Fig. 7. Quantification of category and direction encoding in LIP.

A Diagram of direction encoding in the cosine-tuned GMLM. Motion stimulus direction is encoded as an ellipse in the low-dimensional stimulus space. The ellipse has major (orange arrows) and minor (green arrows) axes that define its shape. If the axes are of similar length, tuning is approximately circular and the motion directions are evenly distributed in the low-dimensional space (top). If the major axis is elongate compared to the minor axis, the population shows a preferred direction which may be aligned with category (middle) or the null direction (bottom) or anywhere in between. Category is encoded as a vector in addition to the direction ellipse, and the category vector is constant for all motion directions within a category (in contrast to the task-aligned direction tuning which places near-boundary directions closer together). B Bayesian estimate of the geometry of the sample stimulus tuning for the eight LIP populations. Each plot shows the norm of the sample category vector (black) and the norms of the major (orange) and minor (green) axes of the direction-tuning ellipse for an LIP population as a function of time relative to the sample stimulus onset. The solid lines denote the posterior median at each time point, and the shaded regions denote 99% credible intervals. If the major and minor axes have equal norms, then direction would follow a circle in the low-dimensional space.