¹ Many, including Gerald Postema, Bernard Boxill, Jan Boxill, Geoffrey Sayre-McCord and Norman Dahl, provided me valuable advice in the preparation of this paper. Special thanks to Thomas Hill, whose wisdom and encouragement are boundless.

² See, e.g., Benn, S. I., ‘Punishment’, The Encyclopedia of Philosophy, 8 vols., ed. Edwards, Paul, New York, 1972, vii. 30Google Scholar; Hart, H. L. A., Punishment and Responsibility, Oxford, 1982, pp. 231–2Google Scholar.

³ See, e.g., Byrd, B. Sharon, ‘Kant's Theory of Punishment: Deterrence in its Threat, Retribution in its Execution’, Law and Philosophy, viii (1989), 151–200CrossRefGoogle Scholar; Scheid, Don E., ‘Kant's Retributivism’, Ethics, xciii (1983), 262–82CrossRefGoogle Scholar.

⁴ See Murphy, Jeffrie G., ‘Does Kant Have a Theory of Punishment?’, Columbia Law Review, lxxxvii (1987), 509–32CrossRefGoogle Scholar.

⁵ I do not claim that Kant is in no sense a retributivist, for the term is used in widely varying ways. Jean Hampton, for example, provides a Kantian account of the idea of retribution that squares well with the interpretation of Kant I will provide here. See Murphy, Jeffrie G. and Hampton, Jean, Forgiveness and Mercy, New York, 1988, pp. 124–43CrossRefGoogle Scholar.

⁶ For a full discussion of ‘thoroughgoing’ retributivism see Hart, , pp. 231–2Google Scholar. The thoroughgoing retributivist also requires that we punish only those who voluntarily have committed a crime and that punishment be in proportion to the wickedness of the offence.

⁷ Byrd argues, for example, that the justification for making the threat of punishment part of the criminal law on Kant's account is general deterrence. Retribution, while it does not justify punishment, places limits on its just execution. Scheid contends that Kant's retributivism is partial rather than thoroughgoing because its justification is deterrence, while its answers to questions of whom to punish and how much are retributive. Neither interpretation clearly provides the kind of underlying explanation for all commitments that I will suggest here.

⁸ I do not pretend to offer an interpretation the historical Kant would have accepted without pause. After all, I will argue that some of Kant's more concrete commitments are inconsistent with his larger theory. I do intend an interpretation fully in the spirit of Kant's basic moral and political commitments.

⁹ Kant, Immanuel, The Metaphysics of Morals, trans. Gregor, Mary, New York, 1991CrossRefGoogle Scholar. All initial page references to The Metaphysics of Morals (MM) are to this edition. Bracketed numbers refer to vol. vi of the Prussian Academy Edition.

¹⁰ It is reasonable to conclude that both limitations on action and on choice may be unjust given, as we will see, that coercion is just only when it hinders hindrances to freedom. Coercion operates in large part as a threat, influencing decisions about how to act. When it interferes with just decisions, the coercion is unjust.

¹¹ Also see Kant's, Lectures on Ethics, trans. Infield, Louis, Indianapolis, 1979, p. 55Google Scholar.

¹² Of course coercion might take forms other than the criminal punishment with which we are familiar and which Kant discusses in some detail. For example, compliance with some legal standards might be encouraged through the kind of civil penalties imposed by our own systems of tort and contract. Although I will not take up the question here, the decision as to what other forms incentives should take presumably would depend on what best will promote the fundamental concerns for freedom, autonomy and equality that we will discuss below.

¹³ Kant does not claim that there ever was such a contract. The notion of the original contract is simply a tool to aid his discussion of justice and the legitimate state.

¹⁴ We will consider the significance of rational nature more fully in sect. C.

¹⁵ That Kant intended some such relationship is suggested by his references, throughout the Rechtslehre, to maxims consistent with the freedom granted by universal law (MM 56 [230]), or to ensuring that one is an end and not a mere means for others (MM 62 [236]), or to punishments that make humanity into something abominable (MM 142 [333]). Each of these harks back to some formulation of the categorical imperative and suggests that less fundamental components of Kant's system are charged with giving effect to this most basic principle in various contexts and at various levels of generality.

