Encoding modeling of brain activity during naturalistic behavior

概要

One of the central aims of computational cognitive neuroscience is to understand brain functions via modeling of brain activity during naturalistic behavior. Numerous important findings in neuroscience have been derived by taking hypothesis-driven approaches that impose strict conditions to control experimental parameters. However, active discussions have recently occurred over the generalizability of such findings emanated from ecologically invalid experiments. To test the generalizability of previous findings as well as further extend the understanding of the brain, I used encoding modeling in companion with naturalistic, ecologically valid experiments. This paper consists of two studies: 1) An analysis of the effect of selective attention on linguistic semantic information representations in the brain during reading and listening to a story. By utilizing unimodal encoding models to conduct intramodal and crossmodal prediction analysis, I found that selective attention exclusively directs modality-invariant cortical representation of linguistic semantic information towards the contents of the attended story.; 2) An analysis of multimodal information representations in gaze-related areas during freeviewing of naturalistic audiovisual movies. By conducting a joint encoding modeling analysis that combined retinotopic visual, spatiotopic visual, auditory, visual semantic, and eyemovement-related feature spaces, I found multimodal representation in frontal and parietal cortices, which are thought to be involved in gaze navigation. These findings support the usefulness of encoding modeling and naturalistic experiments for understanding human brain functions.