Acquisition and contextual switching of multiple internal models for different viscous force fields

Abstract

Humans can learn an enormous number of motor behaviors in different environments. To explain this, the MOSAIC model proposes that multiple internal models are acquired in the brain, which can be switched. However, previous behavioral studies that examined arm-movement adaptations to multiple environments reported a rather limited learning capability. Hitherto, humans have been believed incapable of learning two opposite viscous force fields, which are both dynamic transformations and depend on the same state variable, presented in a random order with only a visual cue. In contrast, this study found that humans are capable of this. Elbow joint movements to specified targets were perturbed by either resistive or assistive viscous force fields generated by a single degree-of-freedom manipulandum. The resistive or assistive viscous force fields were cued by a blue or red color on a CRT screen, respectively. The squared distance between the end point and the target, and the variance of the joint angular velocities were used as kinematic performance indices. These movement errors decreased significantly as a function of the training days. Aftereffects and learning consolidation were demonstrated in the random presentation of the two force fields. Consequently, humans were able to learn the multiple and distinct internal models of the two force fields and appropriately switch them even for a random presentation cued only by color after several days of training. This study suggests that none of the previously proposed conditions for multiple internal model learning are necessary prerequisites, and indicates that the difficulty in learning is determined by the balance between the effectiveness of contextual information and the similarity of force fields.