Perceptual Science Series
C. Edward Connor (Johns Hopkins University, Department of Neuroscience)
Shape Information Processing in the Ventral Visual Pathway
April 14, 2003
C. Edward Connor
John Hopkins University, Department of Neuroscience
Shape Information Processing in the Ventral Visual PathwayObject shape is represented by large populations of neurons at intermediate and higher stages in the ventral pathway of primate visual cortex. The nature of the shape information encoded by higher-level visual neurons has not been quantitatively characterized. Moreover, we do not know how complex shape information is derived from simple, lower-level shape signals. Finally, we do not know how populations of neurons cooperate to encode object shape in a distributed fashion. We approach these issues by testing higher-level visual neurons (using microelectrode recording in awake monkeys) with large sets of parametrically varied shape stimuli. We model the neural response patterns with tuning functions in high-dimensional shape space. We combine data across cells to estimate population responses post hoc. We use these population activity patterns to reconstruct the shapes and measure similarity to the original stimuli. Our results in areas V4, PIT and CIT (posterior and central inferotemporal cortex) indicate a parts-based or structural coding scheme, with a gradual increase in parts complexity at higher stages in the ventral pathway.