A limitation on the law of effect.

Nineteen mildly or moderately retarded subjects were presented 32 oddity-training trials per day for 10 days with all new etimuli presented on each trial. Six subjects learned oddity, approaching an asymptote of 100 percent correct. Six others showed a strong preference (78 percent) for the odd stimulus but failed to improve. Seven remained in the performance range of 50 to 60 percent without improving. On eight trials per day, the two specific cues of an oddity display, chosen randomly, were presented alone as probes for specific-cue learning. Probe trials were above chance for all subjects. These results show that attention to the relevant dimension, demonstrated by oddity preference, plus adequate reinforcing conditions, indicated by specific-cue learning, did not ensure oddity learning. We concluded that subjects do not necessarily learn about the cues that control behavior. The operation of differential forgetting and/or rehearsal of relative vs. specific vs. specific cues was proposed as a likely explanation.