1.1 Foundationalism in Epistemology and Ethics

The idea of giving a foundation to our commitments seems naturally at home in the theory of knowledge, but the idea also applies to ethics. A book that influenced Williams, Alasdair MacIntyre’s After Virtue, exemplifies it (Reference MacintyreMacIntyre, 1981). MacIntyre argued that ethical thought was once in rational good order: It related a conception of human nature as it is to an ideal of how it might be developed. It proceeds from the former to the latter via a set of rational injunctions. (Together these constitute a ‘tripartite schema’.) Then there was a ‘fall’ – a modernity crisis, even though the word ‘modern’ here picks out an event in the ancient world. For MacIntyre, every ethical thinker since that time has engaged in a forlorn, ahistorical, task of reconstructing this tripartite schema.

After Virtue argues that, in our modern condition, the failure of this project has left ordinary ethical thinking fragmented and disordered. Only one outlook avoids this predicament: the Neo-Thomist Catholicism that reflects MacIntyre’s own views (Reference MacIntyreMacIntyre, 2016). If you share his diagnosis, but not his outlook, then the appeal of the rationalism endemic to the Morality System becomes an obvious response to this predicament.

Williams certainly shared MacIntyre’s view that any worthwhile ethical philosophy must demonstrate an awareness of the historical and social structures within which it is embedded. Did he also share his foundationalism? MacIntyre’s tripartite schema is most clearly to be found in the ethical thought of Aristotle, a philosopher with whom Williams engages repeatedly – and to whom he is broadly sympathetic.Footnote ⁴

In his discussion of Aristotelian ethics, we have a test case for the presence of foundationalist commitments in Williams’s thinking:

Aristotle’s search for foundations, then, must fail – we cannot get outside the ethical.Footnote ⁵

Equally problematic, for Williams, is the historically important paradigm of Plato’s ethics. It, too, tries to find an ‘Archimedean point’ for our ethical ideas; it fails for the same reason but leaves a powerful historical legacy. That legacy is a form of moral rationalism, which invokes a ‘characterless’ view of the self that Williams calls ‘ethical Cartesianism’ (Reference WilliamsWilliams, 1993, p. 99). It represents a form of substractionism about the self: Even if one’s self is partly composed of a purely rational self, Williams argues that one can never be merely that taken in isolation (Reference WilliamsWilliams, 1985, p. 73). This false conception of the self is an important element of the Morality System.

Ethics and the Limits of Philosophy sets up its argument, at least superficially, as a search for foundations (Reference WilliamsWilliams, 1985). As the book unfolds, it appears that foundationalism underpins Williams’s understanding of how we are to address what he calls ‘Socrates’ question’: ‘How should one live?’. Now, Williams might well respond that the foundationalism comes not from him, but from the traditions with which he engages: the Kantian and the Aristotelian.

In ELP’s methodological preamble, Williams justifies this assumption. He argues that there are prominent traditions in ethics that start from a minimal conception of rational agency. From that Archimedean point, these traditions develop an argument that tries to propel any agent into ethical life by the force of rational argument alone.Footnote ⁶ Williams diagnoses this as a political aspiration: The goal of making abstract, academic, moral philosophy itself a source of power in the world. He is unconvinced: ‘What will the professor’s justification do, when they break down the door, smash his spectacles, take him away?’ (Reference WilliamsWilliams, 1985, p. 23).

Foundationalism, I will argue, underpins Williams’s arguments and is more than merely a rhetorical device for framing the issues. Specifically, it is how he addresses his recurrent concern with intellectual authority in ethics. This implicit commitment reflects MacIntyre’s influence: If, as Williams believed, ordinary experience contains a reasonable pluralism of correct ethical outlooks, at what point does plurality unravel into disorder in the way MacIntyre diagnosed? A rational foundation would allow us to stabilise this structure – but what else could do so if the foundationalism fails? As we will see, Williams thought that the answer ‘an ethical theory’ was demonstrably unconvincing.

MacIntyre’s diagnosis that ethical thought in modernity is in a state of disorder is bad news for those of us who do not share his Roman Catholic faith (Reference MacIntyreMacIntyre, 2016). Williams’s outlook, in contrast to MacIntyre’s, is resolutely secular, liberal, and attuned to the inherent reflectiveness of an ethical life lived under the special conditions of a modern society. Does anything survive to sustain the hopes of that which – in a revealing epigraph to ELP – Williams characterises as the perspective of the ‘chastened realist’? (The phrase is taken from a poem by Wallace Stevens.) From amidst the ruins, as it were, positive commitments do survive. That is because, Williams argued, we see them for the first time when they have been freed of distortions imposed on them by both the Morality System and its intellectual shadow, the mainstream of academic moral philosophy.

What are these commitments? In a crucial passage of ELP, Williams contrasts the search for ‘social and political honesty’, as Enlightenment ideals, with a much more ambitious conception of them foisted on us by ‘a rationalistic metaphysics of morality’ (Reference WilliamsWilliams, 1993, p. 159). The core Enlightenment value, Williams claims, is critique – ‘the spirit of political and social truthfulness’ (Reference WilliamsWilliams, 2002, p. 4).

Williams viewed this task of critique as identifying, articulating, and metaphorically ‘placing’ our ethical commitments in relation to other conceptions of objectivity. This is pursued via Williams’s historical writing and his late commitment to the method of genealogy. The most important act of ‘placing’ is to understand our ethical ideas in relation to another important expression of our rational nature, namely, scientific understanding. There is no detaching Williams’s conception of the objectivity of the ethical from his conception of what he called ‘the scientific’ (Reference Moore and ThomasMoore, 2007). Modern science has given us a glimpse of a conception of a world maximally – but not totally – independent of our distinctive particularities as knowers (Reference ThomasThomas, 2006, chapter six).

While it makes sense to understand the scientific as a way of finding our way around a unitary physical world, the ethical, for Williams, is a way of structuring a plurality of ways of living internal to the ethical. But within each of those ways of life, there is something to which one can appeal: An ideal of mature ethical agency that is shaped by the agent’s acknowledgement of the demands of the social. Those demands are reflected in an internalised ‘other’ who is a person one can respect. It would be shameful not to live up to the standards shared with this internalized figure (Reference WilliamsWilliams, 1993, p. 98).

In Section 4, I discuss whether Williams’s contrast between the absolute conception of the world implicit in our scientific understanding of it is intended to contrast with the pluralism internal to our ethical lives – to problematize the latter. In fact, Williams did not believe this was the most interesting way in which one could be a sceptic about ethics (Reference WilliamsWilliams, 2016). If modern ethical scepticism characteristically takes the form of asking how moral values ‘fit in’ to a modern, disenchanted, scientific worldview, Williams is more interested in a local form of scepticism within the ethical. He thinks that we ought to focus our scepticism on the Morality System.

Given his scholarly interest in Plato, Williams identifies the origin of this System in Plato’s rationalism (Reference WilliamsWilliams, 1998). This rationalism finds canonical modern expression in the work of Kant (via the ancient Stoics) and has a strong presence in the recent development of ethical theories. Strikingly, Williams traces all these expressions of rationalism to a set of nine commitments.

They are as follows: The Morality System is focused on practicality, in the sense that one’s obligations must always be something the agent can do (Reference WilliamsWilliams, 1985, p. 175); such obligations cannot ultimately conflict (Reference WilliamsWilliams, 1985, p. 176); all specific obligations are grounded on general obligations (Reference WilliamsWilliams, 1985, p. 175, pp. 185–186); moral obligations are ‘inescapable’ – an agent is always within the scope of blame (Reference WilliamsWilliams, 1985, p. 177); obligations can be trumped only by other obligations (Reference WilliamsWilliams, 1985, p. 180); all important ethical considerations are translatable into obligations (Reference WilliamsWilliams, 1985, p. 179); moral luck is impossible because of the purity of morality (Reference WilliamsWilliams, 1985, pp. 195–196); the characteristic reaction of the system is blame, always ‘directed to the voluntary’ (Reference WilliamsWilliams, 1985, pp. 177–178); morality is essentially impersonal in the negative sense that ‘the thought I did it has no special significance’ (Reference WilliamsWilliams, 1985, p. 177).

Revealingly, when Williams states that he takes the primary exposition of the system to be the moral philosophy of Kant, he adds that ‘the purity of morality itself represents a value … . the ideal that human existence can ultimately be just’ (ibid., p. 195) Williams interprets this as Kant’s secularised version of the Pelagian heresy ‘which … adjust[s] salvation to merit’ (Reference WilliamsWilliams, 1985, p. 38, p. 224, fn. 20). The Pelagian heresy, propounded by Pelagius (c355–c420 AD) and condemned by the Catholic Church, denies original sin, believes in the fundamental goodness of humanity, and sets humanity the task of freely achieving an ethical life without the need for God’s grace.

Williams believed that if we can free ourselves from the snares of the Morality System, we achieve a conception of the ethical as reasonably pluralist, historically informed, genealogically vindicated in its core commitments and centred on an ideal of mature ethical agency. It is this positive conception that I will now exposit in more detail.

1.2 The Ethically Mature Agent

Williams’s basic thesis is that a mature ethical agent has a fully developed (but not ‘perfected’) character. Once an agent has a mature ethical identity, then he or she acts from it in a way that expresses their practical identity. Practical identities are constituted by fundamental ground projects: ‘Some project or objective with which the agent is deeply identified’ (Reference WilliamsWilliams, 1995a, p. 5).

There is a tight circle of interconnected ideas here, but the most fundamental is that of character. It allows us to explain further concepts important to Williams such as practical necessity, an agent’s (in)capacities, and what it is to act from integrity. If we must use the classificatory terminology of academic writing about ethics, then Williams is primarily a virtue ethicist.

The fundamental idea is that, as Elizabeth Anscombe originally claimed, virtues of character are central to any defensible moral psychology (Reference AnscombeAnscombe, 1958). Williams concurred that the concept ‘has to be used in moral philosophy’ (Reference WilliamsWilliams, 1985, p. 9). A virtue is ‘a disposition of character to choose or reject actions because they are of a certain ethically relevant kind’ (Reference WilliamsWilliams, 1985, p. 9).

One of Williams’s most insightful remarks about the virtues figures obliquely in his critique of ‘two level’ forms of utilitarianism:

One’s character is, as it were, fixed by triangulation: It emerges from coordination with two other points (Reference ThomasThomas, 2005, Reference Thomas and Webber2010, Reference Thomas2024). First, you identify that on which it is a point of view, namely, the values to which it is a response via ‘feeling and judgement’. Second, it is characteristic of the virtue terms that typically are ascribed from an interpreter’s point of view. It is a community that shares the understanding of that concept which, for example, judges a person to be kindly in her dealing with others.

This account of ethical judgment is essentially complemented by Williams’s explanation of whether it is a form of heteronomous motivation (in Kant’s sense) to act merely in this third personal light of how others see you. Williams argues not: In his positive conception, the correlate of the third person is an internalized other with a distinctive characterisation:

Specifically, this ‘other’ represents a person whom one respects such that it would be shameful not to live up to their expectations (Reference WilliamsWilliams, 1993, p. 98).

Triangulation from values, and any given community of interpretation, allows us to identify the ethical self who ‘sees the world from the point of view of that character’. But, as Mark Sacks puts it, this self is always ‘behind the lens’ (Reference SacksSacks, 2005). This looks like a conception of virtue as a form of knowledge which, to a point, it is – Williams’s ‘non-objectivism’ about the ethical which I will discuss in Section 4 explains why the qualifier is apposite.

There is a similarity between Williams’s views and those of French existentialist Jean-Paul Sartre (Reference SartreSartre, 1956; Reference Thomas and WebberThomas, 2010, Reference Thomas2024). They agree on the systematic elusiveness of the self (Reference SacksSacks, 2005). Both also think that there is an implicit hierarchy to our practical commitments structured via fundamental ground projects (Reference Thomas and WebberThomas, 2010). At the apex of this hierarchy are the most important ethical identities. An agent is identified with them; they are the perspective from which she acts. Williams adds that they answer the question of why one goes on at all (Reference WilliamsWilliams, 1981b, p. 13). It is the idea of acting from your own ethical identity which explains Williams’s use of the term ‘expressive’: A virtuous agent’s actions are expressive of their character.

If ethics is a way of navigating a social world, we must presuppose that in characterising this agent we are talking about someone who inhabits our world – where ‘our’ is always a problematic expression in Williams given his attentiveness to social and political context. Our envisaged agent lives in modern liberal society which has a certain conception of law (Reference WilliamsWilliams, 1993, pp. 65–68). That conception extends, importantly, beyond the idea of penalising crimes to the broader idea of civil wrongs (‘torts’). Indeed, it extends to the very idea of a self who enters into the kinds of commercial contracts that constitute many of our impersonal relations to others (Reference Thomas and LeBarThomas, 2019). In particular, generalised social trust requires the stability of the self (Reference WilliamsWilliams, 2002, pp. 200–202). This matters, because our envisaged subject is embedded in a social context that negotiates ways in which the ideas of the voluntary and the responsible are related. In so doing, it identifies one way of legally ordering social life – not the way. The Morality System implies that there is only one such correct ordering – one which assumes a metaphysically ambitious notion of the voluntariness of action.

If this agent is reflective and historically informed, they may have the thought that the ways in which these basic concepts of freedom, law, voluntariness, and ethical reactions relate to each other can be found in some canonical texts of the Western cultural tradition. They can find this out via reflection on ordinary ethical practice and its persistent defiance of the tenets of the Morality System. In S&N, Williams argues that the apparently ‘archaic’ emotion of shame still does a lot of ethical work for us – for example, in his model of the internalized other (Reference WilliamsWilliams, 1993). That is a truth of our experience: S&N does not teach it to our envisaged agent but reminds her that she uses this concept already to make sense of her ethical experience.

