Abstract
Neuroimaging studies have revealed strong selectivity for object categories in the ventral visual pathway. However, the extent to which this selectivity is based on high-level representations, or on the tuning for low-level or mid-level features that are common to images from a particular category remains unclear. To address this issue, we measured the neural response across the ventral visual pathway to five object categories (bottles, chairs, faces, houses, shoes) using fMRI. Using correlation-based MVPA, we found that each object category elicited a distinct pattern of neural response. To determine the extent to which these patterns of response might reflect more basic properties, we used a low-level image descriptor to measure the objects from each category. The average image descriptor for each category was then used to find five clusters of objects from a large database that had similar image properties, but were categorically and semantically different to the five object categories. We then measured the pattern of neural response to the five clusters of objects. We found that each object cluster elicited a distinct pattern of neural response. Critically, we found similar patterns of response to objects defined by either category or image. For example, the pattern of response to bottles was similar to the pattern of response to objects with similar image properties to bottles. Moreover, the similarity between patterns of response was dependent on the similarity in the image properties. These results suggest an important role for image properties in the neural representation of objects in the ventral visual pathway.