Whole-report procedures reveal bimodal distribution of visual memory precision within a single array.

Most studies testing models of visual working memory capacity have relied on partial report procedures that probe the observer’s knowledge of a single item from a multi-item array. Whole-report procedures have been less common in tests of visual memory, perhaps because of concerns about output interference, but such procedures may provide richer information about within and between-trial variations in memory performance. Here we measured memory performance in a whole-report procedure to test whether output interference was a concern, and to examine the variations in memory quality between items from a single sample array. In a continuous whole-report recall task, participants recalled the color of each item in the array. In Experiment 1, participants freely recalled the color of presented items (set sizes 1,2,3,4, and 6) in any order they wished by selecting each object and then clicking a color on a surrounding color wheel. In Experiment 2, participants responded to the colored items in an order randomly determined by the computer. When subjects were free to recall the items in any order, we observed a bimodal distribution of high and low precision for the first three versus the last three reported items, respectively. Importantly, these differences in representational quality were not due to output interference: In Experiment 2, participants responded to all items in a random order, and no reliable effects of response order were observed. Thus, continuous whole-report reveals strong variations in the quality of memories from a single array. Moreover, the order of observers’ free recall is highly sensitive to these variations, suggesting that observers have reliable meta-knowledge of which items are best represented. Whole-report procedures may provide richer information about variations in mnemonic storage both within and across trials. In turn, data from such procedures will provide stronger constraints for formal models of memory capacity. Meeting abstract presented at VSS 2013