Fig. 9. Analysis to detect potential gaze confounds.

Red bars: proportion of neurons tuned for three key world-centric variables using the standard GLM described above. Blue bars: same results, but this time from a version of the GLM that included eye position as a regressor. That version is constructed so that all variance possible associated with eye position is assigned to eye position first and only residual encoding of task variables is counted toward those variables. All three variables are still significantly observed in the population when including gaze position. Significance was tested by two-way binomial test (n = 10/37, p < 0.001).