons to do what would be best for ourselves and our loved ones. The existence of these personal reasons explains why, according to CSM, we may not always have most reason to do what would be impartially best.

Parfit suggests that CSM is right to affirm the existence of personal reasons, but it is wrong to think there are any purely personal reasons. Thus, Parfit defends what he calls the Any-All Thesis, according to which, “if anyone has a personal reason to want something to happen, everyone has a corresponding but perhaps much weaker impartial reason to want this thing to happen” (p. 321). Given that the impartial reasons generated by personal reasons may be substantially weaker than their corresponding personal reasons, AC and CSM may still give us conflicting action guidance. But nonetheless, according to Parfit, “these conflicts are easier to understand, and may be easier to resolve, if there are not purely personal reasons” (p. 334).

Parfit's second major move to bridge the gap between AC and CSM is to try to undermine the force of (what he argues is) the most plausible deontological principle, which he labels the Means and Side-Effect Principle, and which tells us that, “though we could sometimes justifiably save several people's lives in some way whose foreseen side effect would be to kill one other person, it would be wrong to kill one person as a means of saving several other people” (p. 374). Consider, for example, that many think it was wrong to drop an atomic bomb on Japanese civilians to end World War II rather than risk a land invasion. Targeting civilians with atomic bombs involved harming them as a means to ending the war. According to deontologists such as Warren Quinn and Thomas Nagel, this is wrong just insofar as one may justifiably object to being sacrificed, without one's consent, as a means to the promotion of some goal to which one is not committed. In contrast, the civilians who would have died in a land invasion would not have been targeted as a means to end the war; their deaths would have been the foreseeable, but unintended, side-effect of military action. This would have been unfortunate, but permissible. The Means and Side-Effect Principle explains this set of judgments.

But, Parfit objects, this argument is not convincing. It assumes there is something distinctly objectionable about being non-consensually harmed as a means to, rather than as a side-effect of, the promotion of a goal to which one is not committed. But why should that be the case? Someone who will die as a result of the land invasion of Japan might just as well object to being harmed, without her consent, as a side-effect of some American project for which she has no sympathy. And the mere fact that her death was not causally necessary for the war to end will be cold comfort to her relatives. Parfit thus concludes that CSM can be brought to see that superficially attractive deontological principles are not as compelling as they initially seem.

But CSM doesn't just contain act-centered restrictions, it also contains act-centered prerogatives. Should these, too, be modified or abandoned in pursuit of promoting the best consequences for all? Not according to Parfit, who argues as follows: Act consequentialism is based on the thought that what matters most is that everything goes as well as it possibly could. Among those things most central to making things go as well as they possibly could are strong loves and attachments to particular friends and family. But such loves and attachments require us to act in ways that favor our friends and family, and are therefore incompatible with accepting and acting on AC. Given this, Parfit concludes that the very reasoning underlying AC requires all of us to accept and act not on AC, but rather on the most optimific rules and motives available. One thing those motives will include is strong love for particular family and friends, and one thing those rules will permit is for us to favor them in various ways. Thus the reasoning behind AC itself leads us to accept and try to act on motives and rules that accommodate agent-centered prerogatives.

With this summary in hand, we can now ask how successfully Parfit reconciles AC and CSM. I think the results are mixed. Consider Parfit's treatment of the different pictures of practical reasoning underlying AC and CSM. Here, I had two worries. First, it seemed to me that Parfit didn't confront the strongest possible counterexamples to the Any-All Thesis. Suppose, for example, that you and I have both trained long and hard for the Boston Marathon, and both of us have a reasonable chance of winning. I think we each have personal reasons to win. But do I have some impartial reason (even if weaker) for you to win the Boston Marathon? And do you have some impartial reason (even if weaker) for me to win the Boston Marathon? Assuming we are strangers to one another, I don't see why we would. The general point is that certain competitive enterprises seem not only to give one a reason to win, but also seem to preclude one from a having a reason for others to win (and thereby undermine one's own chance at the crown).

