Abstract
The cognitive ability to group information into “chunks” is a well known phenomenon in verbal working memory paradigms. However, the effects of chunking within visual memory are not well understood. Previously, we have demonstrated how the bottom-up Gestalt principle of good continuation affects the “online” memory representations for elements in Kanizsa figures. We have further demonstrated the effects of common fate or connectedness on the online representation of items during a Multiple Object Tracking task. Here, we investigate how learned arbitrary relationships between items in a visual display may be used to group the items, in the absence of bottom-up grouping cues.To do this, subjects learned patterns of dots related by color set, locations or both color set and locations, indicating the appropriate category for each novel exemplar. Following the learning of the stimuli, the subjects performed a color change detection task using the learned stimuli. ERPs were recorded during the experiment, in particular, we measured the contralateral delay activity. The amplitude of this ERP component is sensitive to the number of items being held in visual working memory. We were able to demonstrate an effect of the learning of color and pattern categories both on behavior and online memory representations during learning, as well as during subsequent performance of a change detection task using the learned stimuli. These results are relevant to our understanding of the representation of visual information and the effects of recently learned information on the encoding and maintenance of information online.