Two classes of experiment on the role of attention in discrimination learning are reviewed: (a) Investigations of the effect of attention on the amount learned about different cues have been interpreted as disproving noncontinuity theory (according to which animals attend to only one cue at a time). The fact that animals learn something about a second cue, however, does not prove that attention has no effect on learning, and more recent evidence shows that it does, (b) If animals do not automatically attend to all cues, part of what they must learn in order to solve a discrimination problem is to attend to the relevant cue. Experiments on the acquired distinctiveness of cues, transfer along a continuum, and reversal learning provide evidence for the importance of such classificatory learning.