Explaining Learned Predictiveness: Roles of Attention and Integration of Associative Structures

In 3 experiments, participants were trained in an associative learning paradigm in which they learned the relation between consumption of certain foodstuffs and the type of allergic reaction shown by a fictional patient. Experiment 1 demonstrated the learned predictiveness effect, showing that cues that had served as good predictors of outcomes in an initial phase of training were especially effective in a test given after a second phase of training in which learning about the same cues, but with different outcomes, had been required. Experiment 2 showed that this effect could be obtained when the two phases of training occurred in reverse order, so that the critical cues were established as good or bad predictors only after the associations tested in the final test had been acquired. This learned predictiveness effect cannot be explained by an enhancement of the associability of the predictive cues that facilitated learning about them in phase two. This encouraged us to consider 2 alternatives to associability for explaining learned predictiveness: (a) that training a cue as a good predictor increases its effective salience, thus enhancing its power to evoke responding on test and (b) that learned predictiveness is the result of a nonattentional process in which subjects integrate information acquired in the separate phases of training. Support for the latter came from Experiment 3. which showed that a modified test procedure. designed to reduce the tendency to integrate across phases, eliminated the learned predictiveness effect.