Is visual working memory modulated by prior knowledge about probability? An ERP study.

Prominent theories of visual attention claim that the contents of visual working memory (VWM) can guide attention during visual search. This is supported by numerous experimental observations showing that information held in VWM automatically captures attention (e.g., Soto et al., 2008, Trends Cong Sci). Recently, Carlisle & Woodman (2011, Acta Psychol) reported that the effects of attentional capture varied according to prior instructions about the probability of memory-match trials, which suggests that strategic use of the contents held in VWM may alter visual search performance. However, it remains unclear how these instructions influence the contents stored in VWM. In the present study, we used contralateral delay activity (CDA) as an electrophysiological index of VWM load. The task was to conduct visual search while holding an item in VWM. There were three within-subjects conditions, across which we pre-informed participants about the probability (20%, 50%, or 80%) of memory-match trials. We hypothesized that VWM would be more loaded with more probability of memory-match trials. We found that RT costs (invalid minus neutral trials) in the 80% condition were larger than in the 20% and 50% conditions, but there were no between-condition differences in RT benefits (valid minus neutral trials). These results indicate that prior knowledge about probability affected RT costs only. Unexpectedly, CDA was only obtained in 50% condition. These results imply that knowledge about biased-probability (20% and 80%) leads to a decrease in the demands to attend to or hold to-be-memorized item. In conclusion, the present ERPs findings, combined with the RT results, suggest that changes in RT costs across conditions by instruction may not depend on VWM activity as measured by CDA. Meeting abstract presented at VSS 2015.