Figure 3.

Coding of spatial quantities in multivoxel patterns in PPA and RSC. A, Multivoxel pattern correlations between each pair of views were sorted into four groups based on shared spatial information between the views. Here we show example pairings for each group. B, Condition matrix showing all pairings between views, with assignment to groups. Numbered circles refer to campus locations in Figure 2; locations to the left of the slash were shown in the first half of the experiment and locations to the right of the slash were shown in the second half of the experiment. Categorical regressors for direction, location, and view were created by contrasting these conditions as shown. Each regressor was used in a separate multiple regression analysis (see Materials and Methods). C, When conditions were compared directly without controlling for visual similarity, multivoxel patterns in PPA distinguished between locations and also distinguished between views, and multivoxel patterns in RSC distinguished among directions, locations, and views. Bars represent the difference in pattern similarity between the conditions that constitute each categorical regressor, as measured by the β weight on that regressor in the multiple regression; error bars indicate mean ± SEM. D, When visual similarity was controlled using a parametric regressor, multivoxel patterns in PPA distinguished between views, but not locations or directions, whereas multivoxel patterns in RSC distinguished between both views and locations. ***p < 0.001; **p < 0.01; *p < 0.05; ^†p < 0.10.