Abstract
LOC localizers compare pictures of objects with gird-scrambled versions producing a comparison of object vs. texture. However, this results in a markedly different set of local features: an increased number of line endings (creating new T-junctions even when using a superimposed grid) and a reduction in longer contours. Here we present a new localizer based on local feature deleted (LFD) line drawings of objects in which every other vertex and line was deleted from each part. These LFD stimuli remain readily interpretable, and complementary LFD stimuli prime one another (Biederman & Cooper, 1991) and produce equivalent adaptation in LOC (Biederman & Hayworth, 2005). Because LFD stimuli have undergone contour deletion, they can be scrambled (sLFD) to produce an image that is incoherent without any change in the number or nature of the local features. In theory, the LFD and sLFD stimuli should be processed similarly by the visual hierarchy, until a level is reached where the parts of an object are encoded. The comparison of fMRI activity for LFD > sLFD showed similar regions of activation as a grid-scrambled control but with differences in the region of maximal activation: LO for the grid-scrambled control, pFs for the LFD localizer. These results suggest greater sensitivity to object parts vs. local features in pFs than in LO. The opposite ordering (sLFD > LFD) was observed in area V2.
Supported by NSF 0420794, 0426415, and 0531177