The effectiveness of the inductive instructional method of general case programming was compared to the deductive approach of rule learning, in attempting to facilitate generalized learning and maintenance of a community living skill for adults with mental retardation. The impact of functional ability on the type of instruction was also assessed. Thirty-six adults were randomly assigned to one of three conditions (rule, general case, or control) and were designated as high or low functioning based on a median split from their pretest performance on a tablesetting task. The findings, following training on tablesetting tasks, indicate that both general case programming and rule learning were effective for those in the higher functioning group, but only general case programming was effective in promoting generalization and maintenance for those in the lower functioning group. Generalized skill performance was maintained at a 1-month followup. Performance of trained participants in the high-functioning group was comparable to the tablesetting ability of the nondisabled. Those in the low-functioning group, however, did not show comparable performance irrespective of training. (JDD)