Neural encoding of head-free gaze shifts in monkey superior colliculus

11. April 2013


We recently proposed a population-encoding model that explains how cells in the midbrain superior colliculus contribute to the trajectory and kinematics of saccades (Goossens & Van Opstal, PloS Comp Biol 2012). Briefly, the model holds that each spike from each recruited cell contributes a tiny, but fixed, contribution to the saccade (“spike vector”) that only depends on its location within the motor map. The saccade results from dynamic linear summation of all spike vectors. This extremely simple model predicts a linear relation between a cell’s cumulative spike count, CS(t), and the straight, imaginary trajectory of the eye, S(t-τ), between initial and final positions (τ: lead time). We tested this model for >20000 head-fixed saccades to visual targets across the oculomotor range. Interestingly, when using  measured spike trains to simulate saccades, the model produced straight eye-movement trajectories with the correct velocity profiles and nonlinear main-sequence properties, although horizontal and vertical brainstem burst generators were kept linear and uncoupled.
I will also discuss and test a strong prediction of the model: the linear relation between dynamic spike counts and instantaneous straight gaze-displacements should also hold for head-free gaze shifts, in which kinematics and relative eye- and head contributions vary substantially from trial to trial. We recorded from ∼40 single units in two monkeys trained to make large gaze shifts (up to 85 deg amplitude) to flashed targets that were elicited throughout a cell’s movement field. The data support the simple linear spike-count model.