¹ In other words, principles of justice might not always entail ought statements. See Gilabert, P., ‘Feasibility and Socialism’, Journal of Political Philosophy 19 (2011), pp. 52–63CrossRefGoogle Scholar, at 56. More generally, it is simply not plausible to argue that all normative statements are prescriptive: see Cohen, G. A., Rescuing Justice and Equality (Cambridge, Mass., 2008), pp. 251–2CrossRefGoogle Scholar.

² Cohen, Rescuing Justice and Equality, ch. 7. Cohen does not define ultimate principles of justice as principles that are fact-insensitive. Rather, that is an additional substantive claim that he makes: see Rescuing Justice and Equality, pp. 280–1 and ch. 6. For relevant discussion of the claim that ultimate principles of justice are not grounded in facts, see Miller, D., ‘Political Philosophy for Earthlings’, Political Theory: Methods and Approaches, ed. Leopold, D. and Stears, M. (Oxford, 2008), pp. 29–48Google Scholar; Pogge, T., ‘Cohen to the Rescue!’, Ratio 21 (2008), pp. 454–75CrossRefGoogle Scholar; Faik Kurtulmus, A., ‘Rawls and Cohen on Facts and Principles’, Utilitas 21 (2009), pp. 489–505CrossRefGoogle Scholar; Jubb, R., ‘Logical and Epistemic Foundationalism about Grounding: The Triviality of Facts and Principles’, Res Publica 15 (2009), pp. 337–53CrossRefGoogle Scholar; Ypi, L., ‘Facts, Principles, and the Third Man‘, Socialist Studies/Etudes Socialistes 8 (2012), pp. 196–215.Google Scholar

³ Cohen sometimes expresses his thesis that ultimate principles are not grounded in any facts by saying that these principles are fact-insensitive, but it is important to recognize that when he does so he is using the expressions ‘fact-insensitive’ and ‘fact-sensitive’ in a specific sense. In his sense, a fact-sensitive principle is simply one that is (partly or wholly) grounded in or justified by facts, whereas a fact-insensitive principle is one that is not (partly or wholly) grounded in or justified by facts (Cohen, Rescuing Justice and Equality, p. 231). So, for example, he would not regard a principle whose applicability depends on the facts as fact-sensitive in his sense: see Cohen, Rescuing Justice and Equality, pp. 335–6. (Both Miller and Pogge seem to misunderstand Cohen's thesis in this respect: see Miller, ‘Political Philosophy for Earthlings’, pp. 34–8; Pogge, ‘Cohen to the Rescue!’, especially pp. 466–7. Contrary to what Pogge claims, the search for fact-insensitive principles, in Cohen's sense, is not necessarily a search for principles that would be applicable in all possible worlds.)

⁴ See Rawls, J., ‘Kantian Constructivism in Moral Theory’, John Rawls: Collected Papers, ed. Freeman, S. (Cambridge, Mass., 1999), pp. 303–58Google Scholar, at 306. What this practical task involves may depend in part on the nature of the society for which these principles are intended: Rawls, J., ‘The Idea of an Overlapping Consensus’, John Rawls: Collected Papers, ed. Freeman, S. (Cambridge, Mass., 1999), pp. 421–48Google Scholar, at 421. In modern democracies, they must enable those who are profoundly divided by different religious, philosophical and moral doctrines to live together over time as free and equal citizens: see Rawls, J., Political Liberalism (New York, 1996), p. 4Google Scholar; see also p. xxxix. Andrew Williams is one of the few theorists who in the context of the debate between Cohen and Rawls has emphasized the point that Rawls believes that principles of justice must play a particular social role: see Williams, A., ‘Justice, Incentives and Constructivism’, Ratio 21 (2008), pp. 476–93CrossRefGoogle Scholar, at 485. See also Stemplowska, Z. and Swift, A., ‘Ideal and Nonideal Theory’, Oxford Handbook of Political Philosophy, ed. Estlund, D. (Oxford, 2012)Google Scholar, section entitled ‘Is Ideal Theory Too Realistic?’.

