Visual training for population vector based cortical control

We have developed a method for training animals to control a variety of devices from cortical signals (Helms Tillery et al., 2003). In this report, we describe a protocol to parameterize a cortical control algorithm without an animal having to move its arm. Instead, a highly motivated animal observes a computer cursor moving towards a set of targets once each in a center-out task. From the neuronal activity recorded in this visual following task, we compute the set of preferred directions for the neurons. We find that the quality of fit in this early set of trials is highly predictive of each neuron´s contribution to the overall cortical control.