The paper argues that the capacity of practical rationality can be best understood by investigating the conditions under which we would regard human behavior irrational. Practical rationality does not, in general, require that agents do what they judge best, since many cases such judgements are not available. But, contrary to some recent opinions, actions not controlled by agents’ judgement about the value of their own (future) actions when such a judgment is available cannot be practicaly rational. What practical rationality requires is that agents do not revise their intentions unless they have sufficient reason for doing so.