Differential rate of temporal processing across category-selective regions in human high-level visual cortex

Abstract

What is the speed of processing in high-level visual cortex? Previous research (McKeeff et al. 2007) suggests that responses in human face- and scene-selective regions peak for stimuli presented at 4-5 Hz, unlike early visual areas that respond maximally for faster rates of presentation (18-25 Hz). However, these results are hard to interpret because presentation rate and the number of images in a block were confounded. Here, we used a novel paradigm to measure the rate of processing across human high-level category-selective regions. Twelve participants were scanned with fMRI while viewing images of faces, bodies, objects, scenes, and written characters presented at 1, 2, 4, or 8 Hz. Stimuli were presented in 8-image blocks at all rates to equalize the number of stimuli in a block across rates. Results indicate that networks of category-selective regions are optimized to process stimuli at different rates, with regions selective for faces, scenes, and characters preferring slower rates (2 Hz), and body-selective regions preferring faster rates (4 & 8 Hz). This pattern of results manifests in three ways: (i) The proportion of VTC voxels showing selectivity for a category peaked at a certain rate. Specifically, the proportion of VTC voxels selective for faces, scenes, and characters was greatest at 2 Hz, and the proportion selective for body parts was greatest at 4 Hz (Figure 1). (ii) Independent ROI analyses revealed that selectivity for the preferred category was highest at the optimal rate associated with a region. (iii) Heightened selectivity resulted from a specific increase in the magnitude of response to a region's preferred category at the optimal rate. These results demonstrate differential temporal processing of category information across high-level visual cortex and that bodies are processed at faster rates than other types of stimuli, perhaps as a result of their mobility and non-rigid nature. Meeting abstract presented at VSS 2014