GLM analysis was conducted via a mixed effects approach using FLAME (FMRIB's Local Analysis of Mixed Effects; Beckmann 2003; Woolrich 2004). Parameter estimates of the regressors representing the stimulus categories were obtained individually for each run of each subjects under fixed effects assumptions. To represent face-specific changes in activation, a contrast of parameter estimates (COPE) was defined as \(faces>\left(houses+objects\right)\). In turn, Scene-specific changes in activation were definded as \(scenes>\left(objects+faces\right)\). Both contrasts were tested for significance via T-Tests. The aggregation of the resulting parameters across runs for each subjects was performed using a fixed effects model higher-level analysis (single group average). The resulting statistical images were passed onto a random effects higher-level analysis for the aggregation across subjects, thereby accounting for inter-individual variance. The resulting statistical z-maps were thresholded at z > 2.3 and corrected for multiple comparisons on a cluster level at p < .001 as recommended, among others, by Eklund 2016. Brain structures corresponding to these clusters of significant voxels were identified with the Harvard–Oxford cortical atlas (Desikan 2006) and the Juelich Histological Atlas (Eickhoff 2007), both provided by the FSL package.