Abstract
In visual search, targets that appear infrequently are missed disproportionately often, relative to those that appear more frequently. This "low-prevalence effect" (LPE) has significant societal consequences (e.g., missed tumors or weapons in applied screening contexts), and is challenging to alleviate. Previous eye-tracking research has demonstrated that low-prevalence targets are often missed even when they are directly fixated. The LPE has been attributed to several non-mutually exclusive mechanisms, including criterion shifts, search termination errors, and reduced priming for lower-prevalence targets. However, using passive (RSVP) search tasks, we previously found that the LPE persists even when searchers do not terminate search on their own. This suggests that low-prevalence targets have a weaker target template than their higher-prevalence counterparts (Hout et al., 2015). Here, we investigated low-prevalence misses by adopting a new method for examining target template strength: Pupillometry. Participants completed a relative-prevalence search task, wherein they searched for two target categories, butterflies and teddy bears. Although targets appeared in 50% of the trials, one category appeared in 5% of trials, and the other appeared in the remaining 45%. We used an RSVP paradigm to ensure that target objects were examined directly: Target objects were presented amid streams of rapidly presented pictures. After the stream concluded, participants indicated whether a picture from one of the target categories was present, or if both categories were absent. We tracked pupil dilation (a marker of memory strength and specificity) over three blocks of 100 trials. Across blocks, we observed a pupillary prevalence effect, wherein high-prevalence targets elicited greater pupil dilation, relative to low-prevalence targets. These findings suggest that high-prevalence targets resonate more strongly with mental representations of target templates used during search, and provide further evidence that low-prevalence targets are missed (at least in part) due to weaker templates, reducing "perceptual readiness" for their detection.
Meeting abstract presented at VSS 2016