Abstract
The low-prevalence effect is the visual search finding that target detection is dramatically reduced for rare targets. This effect is robust, difficult to mitigate, and has implications for real-world searches (e.g., baggage screening; radiology). We previously showed that distributing probe trials which involved feedback throughout a search task (rather than in a miniblock as done by Wolfe et al., 2007) improved accuracy when compared to control condition without probes. But it is unclear whether this benefit requires probe targets that are identical to the rare targets. To address this issue, we asked participants to search for a T or O among arrays consisting of L and Q distractors. They performed a control block of low-prevalence (10%) search and another block that was identical but included 50 probe trials scattered throughout. In Experiment 1, 80% of the probe trials displayed of one target with 20% displaying the other target. Compared to the control block, probes improved accuracy and produced longer target absent reaction times, suggesting that the probes produced higher quitting thresholds. Additionally, both the highly and lowly probed targets improved in accuracy to roughly the same extent. To push the limits of this finding, in Experiment 2 we attempted to replicate the findings with 100% of the probes matching a single target. While we still found a probe benefit (higher accuracy and longer target absent RTs) when compared to the control, only the probed target showed an improvement in accuracy. These findings suggest that while not all targets need to be probed at the same rate to improve their detection, in order to produce a benefit some of the probes must match the target. Follow-up research will more closely examine the limits of this probe benefit.