In a fraction of a second, an image of an object or scene--one never experienced previously--can be easily comprehended by a human observer. The functional magnetic resonance imaging (fMRI) study of the neural basis of this remarkable feat has been greatly advanced by the discovery of a cortical area that responds well to intact objects but not to scrambled versions of these same images. This region, termed "the lateral occipital complex" (LOC), runs ventrally from the occipital lobe to the posterior regions of the temporal lobe. LOC is not merely activated by familiar objects in that unfamiliar objects, such as abstract sculptures, also produced greater activation in LOC compared to their scrambled versions. Nor is activation in LOC merely a feed forward effect from earlier visual stages in that these stages show greater activation for the scrambled compared to the intact versions of the images. LOC responds equivalently to a photograph and a line drawing of the same object. Bilateral lesions to LOC have been shown to produce a complete inability to recognize objects on the basis of their shape, while leaving visually guided motor interactions unaffected. LOC thus represents physical shape to which, presumably, semantic information, including a name, can be associated. This stage thus appears to be at the threshold of cognition. Despite the advances enabled by the LOC localizer, there is considerable uncertainty as what this localizer is actually localizing. The scrambling operation affects both low-level measures of the image as well as higher-level aspects associated with the interpretation of the object. The low-level measures include an increase in the extent of the display produced by the scattering of the elements, a loss of low frequency information, and a marked increase in the clumps from one (the intact object) to over 20. Of course, with the scattering, the sizes of the clumps are similarly reduced. With National Science Foundation funding, Dr. Irving Biederman will conduct fMRI research designed to clarify just what the localizer is localizing and in doing so, provide some clues as to how shape is abstracted from an image. The research would elucidate the neural coding of object vs. texture, the effect of the number of elements in a display and whether these elements comprise an intact object or scene. Additionally, attention will be paid toward defining those cortical areas that might be more responsive to aspects, such as texture rather than object, in the scrambled images.
The immediate broader impacts will be to contribute to advancing our knowledge as to the neural representation of shape for purposes of recognition. The research will fill a critical gap between psychophysical (behavioral) research on object recognition and single-unit recording of shape variations in the macaque. Although the homologues between human and macaque of primary sensory and motor cortices are well established there are significant questions concerning the homologues of later, perceptual "association" areas. Extensive plans are already in place for an outreach program involving USC's NSF-funded 3T scanner. Because of the accessibility of the problem of real-time object recognition to a non-technically advanced audience, this research will be one of the featured research projects for exposing cognitive neuroscience to students from USC's largely minority neighborhood and to recruit some to participate in the research itself.
