Analysis of learning-related brain activity by model-based state estimation

Abstract

A big challenge in analyzing the brain activity related to learning is its non-stationarity; neural responses to the same stimulus should change over the course of learning. Thus conventional block-design or event-related analyses that assume the same responses to the same condition or event are not sufficient. In this talk, we propose a new approach of estimating the time course of learning-related variables and parameters based on each subject's behavioral data. We applied the method to estimation of reward-expectaion variables and found a ventro-dorsal map of prediction time scale in the striatum. [Jpn J Physiol 55 Suppl:S21 (2005)]