Abstract
Several recurrent models of vision propose that low and high levels of the ventral visual hierarchy interact over the course of processing to build up progressively finer representations of the visual input. Previous research (Goffaux et al., 2010) suggests that high-level face-preferring regions initially process coarse information (low spatial frequencies, LSF) and later process finer details (high spatial frequencies, HSF). We conducted a block-design fMRI experiment with 18 subjects to explore whether V1 contributes to the coarse-to-fine build-up of face representations, as predicted by recurrent theories. We expected LSFs to be encoded early and rapidly, and HSFs to be encoded more slowly and progressively over the course of processing. Broadband unfamiliar face images were shown for either 50, 83, 100 or 150ms, masked by their phase-scrambled counterparts. The scrambled masks were filtered to selectively contain either LSF (< 11.5 cpi) or HSF (>11.5 cpi), targeted to selectively interfere with the processing of the LSF or HSF of the broadband faces, respectively. In early stages of processing (50ms), V1 showed a preference for LSF compared to HSF processing. Over time, responses to LSF decreased, while progressively increasing for HSF. By the later stage of processing (150ms), responses were higher for HSF than for LSF processing. Similar results were found in the right fusiform face area, agreeing with previous research (Goffaux et al., 2010). Taken together, we provide evidence for coarse-to-fine processing in both low and high levels of the visual hierarchy. Our findings shed light on the potential role of V1 in the progressive build-up of detailed representations of faces in the visual system. Next, we will explore whether low and high levels of the visual hierarchy interact over the course of processing as predicted by recurrent models of vision.