Abstract
It is well recognized that spatial frequency plays an importance role in face perception revealed as 10 cycle/face optimal frequency (Keil, 2008) and hybrid faces (Schyns and Oliva, 1999). However, the neural mechanism of face spatial frequency processing remains largely unknown. In this study, we recorded intracranial evoked potentials from epilepsy patients with implanted subdural electrodes when 40 face pictures of different identities, genders and emotional expressions were presented in low spatial frequency (LSF) or high spatial frequency (HSF) forms while the subjects were simply required to detect rare oblique pictures as catch trials. We found N200 component on 17 electrodes from 6 subjects. HSF faces induced a longer N200 latency compared to LSF faces and a significant interaction has been found between spatial frequency and gender of the face. Female faces in HSF form generate longest latency and LSF female faces induced the shortest one, leaving the LSF and HSF male faces in between. However, only the differences between LSF and HSF female faces are significant in an ad-hoc analysis. A possible explanation is that female faces are smoother and extending to broader frequency ranges. So the difference between mean frequencies in LSF and HSF forms would be larger for female faces than for male faces. But spectrum analysis of the stimuli rejected this and other simply features statistics possibilities. However, we do find that HSF male faces have more SIFT keypoints (Lowe, 1999) than HSF female faces. Since it is also conflict with the fact that the number of SIFT descriptors of LSF male faces is much lesser than HSF male faces but their N200 latencies are not significantly different, we believe that a high order face invariance somehow related with spatial frequency is necessary to understand the face perception and its neural dynamics.
The Ministry of Science and Technology of China (2005CB522800, 2004CB318101) and The National Natural Science Foundation of China (30621004, 90820307).