Assembling the pieces, then: The basic idea for making sense of our ethical lives is that of virtue. This is a multitrack character disposition that goes beyond a mere skill because it involves characteristic patterns of motivation (Reference WilliamsWilliams, 1985, p. 9). Specific practical questions focus a person’s virtues on a course of action. From the third personal perspective, one of the most interesting aspects of this is that a person can manifest character via the category of the unthinkable – what it does not even occur to that person to do (Reference Williams, Smart and WilliamsWilliams, 1973b, p. 93; Reference Williams1981f, p. 129). Furthermore, specific practical questions focus on a range of virtues in a way that can be surprising to the deliberating agent, can create new forms of (fallible) self-knowledge, and give substance to that which Williams calls moral incapacities (Reference WilliamsWilliams, 1995d). The latter explains those things an agent cannot knowingly do.

Williams believed that there was ethical significance to agency because the ethically mature agent knows that she will be held answerable for what she does – not simply what she does voluntarily. We have a workable idea of the voluntary: It is no more than a mature ethical agent deliberating and acting from a normal state of mind (Reference 72WilliamsWilliams, 1995b). As in other cases we may have only an implicit grasp of normalcy here, but we have an explicit grasp of non-normal conditions on deliberation and action (Reference ThomasThomas, 2011, p. 158).

If Williams is right, this characterisation of an ethical agent ought to strike us as truistic. It will draw us back to facts evident either from our experience or from those of others known to us through testimony. It is the removal of theoretical distortions of our ethical experience imposed by philosophical assumptions embedded in the Morality System.

Yet it is important not to forget that ‘ethical Cartesianism’ is one strand of the historical traditions that inform common-sense morality. It is not as if there is nothing in our ordinary conceptions to which these theoretical reflections can appeal. Furthermore, as we begin to explore Williams’s approach to issues such as moral luck, we are likely to feel ourselves torn in our allegiances. This reinforces Williams’s argument that the Morality System is reflected in what we already think: Giving up these ideas imposes costs on us. The controversy that Williams’s views have occasioned offers ample evidence of his point – as I will illustrate in the next section.

2.1 The Importance of the First Person

A continual interlocutor with Nagel, Williams’s critique of impartialism seems closely to track Nagel’s different formulations of that ideal. Essential first personality appears in ELP immediately after Williams’s critique of Kant’s failed attempt to ground ethics in an impartial perspective:

What notion of impartiality could help the contemporary impartialist to conduct a salvage operation from this Kantian wreckage?

The first step would be to note that a rational agent does not merely act from reasons but ‘on reasons’ (ibid). Call this the fact of cognitive autonomy. Williams asks whether that is enough to bring in an impartial perspective on one’s first-order beliefs and desires such as to generate a standpoint from which they can be assessed:

If we can motivate this step, then we do seem able to generate what the neo-Kantian needs: Individual agents will become legislators for all other agents (Reference WilliamsWilliams, 1985, p. 66).

But Williams is sceptical. He begins his critique by pointing out that Kantian cognitive autonomy cuts across the theoretical and the practical uses of reason: That is a problem for the impartialist about ethics. Rational freedom involves ‘standing back’ from both beliefs and desires; in the latter case, reflection itself may generate new desires. Williams concedes that this may well be an insight into theoretical reasoning but, for that reason, it helps to identify an asymmetry between theoretical and practical reasoning. This asymmetry undermines the impartialist project of reconstruction:

Theoretical reflections about the word are essentially world directed which explains why ‘I occur in them, so to speak, only in the role of one who has this thought’ (Reference WilliamsWilliams, 1985, p. 67). By contrast, practical reflection isolates the essential first personality of practical thinking:

Stepping back in thought, as an expression of rational freedom, does not commit the agent to take the results of other people’s similar reflections as a datum to be accommodated. Nor, crucially, does it introduce the aim of the ‘harmony of everyone’s deliberations’ – the agent is not forced to become a legislator for others.

Those parallel thoughts do apply to theoretical reflection. But there is a deeper explanation of why this should be so. The constitutive aim of theoretical reflection is truth – and that is not simply because it is a form of reflection (Reference WilliamsWilliams, 1985, p. 69). I should take seriously the perspective of those rational agents who also want the theoretical truth, because it is a constraint on our practice of finding things out that we are all investigating a unitary physical world (Reference ThomasThomas, 2006, p. 93; Reference Moore and ThomasMoore, 2007). The basis of the asymmetry between theoretical and practical reasoning is that because belief constitutively aims at truth, theoretical reflection about beliefs has the same interest merely at one remove – truth. In the case of desire, practical reflection on our conduct need not share the same constitutive aim as the desires expressed by it.

If we mistakenly construe reflection as detachment, as the impartialist does, then Williams claims that the disengaged self would have no reason at all to see any of its desires as objects to be satisfied. They become, as it were, merely data (a point that will prove to be important in Section 5). From the first order, engaged, perspective, I act from my desires; if reflection on them is to detach from them, then there is a radical break between reflection and agency.

The target here is clearly Nagel, whose commitment to the intellectual authority of the impartial point of view evolves, but is never abandoned (Reference NagelNagel, 1970, 1986). This is, I think, clear from what Williams goes on to say next. Suppose reflection forces the practically rational agent to become a legislator: He or she has to accommodate the symmetrical perspective of all other agents and seek a harmony of practical interests. What is left, Williams asks, of the permission to live one’s own life (Reference WilliamsWilliams, 1985, p. 70)? One response is that we are allowed to put those permissions ‘back in’. However, that seems to Williams a bad answer to a bad question. The idea that the individual needs permission to re-insert a concern with her organising ground projects shows that our enquiry has already gone awry.

His challenge to the impartialist is this: If the underlying moral psychology is that of ‘ethical Cartesianism’, as Williams puts it, ‘does it leave anyone in particular for me to be?’ (Reference WilliamsWilliams, 1985, p. 70). In the Procrustean bed of the impartialist, one’s concerns are either fully altruistic or fully egoistic with no space in-between for, for example, Williams’s ground projects interpreted as ‘personal commitments that are not necessarily egoistic but are narrower that those imposed by a universal concern or respect for rights’ (Reference WilliamsWilliams, 1985, p. 70).

2.2 Cognitive Autonomy and ‘Internal’ Reasons Ascriptions

There are often multiple routes to the same conclusion in Williams’s work. The critique of impartialism set out in ELP is a culmination of arguments dating back to ‘Internal and External Reasons’ (Reference WilliamsWilliams, 1981e). Williams there argued that all ascriptions of reasons to an agent must bear what he called an ‘internal’ interpretation. The basic idea is that given that reasons are fundamentally both normative and explanatory, then any reason ascribed to an agent must potentially explain some action of hers in a distinctive way (Reference WilliamsWilliams, 1981e, Reference Williams1995c; Reference ThomasThomas, 2006, chapter 4; Reference FinlayFinlay, 2009).

Add to this Williams’s further thesis that the fundamentality of ground projects implies that all the reasons potentially available to an agent who deliberates soundly are dependent on their deliberative starting point. Combine those claims and we must forgo a particular guarantee integral to the Morality System. That guarantee is that moral reasons must form part of every agent’s reasons – in the expanded sense of that word which Williams calls the agent’s ‘subjective motivational set’.

For the rationalist, altruistic reasons are already in every agent’s subjective motivational set or sound deliberation can put them there. No matter what your starting point, then, it is an a priori truth that morality is everyone’s ground project. Ascriptions of reasons for action to an agent must explain specifically why that agent acts for that reason and, in so doing, explain why they saw that action in a favourable light (Reference WilliamsWilliams, 1981e). We want to contrast correct deliberation with cases of error or action from ignorance. However, ‘good’ cases where reasons both justify and explain an action need share no interesting similarities with the failed cases (Reference Hornsby, Haddock and MacphersonHornsby, 2008). Williams thinks that it is an a priori truism that correct reasons ascriptions also explain why an agent acted as she did (Reference ThomasThomas, 2002, p. 137). This paper also foreshadows Williams’s later concerns given the way in which this comparatively early, and very abstract, discussion of reasons connects to his later critique of the Morality System.

If there is a characteristic Williams’s word it is ‘bluff’. ‘Internal Reasons’ gives the example of Owen Wingrave, the eponymous hero of Benjamin Britten’s opera, who is from a military family. The family insist that Owen has a reason to continue its traditions and join the army. He is equally convinced that he does not. Williams comments that Owen’s family are ‘bluffing’ in a particular way; the family’s charge against Owen is that he is irrational (Reference WilliamsWilliams, 1981e, p. 110).

I think Williams’s paper is best interpreted as a rejoinder to Nagel’s influential book The Possibility of Altruism (Reference NagelNagel, 1970). Nagel argued that the method of transcendental argument can show that every practically rational agent has an a priori interest in acting altruistically.Footnote ⁸ For it to be possible for one person to act directly to promote the interests of any other (the ‘altruism’ of his title), the former must have a conception of herself as one person extended through time, and as one person amongst other people equally real.

This is not the place to dissect the internal complexities of Nagel’s argument (Reference ThomasThomas, 2009, chapter four). But its key features are these: Its commitment to the method of transcendental argument; the ambition of the conclusions it draws (a priori) from the very concept of a practically rational agent; and its defence of the overriding rational authority of moral reasons. They are, as Nagel memorably puts it, reasons from which one cannot ‘beg off’ (Reference NagelNagel, 1970, p. 4).

The Wingraves think that Owen has a moral reason from which he cannot ‘beg off’. It is not merely a good reason for Owen: His family argue that it is one of Owen’s reasons ‘already’ (without the temporal implication). That makes sense if Nagel is right – If moral reasons are inescapably present in everyone’s subjective motivational set. Furthermore, and this is the point that interests Williams, were Nagel correct then the Wingrave family could substantiate their charge of irrationality against Owen.

Williams inverts this argument: Your starting point in deliberation determines where sound practical deliberation can take you. So, there is no guarantee of convergence on moral reasons as a presupposition of everyone’s ground project. From Williams’s perspective, Nagel’s neo-Kantian style of argument creates a gap between two ideas: that of reasons for an agent and an agent’s reasons. Ironically, at this point, Williams claims that Kant is on his side in this argument: The latter reciprocally stipulates an account of practically rational agency and an account of reasons for an agent such that the two ideas – reasons for an agent and an agent’s reasons – cannot possibly come apart (Reference Williams, Altham and HarrisonWilliams, 1995k, p. 220, fn. 3). For Williams the question is: Why accept Kant’s stipulation?

I have already noted the deeper background issue: the fate of psychologism. Nagel maintains that the science of psychology can, in part, be conducted a priori: Human psychology must be aligned with an order of reasons (Reference NagelNagel, 1997). By contrast, Williams thinks that the psychology of human action is not trying to align itself with a metaphysics of reasons. That is why, for Williams, the direction of explanation runs from an agent’s reasons to the idea of reasons for an agent.

On some issues Williams and Nagel share common ground; on others they are far apart. The basic picture of action in The Possibility of Altruism is Aristotelian: The idea of an action explanation that did not mention a ‘desire state’ is, for Nagel, absurd. The distinction he wants to draw is between chains of practical reasoning that begin from a desire, and others that begin from the acceptance of the truth of a proposition that, in turn, motivates both an action and the desire to perform that action. This distinction between unmotivated and motivated desires is one of the linchpins of the book and Williams (as a fellow Aristotelian on this point) has no disagreement with it.

However, this threatens to make Williams’s critique of external reasons become elusive. If, by coming to believe that she has a reason, an agent can thereby acquire that which Nagel calls a ‘consequentially ascribed’ motive – then does not that claim automatically make all reasons ascriptions ‘internal’ in Williams’s sense? If a person lacks a desire, then by coming to accept a truth, she acquires one anyway and hence also the reasons that this desire makes available.

This point helps to identify Williams’s challenge: What is it to come to accept the truth of a proposition about action in the first place (Reference WilliamsWilliams, 1981e, p. 108)? For Nagel, you write the theory of reasons for an agent first, and then the theory of an agent’s reasons follows from it: That is the form taken by his radical anti-psychologism. For Williams, the direction of explanation is reversed: We begin from the idea of an agent’s subjective motivational set and enrich its contents by sound practical deliberation (including our use of the imagination). We have no a priori guarantee that all agents will converge on what Nagel presents as the truth of basic moral principles guaranteed to be in everyone’s subjective motivational set.

This focus on an agent’s capacities follows from Williams’s thesis that all reason statements true of an agent must be potentially explanatory of that which an agent may come to do after sound deliberation (Reference ThomasThomas, 2006, chapter four; Reference FinlayFinlay, 2009). The deeper issue is Kant’s democratisation of morality such that the requirements that moral truths place on agents’ psychological capacities are undemanding. A psychologically realistic account of agency shows us, in Williams’s view, that Kant’s optimism about the universal accessibility of moral motivation is false (Reference Skorupski, Thomas and WilliamsSkorupski, 2007). Williams’s argument is a normative one, grounded on a view of sound practical deliberation. An apparently abstruse discussion of the nature of reasons leads directly, in his view, to a diagnosis of the Morality System’s errors in its treatment of blame.

