Physiological Study of the Striatum Based on Reinforcement Learning Models of the Basal Ganglia 

Tomohiko Yoshizawa ( 1461011 )


Many lines of research, including functional brain imaging and neural recording, have demonstrate that the basal ganglia plays a critical role in decision making and reinforcement learning. Especially, the striatum, which is a major input site of the basal ganglia, is known an important brain region for reward prediction.


In order to physiologically examine specific roles of stratal compartments (striosomes and matrix) and subregions (dorsolateral, dorsomedial and ventral striatum) for reward-based learning, I designed classical- and operant-conditioning tasks for rodents based on the reinforcement learning theory, then recorded their neural activities by a deep-brain in vivo calcium imaging method using a GRIN lens or an electrophysiological method during the tasks. The recorded neural activities were fitted to various linear regression models and examined what types of information were encoded.

The most important findings of this study are follows;

1) Striosomal neurons showed reward-predictive activities proportional to reward size expected from presented odor cues; therefore, they encoded the values of odor stimuli.

2) Compared with neurons in the dorsolateral striatum, those in the dorsomedial striatum encoded higher-level (complex) information.


In my talk, I will show other important findings and discuss the results in relation with the reinforcement learning theory.