Book Reviews

David Shoemaker, Responsibility from the Margins, Oxford University Press, 2015

David Shoemaker’s highly innovative and intricately argued new book draws on much of his previous work together with substantial original material to form a detailed and cohesive treatment of responsibility. The book is engaging, crisp, and admirably clear. It is marvelously ambitious in its strategy and framework, engagement with multiple literatures, and decidedly novel approach to Strawsonian theory. Moral philosophers, psychologists, clinicians and practitioners, and anyone who has ever wondered about “marginal agents” – people with dementia, autism, (manic-)depression, OCD, and psychopaths – will find much to entice them in this thorough and accessible treatise.

Shoemaker’s starting point – the phenomenon he aims to explain – is the observation that many of us feel a certain ambivalence toward marginal agents of the sort mentioned above. When we interact with or read case studies of marginal agents, we feel a “profound unease,” which Shoemaker diagnoses as caused by the fact that “these agents seem worthy of some responsibility responses but not others, which suggests that they are responsible in some ways but not in others” (3). This is the foundational premise of the entire book. For Shoemaker, responsibility responses include but are not at all limited to the standard praise, blame, and resentment; admiration and contempt, approval and disapproval, pride and shame, anger and regret are all brought in under this unusually wide conception of responsibility responses (a point I’ll return to in my final remarks). The idea is that psychopaths may deserve contempt but not anger, patients with dementia may deserve admiration but not resentment, and so on and so forth.

Starting from responsibility responses puts Shoemaker squarely in the Strawsonian camp of moral responsibility theorists. Put in the simplest of nutshells, P.F. Strawson’s revolutionizing argument for sidestepping the free will problem in “Freedom and Resentment” was that we should stop thinking of responsibility responses as only justified by certain metaphysical facts, i.e., by the fact that we are agents with free will, and instead think of our responsibility responses as justified by their role in (unavoidably necessary but also intrinsically valuable) social practices of caring about and responding to others’ manifestations of goodwill or ill will toward us (Strawson 1962). In Part I, Shoemaker introduces three main modifications[1] to Strawson’s view. The first is that Shoemaker draws a tripartite distinction amongst the “attitudes and intentions towards us” or the “quality of others’ wills towards us” that Strawson takes to be the object of reactive attitudes (Strawson 1962 48, 56); he distinguishes between the quality of others’ character, the quality of others’ judgment, and the quality of others’ regard for us. The second is an account of responsibility responses in terms of “sentiments” (along the lines of metaethical and aesthetic sentimentalism), where sentiments are dispositions to feel a special type of emotions that are culturally universal and recalcitrant to judgment.[2] The third, following closely from the second, is slipped into an unobtrusive argument that if the appropriateness of our responsibility responses are not justified by metaphysical facts then “it must be a matter merely of how the responses somehow fit with their objects” (19). In other words, Shoemaker understands the justification of responsibility responses in terms of fittingness. Responsibility responses are justified when they are fitting to their objects, just as emotional and aesthetic responses are appropriate when they are fitting to their objects.[3]

Shoemaker thus identifies three paradigmatic pairs of sentiments (and other related emotions) that correspond to the three different objects of our responsibility responses: agential disdain and admiration (along with contempt, abhorrence, awe, veneration, etc.) are the fitting positive and negative responses to good and bad character; agential regret and pride (etc.) are the fitting responses to good and bad judgment; and agential anger and gratitude (etc.) to good and bad regard. These correspond to what Shoemaker calls responsibility as attributability (directed at character), answerability (directed at judgment), and accountability (directed at regard). Part I of the book is devoted to laying out this theory and defending the distinctiveness of each of the three types of responsibility.