2.3 Morality and the Social Function of Blame

Williams argues that the Morality System is unhelpfully obsessed with blame. For him, it is just one ethical reaction we can have towards others amongst many possible reactions, one where we acquire our understanding of what it is to be blamed via the experience of being unjustly blamed (Reference WilliamsWilliams, 1995a, p. 16):

The Wingrave family blamed Owen for being irrationally at fault – for not having a reason he ought to have. But, according to the internal reasons thesis, Owen had no such reason; his family were bluffing. He was being confronted by a socially sustained fiction: His family are attempting to ‘recruit’ him into ‘their’ deliberative community:

This ‘proleptic’ theory of blame, as Williams calls it, invokes the contrast between the first person and the social point of view (Reference WilliamsWilliams, 1995c).Footnote ⁹ If this institutionalised practice of blame is regarded merely as a ‘device’ – wholly third personally – we would be alienated from it as merely a form of social control. But it appeals to our reasons in just the way that Kant thought that we must in order to cultivate our characters as effective instruments of the moral law. The proleptic theory appeals to ‘the desire to be respected by people whom, in turn, one respects’ (Reference WilliamsWilliams, 1995c, p. 41). This mechanism works, then, only if we have other ethical dispositions and motivations beyond blame as the Morality System understands it.

Given that there literally cannot be any external reasons in Williams’s sense, in our actual practice we must be witnessing the deployment of what he called ‘optimistic’ internal reasons statements with this feature: They seek to make it true that people ‘have’ the reasons that we take as the ground for blaming them (Williams, 1981, p. 111; Reference WilliamsWilliams, 1995c, p. 40; Reference Thomas, Himmelmann and LoudenThomas, 2015, p. 254). The proleptic theory describes how this mechanism operates: A person who lacks an appropriate reason to be blamed in a ‘focused’ way nevertheless has this other desire-cum-disposition, which invokes respect. Blaming triggers the mechanism such that the agent has now made it true that she has reason to avoid the action that would lead those others (whom she respects) to blame her. If this mechanism succeeds, the recruit sees the process of conversion as the acquisition of a new set of reasons:

This appeal to a proleptic mechanism, importantly, does not restore Kantian universalism about blame; it is a contingent question of whether people have this ‘ethically important disposition’ or not (Reference WilliamsWilliams, 1995c, p. 41; Reference Thomas, Himmelmann and LoudenThomas, 2015, p. 254). Thus, Kant is once again key to this discussion: He thinks we directly experience ‘a conclusion of practical necessity’ in which motivation is given by reason alone that need not involve any desire (Reference WilliamsWilliams, 1995a, p. 17). Hence, for Kant, true freedom is action done from duty – the guise in which the moral law presents itself to beings with our finite nature.

Williams certainly does not deny the existence of an agent’s particular experience of acting from practical necessity. His disagreement is with Kant’s explanation of it:

Kant’s claims about necessity ground his claims about universality, and it is the latter that is the focus of Williams’s doubts.

As Williams notes in his account of the Morality System’s attempt to make morality luck free that aspiration appears in ancient Stoicism. But the Stoics never doubted that it was a matter of constitutive luck whether or not one could be a moral agent in the first place. For Kant, that deeper issue of luck must be resolved such that morality is luck free ‘all the way down’. Yet this must also be true of people for whom it cannot also be a matter of luck that they have the capacities to respond to those demands at all. The next section continues Williams’s diagnosis of how, in the hands of the rationalist impartialist, the attempt to ‘purify’ morality distorts both the ideas of immunity from luck and the nature of voluntary action.

3.1 Existential Decision and the Grounds of Morality

Suppose you are a wealthy person – a successful trader, for example. But it is not how you see yourself; you see yourself as a creative artist. However, you also chafe within what seems to you a competitive and claustrophobic artworld. You both desperately want peer approval but think that your art will flourish only away from those people – (provided enough people in your destination speak French). This is a thumbnail sketch of the nineteenth-century French artist Paul Gaugin. Footnote ¹⁰ He is the subject of Williams’s famous example of an individual faced with a life-defining choice between a structuring identity and the demands of morality (Reference WilliamsWilliams, 1981c).

This example was Williams’s vehicle for exploring the tension between the importance of ground projects and the Morality System’s aspiration to make morality luck free. Williams thinks that, to buy plausibility, the thesis that morality is immune to luck must narrow its scope. Paradigmatically, Kant restricts it to the thesis that the luck-free part of morality is ‘unconditioned’ moral value: When it comes to character, all that counts is motive; and when it comes to action, all that counts is intention (Reference WilliamsWilliams, 1981c, pp. 20–21). There is one element of constitutive luck that the ancient antecedents of the view included, but which Kant excluded, namely, that there is no luck in being able to be a moral agent at all. In that sense, as Williams notes, the Kantian refinement of the ancient view develops it into a view of the world as fundamentally just: Pelagianism, once more.

One reason why the Morality System is deeply appealing is that if your aim is to make your life immune from bad luck, and morality is exempt from luck, then you have a strong motivation to live a moral life (Reference WilliamsWilliams, 1981c, p. 21):

There is a tension between the Nagelian claim that moral reasons are already in everyone’s motivations and the view that morality offers an inducement to agents to embrace it and thereby liberate themselves from luck. We are all so Kantian that the phrase ‘moral luck’ does indeed sound like an oxymoron, but that is puzzling given that, as Williams also puts it, the goal of making morality luck free ‘is bound to be disappointed’ (ibid).

The distinction between the ethical, broadly conceived, and the Morality System narrowly conceived makes Williams’s thesis about this inevitable disappointment expressible: Ethics, in the broad sense, is more honest about its relationship to luck than the Morality System (Reference WilliamsWilliams, 1995j, p. 241). Furthermore, we can find this honesty in our own tradition because Aristotle’s ethics exemplifies it (ibid, p. 241).

For Williams, it is a ‘bitter truth’ that ethics is subject to constitutive luck (ibid). However, that is not his primary concern when he discusses moral luck: His focus is Kant’s view of how immunity from luck impacts an agent’s understanding of her own actions. Kant guarantees that ‘at the ultimate and most important level, it cannot be a matter of luck’ whether a person was justified in acting in the way she did (ibid). Why does Williams think that this guarantee is unavailable? The several examples in Williams’s ‘Moral Luck’ paper constitute his explanation. He begins with a prior question: For which human activities do we have reason to be grateful even when their exercise imposes costs on others?

This is a characteristic diagnosis: The asymmetric dependence of commitments of the Morality System on a broader conception of the ethical on which they depend, but can only, in its own terms, repudiate. The Morality System, far from being a rational purification of the ethical (as it pretends to be) exists in this unstable relationship to it.

Williams appeals to the point repeatedly: For example, he thinks that the Morality System embraces a very strong claim to intellectual self-sufficiency. It has a self-conception as inevitably higher or noble (such that it could not possibly be explained in terms of ‘lower’ values). Williams thinks both that this claim is false, and that this point further explains why Nietzschean genealogy is such an effective weapon to deploy against the Morality System.

Nevertheless, if Williams’s reflections force us to give up on something – for example, morality’s immunity to luck – we will experience that as a loss and not merely an intellectual one. Our starting point can be only the ‘feelings and dispositions of judgement’ that we, as historically situated selves, actually have: We are strongly motivated by the thought that justice requires us to ‘exempt agents from (some) blame for (some) things done involuntarily’ (Reference WilliamsWilliams, 1995j, p. 243). The Morality System attaches considerable force to the observation that even if one is ‘shunned, hated, unloved, and despised’ as Williams puts it, at least one is not being blamed. Williams does not think that constitutes much of a reassurance.

Williams’s answer to the question of why blame matters so much to us, and is so hegemonic over our other ethical reactions, appeals to its fundamentally universalising and democratising treatment in Kant’s hands. But Kant’s aspirations are ours as well: Williams argues that it is an attractive political project to remodel our social world as if Pelagianism were true. Blame, in this conception, focuses on ‘what is available to the rational deliberator’ and operates under the guarantee that ‘the correct ethical demands are …. available to any rational deliberator’ (Reference WilliamsWilliams, 1995j, p. 246). In that sense the Morality System ‘reassures’ us. But Williams thinks that this reassurance is bogus.

Williams’s critique begins from the phenomenology of how people think and feel about some ‘usual’ cases and extrapolates from that to some more dramatic, and unusual cases. In both kinds of case, these examples are freed from prior theory that dictates how we ought to interpret them. In other words, we are not as Kantian as neo-Kantian moral philosophers think we are or, perhaps, hope we are.

The point of the ‘Gauguin’ case, then, is to represent any person with ‘definite and pressing’ moral claims in his or her life who ‘turns away’ from them to pursue a ground project that seeks the genuinely worthwhile. ‘Turns away’, and not ‘ignores’: If this Paul Gauguin simply did not care that he was married, had children, and familial obligations, it would not be the kind of case that interests Williams. That would distort the example in line with the Morality System’s insistence that the only two kinds of motivation are altruistic or egoistic. Gauguin the rampant egomaniac would not be a very interesting character. (Reference WilliamsWilliams, 1995j, p. 244).

The conditions of Gauguin’s success are clear, but ‘whether he will succeed cannot, in the nature of the case, be foreseen’ (ibid, p. 23). You might describe the choice as risky:

Suppose we take ethics to have two core concerns: One’s relation to oneself and the expression of one’s values, and a general disposition to be able to explain and justify one’s actions to others. For this fundamental project, Williams does not think that the second aim can be met. Even if Gauguin made the right choice, and was ‘ultimately’ justified, he could not have justified himself to others (‘or at least to all others’). Those who were hurt by his choices will still have grounds for reproaching him even if he succeeds in terms internal to the project. Failure means that there is no justification for his choice; yet success does not let him off the hook, either.

Williams argues that this example threatens a general model of how we reason about practice (Reference WilliamsWilliams, 1995j, p. 245). There is ex ante deliberation, then the problem of how, ex post, an agent may change their view of themselves in ways that also changes their view of that original deliberation. The question then is not, simply, how one might have done better: ‘[T]he question I want to press comes before that reflection’ (Reference WilliamsWilliams, 1995j, p. 245).

It is only in so far as one can control an outcome via one’s deliberation that the agent can be held to the standards of rationality. If an outcome emerges that is not responsive to that, then from the standpoint of the Morality System it might as well be the product of the agency of another person – or akin to a natural event (Williams, ibid). Williams thinks that this restriction is ‘importantly wrong’ and a person’s ‘involvement’ in an action goes beyond the quality of his or her ex ante deliberation. If the question is a completely general one about practical agency, then blame is key again, because the Morality System treats guilt as self-directed blame, which is restricted to ‘the rational self-criticism of a deliberator’ (Reference WilliamsWilliams, 1995j, p. 246).

That does not match up to our ordinary ethical experience. In particular, it falsifies our ethical emotions: We criticise those whom we take to have ‘the wrong ethical sentiments or none’ (Reference WilliamsWilliams, 1995j, p. 246). Morality tells us to discipline those reactions to align them strictly with the scope of blame. Blame is restricted to those not merely at fault, but at fault for being at fault: The reason on which they failed to act is supposedly in everyone’s ‘subjective motivational set’. The Morality System pressurises the very idea of agency to re-shape it as luck free – as ‘escap[-ing] as far as possible from contingency’ (Reference WilliamsWilliams, 1995j, p. 246). We are, once again, identifying why the Morality System is obsessed with blame and cannot make sense of agent regret.

Williams thinks that his Gauguin example demonstrates that a person’s projects may unfold through time making justification available only retrospectively. At the time of deliberation and decision, the justificatory reasons lie in the future. What counts as failing to achieve a given outcome has criteria for success and failure that are internal to the project concerned. It makes a difference, in Williams’s view, if there is an outcome which contains no good art by Gauguin because he tried and failed to produce it, or because he sustained an injury which meant he never tried to paint at all (Reference WilliamsWilliams, 1981c, p. 25).

This is another aspect of the example that stresses the importance of agent regret: Williams argues that failure, in cases such as these, is something that the agent has this special reason to regret in a way that marks off this kind of regret (Reference WilliamsWilliams, 1981c, p. 27). A person can feel such regret ‘only towards his own past actions (or, at most, actions in which he regards himself as a participant)’ (ibid). First personally, things could have been different, because you could have acted differently. You can, indeed, regret things done in the past and that would put you in the same position as a third-personal observer of your life. But this is not agent regret, which involves the first person in how it is distinctively expressed.

The scope of agent regret goes beyond what one intends to the outcomes that one brings about as a result of what one intends. Even in the case of an accident, agents can – and we expect them to – have ethical sentiments that reflect their causal involvement in producing that outcome. A natural impulse is to want to make amends, but even here this general desire does not reflect one’s first-personal involvement in the right way. Williams’s test is whether or not an insurance payout would cover the case. If not, then there is some ‘reparative significance’ to trying to make amends. In any case, he takes the existence of agent regret to be a phenomenological insight into broader issues about the structure of freedom and responsibility.

3.2 Agent Regret and the Shape of a Life

Human agency reaches full rational expression in that which one’s actions bring about; exposure to luck, and the possibility of agent-regret, both reflect that underlying fact:

Williams thinks that the Kantian attempt to ‘purify’ the concept has demonstrably failed. That leaves the second option: a consciously ‘superficial’ account of responsible agency. One of Williams’s strongest motivations here is to make sense of ancient tragedy: who you are is shaped by what you have done – ‘what in the world one is responsible for’ (ibid, p. 30).

In the case of ancient tragedy, people are victims of tragic fate: By contrast, Gauguin (and Anna Karenina in the ‘Moral Luck’ paper) make voluntary decisions. That which unites these examples is that how things turned out, something beyond the agent’s control, determines whether the agent ought to regret them. Thus, Williams denies the appropriateness of regimenting the ethical sentiments so that regret is the canonical third-personal emotion, remorse the canonical first-personal emotion, where the latter is appropriate only for what one has done voluntarily (Reference WilliamsWilliams, 1981c, p. 30). It is the Morality System which insists on this regimentation.

In the case of moral conflicts, Williams argues that there can be rational regret for the action not carried out even if what one did was for the best – another denial of the Morality System’s thesis that obligations cannot ultimately conflict. Regretting how things turned out is not the same as regretting one’s action: In the conflict case, there is no impugning what one did, simply that it was impossible to respond to two conflicting demands. The same distinction applies to agent regret: You can feel agent regret towards an outcome and one’s role in bringing it about without regretting one’s action.

