Fig. 6.

CSS model outperforms linear model. A: cross-validation accuracy in different visual field maps. Vertical bars indicate the median accuracy across voxels in each map. Horizontal lines indicate the median noise ceiling. The CSS model outperforms the linear model in all maps (P < 1e−8, 2-tailed sign test). Increases in accuracy are more substantial in extrastriate maps than in V1. B: cross-validation accuracy for individual voxels. Each dot compares the accuracies of the 2 models for a single voxel.