Musculo-skeletal geometry and the control of single degree of freedom elbow movements

Empirical and modelling studies are reported which explore ways in which the central nervous system might consider musculo-skeletal geometry when generating commands for single degree of freedom elbow flexion and extension movements. In a series of experiments it is shown that subjects do not perform rapid, goal-directed flexion and extension movements equally accurately in different parts of the elbow's workspace. In these experiments, movements of 10, 20 and 30 degrees in amplitude were tested using up to five different initial elbow angles. When performing flexions, subjects tended to overshoot targets when starting from extended positions, to undershoot targets when starting from more flexed positions, and to perform relatively accurate movements when starting from the centre of the workspace. Final position accuracy was more variable for extensions. When reliable differences existed for extensions, subjects tended to produce a pattern of results opposite to that of flexions: subjects overshot targets when starting from flexed positions and undershot targets when starting from more extended positions. A model of elbow movement based on the $ lambda$ version of the equilibrium-point hypothesis was used to assess the extent to which the pattern of errors obtained in the empirical studies could be reproduced by a control scheme that does not adjust commands in response to changing musculo-skeletal geometry, but rather uses one single invariant command throughout the workspace. The motivation for testing the invariant command notion was to explore the possibility that motion planning might be achieved without an explicit representation of musculo-skeletal geometry. Predicted patterns of final position errors across the workspace matched empirically obtained error patterns for flexions, but the model performed less well when predicting the pattern of errors observed for extension movements