What is represented in memory after statistical learning?

AbstractStatistical learning is a powerful mechanism that allows us to rapidly extract structure from the environment. However, nuances of what structure is extracted—for example, whether reliable groups are stored without knowledge of their constituent item order—are not well understood, leaving us with open questions about how this mechanism supports behaviour. Here, we extend prior work on the representation of statistical structure by asking what specific aspects of structure matter for memory judgments. We consider three candidates for memory representation: transitional probability, order-independent group information, and position tags. Participants watched a stream of shape triplets and then completed a recognition memory test designed to isolate contributions of transitional probability, group, and position. We demonstrate that although memory for transitions alone would be sufficient for knowledge of triplets, participants showed evidence of representing both transitional probability and group. Our data highlight statistical learning as a mechanism enabling generalization across experiences.

Return to previous page