Learning to use functional rules in inference tasks

Abstract.— Earlier studies have established that subjects perform less well in inference tasks with nonlinear rules than in tasks with linear rules, and that one source of the lower level of performance in nonlinear tasks is that the subjects cannot utilize a nonlinear rule as well as they utilize a linear rule. The three experiments in this paper investigate whether utilization of a nonlinear rule can be improved by training. The results show that there is some improvement with training, but the improvement could not be attributed to feedback, or to the learning of specific cue and criterion values. Cognitive feedback did not produce higher performance than ordinary outcome feedback.