¹ I personally would not include condition (4), for it imports non-normative considerations into the definition of what is essentially a normative view. But I will follow Sumner's more restrictive usage in what follows.

² Actually, Sumner denies that natural rights theories treat rights as absolutely basic; see 127 below.

³ The “just about” qualification is needed to exclude meta-ethical views that propose naturalistic definitions of moral terms. The view that “right” just means “produces the most happiness” obviously has normative consequences. What can be said, I think, is that any normative position can be combined with any plausible meta-ethical position.

⁴ Although Sumner's formal presentation of Bentham's argument talks of the existenceconditions of rule-systems conferring rights, he often talks interchangeably—or what seems to me interchangeably—of the existence-conditions of rights themselves. I myself cannot see a substantive difference between the two modes of talking, and will therefore speak henceforth just of the existence-conditions of rights, assuming that this covers the existence-conditions of rules as well.

⁵ In different places in the book, Sumner uses “duty” in different senses. In a broad sense, there is a duty whenever a person ought to do something. In a narrower sense duties are essentially relational; a person only has a duty when he owes something to someone else. Now, the duties generated by rights are relational, while the consequentialist duty to maximize is not. But this does not affect my ad hominem argument. As Sumner explains in Chapter 2, the “protected choice” model of rights gives an explanation of the relationality of right-generated duties, an explanation that turns on one person's having control over another's duties. This explanation is incorporated in my analysis of the right to life, even though the clauses of that analysis use only the same deontic concepts as a statement of Sumner's consequentialist duty.

⁶ This condition omits reference to the enforcement clause (3) in the original analysis, but only for the sake of brevity. Also, the condition covers only the case where X transfers his right to life to someone else, as in the contract example above, and not the case where he simply renounces his right. Obviously, there will have to be slightly different analyses of all the different powers X has over his right to life.

⁷ The first model is associated with Rawls and Daniels. The second is defended by Peter Singer, who attributes it to Sidgwick (Rawls claims Sidgwick for the first model); it is also implicit in Ross. The third model occurs in Prichard and Bradley. It is worth noting that the choice among these models is independent of one's stand on moral realism/ anti-realism. Sumner says that different views about the ontological status of rights determine different methods of supporting claims about them (107), and in a sense this is true. A moral realist will describe, say, the coherence model as relying on partial insights into moral reality at different levels of generality, while an anti-realist may describe it as adjusting attitudes at different levels. But if we consider the models from the inside, and ask what it is like actually to apply them, we will find no difference. The normative writings of realists and anti-realists who share the same model should be, and usually are, indistinguishable.

⁸ One of Sumner's aims in characterizing natural rights theory is to reflect the natural rights tradition in ethics, and one of his reasons for making the theory realist is that this is what it has, historically, most often been. I agree that most natural rights theories have been realist, but would say that this is because most moral theories period have been realist. Hume and a few others aside, moral philosophy before this century was primarily realist; so the fact that rights theory has generally been realist is not something distinctive of it.