Abstract
Does a target representation maintained in visual working memory (VWM) change with the expected difficulty of a search task? Previous work suggested that it does not (Gunseli, Olivers, & Meeter, 2014). We investigated this question, and how expected search difficulty impacts later search guidance, by having participants search for teddy bear targets among either other teddy bears (difficult search, high target-distractor similarity) or random non-bear objects (easy search, low target-distractor similarity). Target previews were identical in these two conditions. We measured target-related VWM load using contralateral delay activity (CDA), an event-related potential indicator of VWM load, after target designation but before search display onset. Expecting a difficult search produced larger CDA compared to when an easy search was expected (p < .01). Moreover, we found an interaction between time and task difficulty (p < .01). Comparing difficult to easy search, difficult search resulted in stronger CDA shortly after preview offset (p ≤ .05), with this difference disappearing shortly before search onset (p > .37). We also found a cross-over interaction (p = .05) in CDA between expected search difficulty and search guidance. For difficult searches, when the initial search saccade was directed to the target (strong guidance) target-related VWM load was higher than when the initial search saccade was directed to a distractor (weak guidance). The opposite pattern was found for easy searches; target-related VWM load was higher for weak guidance trials. These findings demonstrate that target representations, and their subsequent impact on search guidance, are modulated by expected search difficulty. Constructing target representations may entail the extraction of a large pool of target details that are rapidly pruned away to create a smaller set of discriminative target features (Schmidt, MacNamara, Hajcak, & Zelinsky, 2014), suggesting that search target representations are fluid constructs that change over time to adapt to task demands.
Meeting abstract presented at VSS 2016