In Part II, Shoemaker shows how this tripartite account accommodates our ambivalent responses to marginal agents; in each case, he argues that the condition in question disrupts some of the capacities or psychic elements required for one form of responsibility, but leaves intact others required for other forms of responsibility. People with depression have mitigated attributability for their actions, because the depression has interfered with the cares and commitments that constitute a person’s character (or “deep self”), but they may still be capable of responding to reasons and experiencing empathy in the ways required for answerability and attributability. Psychopaths lack the capacities for experiencing empathy (or have severely mitigated capacities to do so) required to sustain properly moral regard for others, so they are exempt from accountability, but they still have the capacities for character and judgment[4] that are required for attributability and answerability. And people with mild intellectual disabilities are exempt from (or have mitigated) answerability, because they lack the cognitive and adaptive capacities – reasoning, planning, learning – that are required for properly reasons-responsive evaluative judgment; but they may still have cares and commitments that constitute their distinctive selves and they may still be capable of empathy, especially for those with whom they have close relationships, in such a way as to preserve their attributability and accountability. Shoemaker provides similar accounts for people with mania, scrupulosity,[5] high-functioning autism, dementia, and poor formative circumstances (i.e., morally deprived childhoods), demonstrating impressive mastery of the empirical literature around each of these marginal cases. The book really shines when Shoemaker draws on the resources of his tripartite theory to make illuminating proposals for further empirical and clinical investigation. He writes: “[O]nce we move beyond the deeply entrenched all-or-nothing model of responsibility, the tripartite theory opens us up to a new, nuanced set of tools for both those who treat and those who suffer from (at least some of) these disorders” (145). In short, a more fine-grained moral theory can help us better conceptualize complex psychological phenomena, while in the other direction, diving into real-life details prompts the development of more sophisticated moral theory.

In the remainder of this review, however, I will raise three worries for the project: one regarding the general strategy, and two regarding Shoemaker’s sentimentalist take on Strawsonian theory. My first worry is that, insofar as Shoemaker claims that the puzzle to be solved is the fact of our ambivalent responses to marginal agents, it is not clear that the tripartite theory is really necessary. Shoemaker says very little to characterize the phenomenon or what he actually means by “ambivalent” responses, other than that the “unease of ambivalence” is not the unease of uncertainty, before diagnosing it in (the already quite theory-laden) terms of feeling that agents merit some responsibility responses rather than others (3). There might be alternative ways of explaining this ambivalence, however. It might not be genuine ambivalence, for instance, if our responsibility responses are simply keyed to certain facets of the cases which feel warranted or not depending on whether those facets are made salient at a given moment or not. In other words, like optical illusions that flip from duck to rabbit, young girl to old woman, depending on which cues we focus on – but where the picture is either of a duck or a rabbit, and not a part-duck/part-rabbit – perhaps, too, a marginal agent could be responsible when her case is viewed under a certain light, and not responsible under a different light, but it not be the case that under a single description[6] some responsibility responses are appropriate while others are not.

Or, to take another visual metaphor inspired by Shoemaker’s own discussion of “local blindnesses” whereby an agent may be responsive to reasons only in some domains but not others (82), perhaps we could think of our ambivalent responses as picking up on domain-specific gaps in more unified responsibility-conferring capacities that are overall intact. For example, on accounts that would deny Shoemaker’s attributability-answerability distinction,[7] it might be that a marginal agent is a responsible agent overall for her actions and attitudes in virtue of how it is possible for them to flow from her evaluative judgment, but that she lacks responsibility for specific actions and attitudes flowing from certain judgment “blindnesses,” or domains over which her capacity for judgment is blocked or lacking. Thus a person with depression might still count as a responsible agent in virtue of the fact that her evaluative judgment is intact overall, as evidenced by the fact that she manages to sustain quite a few minimally requisite actions, e.g., feeding herself, keeping her job; but there are large and important swaths of a full human life, e.g., her relationships, her activities, in which the operation of her evaluative judgment is too distorted for her to be responsible for them. This last observation might be related to the fact that Shoemaker claims to be giving an account of responsible agency, rather than responsibility for particular actions, which suggests, finally, that some of our ambivalence toward marginal agents might also be handled by distinguishing between how we feel toward them as agents versus how we feel toward them as agents in light of specific instances of behavior.

I suspect, however, that the puzzle of ambivalent responsibility responses might be (at least somewhat) less central to the motivations behind the tripartite theory than suggested, and we could simply think of marginal agents as exemplary case studies for illustrating its power. This brings me to my second set of worries, however, which is that the trifold sentimentalist strategy and framework – marvelously conceived though it is – may not be enough to sway those who prefer unified accounts of responsibility that can still save the phenomena. The costs of relatively minor or esoteric counterexamples may not outweigh the benefits of simplicity. Out-of-character actions and whims,[8] for instance, which Shoemaker takes to be instances of answerability without attributability (59, 83), seem to me still attributable to agents as manifestations of some kind of character trait, even if not the obvious ones: a wandering mind, an active curiosity, sublimated anxieties and desires, a love of spontaneity, and so on. (This move would be akin to one Shoemaker makes in his own analysis of scrupulosity, according to which such agents’ general “moral orientation” and perfectionist stance renders many of their thoughts and attitudes still properly attributable to them.) Or they may manifest pro tanto evaluative judgments rather than all-things-considered evaluative judgments[9] about the worth of some action.

