Independent processing of statistical regularities in different hierarchical levels
26.4092, Saturday, 16-May, 2:45 pm - 6:45 pm, Pavilion
Jihyang Jun1, Sang Chul Chong1,2; 1Graduate Program in Cognitive Science, Yonsei University, 2Department of Psychology, Yonsei University
Visual statistical learning (VSL) refers to an ability to extract statistical regularities that exist in our environment. When we perceive a natural scene, a number of statistical regularities appear simultaneously in different hierarchical levels. The current study investigated how such regularities could be acquired. Specifically, we examined the learning of temporal regularities that simultaneously occurred in the local and global levels. Participants passively viewed a stream of hierarchical scenes, in which two different shapes were structured into the local and global levels as in Navon letters. In this stream, the temporal regularities existed among three shapes (i.e., triplet) that always appeared in the same order in each of the levels. Following familiarization, participants performed 2AFC of familiarity judgments between old triplets in the same orders and new triplets in changed orders. In Experiment 1, triplets were presented in one level so that participants could only rely on the temporal regularities in the single level. We found that temporal VSL occurred similarly in each of the two levels. In Experiment 2, participants could rely on the temporal regularities either in the single levels (local or global), or in both levels, since we maintained the hierarchical structure in the test unlike in Experiment 1. The extent of learning was higher in the latter condition than in the former condition, suggesting that participants additively used temporal regularities in each level. Importantly, the summed probability of correct judgments in the local level and that in the global level was higher than the correct percentage in both of the levels. This analysis suggests that temporal VSL in each of the two levels occurred independently. Taken together, our results suggest that human observers can extract the statistical regularities that simultaneously appear in different hierarchical levels, and that the regularities in the two levels are processed independently.