Abstract
Inferences from desired ends to intended necessary means seem to be among the most unproblematic elements of practical reasoning. A closer look dissolves this appearance, however, when we see that such inferences are defeasible. We can nevertheless understand such inferences as leading to the adoption of plans, by analogy with inferences leading to explanations. Plans should satisfy at least some important ends desired by the agent, be consistent with the satisfaction of other desired ends, and be inconsistent with as few desired ends as possible. A rational plan may rule out the satisfaction of some desires, however, and this feature explains the defeasibility of such inferences.