I also found myself worried at times about proposals that, while innovative, struck me as potentially ad hoc or insufficiently grounded in deeper justification. If debates about the deep self have centered around Platonic versus Humean accounts, for instance, is it a satisfying resolution to simply combine them into an ecumenical position, without further diagnosis, explanation, or motivation? If the dominant philosophical and psychological view is that anger fundamentally involves revenge, is it enough to point to the existence of some cases in which people feel frustrated if they cannot communicate to the objects of their anger that “This is for what you did to me!” (105)? (For surely we can imagine agents whose anger leads them to want simply to annihilate and destroy the objects of their anger, or whose anger is satisfied when revenge can be carried out anonymously such that their objects “never even knew what hit them”.) Is it fair game to simply declare that we have “empathic control” when identifying with others’ perspectives leads us to certain kinds of emotional or reasons-responsiveness, such that these are “up to us” in the way that things over which we have voluntary control or rational control are up to us? (For the latter have been understood as “up to us” because they derive from our wills or our evaluative judgment, which have been argued to be constitutive of moral agency; but such an argument has yet to be made for empathic capacities.)

Finally, there were some moves that seemed to betray some artificiality in the tripartite taxonomy of sentiments. In the case of admiration (one of the emotional syndrome pair for attributability), for instance, Shoemaker restricts the sentiment to agential admiration, setting aside cases in which we admire non-agents, e.g., the Grand Canyon; but when it comes to finding the emotional syndrome pair for answerability, Shoemaker simply sets aside disappointment on the grounds that it is not restricted to agents. This leads him to adopt a first-personal emotional syndrome pair for answerability (pride-regret) that does not match the third-personal pairs for attributability and accountability. And although Shoemaker acknowledges that we may sometimes feel regret even when we absolutely do not manifest any flaw in evaluative judgment, he dismisses these as non-paradigmatic cases of regret which do not count as genuine responsibility responses. Of course, any theory – especially new and ambitious ones – will contain anomalies. But since Shoemaker’s job is to convince us that there really are three fundamental pairs of sentimentalist responsibility responses and since his work is otherwise so thoroughly grounded in empirical evidence, I am hopeful for more reason to think that the tripartite theory has successfully latched onto natural kinds in our actual practice rather than being a theoretical framework imposed from the armchair.

This brings me to my final worry, about the sentimentalist fittingness approach to justifying responsibility responses. In developing his tripartite theory, Shoemaker has broadened his responsibility responses beyond the responses of resentment and indignation on which other theorists have focused.[10] To be sure, Strawson himself does identify a wide range of reactive attitudes, including “gratitude, resentment, forgiveness, love, and hurt feelings” (49). But this is in the first stage of Strawson’s argument, in which he is pointing out the existence of reactive attitudes in general; the second stage of his argument is to explore whether an understanding of specific reactive attitudes can “bring us, if possible, nearer to a position of compromise in a more usual area of debate” (56), where the usual debate, of course, centers around “desert, responsibility, guilt, condemnation, and justice” and the problem of free will (64). Moreover Strawson is clear that he is interested primarily in moral responsibility,[11] that is, in the “concept of moral responsibility and of the practices of moral condemnation and punishment” (62). It is no accident that desert, condemnation, justice, and above all free will – precisely those concepts that Shoemaker sets aside in his last chapter – are concepts that intimately involve concerns about fairness. Thus when Shoemaker declares that appeals to unfairness constitute “the wrong kind of reason,” since they are orthogonal to questions of whether some responsibility response is fitting to its target in the way that aesthetic responses are (201), this seems to me a significant deviation from the Strawsonian project. Even if it is (quasi-aesthetically) fitting to feel contempt toward a vicious agent, concerns about determinism and free will – Strawson’s target – arise only when we worry about whether it is morally appropriate for us to feel that way toward an agent who did not have, e.g., the (fair) opportunity to avoid becoming vicious. Similarly, when Shoemaker disavows the centrality and relevance of harsh treatment to responsibility, this seems to depart altogether from the Strawsonian starting point that begins with our practices of moral condemnation and punishment.

