Abstract
Purpose. A Glass pattern consists of randomly distributed dot pairs (dipoles) whose orientations are determined by a geometric transform. To perceive the structure in a Glass pattern, an observer needs to perform local grouping to find dipoles and global grouping across dipoles to get an overall shape. We investigated cortical mechanisms for local and global grouping by observing BOLD activation to Glass or texture patterns.
Method. Each dipole contained two 5.4′ square dots separated by 27′. The Glass patterns contained randomly distributed dipoles covering 2% of image. The coherence of a Glass pattern was defined as the proportion of dipoles oriented tangent to a concentric global form. To eliminate grouping at the dipole level, we also used line element textures which had bars (5.4′x27′) in place of dipoles. The BOLD activation (N=6) was collected on a Bruker 3T magnet (EPI, TR=3s, TE=60ms, flip angle=90o). Each 252s block design run had two test conditions alternating with each other in a 36s period.
Result. Compared with a zero coherent pattern, the 100% coherent Glass patterns activated the lateral occipital complex (LOC) and the inferior temporal (IT). The Glass and the texture pattern of the same global structure showed differential activation in V2 and V3 which was posterior to the areas responding to Glass patterns. A rotating wedge of coherent dipoles showed that the dorsal LOC had a retinotopic property for Glass patterns. A rotating concentric pattern produced only weak retinotopic activation at the ventral LOC.
Conclusion. The contrast between line texture and Glass patterns suggests that local grouping occurred in V2 and V3. The rotating wedge suggests the dorsal LOC is for analyzing curvatures while the rotating circle hints that the ventral LOC is responsible for global forms. The weak activation for rotating circles may suggest that the global form processing is less retinotopic mapping than local features.