Okihide Hikosaka
National Institute of Health
National Institute of Health
Learning by parallel neuronal circuits for motor skill
Motor Skill is very important for many animals (including humans) to survive and enjoy their lives. To study the neuronal mechanism, we created a sequential behavioral task, in which the final goal (reward) can be obtained after choosing the correct choice in all sequences (e.g., n=5 or 10). The subject (mainly monkey) learned many sequential behaviors (e.g., n=30) across many days (e.g., 100 days), and then maintained the behavior for a long time (> 1 year) without further learning.
We found that the learning is based on 2 different actions: Eye movement to choose the good object in each step, which is followed by Hand movement to use the good object. Even after stopping learning for a long time (e.g., > 1 year), Eye can choose all good objects so that Hand can work well. Then, Hand can start learning and generate quick actions to reach the goal (reward). These data suggest that Eye can teach many actions (e.g., right & left Hand, right & left Leg), selectively or together. In fact, Motor Skill in real life can be complex based on many actions together.
We then found that neurons in various brain areas (e.g., cerebral cortex, basal ganglia) were selectively active during the learning: Rostral areas (early stage) and Caudal areas (late stage). Our recent studies have shown that there are 2 parallel neuronal circuits in the basal ganglia, both of which control Eye Movement (Saccade) through the superior colliculus (SC) based on the value of object. The rostral circuit from caudate head (CDh) uses Short-term Memory selectively, while the caudal circuit from caudate tail (CDt) uses Long-term Memory selectively. Then, the rostral and caudal parts of the brain (shown above) may work selectively with CDh-circuit and CDt-circuit to control early and late stages of Motor Skill learning. These mechanisms may also be critical for Motor Skill which may work almost forever after the long-term learning.