Fig. 5: Strategy-constrained model (Model 4) recapitulates behavioral data.

a Comparison of the radial distance (from origin) of the model’s stopping position against the radial distance of the target (left), as well as the angular eccentricity of the model’s stopping position versus target angle (right) across all trials. Black lines have unit slope. Visual (cyan) and vestibular (red) conditions are simulated by changing the variance of observation noise during testing (see “Methods”). b Activity dynamics of the two OC units that were constrained to encode the x and y components of the relative target position in a subset of test trials. Horizontal line denotes zero. c Coefficients corresponding to the target and stopping position from the linear regression model that best explained the OC unit activity (averaged across the two units). Error bars denote standard errors estimated by bootstrapping. d Error in stopping position, on typical trials (control) and trials in which the model OC units are inhibited by clamping their activity at zero (inh). For comparison, the steering error in free-gazing (control) and inhibited gaze (inh) trials in humans is shown on the right. Error bars denote standard errors of the mean estimated by bootstrapping across 50 realizations of the model or across participants, as applicable.