⁵ Rawls, J., A Theory of Justice (Cambridge, Mass., 1971), p. 7Google Scholar; Rawls, J., A Theory of Justice, rev. edn. (Oxford, 1999), p. 6Google Scholar. (In the citations that follow, references will be given in the form ‘p. x/y’, where ‘x’ denotes the page number in the 1971 edition and ‘y’ denotes the page number in the 1999 edition.) Rawls's claim that the primary subject of justice is the basic structure of society has been taken by many to imply that Rawls thinks that it is the only subject of justice: see e.g. Abizadeh, A., ‘Cooperation, Pervasive Impact, and Coercion: On the Scope (not Site) of Distributive Justice’, Philosophy and Public Affairs 35 (2007), pp. 318–58CrossRefGoogle Scholar, at 323. Whether that follows and, if it does, whether the restriction of principles of justice to the basic structure of society could be justified, will be considered further in section IV.

⁶ In fact, as Arash Abizadeh points out, the basic structure might be understood in at least three different ways, as comprising ‘(1) the institutions that determine and regulate the fundamental terms of social cooperation; (2) the institutions that have profound and pervasive impact upon persons’ life chances; or (3) the institutions that subject people to coercion’ (Abizadeh, ‘Cooperation, Pervasive Impact, and Coercion’, p. 319). The differences between these interpretations do not matter for my purposes.

⁷ See Cohen, Rescuing Justice and Equality, pp. 21, 267–8, 278. This leads Cohen to endorse what Adam Swift calls an epistemological conception of political philosophy: ‘the question for political philosophy is not what we should do but what we should think, even when what we should think makes no practical difference’ (Cohen, Rescuing Justice and Equality, p. 268; Swift, A., ‘The Value of Philosophy in Nonideal Circumstances’, Social Theory and Practice 34 (2008), pp. 363–87CrossRefGoogle Scholar, at 366–8). Contrast this conception of political philosophy with the one endorsed by Robert Goodin and Philip Pettit when they maintain that the concern of political philosophy is ‘to identify the sort of political institutions that we should have’ (A Companion to Contemporary Political Philosophy, ed. R. E. Goodin and P. Pettit (Oxford, 1993), p. 1).

⁸ As Andrew Williams observes, Cohen does not provide us with anything more than a rudimentary answer to the question of what justice is (Williams, ‘Justice, Incentives and Constructivism’, p. 491). Cohen does in fact offer a general characterization of what he understands by justice: ‘if, as some of my critics insist, I simply must say what I think justice is, in general terms, then I offer, for those who will be content with it, the ancient dictum that justice is giving each person her due’ (Cohen, Rescuing Justice and Equality, p. 7). Williams regards this as inadequate, however (Williams, ‘Justice, Incentives and Constructivism’, p. 491, n. 39). We should at least question whether there is a useful concept of justice sans phrase, as opposed to concepts of (say) distributive, retributive and reparative justice, perhaps with family resemblances between them.

⁹ We might also say that a principle P1 is more ultimate than P2 if P2 can be derived from P1, perhaps in conjunction with various other premises, factual or otherwise. Somewhat confusingly, Cohen also speaks of fundamental principles of justice, which he defines as principles of justice that are not applied principles of justice, where an applied principle of justice is ‘a principle of justice that is derived from . . . a principle of justice together with something other than a principle of justice, such as a set of empirical facts, or a value other than justice, or a principle that is not a principle of justice’ (Cohen, Rescuing Justice and Equality, p. 279). As Cohen points out, this allows that a fundamental principle of justice might be derived from another principle of justice (Rescuing Justice and Equality, p. 280). So, paradoxically, a fundamental principle of justice in this sense need not be an ultimate principle. Even more confusingly, Cohen sometimes seems to use the term ‘fundamental’ to mean ‘ultimate’: for example, he says ‘I argued . . . that fundamental principles, that is, principles that are not derived from other principles, do not rest on factual grounds’ (Rescuing Justice and Equality, p. 278, emphasis added). In what follows, to avoid confusion I shall use ‘ultimate principle’ to refer to a principle that is not derived from any other principle but I shall refrain from using the expression ‘fundamental principle of justice’.

