Abstract
It remains unknown whether feature-integrated object representations are maintained in visual working memory. Previous work suggesting feature-based memory maintenance may be confounded with effects in memory retrieval and matching. The current study used a task without direct judgment on location-specific feature-based matching, and provided evidence for feature-integrated representations. Observers saw a preview display with two colored shapes, followed by a target object at one of the preview locations, and judged whether the target contains any color or shape of preview objects regardless of their locations as quickly as possible. When both color and shape of the target were included in preview objects, task irrelevant feature-location correspondences were systematically manipulated, and their effects were evaluated. Mean RT data showed a pattern supporting feature-based preview benefit by shared location. However, when RT data were decomposed into decision and non-decision components by a diffusion model analysis, estimated non-decision components, presumably reflecting stimulus encoding, revealed a preview benefit based on feature-integrated object representations; a benefit only when both target features shared the location with preview features. This conjunction-specific preview benefit was observed even with moving objects, suggesting that the effect reflects feature-binding in perceptual objects, not only binding at the shared location. Furthermore, to investigate neural correlate of the conjunction-specific preview benefit, an ERP experiment was conducted. Using a display with two placeholders located left and right of fixation, ERP amplitude of N1 component triggered by the target onset at parietal electrodes contralateral to the target reduced only when both target features shared the location with preview features, consistent with the non-decision time data. Taken together, these data indicate that color-shape conjunction, not independent features, facilitates encoding of the subsequent target object, suggesting that a color-shape conjunction bound to a spatiotemporal location is maintained as a unit in visual working memory.
Meeting abstract presented at VSS 2012