Abstract
Extensive research over the last decade has characterized the neural response to objects in the ventral visual pathway in humans. However, in most of this work, single objects are presented in isolation, a situation rarely if ever found in real-world scenes. Here we explored the neural representation of displays containing multiple objects by asking whether responses in the ventral visual pathway are sensitive to i) the number of identical copies (“tokens”) of an object in a given display, and ii) the number of different kinds (“types”) of objects. We measured the fMRI responses in LOC, FFA, and PPA while subjects performed a dimming-detection task at fovea while one or four faces, scenes, or objects were presented in the periphery. All ventral ROIs showed sensitivity to copy/token information, as mean response in each ROI to four identical copies of the same item, one in each quadrant (the four-same condition) was nearly twice as high as the response to a single item in one quadrant (the single-object condition). The token sensitivity does not merely reflect the existence of separate pools of neurons that respond to each quadrant, as activation in the single-object condition was not significantly lower than the four-different condition, where different objects from the same superordinate category were presented, one in each quadrant. The ROIs also showed sensitivity to kind/types information because they responded much more strongly in the four-same condition than the four-different conditions. Similar patterns of response were observed in retinotopic visual areas whose receptive fields were confined to a single visual quadrant although the difference among conditions was smaller. These results may reflect two distinct phenomena, a gain in representation strength from multiple tokens and a competition in representation from multiple types.
NIH 071788, ARO 46926-LS, and NSF 0345525.