A theory recently proposed by Whitman postulates that self-instructional training procedures should have special utility for low ability individuals. Although past studies have shown that self-instructional training programs can be employed to improve the performance of mentally retarded individuals in work situations, research has not examined whether this training format is superior to external instruction. To test Whitman's theory, mentally retarded adults were taught to perform a complex sequencing task through one of two training formats: self-instruction or external instruction. Performance was evaluated through the examination of accuracy measures obtained during training, maintenance, and generalization assessments. Results indicated that participants receiving self-instructional training were able to achieve and sustain a higher level of performance than participants receiving external instruction. While self-instructionally trained subjects also showed superior performance on the generalization assessment, these results were not statistically significant. Secondary analyses revealed that higher ability participants, self-instructionally trained participants, and participants who reached criterion on the training and generalization tasks self-verbalized more frequently during the various assessments. In contrast to prediction, there was no significant difference in the length of time required to train individuals in the two instructional groups.