Abstract
There is substantial evidence that the capture of spatial attention by salient, irrelevant stimuli is subject to modulation by top-down task set. To accommodate such findings, the theory of Contingent Attention Capture (CAC) proposes that the attention allocation system is "configurable" such that a given stimulus will only capture attention if it is consistent with current top-down "attentional control settings" related to behavioral goals. Most of the previous research on CAC has focused on capture by salient, low level properties, implying that attentional control settings are limited to working memory templates for simple features such as color, orientation, motion, singleton-ness, etc. A recent study, however, found that relatively high-level, categorical content of an image can capture attention if the task requires monitoring for exemplars from that category (Wyble, Folk, & Potter, 2013). This suggests that the activation of categorical representations in semantic memory can also serve as top-down attentional control settings. The present study tested whether these effects are truly the result of semantic memory activation by testing whether images that are semantically related to the task-relevant category also capture attention. Subjects monitored a central RSVP stream for any exemplar (e.g., a ferris wheel) from a pre-specified semantic category (e.g., amusement park rides). Two frames prior to the target image, irrelevant distractor images appeared above and below the stream, one of which could be another exemplar from the relevant category (e.g., bumper cars), an image of a related item (e.g., cotton candy), or an image of an unrelated item (e.g., a hammer). Distractors consisting of exemplar and semantically related images both produced a decrement in target report relative to semantically unrelated distractors, suggesting that the attention allocation system can be configured not only through templates in working memory, but also through the activation and priming of long-term semantic memory representations.
Meeting abstract presented at VSS 2014