Abstract
The characteristics of stimuli from previous trials can distort visual short-term memory for the current trial’s stimulus (Huang & Sekuler, 2010; Kang, Hong, Blake, & Woodman, 2011). However, the degree to which prior information influences memory, and the variables that control reliance on this information remain unknown. In the current study, we explored prior information’s influence on memory for briefly-viewed random dot cinematograms (RDCs). Subjects viewed RDCs at two levels of coherence (50 and 90%). Following each presentation, subjects used a cursor to indicate the overall direction in which the dots had moved. Within a block of trials, RDC directions were drawn randomly from a narrow distribution located within a particular quadrant. The distribution’s central direction defined a block’s prototype direction, which changed from block to block. Subjects’ errors were sign corrected such that errors in the direction of the prototype stimulus (the block’s mean direction) were positive, and those in the opposite direction were negative. The results revealed a prototype effect: errors were shifted toward the mean direction of motion within a given block of trials. We analyzed the time-course of errors within a block of trials in which the prototype (mean) direction was fixed. The prototype effect was limited to the low coherence RDCs, and was strongest early in a block, when the prototype direction was novel. When errors were scored relative to the mean direction of the immediately preceding block, no prototype effects were observed. These results are inconsistent with an account in which all previous trials are equally-weighted in computing priors for visual motion (Morgan, Watamaniuk, & McKee, 2000). A simple model incorporating the idea that reliance on prior information is strongest in the face of uncertainty about the contents of memory provides a good account of the critical data.
Meeting abstract presented at VSS 2012