Fig. 3: Positive correlations between “Max” vs “Avg” activation differences and encoding model accuracy.

The “Max” vs “Avg” activation differences are captured via Cohen’s d between activations from “Max” images and “Avg” images. The regional encoding model accuracy is determined using all session 1 images that have at least two fMRI measured activations, regardless of condition ("Mat” or “Avg"), source ("Nat” or “Syn") or target region (FFA1, EBA, VWFA1). The positive associations are observed across the six NeuroGen subjects for all brain regions, where each subject is the same color across all scatter plots. Squares indicate the “Nat” image results where the line of best fit is drawn with a solid line and circles indicate the “Syn” image results where the line of best fit is drawn with a dashed line. Note: some regions were not identifiable in some subjects using the localizer scans. Scatter plots (a, b) and (c) show the relationship between encoding model accuracy (x-axis) and the Cohen’s d calculated by contrasting “GroupMaxNat” vs “GroupAvgNat” or “GroupMaxSyn” vs “GroupAvgSyn” for face, body and word perception regions respectively.