¹⁰ See Cohen, G. A., ‘Facts and Principles’, Philosophy and Public Affairs 31 (2003), pp. 211–45CrossRefGoogle Scholar, at 241.

¹¹ Cohen and Rawls are sometimes regarded as providing us with different conceptions of ideal theory. But in so far as ideal theory is characterized as theory which aims to formulate the principles of justice for governing a perfectly just society in which there is full compliance with those principles, this misunderstands the role that Cohen assigns to ultimate principles of justice. Ultimate principles (including those of justice) are what justify the rules that ought to govern the institutions, practices or behaviour in a society, depending on (amongst other things) the nature of that society, the options available within it, and the degree of non-compliance that is likely to result from the adoption of different rules. See Stemplowska and Swift, ‘Ideal and Nonideal Theory’, section entitled ‘Is Ideal Theory Too Realistic?’ For other recent discussions of ideal theory, see Farrelly, C., ‘Justice in Ideal Theory: A Refutation’, Political Studies 55 (2007), pp. 844–64CrossRefGoogle Scholar; Sen, A., ‘What Do We Want from a Theory of Justice?’, Journal of Philosophy 103 (2006), pp. 215–38CrossRefGoogle Scholar; Robeyns, I., ‘Ideal Theory in Theory and Practice’, Social Theory and Practice 34 (2008), pp. 341–62CrossRefGoogle Scholar; Valentini, L., ‘On the Apparent Paradox of Ideal Theory’, Journal of Political Philosophy 17 (2009), pp. 332–55CrossRefGoogle Scholar; Simmons, A. J., ‘Ideal and Nonideal Theory’, Philosophy and Public Affairs 38 (2010)CrossRefGoogle Scholar, pp. 5–36; Mason, A., ‘Rawlsian Theory and the Circumstances of Politics’, Political Theory 38 (2010), pp. 658–83CrossRefGoogle Scholar; Hamlin, A. and Stemplowska, Z., ‘Theory, Ideal Theory and the Theory of Ideals’, Political Studies Review 10 (2012), pp. 48–62CrossRefGoogle Scholar.

¹² See Rawls, A Theory of Justice, pp. 130/112–13, 579/507.

¹³ See Rawls, Political Liberalism, p. xix. In Rawls's view, even ideal theory is subject to this feasibility constraint. Indeed he sometimes characterizes the form of ideal theory he defends as ‘realistically utopian’: see, for example, Rawls, J., Justice as Fairness: A Restatement (Cambridge, Mass., 2001), p. 4Google Scholar; Rawls, J., The Law of Peoples (Cambridge, Mass., 1999), pp. 4–6Google Scholar, 11–12. Rawls's feasibility constraint would be regarded by some as insufficiently realistic. John Dunn and Raymond Geuss, for example, argue that political theorizing needs to start from an analysis of our concrete historical circumstances (see Dunn, J., Interpreting Political Responsibility (Cambridge, 1990), ch. 12)CrossRefGoogle Scholar, including the way in which our actual institutions operate and what actually motivates us (see Geuss, R., Philosophy and Real Politics (Princeton, 2008), p. 9)Google Scholar. A feasibility constraint may take a variety of different forms. Although the Rawlsian version will be my focus, nothing in my argument turns on its particular form. David Miller distinguishes between technical and political feasibility, where technical feasibility concerns whether ‘a proposal contravenes physical laws or rock bottom social or psychological laws’, whereas political feasibility concerns ‘whether it can command sufficient political support to be adopted’ (Miller, ‘Political Philosophy for Earthlings’, p. 46). He then claims that any defensible feasibility constraint must fall somewhere between the two. For further relevant discussion, see Raikka, J., ‘The Feasibility Condition in Political Theory’, Journal of Political Philosophy 6 (1998), pp. 27–40CrossRefGoogle Scholar; Swift, A. and White, S., ‘Political Theory, Social Science, and Real Politics’, Political Theory: Methods and Approaches, ed. Leopold, and Stears, , pp. 49–69Google Scholar.

