II-67. Dimensionality in motor cortex: differences between models and exper- iment

produces a saccade (GO process) and the other that prevents the GO process from finishing (STOP process; Logan & Cowan, 1984, Psych Rev). An interactive race model was formulated to explore how the STOP process can prevent GO from finishing through lateral inhibition, motivated by the finding that firing rates in presaccadic movement cells in the saccadic generator network, including the frontal eye field (FEF), superior colliculus (SC), basal ganglia, etc., which may instantiate the GO process, decline markedly when saccades are successfully inhibited (Boucher et al. 2007, Psych Rev). However, the interactive race model did not take into account the full temporal dynamics of fixation cells in the network, which may instantiate the STOP process. We incorporated this aspect of physiological data in revised interactive race models and tested several reformulations to examine core assumptions of the original model. We found that models that assumed an external inhibitory control (e.g., blocking input to the GO unit, boosting input to the STOP unit, or strengthening the inhibitory connection from the STOP unit to the GO unit) after the presentation of a countermanding signal fit the behavioral data better than models that did not. The former models also exhibited the activation functions of the GO and STOP units that resembled the activities of the movement and fixation cells. We conclude that the source of inhibitory control exists outside of the site (i.e., FEF, SC) of saccadic countermanding.