Properties of the visual and semantic attribute value computations. A, Activity in left pSTG correlated with semantic ratings in the semantic–visual condition. A right-side small figure shows English (vs scene)–Korean (vs scene) contrast from localizer trials. B, Activity in FG correlated with semantic ratings in the semantic–visual condition. A right-side small figure shows Korean (vs scene)–English (vs scene) contrast from localizer trials. The conjunction of the whole-brain and the localizer contrasts is depicted in green. C, Beta plots comparing the responsivity of left pSTG to semantic and visual ratings during the semantic–visual condition. D, Beta plots comparing the responsivity of left pSTG to the valences of the semantic and visual ratings during the semantic–visual condition. E, Beta plots comparing the responsivity of right FG to semantic and visual ratings during the semantic–visual condition. F, Beta plots comparing the responsivity of right FG to the valences of the semantic and visual ratings during the semantic–visual condition. Whole-brain analyses were thresholded at p < 0.05 (corrected). Right FG was plotted in the ROI plots. For D and F, semantic and visual ratings were divided into three levels. pSTG/MTG, posterior superior/middle temporal gyrus. Coordinates are in Talairach space.