In an investigation of social inference (the ability to decode and comprehend cues), prompt questions of decreasing cognitive complexity were employed to train 60 mildly mentally retarded adolescents to logically interpret pictorial stimuli. Cartoon line drawings were used illustrating social situations representing a variety of family and peer interactions. Three levels of prompts, beginning at a low level and proceeding to higher levels, were employed. Analysis of pre- and post-training scores on the Test of Social Inference indicated positive changes as a result of training. The main effect for number of sessions was not significant, nor was the interaction of condition by number of sessions. It appeared that only the middle and high level mental age groups benefited from training. (CL)