We have used functional magnetic resonance imaging (fMRI) to study the changes in cerebellar activation that occur during the acquisition of motor skill in human subjects presented with a new task. The standard paradigm consisted of a center-out movement in which subjects used a joystick to superimpose a cursor onto visual targets. Two variations of this paradigm were introduced: 1) a learning paradigm, where the relationship between movement of the joystick and cursor was reversed, requiring the learning of a visuomotor transformation to optimize performance and 2) a random paradigm, where the joystick/cursor relationship was changed randomly for each trial. Activation in the cerebellum was highest during the random paradigm and during the early stages of the learning paradigm. In the early stages of learning and during the random paradigm performance was poor with a decrease in the number of completed movements, and an increase in the time and length of movements. With repeated practice at the learning paradigm performance improved and reached the same level of proficiency as in the standard task. Commensurate with the improvement in performance was a decrease in cerebellar activation, that is, activation in the cerebellum changed in a parallel, but inverse relationship with performance. Linear regression analysis demonstrated that the inverse correlation between cerebellar activation and motor performance was significant. Repeated practice at the random paradigm did not produce improvements in performance and cerebellar activity remained high. The data support the hypothesis that the cerebellum is strongly activated when motor performance is inaccurate, consistent with a role for the cerebellum in the detection of, and correction for visuomotor errors.
- error correction
- functional magnetic resonance imaging
- motor learning
- step tracking