¹⁴ Cohen, Rescuing Justice and Equality, p. 251.

¹⁵ David Miller distinguishes a number of different ways in which a principle might be grounded in the facts: it might be entailed by some further principle in conjunction with facts; the facts might provide evidence for the principle; or various facts might be presuppositions of the principle, for example, the fact of human self-consciousness might be a presupposition of a principle that everyone should have an equal right to liberty, in the sense that it would not apply if human beings lacked self-consciousness in the way that non-human animals arguably do: see Miller, ‘Political Philosophy for Earthlings’, sect. 2. (For Cohen's response to the very idea of a principle's being presuppositionally grounded, see Rescuing Justice and Equality, pp. 335–6.) If we think of principles as being grounded in facts in virtue of satisfying constraints that are sensitive to the facts, such as a feasibility constraint, then this would provide us with a further ‘indirect’ way in which principles might be grounded in facts.

¹⁶ See Cohen, Rescuing Justice and Equality, pp. 236–7.

¹⁷ See Cohen, Rescuing Justice and Equality, pp. 233–6.

¹⁸ Cohen, Rescuing Justice and Equality, pp. 251–2.

¹⁹ See Stemplowska and Swift, ‘Ideal and Nonideal Theory’, sect. ‘Is Ideal Theory Too Realistic?’.

²⁰ Cohen, Rescuing Justice and Equality, p. 279.

²¹ Cohen, Rescuing Justice and Equality, p. 233. Cohen refers to this as ‘the clarity of mind requirement’.

²² See Scheffler, S., ‘Is the Basic Structure Basic?’, The Egalitarian Conscience: Essays in Honour of G. A. Cohen, ed. Sypnowich, C. (Oxford, 2006), pp. 102–29CrossRefGoogle Scholar, esp. 102–9.

²³ Rawls, A Theory of Justice, ch. 6.

²⁴ Rawls, A Theory of Justice, p. 334/294.

²⁵ See Murphy, L., ‘Institutions and the Demands of Justice’, Philosophy and Public Affairs 27 (1999), pp. 251–91CrossRefGoogle Scholar, at 257; Abizadeh, ‘Cooperation, Pervasive Impact, and Coercion’, p. 328. It is striking that when Rawls contemplates the issue of how to ensure that the principles of justice which apply to individual behaviour cohere with the principles of justice which govern institutions he does not even entertain the idea that the same principles might apply to both: ‘The simplest thing to do . . . is to use the principles of justice as part of the conception of right for individuals. We can define the natural duty of justice as that to support and further the arrangements that satisfy these principles; in this way we arrive at a principle that coheres with the criteria for institutions’ (Rawls, A Theory of Justice, p. 335/295).

²⁶ Rawls, Political Liberalism, p. 266.

²⁷ For further discussion, see Mason, A., Living Together as Equals: The Demands of Citizenship (Oxford, 2012), pp. 57–62CrossRefGoogle Scholar.

²⁸ See Cohen, Rescuing Justice and Equality, p. 61. Otsuka, Michael expresses scepticism about the existence of such prerogatives in his ‘Prerogatives to Depart from Equality’, Political Philosophy, ed. O'Hear, A. (Cambridge, 2006), pp. 95–112Google Scholar.

²⁹ This could provide a basis for justifying the conclusion that different principles of justice apply to personal behaviour than apply to the basic structure of society: if the principles of justice that apply to the basic structure would place unreasonable demands on citizens were they to be applied to personal behaviour then they need to be adjusted. See Murphy, ‘Institutions and the Demands of Justice’. For different grounds for rejecting the idea that the same principles of justice must apply to both sites, see Shiffrin, S., ‘Incentives, Motives, and Talents’, Philosophy and Public Affairs 38 (2010), pp. 111–42CrossRefGoogle Scholar.

