Abstract
Searching a complex display for a rare target is involved in a number of socially important tasks, such as screening for cancerous cells, or checking passenger bags at airport security. Unfortunately, the probability of missing a target increases as target prevalence decreases. For example, in a search for weapons in a baggage-screening simulation, observers missed 40% of low prevalence targets but just 7% of high prevalence targets (Wolfe, Horowitz & Kenner, Nature, 435, 2005). These high error rates seem to occur because participants dismiss a display as ‘target absent’ too quickly, resulting in an error when the target is present. This might occur because in low prevalence search, participants only search a subset of the display before they make a decision. This would result in fewer fixations when targets are infrequent. Alternatively, participants might spend less time looking at each item in the display (and hence less time at each fixation), and simply fail to recognize the target item. We replicate the low prevalence result using a simple visual search task in which the target is a ‘T’ among ‘L’s, and measure eye movements on both target-present and target-absent displays. Participants make fewer fixations when targets are rare than when they are frequent, but there is little difference in fixation duration, suggesting that participants are less likely to search the whole display in low prevalence searches. These data have implications for strategies to improve the accuracy of searching for rare targets.
Supported by the National Health & Medical Research Council, Australia (ANR), and the FAA/DHS (JMW)