¹⁶ Kant, Immanuel, Groundwork of the Metaphysic of Morals, trans. Paton, H. J., New York, 1964Google Scholar. All initial page references to the Groundwork (GW) are to this edition. Bracketed numbers refer to Prussian Academy vol. iv.

¹⁷ I take seriously Kant's Groundwork statement that the various formulations of the categorical imperative ‘are at bottom merely so many formulations of precisely the same law, one of them by itself containing a combination of the other two’ (GW 103 [436]). Among the formulations, that of the kingdom of ends, with its legislative perspective, seems best designed to address questions of social justice. With its focus on respect for rational nature, the formula of the end in itself is in turn well designed to help us determine what laws would govern a kingdom of ends.

¹⁸ A subjective end is whatever gives a particular rational agent reasons for acting one way rather than another (GW 96 [428]).

¹⁹ I discuss this reading in much greater detail in ‘Kant's Formula of Humanity and the Pursuit of Subjective Ends’, Proceedings of the Eighth International Kant Conference, ed. Robinson, Hoak Milwaukee, 1995, pp. 697–704Google Scholar.

²⁰ In a lengthier discussion, one would want to consider objections some (e.g., Onora O'Neill) advance against accounts of moral duty that rely on hypothetical consent. See, e.g., O'Neill, Onora, Constructions of Reason, New York, 1989, pp. 109–10Google Scholar. As that consent is such a small part of the present discussion, I do not take up those issues here.

²¹ Each freedom is significant because each emphasizes a different concern that we must consider in deciding what morality requires. We must ensure not only that fundamental interests are duly protected, but that these protections do not curtail the pursuit of subjective ends too severely.

²² For the more extensive discussion on which I base my brief consideration of price and dignity, see Hill, Thomas E. Jr, ‘Humanity as an End in Itself’, Dignity and Practical Reason in Kant's Moral Theory, Ithaca, 1992, pp. 41–2Google Scholar.

²³ Of course it is these passages, in particular, that have influenced those who classify Kant as a thoroughgoing retributivist. See, e.g., Pincoffs, Edmund L., The Rationale of Legal Punishment, Atlantic Highlands, N.J., 1966, p. 9Google Scholar; Brandt, Richard, Ethical Theory, Englewood Cliffs, N.J., 1959, pp. 496–8Google Scholar.

²⁴ The most familiar example of an ideal theory of justice is Rawls's, John A Theory of Justice, Cambridge, Mass., 1971Google Scholar.

²⁵ For more and less friendly criticisms of ideal theory see, e.g., Baier, Annette, ‘Theory and Reflective Practices’, Postures of the Mind: Essays on Mind and Morals, Minneapolis, 1985, pp. 207–27Google Scholar; Feinberg, Joel, ‘Rawls and Intuitionism’, Reading Rawls, ed. Daniels, Norman, New York, 1989, pp. 108–24Google Scholar.

²⁶ For Rawls's, discussion of the ideal components of his own account of justice see Theory, pp. 8–9, 245–8Google Scholar.

²⁷ See Theory, pp. 60, 302 for the first and final statements of these principles.

²⁸ Kant, Immanuel, ‘On the Common Saying: “This may be true in Theory, but does not apply in Practice”’, Kant's Political Writings, ed. Reiss, Hans, New York, 1989, pp. 61–92Google Scholar. All initial page references to ‘Theory and Practice’ (Theory) are to this edition. Bracketed numbers refer to Prussian Academy vol. viii.

²⁹ Of course, Kant does not announce his use of idealization in the way, for example, Rawls does. It is a feature of his theory that we recognize in examining its various components.

³⁰ This is not to say that no one who is convicted of a crime could display the characteristics of a citizen. After all, some are convicted, though they have done no wrong. The idea is that one who voluntarily commits an injustice cannot provide an appropriate example of the standpoint from which to determine what justice requires.

³¹ By this Kant clearly does not mean that the criminal loses human dignity, therefore freeing us from the requirement that we treat her as an end. If so, he would not worry, for example, about ‘mistreatment that could make the humanity in the person suffering [punishment] into something abominable’ (MM 142 [333]). Rather, Kant says, she has lost ‘the dignity of a citizen’. A citizen chooses to use her freedom independently; she is not forced by state power to submit to punishment.

