in collaboration with mike mozer

Visual attention research has suggested two types of attentional selection. Space-based selection accounts propose that the location of a stimulus is selected. Object-based selection accounts propose that organized "chunks" or groups of information are selected; these groups of visual information correspond to objects. How do space-based and object-based accounts relate to one another, and what are the neural mechanisms that mediate attentional selection?

Several relationships exist between these forms of attention: (1) object-based grouping could occur before spatial selection, (2) spatial selection could occur before object-based grouping, or (3) object-based grouping could interact with spatial selection. There are behavioral and neuroscientific data both consistent with and inconsistent with each of these alternatives. In collaboration with Mike Mozer, we are exploring a computational model that integrates both space-based and object-based accounts of visual attention. We propose that object-grouping and spatial selection can be combined in a single spatial representation, which permits these two processes to mutually influence one another without the computational complexities associated with interactive connectionist models that involve search through multiple representations.

In addition to these computational explorations, we are conducting behavioral studies that address basic questions regarding object-based attention. Can attention select the parts of an object? Can object attention be precued based on shape alone, not spatial location? Answers to these questions have implications for our understanding of object selection processes.