There has been a recent upsurge of interest in exploring how choices of methods and timing of instruction affect the rate and persistence of learning. The authors review three lines of experimentation--all conducted using educationally relevant materials and time intervals--that call into question important aspects of common instructional practices. First, research reveals that testing, although typically used merely as an assessment device, directly potentiates learning and does so more effectively than other modes of study. Second, recent analysis of the temporal dynamics of learning show that learning is most durable when study time is distributed over much greater periods of time than is customary in educational settings. Third, the interleaving of different types of practice problems (which is quite rare in math and science texts) markedly improves learning. The authors conclude by discussing the frequently observed dissociation between people's perceptions of which learning procedures are most effective and which procedures actually promote durable learning. (Contains 4 figures.)