The authors used a task-switching paradigm to investigate set shifting ability in schizophrenia. This paradigm included 2 choice reaction time (RT) tasks: up-down and right-left. Switching tasks were associated with costs (i.e., longer RT in task-switch trials than in task-repetition trials); patients responded more slowly than controls and suffered greater switching costs, were as efficient as controls in engaging in an upcoming task set, and were faster than controls in disengaging from the previous task set. There were indications that patients quickly forgot what each keypress indicated, making it necessary for them to acquire response meaning information anew in each trial. To test this notion, the authors subsequently tested normal participants in conditions in which response meaning information needed to be acquired anew in each trial. These participants produced a pattern of switching costs resembling that of patients. Results suggest that set switching difficulties in schizophrenia, as exhibited in the present paradigm, reflect poor memory for task context information.