An experiment is reported that attempted to dissociate the effects of uncontrollability from those of failure in the standard learned helplessness induction procedure involving instrumental tasks. The task was administered to both university and high school students. The primary finding was a grade by controllability interaction, whereby uncontrollability had a larger effect for university students who became much more helpless. Further analysis revealed that while controllability or uncontrollability was the main determinant of university student performance, the performance of the high school students depended primarily upon success or failure. Perhaps we have previously generalized too widely from a restricted subject population.