Fig. 8. LDR allows richer outputs than rotational dynamics.

This figure is entirely schematics of hypotheses. a Illustration of proposed model. Preparatory activity moves the population state to the necessary state space location, where local rotational dynamics cause the population state to rotate in the correct planes upon movement onset. b In a dynamical system, output-null dimensions act as supporting dynamics for producing activity along output-potent dimensions. This means activity in output-null dimensions, though not directly “read out” by the nervous system in driving behavior, is informative of behavior. c Standard decoders attempt to estimate output-potent dimensions, but in the process discard information in output-null dimensions. d LDR-based decoding, by using motor cortex dynamics on each condition, leverages output-null and output-potent activity to refine estimates of reach kinematics. e Rotational dynamics impose strict phase- and magnitude-locking on outputs, such as CSNs. f By contrast, LDR allows for CSNs to break phase- and magnitude-locking, allowing motor cortex to produce a richer set of outputs.