Stochastic variables responsible for observed saccadic variability.

A new stochastic local feedback model of the horizontal saccadic system based on time optimal neural control within the superior colliculus has been previously described. This model uses a premotor neural circuit composed of burst, tonic, and pause cells and innervates a fourth order linear homeomorphic muscle plant. A sequence of saccades recorded from human subjects shows great variability in peak velocity, final position, and amount and type of post-saccadic behavior. This variability is duplicated in a sequence of simulated saccades through the use of random variables within the neural circuitry. The random variables are agonist burst cell magnitude, antagonist post inhibitory rebound burst magnitude and timing, and to a lesser extent muscle saturation magnitude. These four random variables are shown to cause all the observed variability in human saccades, including: trajectory profile, velocity profile, dynamic overshoot, and glissadic overshoot and undershoot.