Derek Parfit is surely right when he says, at the beginning of Reasons and Persons, that many of us want to know what we have most reason to do. Several theories attempt to answer this question, and Parfit begins his discussion with the best-known case: the Self-interest Theory, or S. When applied to actions, S claims that “(S2) What each of us has most reason to do is whatever would be best for himself, and (S3) It is irrational for anyone to do what he believes will be worse for himself” (Parfit 1984, p. 8). Objections to this theory are of many kinds, and in the first part of his book Parfit examines the objection that S is, in various ways, self-defeating. One such objection is that S implies we sometimes cannot avoid acting irrationally, but Parfit claims it is not a good objection to S that it has this implication. I disagree. The purpose of this paper is to introduce the objection in more detail (section 1), and then to argue that each of Parfit's responses to the objection is inadequate (sections 2, 3 and 4).