The y-axis displays feature AMI, which increases when pRF modulations (size and eccentricity changes combined, see Materials and methods) are greater when attending color compared to TF. The x-axis displays feature preference, which increases with higher color compared to TF preference. pRF modulations were greater when attending color in all ROIs in the 'super subject' method, and were unrelated to feature preference (see Supplementary file 1 -Table 9). In the 'over subjects' method, we confirm that feature AMI was not different between ROIs (RM ANOVA, factor of ROI: F(8,32) = 0.631, p = 0.746, η2p= 0.066), and that it was on average 0.059 over ROIs across subjects, F(1,4) = 18.868, p = 0.012, η2p = 0.394). Additionally, we found that for the ‘over subjects’ method, feature AMI was significantly positive in the combined ROI t(4) = 4.316, p = 0.012, Cohen’s d = 2.158 (see Supplementary file 1 -Table 25 for all ROIs). Finally, we also found no correlation with feature preference in the 'over subjects' method (mean Spearman correlation = 0.20, t(4) = 0.850, p = 0.434, Cohen's d = 0.446; mean Pearson correlation = 0.22, t(4) = 0.770, p = 0.476, Cohen’s d = 0.429). Error bars reflect 95% CIs across voxels in the 'super subject' and 'per subject' methods, and across subjects in the 'over subjects' method. See Material and methods section for definitions of 'super subject', 'over subjects' and 'per subject' methods.