Parfit considers a case somewhat like this, concerning two nations fighting a war in which neither was an unjust aggressor. It might appear that citizens of each nation have personal reasons for their side to win the war, but none of them have impartial reasons for the other side to win the war. But, Parfit claims, that is not so. It would have been better, in World War I, had any nation won a quick and decisive victory. So we all, citizens of both nations included, have impartial reasons for each side to win.

Parfit's claim about World War I seems quite plausible. But I think the World War I case is importantly different from the Boston Marathon case. War is full of suffering and death, and almost any end to it (not any end, of course) is preferable to its continuance. Presumably this fact explains why citizens of one nation have some reason (even if not decisive) for their enemies to win the war. But no such considerations apply to the case of the Boston Marathon. The Boston Marathon case is less like the case of two nations fighting a war and more like two nations competing in the Olympics. And there too, the Any-All Thesis is not very plausible. Citizens of both Romania and Russia may well have reasons to root for their respective teams to take gold in the gymnastics final. But it is implausible to think that either they, or we, have reasons to root for both of them to take the gold.

But even if these counterexamples fail, there is a second worry. Parfit claims that conflicts between personal and impartial reasons may be easier to resolve if there are no purely personal reasons; that is why defending the Any-All Thesis is part of his reconciliation project. But it is unclear why that would be true, or which sort of conflicts he has in mind. The core conflict between AC and CSM, both theoretical and practical, is about what we have most reason to do. Suppose we were to grant that every personal reason generates an impartial one, but one that is only a hundredth the strength of the original personal reason. Would the core conflict not remain as intractable as ever, at least in most cases?

So I'm not convinced that Parfit has done much to alleviate the conflict between AC and CSM that arises from their differing pictures of practical reason. The second part of Parfit's reconciliation project, which aims to undermine deontological claims, seems more promising. When focused on patient-centered defenses of deontological restrictions, Parfit's objections seem to me as devastating as they are simple. But defenders of deontological restrictions also offer agent-centered defenses of deontological restrictions, and what Parfit has to say about these is far less convincing.

Consider Parfit's treatment of Nagel's argument for the Means-Side Effect Principle. Parfit rightly notes that one of Nagel's major arguments in defense of that claim is patient-centered: that a person ought to have some protection from being used without his consent to further some aim to which he is not committed. To this, Parfit plausibly replies that persons could equally be entitled to protection from being harmed without their consent as a side-effect of some project to which they are not committed. But that was only half of Nagel's argument in The View from Nowhere (1986). Nagel also offered an agent-centered defense of the Means-Side Effect Principle. Specifically, Nagel observes that the agent of such harm not only intentionally wills the harm, she must also be counterfactually sensitive to any falling off in the level of the evil that the harm constitutes, and be prepared to alter her course of action to increase, rather than decrease, that evil. To this observation, Nagel adds the objection that,

Tolerating evil as a side-effect, Nagel points out, does not have this counterfactual implication for the will. That is why it is permitted, and the doing of harm as a means is not.

As far as I can tell, Parfit never really responds to this half of Nagel's argument. He does note that Nagel focuses not only on the patient but also on the agent, but summarizes that argument as follows:

Parfit quite rightly dismisses this appeal to ignorance as morally unimportant. But Parfit's summary does not fairly represent Nagel's view. Nagel's argument is not that harming someone as a means requires knowledge of that harm, but that it requires a positive orientation of the will to that harm. It is not just that we know we are harming – it is that we both will that harm and commit ourselves to increasing it when and if necessary. No such orientation of the will is necessary if the harm is a mere side-effect. This is a different, and arguably stronger, argument.

Of course, this isn't to say that Nagel's agent-centered justification of a deontological restriction, or any other such justification, will succeed in the end. Many philosophers view agent-centered justifications as doomed for their supposed moral narcissism. But it must be said that Parfit gives such justifications short shrift in comparison with their patient-centered cousins. Given the ambitions of Parfit's aims, and the ubiquity of such arguments, it would have been good to hear more from him on this issue.

I have focused here on points of disagreement with Parfit, or areas in which I think he needs to say more. The first is usual for this sort of review, and the latter is perhaps inevitable given the enormous ambitions that Parfit set for himself. But neither of these should obscure what Parfit has accomplished in Volume III of What We Owe to Each Other. Moral philosophers will be discussing it for some time to come.