³² In a world where knowledge is limited, and our means of determining guilt imperfect, justice presumably demands that we employ reasonably reliable procedures in deciding guilt or innocence. To determine that this is so, and what procedures are adequate, requires an augmented ideal theory of the sort I suggest in sect. C below.

³³ Since punishment works to deter actual and potential criminals, and to protect actual and potential victims, punishment decisions directly concern the freedom of criminals and victims alike. These decisions thus raise the question of whether the freedom of each victim has been equally respected, likewise for each criminal.

³⁴ See, for example, Kant's, Groundwork, pp. 74–5Google Scholar [407] and his Religion Within the Limits of Reason Alone, ed. Greene, Theodore M. and Hudson, Hoyt H., New York, 1960, p. 16Google Scholar; Prussian Academy vol. vi p. 20.

³⁵ Among the special reasons we might have for wanting to reject the law of retribution (beyond the desire for a consistent theory) are its insensitivity to interests in rehabilitation and to mitigating factors including age, prior criminal record and, apparently, mens rea. Considered in light of higher Kantian principles and empirical facts about contemporary society, these interests and factors might warrant variations in the type and degree of punishment allotted, or differentiations among offenders, that the law of retribution cannot countenance.

³⁶ For a similar view, see Jean Hampton's analysis in Murphy and Hampton, pp. 138–47.

³⁷ In a fuller treatment of Kant's penal theory, we would want to ask just what the relationship is, on Kant's account, between justifications for the state's threat of punishment and for the execution of that threat. On my view, of course, each is ultimately justified by appeal to considerations of freedom, autonomy and equality. Threat and execution, however, likely will raise different concerns in the complex circumstances of our world and so deserve separate and detailed consideration. We must take care, for example, that threats do not discourage legitimate action and must ask whether special circumstances such as age, lack of mental capacity, or extreme psychological stress might require us to refrain from executing otherwise appropriate punishments in some cases. B. Sharon Byrd addresses threat and execution in Kant's penal theory, though not with respect to the concrete issues I have in mind. For a more general consideration of the connection between threat and punishment see Farrell, Daniel M., ‘Deterrence and the Just Distribution of Harm’, Social Philosophy and Policy, xii (1995)Google Scholar.

³⁸ Again, this increasing complexity is not a feature of the theory that Kant explicitly acknowledges, but one we discover by comparing the perspectives of the ideal legislator for the kingdom of ends with that of the ideal citizen of a just society. I explore the connection between these perspectives, as well as Kant's social contract account of justice, more fully in my doctoral dissertation. By examining the two perspectives, we can see, for example, that each idealizes the moral rectitude and decision-making capacities of its members. Unlike the perspective of the kingdom of ends, however, that of ideal citizens does not merely focus on fundamental features of rational agency. It emphasizes characteristics basic to citizens associated in civil society. Thus concerns of freedom of individual choice, autonomy and equal dignity become those of freedom and equality under civil laws and of independent participation in civil governance. Issues of adherence to imperatives by beings who can, but might not, comply with the moral law become those of appropriate protections for each in a world where conflict and consequent limits on individual choice and action are inevitable.

³⁹ For a different way of understanding Kant's ideal theory see Korsgaard, Christine M., ‘The Right to Lie: Kant on Dealing with Evil’, Creating the Kingdom of Ends, New York, 1996, pp. 133–58CrossRefGoogle Scholar. I agree with Korsgaard that the formulas of the kingdom of ends and of the end in itself rest on idealizing assumptions. But I argue that Kant makes room for realistic considerations in moral principles that are less fundamental than the categorical imperative; she focuses on the formula of universal law as a standard less ideal than other formulations of Kant's most basic moral principle.

⁴⁰ Mid-level principles would be sufficiently general to apply to many different human societies at different times. They would be appropriate to guide the formation of national constitutions, statutes and other standards themselves designed to address circumstances peculiar to particular societies and historical periods.

⁴¹ In my dissertation I develop many such principles, using various alterations in the perspective of ideal citizens. Among these are substantive principles applicable to issues of criminal justice and social welfare and procedural principles to guide our analysis of the complex issues that confront us in contemporary society. I apply such a principle to the contemporary problem of prison reform in ‘A Kantian Approach to Prison Reform’, forthcoming in Jahrbuch für Recht und Ethik.