Possible output pathway related to learning to use a new tool

Previous neuroimaging studies have suggested that the human cerebellum is active during many kinds of learning tasks. These studies have shown that strong activities are observed in wide area of the cerebellum at initial stages of learning while the activated area becomes small as a subject’s performance improves. Using a functional MRI, we measured human cerebellar activity during visuomotor learning to investigate functional correlates of the cerebellar activity. A task for subjects was to move a computer mouse so that a cursor followed a target randomly moving on a screen. A relationship between the mouse and the cursor was altered during the task so that learning was necessary. We found that activities observed in the early stage of learning were significantly correlated with the subject’s performance error while restricted activity observed in the later stage was not evoked by the error detection and/or correction. These results confirmed computational theories proposing that the cerebellum acquires internal models through error driven learning.