All this phenomenological evidence informs Williams’s conception of the shape of a life. Lives, as temporally extended sequences of choices, are poorly modelled via an idealized sequence of rational choices each of which is justified at the time of decision. Williams argues that, on this view, regret about past cases, where bad decisions turned out well or good decisions turned out badly, can only be instances where the agent resolves to deliberate better in the future. This makes sense, Williams notes, if one is developing a policy for ‘an ongoing class of cases’ – for example, as a blackjack player does. But the sequencing matters: The goal is not to formulate a policy for every hand of cards in one’s future. That is why Williams thinks it is an extra, optional, step to develop a conception of a temporally neutral rational life-plan, as in Rawls’s work, which involves discounting only for distant probabilities (Reference RawlsRawls, 1971/1999, p. 358).

Williams’s fundamental objection to this model is that it overlooks the idea that one’s life has a shape: It instead treats oneself as a trustee for one’s future self, hoping for an outcome where one’s future self cannot rationally reproach one’s earlier self (Reference WilliamsWilliams, 1981c, p. 34). This ignores the insight that the shape of a life is determined by the fact that ‘what one does and the sort of life one leads conditions one’s later desires and judgements’ (ibid.). The life choices of a Gauguin represent, for Williams, a dramatic instance of something he takes to be ubiquitous in human life:

In the ordinary conduct of life, avoiding regret is the same thing as resolving to take good decisions. By contrast, Williams identifies in these existential cases an asymmetry:

Such decisions are not simply very risky. They are focused on conditions of meaningfulness: Such fundamental projects include, amongst their ramifications, crucially the ‘standpoint of retrospective assessment’ itself (Reference WilliamsWilliams, 1981c, p. 36). The agent, in taking her decision, adopted a project that aimed to ground her life as a whole; it follows that intrinsic failure here leads to these ‘most basic’ forms of regret (ibid, p. 36). How do these existential decisions threaten the claims of morality? There are several possible views here; I think it is helpful to identify which one is Williams’s.

At a first pass, it seems that Williams is arguing that morality is less important in our lives than a broader set of claims that he identifies as those of ‘the ethical’. That, however, is a multiply ambiguous claim because of the ambiguities in the word ‘importance’. We might, for example, reduce the scope of moral considerations, giving them importance in that limited domain, while giving them less overall importance in a person’s life. That is not, I think, Williams’s view. I think the best interpretation of what he says is that – once again – there is an asymmetric dependence at work where not only is the scope of morality reduced, but it has the importance it has only because it depends on ethical considerations that are prior to it.

All of these arguments are underpinned by Williams’s claim that the Morality System lives a hypocritical double life. Morality, on this view, can never make anything less than a claim to total supremacy over our ethical reasons (Reference WilliamsWilliams, 1981c, p. 37). It also insists that the only reasons in opposition to its own are egoistic, or amoral (ibid, p. 38). That is why Williams complains that it is an inherently hegemonic idea that must, of its nature, misrepresent the complex phenomenology of our actual ethical experience. It is part of our ordinary experience that our lives have a ‘shape’, in William’s sense, in a way that the Morality System fails to acknowledge.

3.3 Freedom of the Will

A corollary of the foregoing is that Williams thinks that we need to deflate the traditional problem of free will. It is the Morality System that freights our ordinary concept of voluntariness with a weight it cannot carry. It must be, from its perspective, a deep idea – a focal point for the absolute justice of Kant’s implicit Pelagianism. However, when we turn to our ordinary conception of agency, we do not find a deep idea of human freedom; rather, a ‘shallow’ one.

Williams does not deny that it is a constitutive fact about an action that it be voluntary in the sense of ‘an intended aspect of something done in a state of mind that is deliberately normal’ (Reference WilliamsWilliams, 1995j, p. 243). But this characterisation is ‘essentially superficial’ because the work is being done by the contrasting idea of the involuntary. It is the latter that the Morality System wants to ‘deepen’, because some things done involuntarily get a person off the hook from blame (partly or wholly) – again, in the interests of justice (ibid, p. 243).

Mature ethical agency involves this thought: A person ‘recognizes that his identity as an agent is constituted by more than his deliberative self’ (Reference 72WilliamsWilliams, 1995b, p. 32). The relevant connection is between one’s character and what one has brought about through one’s action in a way that extends to the ‘undeliberated’, ‘unforeseen’, and ‘unintended’ aspects of action (Reference 72WilliamsWilliams, 1995b, p. 32). By acknowledging these aspects as one’s own, the agent responsibly acknowledges that some things she did were not what she intended. More ethically fundamental than the social mechanisms that hold one accountable is holding oneself responsible for these actions in a way one must acknowledge. That is a broader, more demanding ideal than that of being held responsible by some mechanism of public accountability.

Those public mechanisms have the aims of social control and cohesion (Reference 72WilliamsWilliams, 1995b, p. 32). It is an evasion of responsibility if an ethically mature person holds themselves responsible only to those reductive public standards. Given his later sympathy with political republicanism, it is interesting that Williams does not think that his ideal of ethical maturity coincides with the highest form of political citizenship as collective self-authorship of the laws of the republic (Reference WilliamsWilliams, 2002, chapter nine). A liberal politics requires only acknowledgement in a weaker sense: of holding each other to publicly known standards.

Beyond the political, and the public, we have the domain of responsibility for character within our personal relations to one another and the extent to which it can be shaped by reason. Here Williams appeals to a distinction between a moralising psychology versus a minimal one distinguishing how each treats the idea of freedom. Aristotle thought that a person who was unfree, for example, because of self-indulgence could be held responsible for the voluntary steps that led, cumulatively, to a loss of voluntariness. Bad character is not, Williams retorts, like addiction. In any case voluntary steps which, iterated, lead to addiction do not make the latter voluntary (Reference 72WilliamsWilliams, 1995b, p. 27). Local control, however, Williams thinks is psychologically realistic. An individual’s concern with ‘mak[-ing] sense of his or her life’ can be developed in the direction of Williams’s own idea of the ethically mature agent.

Does that mean, then, that we are ultimately free? Williams thinks that is a bad question not least because it is framed, as the Morality System frames it, in ‘all or nothing’ terms. Our ordinary understanding is more nuanced than that: It typically treats freedom – rightly in Williams’s view – as a matter of degree (Reference WilliamsWilliams, 1995a, p. 3). Our ordinary concept of freedom seems, in that respect, closer to the philosophical view known as compatibilism.

Compatibilism thinks that free action is compatible with your actions being caused, in the right kind of way – caused by your beliefs, desires and character. Williams argues that any compatibilist view must also accept that freedom comes in degrees because that which vitiates it is constraint via ‘hostile agency’. In such cases person A threatens, or exerts force, on person B such that the latter cannot achieve her aim in action or does so at increased cost. Consider Aristotle’s example of the sea captain who must throw a valuable cargo overboard in a storm or risk losing the ship. Is the captain responsible for his choice? The agent can, after all, choose and will be held responsible for this decision. Williams concludes we do not learn very much from such cases precisely because there is no loss of free will.

For him, the interesting point is that action under constraint is related in a special way to the conclusion that one must or has to do a certain thing (ibid). That is because he believes that any agent wants to protect the core identifications from which expressions of practical necessity flow. Their form expresses ‘some of the most serious and responsible decisions we take’ (Reference WilliamsWilliams, 1995a, p. 5). That is why we are especially concerned when someone tries to substitute their intentions for our own: Being under someone else’s power seems like a paradigm of unfreedom. (But we are still making decisions.)

Williams argues that our problem is not whether compatibilism and moral responsibility can coexist: Our problem is to understand whether choice is, as Williams puts it, a psychological or metaphysical notion. The point of the philosophical challenge of determinism is that it seems to make choice situations a sham: It cuts down the options available to the agent to one. Williams thinks that we do not need the dramatic threat of universal determinism to make the point. All we need is the idea that human actions can be fully explained by a strong form of philosophical naturalism. What range of actions presents themselves to the agent as deliberative alternatives that are open to him or her? Williams thinks that a generally naturalistic account of human action allows us to retain the basic range of the concepts we need. That fact, Williams avers, does not yet solve the real problem: a reconciliation of this range of concepts with our ethical concepts.

The crux of the issue, then, is the relation between our ordinary notion of responsibility and the philosophically inflected concept of blame re-shaped (for its own ends) by the Morality System. This philosophical conception of blame gives it, essentially, a forward-looking function – it must modify an agent’s motives. As Williams notes, this is a curious combination of an intrinsic justification and an instrumentalist one – we value the practice of blaming for its own sake and for its overall point.

His first complaint is that we sell blame short by explaining it via its efficacy. Such an explanation would be inadequate (for legal philosopher Herbert Hart’s reason) that it is only effective if the person blamed thinks that, qua reaction, is merited. If not, the agent merely resents being blamed.

Morality exaggerates the importance of blame and thereby makes it too central to ethical life. In particular, the Morality System has difficulty explaining why practical necessities are so important to us. It delivers us straight into this paradox: Accounts of free action, influenced by the moral psychology of blame, stress the condition that the agent ‘could have done otherwise’. But this is precisely a condition absent from some of the actions most central to our practical identity and with which we are most deeply identified. What is distinctive about cases of practical necessity, when they stem from our deepest identifications, is that we do not want to have done otherwise. There is a general pattern to such claims about necessity:

Williams argues that if a person rightly concludes that he must do a certain thing then it is impossible, not that he should fail to do it, but that he should intentionally fail to do it. The word ‘necessity’, occurring here, expresses the agent’s intention.

It is not the agent, third personally as it were, reporting some antecedent truth that limited her powers to decide. If that were the correct explanation, then the agent could not do the thing in question either intentionally or unintentionally.

In ordinary speech, we might say ‘I am going to do x’ which is tantamount to saying that if I deliberate, and come to a conclusion, then I can reasonably count on that bringing about a change in the world of which my agency is the cause. By contrast, Williams thinks, when I say ‘I cannot do x’ then there is a different implication, namely:

Conclusions of practical necessity, then, express some of the deepest ethical demands on us, yet we do not insist, in these cases, on the condition for free action on which the Morality System most strongly insists.

Williams’s overall diagnosis, then, is that nothing in our ordinary idea of freedom is incompatible with the idea that there is a naturalistic explanation of human action. The real problem is the incompatibility between the pressure that the Morality System puts on our psychological concepts and our ordinary notion of free agency. The important issue is neither naturalism, nor determinism, but the moralisation of basic psychological concepts such as choice.

Freed from the pressures generated by the Morality System, Williams’s conclusion about freedom is optimistic:

Given the centrality of character and its expression by action in Williams’s overall conception, it is unsurprising that necessities grounded in character are of particular interest to him. I will turn next to the connections that he draws not between that idea and freedom, but between practical necessity and the idea of certain (in)capacities of character that are also revealing aspects of our ethical life.

3.4 Practical Necessity and Incapacities of Character

Williams thinks that it is part of our everyday ethical experience that an individual may conclude that ‘a certain action is one that he must, or has to, do’ (Reference WilliamsWilliams, 1981f, p. 124). Williams was here influenced by a paper he admired, namely, Peter Winch’s ‘The Universalisability of Moral Judgement’ (Reference WinchWinch, 1965). While ‘ought’ is a comparatively weak modality it can be used to express a practical conclusion – in Philippa Foot’s helpful terminology, the ‘verdictive’ use of that term expresses an all things considered conclusion to practical deliberation (Reference FootFoot, 1978).

Williams notes that it is not usually true that in successful cases of deliberation, there is only one thing I can do to realise my goal. (Even though alternatives may be more costly or violate some deontic constraint.) However, when they are all ruled out there is one thing I must do. He argues that there are ‘serious cases’ where the concept of necessity applies to our assumptions from which we deliberate: Some alternatives are ‘impossible’ for us. This fact can play an important role in deliberation. The deliberator can either express her conclusion as a practical necessity: ‘I have to do x’ (thus the alternatives are things she cannot do) or as a moral incapacity: ‘I cannot do y or z, hence I must do x’.

This analysis allows Williams to isolate the case that interests him: This special sense of ‘cannot’: ‘The cannot of practical necessity introduces a certain kind of incapacity’ (Reference WilliamsWilliams, 1981f, p. 128). The deliberator can recognise that he or she cannot do a certain thing, but the observer of that person can go a step further and acknowledge the category of the unthinkable:

It does not follow that, if an agent is incapable of doing x in this sense, then she will not do it – but she will not do it intentionally. She might do it unintentionally – for example, through ignorance. These are incapacities of character, not literally what you physically cannot do. Physical incapacity is different; if I cannot lift a ton there is no event in the world that is my lifting a ton. Other psychological incapacities are being unable to do something if I know I am – such as walking on a tightrope over a steep drop. The world might contain this event but not if I know what I am doing:

The necessary asymmetry of most virtue terms means that we, the agent’s community of appraisers, usually apply the virtue term to the agent whose ethical attention is outwards, on the practical question to hand.

But we, as appraisers, can also come to a judgement of your character based on what it occurred to you to do or that which was unthinkable for you. This is, for Williams, an important source of self-knowledge:

Winch focused on Melville’s novella, Billy Budd, and the judgement of the fictional Napoleonic sea captain ‘Starry’ Vere that he could live with the unjust condemnation of the saintly Billy Budd. It was a matter of military necessity that Vere sentence Budd to death for a killing that was not intentional. But, as it turned out, Vere found out via his decision that this was precisely something he couldn’t live with: He dies, guilt-stricken, with Budd’s name on his lips. A decision with which one is identified is one for which a person will be held responsible. In acknowledging that fact, the agent makes the action ‘one’s own’ (Reference WilliamsWilliams, 1981f, p. 130). Vere gets this decision horribly wrong; yet Williams also emphasises that this form of self-discovery is not in spite of, but because of, the agent’s ‘outward’ focus on getting an ethical conclusion right.

