This investigation deals with the active learning hypothesis in Karasek and Theorell's (1990) job demands-control model. The active learning hypothesis holds that high levels of learning and selfefficacy will occur among incumbents of high job demands/high job control jobs, whereas low levels of learning and self-efficacy will be found in low demands/low control jobs. This study tested these notions in the context of a two-wave study conducted over a period of one year among 876 Dutch teachers. Regression analysis revealed that job demands had a lagged negative (rather than a positive) effect on learning and self-efficacy; as expected, job control had a positive effect. Thus, the highest levels of learning and self-efficacy were found among incumbents of high control/low demands jobs (and not among incumbents of high control/high demands jobs). Further, the effects of changes in work characteristics on changes in learning behaviour and self-efficacy were examined, providing additional evidence that especially the transition from a low demands/low control to a high demands/ low control job is associated with a strong deterioration of learning and self-efficacy. It is concluded that future research should address the interrelationships between learning and strain, preferably using longitudinal designs.