³⁰ It can be regarded as a disabler in Jonathan Dancy's sense: see Dancy, J., Ethics without Principles (Oxford, 2004), pp. 38–43CrossRefGoogle Scholar.

³¹ See, for example, Nagel, T., Equality and Partiality (New York, 1991) pp. 178–9Google Scholar.

³² See, for example, Cohen, Rescuing Justice and Equality, pp. 8–11.

³³ See Cohen, Rescuing Justice and Equality, pp. 10–11, 71, 391. (But contrast Cohen, Rescuing Justice and Equality, p. 389, where Cohen seems to endorse the idea that justice itself is a compromise between personal and impersonal standpoints.)

³⁴ Rawls, A Theory of Justice, p. 334, italics added.

³⁵ Similarly, when Rawls discusses the duty of mutual aid he says that it is a duty to help others when they are in need ‘provided that one can do so without excessive risk or loss to oneself’ (Rawls, A Theory of Justice, p. 114) and he observes more generally that ‘while we have a natural duty to bring about a great good, say, if we can do so relatively easily, we are released from this duty when the cost to ourselves is considerable’ (A Theory of Justice, p. 117).

³⁶ In that passage Rawls maintains that ‘the attempt to forestall . . . [the undesirable consequences of economic choices permitted by a just basic structure] by restrictive rules that apply to individuals would be an excessive if not impossible burden’ (Rawls, Political Liberalism, p. 266).

³⁷ I am not here claiming that there is a straightforward inference from facts about what human beings are like to conclusions about what it would be unreasonably or excessively demanding to require them to do. For a critique of that idea, see Estlund, D., ‘Human Nature and the Limits (If Any) of Political Philosophy’, Philosophy and Public Affairs 39 (2011), pp. 207–37CrossRefGoogle Scholar, at 223–4. But I am claiming that in some cases facts about what human beings are like are relevant for determining what it would be unreasonably or excessively demanding to require them to do. In reaching a view about whether a feature of human nature is relevant in determining what it would be unreasonably demanding to require them to do, we have to make a judgement about ‘whether the feature's moral value or significance suits it to have this kind of weight’ (Estlund, ‘Human Nature and the Limits (If Any) of Political Philosophy’, p. 228). The fact that a person (or human beings in general) has projects that give coherence and meaning to her life but which she would have no time or energy left to pursue were she to fulfil the requirements of a principle seems to me have the kind of significance that makes it relevant to determining the acceptability of that principle.

³⁸ In principle it would be coherent to think that a proper integration of the personal and impersonal standpoints requires us both to accept the constraint that adequate principles of justice cannot be unreasonably demanding and to accept the idea that even when justice is not unreasonably demanding, there are moral prerogatives that entitle us to depart from what it requires – to act in a way that is unjust but nevertheless morally justified. I do not rule out the possibility of combining these ideas. But it is not clear that this position would be particularly compelling, for we might plausibly hold that if a principle doesn't place unreasonable demands on individuals, then there is no prerogative to depart from what it requires.

³⁹ Williams, A., ‘Incentives, Inequality, and Publicity’, Philosophy and Public Affairs 27 (1998), pp. 225–47CrossRefGoogle Scholar, at 233.

⁴⁰ Williams, ‘Incentives, Inequality, and Publicity’, pp. 243ff.

⁴¹ Williams, ‘Incentives, Inequality, and Publicity’, p. 244. Williams also thinks that social unity may be non-instrumentally valuable because it is necessary for the achievement of personal autonomy and for the realization of a particular form of community (‘Incentives, Inequality, and Publicity’, p. 244).

⁴² Cohen, Rescuing Justice and Equality, ch. 8.