Vere was not obsessing about his character but with what he had to do; his flawed deliberation led to his discovering something about himself he could not have found out any other way. Williams thinks that Winch identifies something important:

This early paper foreshadowed the contents of Shame and Necessity, published in 1993. This book is an extended argument to the effect that, given that the ancient tragedians and ethicists can make better sense of such ethical experiences of practical necessity and the continued relevance of shame, there is every reason to reject a certain historiography of our own ethics. That claim – that ancient ethical thought was comparatively ‘underdeveloped’ compared to the modern moral consciousness – is yet another manifestation of the Morality System.

3.5 The Recovery of Ancient Ethics

Shame and Necessity is Williams’s most extended act of historical retrieval. He begins by noting the use of the methods of cultural anthropology to make ancient Greek culture more distant from us; he concedes that we cannot share their world. But his study aims to uncover ‘unacknowledged similarities between some Greek conceptions and our own’ (Reference WilliamsWilliams, 1993, p. 2). Our ethical modernity, far from affording us a superior standpoint on basic concepts of moral psychology, rather blinds us to these similarities (Reference WilliamsWilliams, 1993, p. 3).

He thinks that we have sufficient grounds for rejecting the progressivist view that modern ethical conceptions stand to ancient Greek conceptions as sophisticated developments out of the comparatively ‘primitive’. Williams openly acknowledges the difficulty of placing the Sophoclean tragedies at the centre of his discussion: We share neither a Sophoclean conception of the gods, nor of fate. His response is that ‘we should look for analogies in our experience and our sense of the world in the necessities that they express’ (Reference WilliamsWilliams, 1993, p. 19).

Representative in Williams’s discussion is the progressivist treatment of the work of Homer as representing an unsophisticated moral consciousness. Williams responds to the progressivist complaint that there is something ‘missing’ from ‘Homer’s man’:

For Williams it is an insight, not a defect, of these Homeric texts that they do not involve a basic moralisation of psychology – the colonisation of explanatory categories of action by ethical ideas. That is, as we have seen, for Williams one of the primary distortions of our ethical thought by the Morality System.

On the contrary, ancient ethics makes better sense of the very facts to which Williams repeatedly draws attention in his critique of how that System distorts the concepts and categories of our moral psychology: ‘There is not, and there could never be, just one appropriate way of adjusting these elements to one another’ (Reference WilliamsWilliams, 1993, p. 55). The concept of responsibility applies whenever an agent is a cause of an outcome. Williams begins his explanation of responsibility by setting aside cases of strict liability in law: For example, when an employer is held liable for the faults caused by employees. This is a special case where to undertake to do business in this way is to accept this prior and very general presumption of responsibility. For all other cases, we hold responsible only the agent who has caused the outcome.

Our modern conception of law, Williams thinks, plays a role in falsely legitimising the Morality System and its proprietary account of moral responsibility: To a considerable extent, he argues, the idea that the Greeks thought very differently from ourselves about responsibility, and more ‘primitively’, is an illusion generated by thinking only about the criminal law and forgetting the law of torts (Reference WilliamsWilliams, 1993, p. 63).

We do argue, he notes, about harm or damage brought about by what was unintentionally done. The focus is not on fundamental conditions on responsibility; instead, the focus is on that which the agent’s unintentional action caused – and that can simply be a matter of luck. The relevant difference between ancient Greek practice and our own does not depend on some deeper insight we have into moral responsibility but, Williams claims, on our different conception of the role of the state and a different conception of law (Reference WilliamsWilliams, 1993, p. 65).

In particular, we have a political theory of freedom that seeks to preserve an individual’s ‘control over his or her life’, as Williams puts it, that shapes a liberal conception of law. Our notion of responsibility is largely determined by our ideas of what it is for an individual to be treated fairly by the state. By contrast, the focus of ancient tragedy is not on the demands placed on an agent by impersonal third parties; rather, the demands placed on the agent by him or herself:

Williams takes such examples to prove the importance, once more, of agent regret. In our modern world, as Williams notes, if you are affluent then your unintentional harm to others may be insured against (Reference WilliamsWilliams, 1993, p. 70). But that need not exhaust their moral claims on you. He applies his criterial test again: Is an insurance payout enough? You may wreck your own life by things brought about unintentionally, of which you are the cause, where you simply accept that tragic fact (Reference WilliamsWilliams, 1993, p. 70).

Ancient tragedy also gives us insight into the predicament of people who cannot live with what they have done. This is true even of those actions performed ‘when not of sound mind’ – even though in the plays Williams discusses it is divine intervention that causes the agents not to know what they are doing. Ajax, intending to take revenge for a wrong, is deceived by the gods into killing not his human enemies but a flock of sheep. Returned to ‘sound mind’ he is humiliated and shamed by what he has done and cannot live with it. Given his social identity and its need for acknowledgement in the eyes of others, he cannot go on after having carried out the deed, whether he acted intentionally or not (Reference WilliamsWilliams, 1993, pp. 73–74). Motivated by anticipated shame, Ajax is not, Williams emphasizes, in the grip of ‘heteronomous motivations’ as the Morality System would classify them. Ajax fears shame in the eyes of another who is ‘an internalized other’ (Reference WilliamsWilliams, 1993, p. 84). Ajax’s reactions are culturally shared, in that other people would accept the same reason to be ashamed in their own life that Ajax perceives in his. But his life is different: He is a heroic warrior. It is that fact which makes his prospective shame more than he is prepared to live with.

Ajax concludes that he must end his life but, as Williams notes, he might equally think he must not: ‘this is a type of ethical thought as far removed as may be from the concerns of obligation’ (Reference WilliamsWilliams, 1993, p. 74). Ancient tragic heroes, Williams concludes, can exemplify the point that

Ajax’s conclusion is not a categorical imperative of moral obligation; nor is it Kant’s only other option, the hypothetical imperative that Ajax can ‘go as he must go’ if he wants to do so. The Kantian insists that this desire must be presupposed, somewhere, and this deep presupposition explains why, for the progressivist, the ancient philosophers and tragedians are not yet in the domain of modern morality.

Guilt is closer to the concerns of morality; Williams argues that shame and guilt are distinct. The person who is ashamed feels, as a whole person, exposed to the gaze of another and wants to disappear. Guilt, however, seems more pervasive. A person acts, or fails to act, in a way that incurs the anger and indignation of others, and the agent must make reparations or fear punishment. Shame elicits derision or contempt or avoidance in a way that diminishes an agent’s self-respect (Reference WilliamsWilliams, 1993, pp. 89–90). These moral emotions, then, seem to be filling complementary roles.

By one’s actions, one can reasonably incur both shame and guilt; the deed can have ‘authority over our own feelings’ even in the case of the involuntary (Reference WilliamsWilliams, 1993, p. 93). Agent regret can take the form of guilt. Shame, on the other hand, leads the agent to reflect on her moral identity: Who she is and what standards she seeks to live by. While the Morality System invites us to examine the relations between guilt and the voluntary, our ordinary ethical thinking takes it that the territory of shame is the aspect of our character that can lead to involuntary harms. For example, it is shameful to be so careless while driving that one involuntarily causes an accident. While guilt focuses on the victims of one’s actions and reparations to them, the moral repair of our social relationship with them depends on shame and its connection to the agent’s conception of herself as she relates to others (Reference WilliamsWilliams, 1993, p. 94).

This contrast allows Williams to offer the following deracinated conception of virtue offered by the Morality System:

For Williams, the crucial component missing from this picture is that it must include an internalised other that represents the essential sociality of reason. To insist, as his opponents do, that this is a form of heteronomy in motivation is also to insist that being a conduit of reason, alone, suffices for a person’s autonomy. The rationalism underpins the ‘characterless’ moral self:

Williams derives another corollary from ethical Cartesianism: If ‘the criticising self is simply the perspective of reason or morality’ – one way in which it is ‘characterless’ – then this fuels the idea of a perspective of complete rational autonomy (Reference WilliamsWilliams, 1993, p. 159). This ‘total critique’ expresses an ideal of ‘ultimate freedom’ (p. 158). Williams uses this idea insightfully to diagnose the apparently inconsistent claim that ancient ethics is defective both by being too egoistic and yet too heteronomously concerned with the opinion of others.

His response to this charge begins from the point that shame plays the same essential role for ancient Homeric warriors as it does for us:

Tragic characters are represented to us as experiencing a necessity to act in certain ways, a conviction that they must do certain things, and Williams connects this sense of practical necessity to the mechanisms of shame.

The source of the necessity is in the agent, via an internalised other whose view the agent can respect. Indeed, he can identify with this figure, and the respect is to that extent self-respect; but at the same time the figure remains a genuine other, the embodiment of a genuine social expectation. These necessities are internal, grounded in the ēthos, the projects, the individual nature of the agent, and in the way he conceives the relation of his life to other people’s lives (Reference WilliamsWilliams, 1993, p. 103). Mature ethical agency, in responding to this internalised demand, is not guilty of ‘heteronymy’.

Towards the close of S&N Williams generalises not simply over the argument of that book but over his work as a whole. The hero of S&N turns out to be the ancient historian Thucydides (alongside the tragedians). Williams’s discussion identifies a linchpin of his entire philosophy of ethics: What it is to have a naturalistically acceptable view of the basic psychological concepts we need to do ethics. In Williams’s account Aristotle, for all his insights, is classed with Plato as a theorist who ethicises fundamental psychological concepts in a way that is explanatorily inadequate. They both – like Kant – take fundamental categories for the understanding of the mind as themselves basically ethical (Reference WilliamsWilliams, 1993, pp. 43–45, pp. 160 ff; 1995f).

By contrast, Thucydides offers an explanatory perspective that is, on the one hand, not ‘value-free’ – it is an evaluatively engaged perspective. On the other hand, ‘the psychology he deploys in his explanations is not at the service of his ethical beliefs’ (Reference WilliamsWilliams, 1993, p. 161). His goal

There is an element of self-identification in Williams’s admiration for the ironic, realist, historian who, alongside the ancient tragedians, does not in his writings represent an ‘undeveloped’ form of moral consciousness. On the contrary, its particular value is that it has not developed the obsessions of the Morality System.

4.1 Formulating the Issue of Cognitivism/Realism

I have already started to use some of the terms necessary to formulate the issue: objectivity, realism, and cognitivism. I will now explain how these terms are used before introducing a key concept – minimalism about truth (Reference WrightWright, 1992).

The most general term on the list is objectivity. Williams thought that ethical objectivity involved two distinct issues (Reference WilliamsWilliams, 1985, p. 132). One grounds a claim to ethical objectivity in our perspective as practical reasoners: There are some reasons that are necessary for any practically engaged agent. I have documented his multifaceted critique of this claim as developed, for example, by Nagel. The second issue grounds a claim to ethical objectivity in our perspective as theoretical reasoners: Those who seek to know, via inquiry, whether there really are moral facts.

Ought we to be realist about these domains? In speaking of ‘domains’ it has become true that philosophers are rarely realists, tout court, but realists about particular domains. For example, I hope that readers of this Element are realists about astronomy, but not about astrology. The realist claims that when we factually describe a particular domain, its existence and structure is independent of human minds (Reference WilliamsWilliams, 1978, p. 64). The astronomer studies astronomical phenomena, whose existence and nature are independent of our beliefs about them. The astrologer is doing something else: Presenting putative facts not with the aim of informing, but of entertaining, or doing something else that does not involve claims about planetary influence on human character being factually true.

The next term of art here, cognitivism, introduces the idea of the kind of mental content expressed when we make some of our ethical judgements. People seem, on the face of it, to make remarks like ‘I know for a fact that Hitler was a cruel man’. The verb ‘know’ suggests that this sentence, at least across one of its dimensions, expresses the mental state of knowledge. By contrast, a wide range of moral philosophers from emotivists, to expressivists, to prescriptivists deny that ethical judgements express beliefs (or knowledge). These statements do something else: Express simple emotions, or complex structured sets of attitudes, or issue prescriptions and self-addressed imperatives.

Objectivity is the generic idea; realism a species; cognitivism a way of articulating the commitments of realism. What role is played here by the concept of truth? From this necessarily brief sketch, it seems that truth will play an important role in this debate. For example, it seems as though the astronomer’s claims are true to the facts while the astrologer’s claims are not. Williams, however, later endorsed the view called minimalism about truth (Reference WilliamsWilliams, 1995 l). This is, in one respect, a negative thesis: Truth is not the sort of thing about which one can develop a theory (neither explanatory nor metaphysical). When we attach ‘is true’ to a sentence we are using a predicate, but we add nothing to the content of the sentence. All the interest, then, is shifted back onto what it is to assert that sentence in the first place.

In an important late paper ‘Truth in Ethics’, Williams unpacks what he takes the consequences of the adoption of minimalism about truth for the debate over moral realism. It prioritises the point of the practices within which we regiment our statements to substantiate the minimalists’ ambitious view of ‘syntax’ (Reference WrightWright, 1992; Reference WilliamsWilliams, 1995 l, p. 230). Adopting minimalism opens up the broad range of issues that Williams later discussed, notably in T&T, not about truth, rather, the values of truthfulness (ibid.). This relates back to his commitment to Enlightenment values which, in this context, involve such values as ‘the authority of knowledge, the value of honest inquiry’ (Reference WilliamsWilliams, 1995 l, pp. 232–233). One problem for the moral realist is that she thinks what really matters is being able to discern truths about values and that, for Williams, misses the point. The point is whether we live in a culture in which people responsibly tell the truth about anything and not merely ethics.

