Discrimination blocking: acquisition versus performance deficits in human contingency learning.

We compared acquisition and performance accounts of human contingency learning. After solving a discrimination in Phase 1, in which Cue A predicted the occurrence of the outcome and Cue B predicted its nonoccurrence (A+/B-), a new discrimination (X+/Y-) was superimposed in Phase 2 (AX+/BY-). The participants were finally trained in Phase 3 with the added discrimination, which either maintained the same contingencies as those in Phase 2 (X+/Y-; Experiment 1) or reversed the contingencies (X-/Y+; Experiment 2). According to competitive-learning theories, there should be no learning of the added discrimination in Phase 2, so that no advantage or disadvantage for this discrimination should be observed in Phase 3. In contrast, performance theories, such as the comparator hypothesis, contend that learning of the added discrimination in Phase 2 should proceed normally; so, in Phase 3, an advantage for the added discrimination should be observed in Experiment 1, but a disadvantage should be observed in Experiment 2. Our participants learned about the added discrimination and generally showed the effects predicted by the comparator hypothesis.