Assessing the contribution of the top-down processing on the learned predictiveness effect: a strategy based on shortening the duration of the test trials

Abstract

The learned predictiveness (LP) procedure has been interpreted as useful to discuss the mechanisms that regulate a form of attention related to the limited learning resources of the organisms. In this procedure, in Stage 1, half of the cues are established as accurate predictors of their outcomes, and the other half are established as poorer predictors. In Stage 2, all cues are equally predictive of a new outcome. On test, participants rate the likelihood that the cues would produce Stage 2 outcomes. It has been consistently found that participants rate the accurate predictors higher than poorer predictors. This pattern of results has been interpreted as supporting the notion that stimulus associability is directly related to predictive accuracy. We report an experiment investigating this LP effect in which we manipulated the maximum time to respond on each trial: 20 s vs. 4 s. Participants in the long-time period condition showed both a clear discrimination in Stage 1 and a LP effect on the test. Participants in the short-time period condition, however, showed a clear discrimination in Stage 1, but did not show the LP effect on the test. This suggests that the mechanism involved in the LP effect requires more time than the automatic associative mechanisms involved in discrimination learning. These results therefore invite to reconsider the importance of the LP effect in discussing the mechanisms regulating stimulus associability.