In normative economic analysis, it is conventional to treat each person’s preferences as that person’s own standard of value, and as the standard by which the effects of public policies on that person should be valued. The proposal that preferences should be treated in this way is usually qualified by two apparently natural conditions—that preferences are internally coherent, and that they reflect the considered judgements of the person concerned. However, there is now a great deal of evidence suggesting that, in many economic environments, preferences of the required kind simply do not exist. It seems that the preferences that govern people’s actual behaviour are often incoherent and unstable. This prompts the following question: Is there a defensible form of normative economics which respects each individual’s actual preferences, whatever form they take? I shall try to show that there is.