The role of face-selective and object-general mechanisms in the face inversion effect: A simultaneous EEG-fMRI study

The face inversion effect (FIE) refers to the greater difficulty we have in recognizing inverted than upright faces. It has been suggested that inverted faces are processed like non-face objects. Event-related potential studies consistently show an increase in the amplitude of the face-selective N170 component to inverted relative to upright faces (N170-FIE). Here we tested two hypotheses that may account for the increased response of the N170 to inverted faces. According to a quantitative hypothesis inverted faces exploit the same mechanisms used for upright faces but with more computational effort. This hypothesis predicts that the increased N170 amplitude to inverted faces is mediated by face-selective mechanisms. According to a qualitative hypothesis, inverted faces engage additional processing mechanisms such as object-general mechanisms. This hypothesis predicts that the increased N170 amplitude to inverted faces is mediated by object-general mechanisms. To test these two hypotheses we conducted simultaneous EEG-fMRI experiment in which we presented upright and inverted faces in an event-related design. For each subject we also localized face-selective and object-general areas based on a functional localizer scan. Following removal of MR-artifacts from the EEG signal we measured the response of the N170 to upright and inverted faces and computed an N170-FIE index for each subject. Similarly, we measured the response of face-selective and object-general areas to upright and inverted faces and computed an fMR-FIE index for each area of each subject. Correlational analyses revealed that the N170-FIE was strongly correlated (r = 0.8) with the FIE in object areas but not with the FIE in face-selective areas. These findings are consistent with the qualitative hypothesis and suggest that object processing mechanisms are involved in the processing of inverted faces as early as 170 ms after stimulus onset.