![](/wp-content/uploads/2021/03/abstractheadergraphic2015.gif)
Meeting Abstract
Perturbed, transient, and variable movements reveal locomotor dynamics and hence how control is accomplished. Recently a great deal of research has focused on feedback vs. feedforward control strategies. Much less attention has focused on an independent second axis of organization—that of centralization vs. decentralization of the control signals. A longstanding hypothesis of locomotor control is that as speed increases, animals should adopt more feedforward, decentralized control strategies that do not require the processing time inherent in acquiring and coordinating centralized, feedback information. Testing the centralization aspect of this hypothesis has been elusive. Using a large dataset of strides from a running cockroach, I tested whether individual muscles represent control of more local (i.e. decentralized), individual limb states or more global variations in whole body dynamics. To do this I quantified the mutual information between muscle activation and three state variables – 1) limb extension 2) center-of-mass vertical acceleration, and 3) global kinematic phase. Mutual information measures how much one signal constrains the possible states another signal adopts. It avoids assumptions of linearity or specific statistical distributions. I discover muscle activation informs global phase more than the individual leg’s kinematics or the body’s acceleration. A muscle tells us more about what the whole body is doing than it does about the limb that it acts on. This result indicates that even at high speed, variation in neural activity is more centralized. The emerging view is that control is hierarchical with centralized neural feedback operating (when necessary) on top of a feedforward-driven, mechanically decentralized system.