Shoemaker, of course, is aware of this difficulty, stating: “I want to insist on the crucial distinction (a distinction I believe has long been overlooked) between the fittingness of various sentimental responsibility responses and the appropriateness of harsh treatment of offenders” (223). I cannot help but suspect, however, that this final, insisted-upon distinction suggests that the fundamental distinction might hold between just two faces or concepts of responsibility, as defended (in different ways) by theorists such as Gary Watson (2004), Tim Scanlon (1998), Iris Marion Young (2011), and myself. If I am right about this, then Shoemaker will have offered us an exceptionally elegant and richly-detailed Strawson-inspired theory of the first (aretaic or attributability) concept of responsibility, while more or less setting aside the second (substantive or accountability) concept of responsibility with which Strawsonians typically concern themselves. But none of this detracts from the huge contribution that Shoemaker has achieved in this book with his exciting and fruitful new framework for thinking about responsibility.

Robin Zheng
Yale-NUS College


Scanlon, Thomas. 1998. What e owe to each other. Cambridge: Belknap Press of Harvard University Press.

Smith, Angela M. 2012. “Attributability, Answerability, and Accountability: In Defense of a Unified Account.” Ethics 122.3: 575-89.

Strawson, P.F. 1962. “Freedom and Resentment.” Perspectives on moral responsibility. Eds. Fischer, John Martin and Mark Ravizza. Ithaca, NY: Cornell University Press, 1993. 45-66.

Wallace, R. Jay. 1994. Responsibility and the moral sentiments. Cambridge: Harvard University Press.

Watson, Gary. 2004. Agency and answerability: selected essays. Oxford: New York: Clarendon; Oxford University Press.

Young, Iris Marion. 2011. Responsibility for Justice. New York: Oxford University Press.

Zheng, Robin. 2016. “Attributability, Accountability, and Implicit Bias.” Implicit Bias and Philosophy, Volume 2: Moral Responsibility, Structural Injustice, and Ethics Eds. Saul, Jennifer and Michael Brownstein. New York: Oxford University Press, 62-89.


[1] Or, depending on how invested one is in Strawsonian exegesis (the temptation of which I try, but perhaps not hard enough, to resist as far as possible in this review), three main interpretive extensions.

[2] The latter simply refers to the tendency of some emotions to remain even when certain beliefs that constitutively define that emotion are given up, e.g., continuing to fear a spider even after one has acquired reason to give up the belief that it is dangerous.

[3] This sort of fittingness view has been defended with regard to blame in particular; see, e.g., Arpaly (2002) and Hieronymi (2004).

[4] This is true, at least, for nonmoral judgments. But they may lack the capacity for moral judgments that require having moral regard for others, and hence lack answerability in these domains.

[5] A form of OCD involving obsessive thoughts about morality, e.g., that one has committed a moral wrong merely by imagining a certain event.

[6] Whether there is an objectively best way of describing the case might depend on deeper commitments about the nature of responsibility, e.g., whether responsibility is primarily a metaphysical or practical matter.

[7] The difference for Shoemaker consists primarily in the fact that answerability is limited to evaluative judgment, while attributability ranges more widely across commitments and cares that are distinct from or not grounded in specific judgments.

[8] This addresses the second of Shoemaker’s three arguments for the distinction between attributability and answerability. In response to the third counterexample of an agent who holds two conflicting attitudes that are singly unobjectionable but irrational when combined, a unified account theorist might endorse Smith’s (2012) reply and argue that what deserves criticism is not the irrationality of this configuration, but the irrationality of failing to take steps to alter the configuration once the conflict has become so salient. For most of us likely harbor at least some irrational configurations of conflicting attitudes (like contradictory beliefs) which are not worth the time or effort to resolve.

[9] A unified account theorist might use this to argue against Shoemaker’s first argument: cares and commitments might be grounded in pro tanto rather than all-things-considered judgments.

[10] See, most notably, Wallace’s (1994) arguments for narrowing the class of reactive attitudes to those which hold someone to an expectation, e.g., resentment, indignation, and guilt (25-33).

[11] More fully, Strawson writes: “The concepts we are concerned with are those of responsibility and guilt, qualified as ‘moral’, on the one hand— together with that of membership of a moral community; of demand, indignation, disapprobation and condemnation, qualified as ‘moral’, on the other hand—together with that of punishment” (62).