Abstract
Change detection performance is habitually assumed to reflect the capacity of visual short-term memory (VSTM). Accordingly, it has been suggested that only four or fewer items can be simultaneously encoded into VSTM. Recent indirect measures of change detection, however, suggest that explicit reports - typically requested in the change detection paradigm - may underestimate the capacity of VSTM. Studies on implicit change detection, for example, showed that observers remain able to detect a change better than chance would predict even when claiming they failed to perceive it (e.g., Fernandez-Duque & Thornton, 2000). Likewise, studies on eyes movements revealed that when observers fail to explicitly report a change, fixation durations on changed objects are significantly longer (e.g., Hollingworth & Henderson, 2002). Here, we used a cueing paradigm, similar to that used by Becker et al. (2000), to directly investigate whether the capacity of VSTM is underestimated in change detection tasks. The memory array consisted of 6 colored shapes and was presented for 500 ms, with accurate retention required over a subsequent 1500-ms delay. The test array consisted of one colored shape, and participants were instructed to decide whether the shape and the color of that object came from the same object or from two different objects in the memory array. Change detection improved when the location of the two objects that could potentially swap their features was cued during the interval, namely 700 ms after the offset of the memory array. Given that participants must first encode the objects and the associations of features in VSTM in order to make proper use of the cue, these results indicate that we may have a larger storage capacity in VSTM than that indicated by the change detection paradigm. This clearly challenges the hypothesis that explicit change detection provides an exhaustive measure of VSTM.