The present study investigated the effects of verbal feedback combinations on the performance of retarded adolescents on a computer-administered discrimination task. Theoretical explications of such effects on learning originated from studies using nonretarded subjects. Theories of feedback and reinforcement effects with normal children are reviewed, together with empirical reports relevant to such theories. The two basic orientations in this area, reinforcement and information, are considered, with the latter approach being adopted in the present study. Theories of performance differences between normals and retardates are considered. How outcome combinations affect retardate discrimination learning is reviewed and compared to similar data from normal subjects. Since a computer terminal was employed in the present study, data concerning feedback effects in computer-assisted instruction is reviewed. Finally, the variable of mode of presentation of feedback is discussed. The results tended to support the information model presented by Spence (1966), in that neither feedback combination nor mode of feedback presentation produced differences in acquisition of learning. The Rb combination produced poorer extinction performance than did the RW and Wb combinations. Girls overall outperformed boys. These results support the contention that the issue of instruction is critical in demonstrating differences in feedback combination effectiveness, and that mild words of praise or reproof function more on an informational than on a reinforcement level.