The moral cognitivist can, I think, happily accept minimalism about truth. I think he or she can happily accept Williams’s shifting of the focus of the debate to the internal rational discipline of those practices in which our ethical concepts have their home. But, most of all, the cognitivist welcomes Williams’s recognition of an important class of ethical concepts the phenomenology of which makes a prima facie case that cognitivism is true: thick concepts.

4.2 Wittgensteinian Arguments from Thick Concepts

In my extension of Williams’s example of the professor being dragged away by the police I added both her claim to know that her attackers are sadistic and her choice of that concept. Sadistic, we might say, is a concept that tells us more about the actions of the wrongdoers than the fact they are doing wrong. Central to Williams’s discussion of moral realism is his account of thick ethical concepts: An idea that he traces back to Wittgenstein, but where the immediate point of transmission to him was ‘a seminar in the 1950s’ given by Philippa Foot and Iris Murdoch (Reference WilliamsWilliams, 1985, p. 218, n.7).

The point of the distinction is this: Thin concepts, such as good, bad, right, and wrong tell us very little about the circumstances of their application. Sadistic, courageous, and uplifting do give us more determinacy: specificity and concrete detail about the circumstances in which they are applied. Furthermore, when the latter class of concepts are used in judgements, Williams tells us that they can be both ‘world-guided’ and ‘action-guiding’ (Reference WilliamsWilliams, 1985, pp. 140–141).

The rationale for this distinction is to cause problems for any naive distinction between facts and values (Reference WilliamsWilliams, 1985, p. 134). Suppose a guide is describing a cathedral to a group of tourists. In describing the beam across two columns as an ‘architrave’ she is informing them of a fact; in describing the beam as ‘an elegant architrave’ she seems to be doing exactly the same thing. There is a true sentence which ascribes an evaluative (aesthetic) property which is embedded in the speech act of describing.

The anti-realist, of any stripe, will express her unhappiness with this formulation. Yes, those may be the surface appearances of how we think and speak. But there is a difference between descriptive and evaluative properties which maps on to a distinction between descriptive and evaluative meaning. In taking thick concepts at face value, we are buying in to this misleading apparent fusion of description and evaluation. Such concepts ought to receive a philosophical analysis: They are a combination of two factors, one evaluative, and one descriptive. Had the tour guide been more explicit – perhaps a student of Dick Hare’s – she would have added ‘of course, by calling the architrave ‘elegant’, I was prescribing it to you. I was commending it in the light of a presupposed evaluative standard which makes some of its descriptive features relevant to its evaluation’.

The pushback from philosophers influenced by Wittgenstein is this: This proposed two-component analysis cannot be carried out (Reference McDowell and HonderichMcDowell, 1985; Reference WigginsWiggins, 2000, pp. 185–211). For any expression that uses concepts, we can separate out its intension and its extension. The former has various theoretical articulations: For now, let’s call it the expression’s ‘meaning’. The latter are the properties picked out by an expression with that meaning.

Take the thick concept cruel: What do all its instances have in common? The defender of thick concepts adds: Nothing that is not already apparent to those who share the evaluative interest in using the concept with its users. Without this shared interest, the list of things picked out by thick concepts – its extension – will look like an incoherent jumble. The list will seem shapeless unless you share the concept user’s evaluative interest that led them to use this concept in the first place (Reference WilliamsWilliams, 1985, p. 142). The two-factor theorist’s project of analysing away thick ethical concepts into substantive, but descriptive concepts and thin evaluative concepts cannot be carried out.

If minimalism takes us back to the surface appearances of ethical phenomenology, and those appearances include the use of terms like ‘know’, ‘fact’, and ‘assert’, then what more could the cognitivist want? Williams seems to concede to cognitivism everything that it needs simply by recognising the class of judgements that use thick concepts. But Williams does not stop there – in ways I will now explain.

4.3 The Scientific and the Ethical

A trap for Williams’s commentators is that, in spite of his protestations that he wants to break from the traditional question of moral realism, Williams does use the categories of ‘the scientific’ and ‘the ethical’. Furthermore, he does so because he thinks we have moved beyond any naive formulation of the fact versus value distinction – and thick concepts played their phenomenological role in having achieving that. Yet does not that distinction recur in the way in which Williams uses his own pair of phrases? He seems to end up as an ethical sceptic all the same.

The grounds for saying this are that Williams defends a form of scientific realism: Science is an institution for acquiring knowledge where it can reasonably expect to be everything that it seems to be. It offers us a conception of the world ‘to the maximum degree’ independent of our perspective and its distinctive peculiarities (Reference WilliamsWilliams, 1985, p. 139). It offers us a limited degree of self-transcendence. But this is not the chimera of a point of view from nowhere at all. It is simply that the most fundamental sciences transcend that which is distinctive of the human point of view.

Williams does not believe that the concepts and categories we need to conduct the most fundamental scientific inquiries – those of basic physics – are the only concepts and categories we need to explain the reflective activity of doing science (even doing physics). He does not think that the scientific worldview explains itself solely using its own most basic, non-perspectival concepts, in some act of reflexive completion whereby it explains its own generation and acceptance using only its own concepts.

The careful claim Williams makes is about scope: Some of our representations of the world are maximally non-perspectival. That does not mean that our representations about representations have the same feature (Reference WilliamsWilliams, 1985, p. 140). He later came to regret some incautious suggestions in his book on Descartes that an absolute conception implied this kind of physicalist worldview (Reference WilliamsWilliams, 1978, Reference Williams, Altham and Harrison1995k, p. 209). It does not.

Even so, the ethical sceptic thinks that there is enough in Williams’s work to motivate scepticism. True: This characterisation of the contrast highlights how the scientific and the ethical have different purposes. The latter, using repertoires of primarily thick (and some less thick) ethical concepts, is how we construct the social worlds into which we are acculturated. They are ways, as Williams likes to put it, of finding our way around some social world or other. It is important that these socio-cultural worlds, too, are the subject matter of the social sciences. However, the absolute conception of the world claims that, at the limit, we can form a conception of the world, not a world.

For the sceptic, that is all we need to recast Williams as a traditional sceptic about moral realism – forget everything he says about thick concepts and minimalism. All we need, to forge an invidious contrast between the scientific and the ethical, is what he says about explanation. The crucial paragraph is in ELP:

This is why, a page earlier, Williams claimed that ‘science has some chance of being more or less what it seems … … while ethical thought has no chance of being everything it seems’ (Reference WilliamsWilliams, 1985, p. 135). That certainly sounds like a sophisticated form of scepticism!

I do not think, however, that we need to read Williams this way. This argument uses a specific conception of explanation. Suppose that scientific knowledge is of a world that exists anyway. Because, on Williams’s view, this (first order) knowledge is maximally non-perspectival, it is not from any point of view in particular. Of course, it was accessed via a point of view. In explaining that fact, we are going to use all kinds of perspectival knowledge about our scientific practices, institutions and methods. Footnote ¹¹ But the best reflective explanation of scientific knowledge ‘directly vindicates’ it, to use Adrian Moore’s helpful expression (Reference Moore and ThomasMoore, 2007).

By contrast, our reflective understanding of ethical knowledge has to view it as answerable to two dimensions of assessment. One is its articulation, via thick concepts, as both world guided and reason-giving. The other dimension is the social world within which it is articulated. It is accessed via some social world or other, because that is a presupposition of having the sets of concepts available that any given society possesses. That presupposition is the subject matter of our best social scientific understanding of ethical knowledge because it is perspectival through and through: It is always from some point of view.

Furthermore, this explanation takes as its subject matter what ethical agents know, without itself directly endorsing that knowledge. An engaged social scientist stands close enough to our ethical practices to make sense of them while being disengaged enough genuinely to explain them without simply repeating the first-order knowledge being explained (Reference Moore and ThomasMoore, 2007). The scientific case was not like that: A person tried truthfully to know about the world as it is. She succeeded, so the best explanation of her success was that she came to know the world as it is. Her social world enabled her to make this judgement, but it is not part of the best explanation of her making it. Williams is entertainingly scathing about the ‘strong school’ in the sociology of science which, for example, claims that quarks are ‘socially constructed’ via ‘the remarkable assumption that the sociology of knowledge is in a better position to deliver truth about science than science is to deliver truth about the world’ (Reference WilliamsWilliams, 2002, p. 3).

One strategy that the cognitivist can take to undermine this argument is to reject Williams’s view of science as reflecting the aspiration to absoluteness. The other strategy is to accept the contrast, but argue that it is not damaging to the aim of defending the idea of ethical knowledge. If there is merely one core stock of ethical concepts, then we do not need to introduce a plurality of individuable social worlds into our explanation of ethical knowledge. Both Williams, and the cognitivist, have to say more.

4.4 Modernity, Reflection, and the Loss of Moral Knowledge

The idea that our ethical ideas can be indirectly, rather than directly vindicated is one that the moral cognitivist wants to cling to in the face of Williams’s distinction between the scientific and the ethical. But he or she is not out of the woods. Because, on the one hand, Williams acknowledges that there is something distinctive about thick ethical concepts – just as the cognitivist claims. But this distinctiveness creates special problems. If there are such concepts, then how does a social scientist understand them when she explains social and historical change?

Williams consistently argued that how many thick concepts there are for a society to draw upon does not depend on philosophical accounts of them. Everything we know about human cultures tells us that there have been changes in social structures and societal outlooks – including, importantly, evaluative outlooks. To use another characteristic Williams idea, a repertoire of thick concepts expresses such an outlook. And it is the evolution of our societies in history that tell us how many thick concepts we have – not philosophy or ideology.

This point is deployed to attack the cognitivist’s view that the idea of convergence on an ethical truth that can be substantively explained is just as available in ethics as it is in science (Reference WilliamsWilliams, 1985, p. 140). The way this critique proceeds is via a thought experiment: That of a hypertraditional society where people use thick concepts to make ethical judgements and thereby acquire moral knowledge that they then lose (Reference WilliamsWilliams, 1985, pp. 145–147). This is, to put it mildly, a puzzling argument.

To this point Williams has, like his Wittgensteinian interlocutors, ‘start[- ed] with the thick’ (Reference WilliamsWilliams, 1995 l, p. 234). He consistently rejects the two-component reductive analysis of thick concepts offered by the ethical expressivist or prescriptivist (ibid). But he then seems to undercut his own argument using the example of the hypertraditional society where knowledge is displaced by reflection. This process of displacement is always, in Williams’s explanation, a displacement of the thick by the thin where judgements using thin concepts are not world guided and hence not candidates to be known.

We are to imagine this hypertraditional society becoming more modern, in Williams’s proprietary sense, whereby it is more pervasively reflective. Members of this society, Williams argued, now find themselves cut off from the knowledge that they used to have (Reference WilliamsWilliams, 1985, p. 147). Reflection need not have this effect; but, if it does, then these judgers are in the odd situation of being unable to be both reflective and knowers. Yet they can see that their previous judgements were knowledge. The process of reflection may go into reverse: Perhaps the knowledge will be recovered. But not in the light of reflection.

Williams claims we can think about the activities of the ethical thinkers in the hypertraditional society in one of two ways: one, objectivist, construal sees them as finding out a local truth about the ethical – coming to know things. But this local knowledge must have general implications, in just the same way that an anthropologist who concludes that a society’s magical practices really do contain causal claims about the way the world works will see the local claims as having these general implications.

A contrasting non-objectivist view sees these claims as merely ways of finding one’s way around a social world. In a characteristic combination of Williams’s irony and a rhetorical flourish, he argues that only on the non-objective model can we secure the view that the hypertraditionalists know. That is because in the objective model, the local claims have reflective implications that can unseat the original knowledge (Reference WilliamsWilliams, 1985, p. 148). The non-objective understanding of the hypertraditionalists does not attribute to them these implications of reflection. Reflection can, not must, cause more local ethical knowledge to be lost. In the next section, we can see that when the thick is supplanted by the thin, and we try to develop ethical theories on that basis, we do not restore knowledge in that way either.

How could people lose the knowledge they once had? People use repertoires of thick ethical concepts from a specific ‘evaluative outlook’ and loss of that outlook cuts those people off from those concepts; yet as Adrian Moore notes ‘they can still understand the outlook, and indeed the concept, just as an anthropologist or a historian can understand an evaluative outlook, which he or she does not share’ (Reference MooreMoore, 2003, p. 344) Williams, qua philosopher of social science, insisted on the possibility both that judgements using thick concepts can express moral knowledge and that they can be understood by a sympathetic but non-identified observer who takes up an ‘ethnographic stance’. Footnote ¹²

The example of the hypertraditional society pulls apart two of Williams’s thoughts about thick ethical concepts. The outlooks from which they are articulated are ways of finding one’s way around a social world. Yet two people in the same world might become aware of a clash of evaluative outlooks – Williams always cited Oscar Wilde’s remark that ‘blasphemy’ was not a word of his – and now we have to understand how two people with conflicting sets of thick concepts can both have knowledge. Answerability to the world is, after all, one aspect of the use of thick ethical concepts. Williams remarks that historical change can estrange an individual or group from a concept:

One way to explain what is happening here involves the idea of presuppositional failure: Individuals, or societies, are subject to processes of social change, which displace thick concepts, or groups of them, as evaluative outlooks change. Unproblematically, this makes some judgements unavailable that were available before because it was a presupposition of making them that certain concepts were accessible. Footnote ¹³

But another way to make the same point is that Williams agrees with his Wittgensteinian fellow travellers that thick ethical concepts are an amalgam of world-guidedness and action-guidingness. They are not, then, capable of purification into solely theoretical concepts (Moore, p. 353). Accepting them is to accept an evaluative outlook, which is also to undertake practical commitments of reason-giving and, one might add, patterns of feeling.

