Abstract

Recent progress in the understanding of motor cortex function has been achieved primarily by simultaneously recording motor cortex neuron activity and the movement kinematics of the corresponding limb. We have expanded this approach by combining high-quality cortical single-unit activity recordings with synchronized recordings of full-body kinematics and kinetics in the freely behaving cat. The method is illustrated by selected results obtained from two cats tested while walking on a flat surface. Using this method, the activity of 43 pyramidal tract neurons (PTNs) was recorded, averaged over 10 bins of a locomotion cycle, and compared with full-body mechanics by means of principal component and multivariate linear regression analyses. Patterns of 24 PTNs (56%) and 219 biomechanical variables (73%) were classified into just four groups of inter-correlated variables that accounted for 91% of the total variance, indicating that many of the recorded variables had similar patterns. The ensemble activity of different groups of two to eight PTNs accurately predicted the 10-bin patterns of all biomechanical variables (neural decoding) and vice versa; different small groups of mechanical variables accurately predicted the 10-bin pattern of each PTN (neural encoding). We conclude that comparison of motor cortex activity with full-body biomechanics may be a useful tool in further elucidating the function of the motor cortex.