Moral dilemmas in the strict sense - cases where all options are morally forbidden, through no prior fault of the agent's - are dismissed by some philosophers as unintelligible. The author argues that the possibility of such cases is a consequence of a system of social rules that is simple enough to be teachable. The motivational force of the moral judgments pitted against each other in dilemmas can be explained by reference to the role of emotion in ethics.

The argument of the book pushes further some ideas in the work of leading contemporary authors, with particular focus on Bernard Williams. Guilt is seen as the sort of emotional sanction for wrongdoing that might supply the motivational force of moral language. The emotion can be understood as appropriate in dilemmas on the basis of a general account of appropriate emotions as not requiring full warrant for a corresponding belief.

A socially based approach to metaethics is defended as a version of moral realism compatible with dilemmas. Insofar as emotions underpin the teaching of moral language, human emotional capacities impose constraints on the nature of a viable moral code and thus affect the content as well as the structure of morality.

Chapter 1 outlines the contemporary discussion of moral dilemmas and the overall argument of the book. Many philosophers question the intelligibility of dilemmas, as cases where the agent is morally responsible for a wrong act and appropriately feels guilty for it, even though he had no choice but to do something wrong. Work in deontic logic establishes that dilemmas involve no logical contradiction. However, dilemmas raise metaethical questions, about the interpretation of moral ought as action-guiding, or more generally about the motivational force of moral language. Bernard Williams work will be featured in the ensuing argument because of the way it connects all the relevant issues, understanding dilemmas as leaving "residues" of moral emotion for violating an ought.

Chapter 2 addresses issues raised by dilemmas for deontic logic. If we accept the deontic principle that "ought" implies "permissible," dilemmas cannot be characterized as conflicting "all-things-considered" oughts, but rather are best understood in negative form, as exhaustive prohibitions. These "ought-nots" can be seen as all-things-considered, and as having the strong action-guiding force of the moral "must," if we distinguish all-things-considered status from overridingness. As Williams argues, we can retain the principle that "ought" implies "can" if we drop agglomeration, which holds that OA and OB imply O(A&B). The result of a dilemma in logical terms would on this account amount to a fragmentation of ethics into separate logical subsystems

Chapter 3 argues against Williams's apparent view of dilemmas as ruling out moral realism, the view of moral judgments as describing facts about the world independent of our acceptance of them. Some comments from Philippa Foot's early work show how we can allow for the motivational force of moral judgments in general terms, as a result of teaching them in connection with emotions, without adopting internalism, or the view that moral judgments necessarily motivate in any given case just by virtue of their meaning. By modifying features of J. L. Mackie's anti-realist account, which assumes internalism, the resulting view can be used to support a minimal version of moral realism that takes moral judgments as describing hypothetical facts about any set of social rules that would adequately promote group flourishing. The role of emotion as the source of moral motivation explains how conflicting oughts can both be seen as motivating in dilemmas, if we can make out guilt and similar emotions as appropriate even in cases of unavoidable wrong.

Chapter 4 begins a defense of emotional guilt in contrast to Williams's preference for the more general category of "agent-regret" in his treatment of dilemmas. Guilt connects virtue ethics with the ethics of duty, insofar as it amounts to a negative self-evaluative response expected of a virtuous agent in response to violations of duty. It can be understood in terms of emotional identification, or empathy with someone else's anger, typically a victim of one's own misdeeds. It has motivational advantages over shame and other alternative variants of "agent-regret" and can serve as a kind of substitute for action in cases of moral wrong insofar as it essentially inflicts a kind of emotional self-punishment on the agent. Arguments in ethics against requiring an emotion such as guilt of an agent are based on restricting moral requirement to an unduly narrow notion of what is within an agent's control.

Chapter 5 continues the defense of emotional guilt in relation to the question of its intelligibility and appropriateness in cases of dilemma, where wrongdoing is unavoidable. The reaction is first defended as applicable to cases where the agent would not accept the corresponding judgment that he is responsible for a wrong, against Rawls's interpretation of guilt in terms of such a judgment in The Theory of Justice. A "perspectival" interpretation of emotional appropriateness is then introduced that would allow even for appropriate guilt in such cases, where a subset of the available evidence is important enough to warrant holding the content of the corresponding judgment in mind, even in the absence of adequate warrant for belief. This allows for the appropriateness of guilt in problematic cases such as collective guilt and survivors' guilt as well as in dilemmas. An objective correlate of guilt can be built up from the notion of a moral "taint," with self-subjection to the emotion seen as a kind of self-cleansing ritual, to dispel the taint of wrongdoing.

Chapter 6 extends the treatment of guilt in dilemmas to a larger picture of the metaethical role of moral emotion. The version of moral realism that emerged in Chapter 3 is defended in opposition both to anti-realist views such as projectivism that make emotion the sole basis of ethics and to attempts such as intuitionism to understand moral judgment on the model of perception . Emotions function primarily as a source of individual motivation to conform to a moral code set up as an instrument of group flourishing. However, they impose constraints on the content of morality insofar as any viable moral code has to be teachable at an early stage on the basis of emotions, given their role in supplying motivational force. While it allows for social variation, the resulting view of ethics can be defended against charges of relativism.