The ethnographer or historian writing about Edwardian England who seeks to understand both Oscar Wilde’s evaluative outlook and that of the social conservatives whom he scandalised might be able to see the occupants of both outlooks as capable of knowledge. Yet the joint occupation of the two evaluative standpoints is irreconcilable precisely because these are not two ways of carving up a non-evaluative reality or even two ways of carving up evaluative reality. That reduces the exercise to one of pure theory. It would leave out the connection between thick conceptualisations, reason-giving, and action. We are discussing two evaluative ways of life.

That is why our envisaged social scientist must be capable of a sympathetic, but non-identified, point of view from which both claims to knowledge are not known by the social scientist qua scientist. But she knows something – from her perspective in history or sociology – that entails the truth of what each of the two parties knows (Reference MooreMoore, 2003, pp. 348–349). This does not mean that our envisaged social scientist takes up the fully internal perspective of either party – again, not qua social scientist. Her explanation entails what the conflicting parties know and she knows her explanation to be true. But, ex hypothesi, it does not directly vindicate what each party knows. How could it? So there is nothing here to encourage the sceptic: This is merely how any social scientific account of the ethical must work.

The cognitivist remains uneasy: Williams’s position makes sense. But making sense is not the same as being plausible. How can Williams share so many assumptions with the cognitivist yet formulate the challenge of his non-objectivism? Of course, there are reasonable disagreements within ethics which mean that people vary in which concepts they use. But why is this specific point being exaggerated into a fundamental clash of evaluative outlooks – unless the influence of MacIntyre’s After Virtue has led Williams astray?

The cognitivist is worried that Williams’s fable of the hypertraditionalists re-imports the kind of foundationalism about justification that I noted in Section 1 (Reference ThomasThomas, 2006). How, from the fact of piecemeal social change, do we end up with a model of a group of concept users attempting to vindicate their entire ethical outlook from some perspective wholly outside it?

This takes us to the crux of the issue – and why Williams does not really think this is an important issue at all, rather, a candidate for deflation. In a later paper, Williams takes a paradigmatic cognitivist claim: The example of a group of hooligans who set fire to a cat (Reference WilliamsWilliams, 1995 l). He comments that ‘the boys do a wanton and hideous thing to the cat’ – thereby using two thick ethical concepts. David Wiggins, on behalf of cognitivism, claims that there is ‘nothing else to think’ but that the boys’ act was cruel. Williams demurs:

For the cognitivist, this is a puzzling remark. They offer an account of virtuous people, and the evaluative features to which their virtues are a response, that stand in a (virtuously) circular relationship. They see no way of unpacking this further. How could a virtuous person see this situation, be thinking of it correctly, but not see it as cruel?

Williams responds: There is no one set of thick ethical concepts that all ethical judgers, as such, can be guaranteed to have. His earlier argument is restated: We inhabit some social world or other, each structured by its own repertoire of concepts, and there is no core set that all social worlds share. He accepts that the latter claim is the one that the cognitivist is most likely to want to defend. Differentiating two versions of the claim, Williams contests both the ways in which the cognitivist tries to shore up his or her position.

The first version claims that there is a single, unitary, set of thick ethical concepts (correlated with a set of virtues) and when evidence to the contrary is presented proponents of the view simply dismiss it. Williams was personally amused by a researcher he met who was adapting Thomas Aquinas’s list of virtues to our modern conditions. This is the basis of the oblique reference in ‘Truth in Ethics’ to the ‘sociologically optimistic’ view that we can, via a ‘large jump’, move from Aquinas’s social world to ‘late twentieth-century Los Angeles’ – which is where this researcher happened to be located while working with Philippa Foot (Reference WilliamsWilliams, 1995 l, p. 241). Does a table of virtues drawn up by a Dominican monk in the thirteenth century just need ‘adjustment’ to modern conditions? Williams thought not.

The second version of a cognitivist defence claims that while there are, in Williams’s words, ‘undeniable’ facts about ‘historical and social variation’, this has a deep explanation in terms of a core set of concepts, which receive different cultural expressions. This would have to be combined with an account of moral error. That attracts a different sceptical response from Williams:

My own view – both when I worked with Williams and now – is that this represents a set of challenges to cognitivism and not a list of fatal defects (Reference ThomasThomas, 2006). The onus is on the cognitivist to say more to address these well-motivated concerns about the final defensibility of cognitivism.

Whatever one’s view we are at least in a position to identify what is at stake between Williams and the moral realist. The latter thinks that, using the resources of the very social sciences that Williams has identified, there can be a reasonable explanation of moral disagreement framed in terms of basic points of agreement. That explanation need not, itself, simply endorse the knowledge that it explains. Those who believe that evaluative outlooks express sets of thick ethical concepts do not want to give up on the idea that they have some correct social scientific explanation. But nor do they accept the disaster scenario encapsulated in the fable of the hypertraditionalists. They contest Williams’s history: There is enough agreement reasonably to frame our disagreements in just the way that MacIntyre argued was true within the one tradition he favoured. But that is because, for that tradition alone, MacIntyre had a different model for moral knowledge than the foundationalist model he applied to everyone else’s tradition.

I will break off the argument between Williams and the cognitivist at this point: The latter will level against Williams’s argument the verdict possible in Scottish courts of ‘not proven’. In any case – and typically – Williams had moved on. In his later papers on this topic, Williams turned to the question of the value of truthfulness – that is the point that he thinks the moral cognitivist really ought to be concerned about (Reference WilliamsWilliams, 1995 l).

The point of appealing to thick concepts, one that goes beyond the minimalist surface of the sentences that express them, is that they are a natural entry point for the idea that in ethics we have a lot of knowledge. If the point of that concept is to flag up reliability on the part of other ethical judges, we can form a conception of ethical expertise which is not modelled on that of the theoretical expert. Perhaps a better word would be insight: A ‘helpful advisor’, as Williams puts it, ‘who can see that something falls under a certain thick concept’ (Reference WilliamsWilliams, 1995 l, p. 235).

Discussing the enterprise of ethical theory, in the next section, we will see that Williams contrasts it with a ‘phenomenology of the ethical life’. It is part of that phenomenology, he thinks, that an advisor is not really in the business of telling you about the truth directly:

Williams anticipates the realist’s protest against this argument: ‘You may think ‘a life worth living’ is itself an ethical concept that involves the notion of the truth’ (ibid.) I don’t think Williams disagrees: He just holds, simply, that the realist has the argument backwards (in a typical way). The cognitivist is so focused on establishing that we have moral knowledge that she has failed to explain why this is important in the first place.

5.1 Theory, For and Against?

Williams rejects of all forms of ethical theory. Theories contribute to the erosion and displacement of thick ethical judgements; they replace conceptualisations in which we ought to have confidence with deracinated false equivalents in which we ought to have none. Theory takes itself to represent an Enlightenment spirit of critique when it simply perpetuates the false claims of the Morality System. And the enterprise of theory works in just the same way as the Morality System itself: It is asymmetrically dependent on the very ethical materials it falsifies and misrepresents in the name of critique in the guise of ‘scientific progress’. Footnote ¹⁵

Williams’s most extended discussion of the motivations of the theorist is in ELP:

We may be unfortunate and find ourselves in a historical context where ‘thinking’ in ethics has been made impossible. If we are fortunate enough not to be in that position, that is not itself a philosophical issue. This is because it is the very possibility of philosophical reflection about ethics that is at stake. Yet, even so, in those propitious circumstances we need to live reflectively in the light of the truth in a way that does not seem to draw on the distinctive resources of philosophy. Those resources, in the hands of theorist, are primarily those of building abstract models that use only thin ethical concepts.

It follows that, for Williams, whether philosophy supplies any ‘tests’ for ethical thinking is a question that has to be open. That is because the availability of such tests varies from one cultural and historical context to another (Reference WilliamsWilliams, 1985, p. 74). In the right circumstances there can, he believes, be ethical reflection and progress, but not using the methods of philosophy.

At one point Williams describes an alternative to theory, namely, ‘a phenomenology of the ethical life’:

This will strike his readers as a self-portrait: ‘[T]his could be a good philosophy’, Williams continues, ‘but it would be unlikely to yield an ethical theory’ (ibid.). It would have started from the right place, within ethical life as it is lived, but it would not be focused, as theory is, on ‘just one aspect of ethical experience’, namely, its cognitive structure as a set of beliefs. (By analogy with a scientific theory.) Ethical theory, on this understanding, ‘in part provides a framework for our beliefs, in part criticises or revises them’ (ibid). Williams is sceptical that they have the intellectual authority to do so (Reference WilliamsWilliams, 1985, p. 99). In the next sections, I will discuss Williams’s critiques of two major forms of ethical theory in the academic literature. Since both of these theories are impartialist, it is helpful to begin with this general target for his objections.

5.2 Mind the Gap: The Critique of Impartiality

The two most important options for the theorist, at the time when Williams wrote, were consequentialism and contractualism, both of which are impartialist theories. Much philosophical work adjudicates the conflict between them; Williams thinks both are incorrect. Which of his arguments apply to this generic impartialism?

First, we can set aside an irrelevant issue: Williams does not commit the fallacy of equivocation. Adrian S. Piper represents the Kantian impartialist’s concern that critiques of impartialism equivocate between impartiality and impersonality (Reference PiperPiper, 1987). Consider the example of William Godwin, an early exponent of utilitarianism, who claimed that he would save a benefactor to humanity from a burning building rather than a chambermaid – even if the latter were his mother. Utilitarianism is an impartial view, but is it committed to this kind of cold impersonality? The impartialist says ‘no’: That objection confuses the impartial versus partial distinction with the personal versus impersonal distinction.

It seems unlikely that Williams committed this elementary error given that he draws attention to it: ‘Rawls claims that impartiality does not mean impersonality’ – and he does not go on to deny it (Reference WilliamsWilliams, 1981b, p. 5). Williams used the word ‘particular’ – not ‘personal’ – positively to characterise his own views in a contrast with the universal. Kantian impartiality, on this understanding, seeks to abstract from the particular to identify principles that are universal. More substantively, Williams believed in basic (‘not inferred’) reasons of partiality.

As an example of the latter, Williams presented a much-discussed example of an individual who, given the choice between saving one of two people who are at risk of drowning, saves his own wife (Reference WilliamsWilliams, 1981b). Academic commentary has focused on Williams’s claim that any impartialist treatment of the case involves ‘one thought too many’ (Reference WilliamsWilliams, 1981b). This commentary clearly envisages the structure of the problem as whether impartiality can underwrite the claims of the personal.

Williams objects that impartial theories have no business writing back in permissions to be partial to one’s ‘nearest and dearest’. But the further aspect of his discussion is that the phenomenology of our ethical experience involves basic, partial, reasons. ‘That (he or) she is my wife’ is a basic reason that does not need to be inferred from any deeper theoretical justification. Williams is committed to the existence of partial reasons in ethics. An initial problem here is that our ordinary word ‘partial’ equivocates between an ethical idea and a different idea of unfairness, or chauvinism. (John Cottingham calls the latter ‘tainted partialism’ (Reference CottinghamCottingham, 1997)). But that equivocation, once clarified, can be set aside.

For Williams, the fundamental problem with impartialism is its insistence that it represents the claims of morality as both supremely important and rationally authoritative. This, for Williams, renders problematic the relationship between this point of view and others, which also claim ‘significance or structural importance in life’ (Reference WilliamsWilliams, 1981b, p. 2). For which agents are these concerns important and authoritative: Rational agents as they are or rational agents as they might be if we accept the impartialist’s stipulation?

This general thesis underpins the more specific charges that impartialism confines our freedom and alienates us from our ground projects. In his helpful discussion, Paul Hurley notes that the metaphysicians of rightness think they can evade these phenomenological arguments (Reference HurleyHurley, 2009). Moral psychology is not their concern: Modelling the right-making features of actions by explaining the relation between the intrinsic value of outcomes and the nature of rightness is, they think, the only topic in their remit. As Hurley reads Williams – to my mind accurately – this leaves open the relationship between rightness and our ordinary notion of a reason. Therein lies the problem.

This is precisely the gap that we saw Kant close by stipulation: The moral point of view is the supremely important point of view as it reflects our most important values. Its special reasons have rational authority over our other reasons. Throughout this section, we will see Williams asking – once again – why we ought to accept this stipulation. From his perspective, to have an ethical point of view is to be committed to certain ground projects that give substance to the idea of individual character (Reference WilliamsWilliams, 1981b, p. 5). This is not the quest for absolute authenticity in the form of distinctiveness: The content of a ground project can be the same as many other people’s projects. What is distinctive is that it is one’s own (Reference WilliamsWilliams, 1981b, p. 15). Nevertheless, it is important to Williams that one inhabit a social world in which it matters that we are not all the same. This lack of fungibility is particularly important when it comes to personal relations, including friendship: A person does not want to feel that he or she is simply ‘inter-substitutable’ (Reference WilliamsWilliams, 1981b, p. 15).

Even if such a project gives each person a reason to go on, this does not make it a selfish project. However, there is still a distinction to be drawn when it comes to the involvement of the self in the success of a project. It can matter to a person that she be the author of its success: That is true even of moral projects. To take Williams’s example, ending injustice might be reflected in a person conceding that another person’s way of achieving that matters more than her own agency in bringing it about. Yet, a wholly other-regarding aim may still involve the goal of wanting to be the person that brings it about (Reference WilliamsWilliams, 1981b, p. 14).

