Functional correspondence between human and monkey face-selective regions in processing face configuration

The monkey plays a critical role in understanding the neural mechanisms of face perception. However, it is unclear whether monkeys process faces in a similar holistic manner as humans and how monkey and human face-selective regions correspond. To address this, we performed the same fMRI experiment in these two species. We presented human face parts in either their normal composition or deviant compositions in different blocks. Veridical and scrambled faces were created by disassembling the same composite images (four veridical faces positioned around a fixation point) in two different ways. We either presented each full set of face features from one of the four veridical faces separately, or spread them over the four faces (hence showing four different features from the four different faces). This resulted in scrambled faces that fully controls for local low-level stimulus differences compared to the veridical faces while destroying the veridical configurations. To further control for different degrees of stimulus spreading and asymmetry, we created two other conditions with face features arranged in a scrambled configuration at each of the four positions in the initial composite images. Finally, two additional control conditions were created with objects instead of face parts. Three monkeys and twenty humans were scanned while they performed a fixation task. An independent localizer experiment was performed in the same subjects to identify face-selective regions. We found a significant face configuration effect in monkey face-selective regions PL, ML, MF and lMFa, and in human OFA, FFA, rpSTS and rmSTS. In addition, compared to the spreaded version, the scrambled faces and objects elicited a significantly lower response in monkey PL and ML, and in human OFA and FFA, likely due to surround suppression in the latter stimuli, whereas the same contrast revealed a significant positive activation in monkey MF and lMFa, and in human rmSTS. In PL and OFA, the veridical faces elicited a similar response as their spreaded version, whereas the veridical faces elicited a significantly higher response in FFA and all the other regions, suggesting a correspondence between PL/OFA. Furthermore, through an inter-species multiple voxel representational similarity analysis, we found that the pattern dissimilarity between the four types of faces in human rmSTS is more similar to that in monkey MF and rAF, whereas the pattern dissimilarity in human OFA, FFA and rpSTS is more similar to that in monkey PL and ML. Taken together, these results suggest a similar face processing system between humans and monkeys, with possible functional correspondences between PL/OFA, ML/FFA and MF/rmSTS.status: publishe