⁴³ As David Copp argues, ‘justice may . . . be fragile even in the best of circumstances’ (Copp, D., Pluralism and Stability in Liberal Theory’, Journal of Political Philosophy 4 (1996), pp. 191–206CrossRefGoogle Scholar, at 204).

⁴⁴ It is perhaps less clear in the case of Pareto efficiency: see Tomlin, P., ‘Survey Article: Internal Doubts about Cohen's Rescue of Justice’, Journal of Political Philosophy 18 (2010), pp. 228–47CrossRefGoogle Scholar, at 240–6. In the case of publicity, one might think that one type of publicity is conceptually connected to justice, viz. the idea that any adequate principle of justice must be publicly justifiable, that is, justifiable in terms that any reasonable person could accept. This idea has been central to liberal political theory; indeed some have regarded it as constitutive of liberalism.

⁴⁵ Cohen, Rescuing Justice and Equality, p. 365, n. 36.

⁴⁶ J. Rawls, ‘Kantian Constructivism in Moral Theory’, p. 347.

⁴⁷ Rawls, A Theory of Justice, p. 6/6; see also Rawls, J., ‘Reply to Alexander and Musgrave’, John Rawls: Collected Papers, ed. Freeman, S. (Cambridge, Mass., 1999), pp. 232–53Google Scholar, at 237.

⁴⁸ Williams, ‘Justice, Incentives and Constructivism’, p. 489.

⁴⁹ Rawls, A Theory of Justice, pp. 3–4/3–4.

⁵⁰ See Wittgenstein, L., Philosophical Investigations, trans. Anscombe, G. E. M. (Oxford, 1953)Google Scholar, part II, sect. x: ‘If there were a verb meaning “to believe falsely”, it would not have any significant first person present indicative’. Statements of the kind ‘Regardless of whether p is true, I believe that p because it is useful to do so’ would also seem to violate a norm that is constitutive of having the attitude of belief.

⁵¹ See Rawls, A Theory of Justice, p. 4/4.

⁵² I say ‘relative neglect’ because Cohen does implicitly give the issue of whether principles of justice must serve social purposes some partial attention in discussing the question of whether normative principles must guide practice: see Rescuing Justice and Equality, pp. 267–8. But here he is discussing ultimate normative principles in general rather than ultimate principles of justice in particular, and he simply reiterates the point that to suppose that ultimate normative principles, by their nature, must guide is to confuse them with rules of regulation.

⁵³ For all I have argued, it would also be possible to retain Rawls's constructivist methodology and accept that ultimate principles of justice are not grounded in any facts in the relevant sense, but then deny that Rawls is offering a constructivist theory of ultimate principles of justice. (Indeed this is essentially Williams's proposal in his ‘Justice, Incentives and Constructivism’.) But I would resist the idea that ultimate principles of justice must be entirely independent of the facts since in my view they can be constrained by facts, such as facts about what is feasible. That is consistent, however, with supposing that ultimate principles of justice are not directly grounded in facts of any sort.

⁵⁴ On this issue I have not changed my view substantially from the one I defended in ‘Just Constraints’, British Journal of Political Science 34 (2004), pp. 251–68.

⁵⁵ This does, however, raise again the issue I put to one side on p. 533, viz., whether what can count as a conception of justice depends upon facts about what is feasible. I am implicitly denying that this is so. Here the distinction that Hamlin and Stemplowska draw between the theory of ideals (or theorizing about ideals) and ideal theory is relevant: theorizing about an ideal such as justice is concerned with the nature of that ideal and need not be subject to any feasibility constraints at all and (unlike both ideal and non-ideal theory) is not concerned with issues of institutional design: see Hamlin and Stemplowska, ‘Theory, Ideal Theory and the Theory of Ideals’, pp. 52–3, 60.

⁵⁶ I would like to thank Chris Armstrong, David Estlund, Brad Hooker, David Owen, Adam Swift, Andrew Williams, Lea Ypi and participants in the Jerusalem Political Philosophy Forum at the Hebrew University for helpful comments on an earlier version of this article.