Ethical impartialism, then, problematises in a general way this relationship between a person and their ground projects. The most famous of Williams’s arguments are the two memorable vignettes he presents in ‘A Critique of Utilitarianism’. George, a researcher in chemistry, is hampered by ill health from finding work he can do and stays at home to look after his young children. His wife, who holds no strong views about chemical weapons research, works to support the whole family. (So we can presume that she would have no strong principled objections were George to take the job.) This predicament can be eased by George’s mentor who can arrange for him to take a well-remunerated job in chemical weapons research. George demurs, at which point his mentor notes that if he does not take the job, a zealot in the cause of chemical weapons research will do so instead.

Jim, a lost and hapless traveller, finds himself in a South American country where the local representatives of a repressive regime – led by the ‘sweat stained’ Captain Pedro – propose to execute twenty of the indigenous inhabitants to terrorise the local population. Pedro’s offer to Jim is that if he kills one of the captives, the other nineteen will be freed. Failing that, all twenty will die.

I will examine these examples in more detail in the next section. At this point, I want simply to note the fundamental issue they raise about a person’s relation to their self-conception, their projects, and what they are prepared to ‘live with’. These depend, on Williams’s view, on a generalisable error: He thinks that the Kantian impartialist misrepresents the relation between a person and her ground projects, but this is a general failing of all forms of impartialism.

The rational authority of impartial morality is such that in any conflict with a ground project it ‘must be required to win’. The error here, for Williams, is the absorption of the individual into the moral self in a way that loses the particularity of the individual. This is a specific illustration of his more general claim that there are some parts of our ordinary ethical outlook that the impartial perspective must fail to include within its scope:

This is not a denial that there are moral reasons; it expresses the idea that there are moral reasons as the impartialist understands them, but only if there are other reasons, too. This is the asymmetric dependence I have identified between the commitments of the Morality System and those of our ordinary ethical outlook. I will now turn to how this idea plays out in Williams’s critique of a specific form of impartial theory: an act utilitarian form of consequentialism.

5.3 Williams’s Critique of Consequentialism

George and Jim above are central examples in the literature that has evolved in response to Williams’s essay; Williams later expressed some unhappiness, however, with the way he had framed the issue. The most important, methodological, point is that Williams did not take himself to be supplying practical verdicts that differed from those an act utilitarian might give to the two cases. Footnote ¹⁶ The point is not the verdicts, but the kinds of considerations that supported them. Furthermore, given that he did not share the methodological approach of his opponents, Williams did not think he was in the business of supplying counter-examples to putative principles.

In his work as a whole, Williams showed a great deal of sympathy with the view known as ethical particularism (Reference ThomasThomas, 2011) Footnote ¹⁷. This is the claim that ethical thinking can get along perfectly well without being modelled as a finite set of finite principles. He never refined this into a specific objection to principled ethics, perhaps because he thought his critique of the Morality System subsumed it. In the case of his critique of act utilitarianism, however, he protested that he had not invented a counter-example to a principle so as to invite the re-formulation of a refined (and presumably more adequate) principle. Nor, more specifically, did he take himself to have invited the ‘integrity objection’ to act utilitarianism. I will open the discussion, as Williams did, by highlighting some issues of formulation to identify whether we are discussing a distinctive view here at all.

5.4 Williams’s Critique of Contractualism

In his discussion of the project of grounding moral reasons on the very idea of practically rational agency we saw, in Section 1, that Williams thought that the project could not be rehabilitated. He did, however, identify it as a task in which several of his peers were engaged. This reflects the widespread intellectual influence of John Rawls – Williams was prepared to concede the impartialism had some potential as a way of modelling justice (Reference WilliamsWilliams, 1985, p. 64). When he turns to the enterprise of ethical theory, however, he does not directly engage with Rawls’s views. Instead, he focuses on the moral contractualism developed by Thomas Scanlon.

I think Scanlon’s views are foreshadowed when Williams explains his target notion of an ethical theory (Reference WilliamsWilliams, 1985, p. 71):

Williams’s inclusion of Scanlon explains the curious part of this definition – the ‘negative kind of ethical theory’ (ibid.). Scanlon’s contractualism, like the Kantian Categorical Imperative which is its ancestor, rules nothing in. It rules candidate conceptions out.

Scanlonian contractualism proceeds as follows: First, we isolate a narrow part of the moral focused solely on interpersonal justification. The next step is to argue that an action is right if it is not wrong. Wrongness is then defined as being impermissible according to a system of principles that survive a distinctive test: That a person would not reject this candidate set of principles as unreasonable were they put forward as ‘a basis for informed, unforced, general agreement’ (Reference ScanlonScanlon, 1998, p. 193).

Contractualism thus directly addresses the issue of intellectual authority: Our goal of living by a set of principles in the light of which we can offer justifications to each is grounded on the deeper value of mutual recognition. Furthermore, being unjustifiable, in Scanlon’s sense, is not a property that tracks some independent property of wrongness: It is that in which wrongness consists. Wrongness is derived from the wronging of one person by another such that the former cannot justify her actions to the latter. Our shared reasons are grounded on shared principles underpinned by an evaluative vision of our moral world as a place of mutual recognition of each other’s value.

One of the great merits of Scanlon’s position is that the narrowness of its scope is flagged up throughout: ‘what we owe to each other’ maps out a discrete area of ethics and not its full range. It is the conceptual ties between wrongs, wrongings (of one person by another) and interpersonal justification that are its limited focus. Scanlon hopes to avoid circularity by not involving any evaluative commitments when he appeals to reasonableness: Contractualism supplies a negative test applied to sets of principles. It rules some out – but says nothing about how candidate sets are generated in the first place to become candidates for testing. Individual complaints can be brought against candidate sets by appealing to reasonableness. However, that idea presupposes, but does not directly appeal to, the background conception of value that gives the exercise its context and point. There is, overall, value to living with each other on these terms of mutual respect.

Williams notes, in his discussion of the view in ELP, that this is not the foundational impartialism he criticised earlier in the book: Contractualism presupposes agents ‘assumed to be already interested in reaching agreement’ (Reference WilliamsWilliams, 1985, p. 75). Contractualism generally receives a sympathetic treatment in ELP. Williams thinks it does offer an answer to the question of how an ethical theory can have intellectual authority. The main focus of his critique is on the special assumptions he thought it needed for it to exercise such authority. I think the contractualist might reasonably complain that Williams does tend to misrepresent contractualism as a convergent theory which aims at agreement. I think it would be more sympathetic to Scanlon’s project to take agreement as a given such that we work out its presuppositions. (That goes hand in hand with the avowed narrowness of Scanlon’s focus.) Williams assimilates Scanlon’s project to Rawls’s project which has a different subject matter. That latter case better fits Williams’s characterisation of a common project which seeks agreement on fair terms and which uses Rawls’s method of reflective equilibrium (Reference WilliamsWilliams, 1985, p. 99). Williams’s only comment is that ‘the method is appropriate to constructing an ethical theory under these assumptions, but it is very important how strong these assumptions are’ (ibid.).

Williams argues that ‘the theory’ – and now Rawlsian justice and Scanlonian contractualism are not clearly separated – not only starts within the ethical (as the arguments against foundationalism earlier in ELP have shown they must). They are ‘doubly’ in the ethical world: with a method that incorporates substantive commitments combined with what Rawls would call ‘realistically utopian’ ideals:

This is the first appearance in the text of ELP of an important idea that characterises Williams’s approach to ethics. Ours is a modern society – a concept that, as we saw in the previous section, discharges several critical tasks in Williams’s work.

In the present context, it seems as if the same doubts Williams has about Rawls – or ‘Rawlsianism’, an even vaguer target – reflect back on Scanlon’s view. But it is unclear that they do. Williams moves from the claim that Rawlsian justice requires explicit discursive principles, to the claim that it seeks social consensus on contested issues, to the most general claim that ‘the society represents its values in a set of stateable principles’ (Reference WilliamsWilliams, 1985, p. 100). But it seems a stretch to move the expression of conceptions of justice in a principled form to the claim that all (or most?) of a society’s values be expressed in this way. Williams then argues that it is characteristic of a pervasively reflective modern society to seek to represent its values explicitly as a transparent set of principles. It then seeks to rank them in order of priority and further reconcile any conflicts via a ‘rationalistic decision procedure’ that would eventuate in ‘an ethical theory in the fullest sense’ (Reference WilliamsWilliams, 1985, p. 101).

Williams originally made a concession to the contractualist enterprise: That ‘the contractualist enterprise is coherent’ (Reference WilliamsWilliams, 1985, p. 102). But as the argument proceeds, the concession is steadily withdrawn, and Williams uses as his interpretative frame Hegel’s critique of the putatively abstract and formal nature of Kant’s moral philosophy:

How does Williams’s initially sympathetic of contractualism end up at this point? His claim is that the contractualist formula contains an inner drive to universalism: That is why the basis of judgements of ‘reasonableness’ starts to ‘thin’, as Williams puts it. Contractualism is re-interpreted as aiming at the very foundationalist perspective that Williams takes himself to have shown to be impossible.

Yet I think it is an open question whether the analogy with Hegel’s ideas is appropriate here. As Ken Westphal has argued, it is an interpretative mistake to contrast Kant’s ‘empty formalism’ with Hegel’s rich concretisation of ethical life – that is not what Hegel says. He says that, were it not possible fully to specify the content of Kant’s ethics, then it would be empty and formal – but it can be so specified and Hegel shows you how (Reference WestphalWestphal, 2005; Reference Thomas, Himmelmann and LoudenThomas, 2015).

By analogy, then, if contractualism denied its dependence on a presupposed ideal of a community of mutual recognition, then it would be empty, formal, and only able to purchase universal appeal at the cost of making its judgements of reasonableness otiose. However some versions of contractualism do not involve that denial (Reference DarwallDarwall, 2009). In framing his argument in terms of there being no natural stopping point for the contractualist enterprise it is worth bearing mind that Williams repeatedly notes that slippery slope arguments are invalid.

One might reasonably conclude that Williams has some sympathy with what one might call the minimal core of contractualism on the understanding that this a narrow part of the ethical. It is certainly not committed to the complex presuppositions of the Morality System. One might reasonably complain against contractualism – in a way that Williams does not – that the conceptual connections between wrongs, wrongings, and reasonable complaints do not have to be treated reductively. Footnote ¹⁸ The idea that it is part of our conception of an ethical system that mature agents within it want to be able to justify their actions to others would seem to be an idea that Williams endorses – if that claim is understood in a sufficiently minimal way.

Of all of Williams’s discussions, then, his critique of contractualism is the most inconclusive. That said, there are multiple routes to the same destination in Williams’s work. From everything that he says, we might conclude that contractualism is simply irrelevant, because the phenomenology of our moral experience shows us that ethical reasons are fundamentally partial. Another direct path to that conclusion would be to develop, in a positive way, the kind of particularist view of ethical judgment that argues that it need not be principled and, in fact, it is not (Reference DancyDancy, 2004; Reference ThomasThomas, 2011). An indirect path would be to argue that Williams’s internal reasons thesis about the nature of reasons rules out contractualism. I noted that Williams and Nagel agree about motivated desire theory and the fact that any account of the rationalisation of ethical action had to mention a desire. Scanlon is more drawn to a purely cognitivist view where believing that one has a reason suffices for an agent to act from reasons (Reference ScanlonScanlon, 2014).

It is worth recalling at this point Williams’s surprising recruitment of Kant to the ranks of internal reasons theorists. While Williams may have his doubts about Kant’s founding stipulation, as John Skorupski notes there is a concern that Kant and the internal reasons theorist undoubtedly share:

It seems, then, that there is a minimal core commitment that Williams and the contractualist share. It seems that if contractualism represents one form of the pathology of theory, whose cultural dominance reflects underlying truths about the prominence of the Morality System, or the social pressures of modernity, then it is on the field of those wider engagements that Williams’s conflict with contractualism ought to be fought.

5.5 What Can a Person Live With?

As I have previously noted, Williams was influenced by Winch’s ‘The Universalisability of Moral Judgement’ (Reference WinchWinch, 1965). In this section, we have concluded that moral philosophy, in the form of theory, gives us no ‘tests’ in either its positive or negative variants. But we are left with this question, which Williams poses more than once, and which seems to take its cue from Winch: What can a person live with?

Is that our substitute for a ‘test’ when we think reflectively about the ethical? In Winch’s discussion, Vere thought he could live with condemning Budd to death: He was doubly wrong both in his verdict on Budd and about himself. When our ethical dispositions are focused on an ethical question we face a dual task: of looking outwards, to the demands of the case, and inwards – to what we know about ourselves. George the potential chemical weapons researcher and Jim, the potentially coerced killer of an innocent victim both have to ask themselves: What can I live with? Charles Taylor notes that these questions pose a special challenge:

Williams envisages a continuity between everyday ethical decisions and these radical choices where our most fundamental identifications are challenged in this kind of disorienting way. Through ethical decision, we find out who we are, or shape who we are, in ways that can be mistaken – disastrously so for Vere. In his culminating discussion, in Truth and Truthfulness, Williams notes that we can describe this task in general terms (Williams, pp. 191–205) But these observations are no substitute for reflective decision.

This late discussion is helpful in several ways: First, Williams remained committed to the idea that reasoning from one’s must fundamental identifications marks off some of our most important ethical deliberations. This is so even if, as Taylor implies, such decisions are dizzying at altitude: These questions about who we are and aspire to be leave behind the familiar criteria of more mundane decisions. Again, concerned to address the charge of unwarranted egotism, Williams argues that we seek to act in the light of identifications that others can acknowledge. And we are persistently hampered, in taking such decisions, by the ease with which belief can be suborned by fantasy and wish (Reference WilliamsWilliams, 2002, p. 197). Self-deception is an obstacle both to taking the correct ethical decision and, in the process, knowing who we are. Here, too, other people play a role as a resource for the ‘steadying’ of our minds (ibid.). We can begin to appreciate why Williams thought that the resources of academic moral philosophy were not of much help in answering questions of this kind.