Statistical Structure and Implicit Serial Learning

Abstract

In serial reaction time (RT) tasks, RT decreases faster with practice when a short sequence of trials is repeated continuously than when the sequence of trials is random (e.g., Nissen & Bullemer, 1987). Learning is implicit because Ss are not asked to learn the sequence and they often seem unaware of it. Relative to a random sequence, the repeating sequence imposes extreme statistical constraints. The present research examined the relation between these statistical constraints and implicit serial learning. Two experiments showed that the greater the statistical structure, the greater the reduction in RT with practice. Statistical structure was manipulated by varying the structure of a repeating sequence (Experiment 1) and by varying the degree to which sequences approximated the statistical structure of a repeating sequence (Experiment 2). The implications of these results for theories of implicit serial learning are discussed.