Abstract
The Low Prevalence Effect (LPE), the increased rate of misses for rare targets, is a stubborn problem with potential consequences for real-world searches. One promising method for mitigating LPE is to add “probe” trials, consisting of a target with feedback, to a low-prevalence search task. However, the original experiments using this method had a single target, thus the target was identical in probe and non-probe trials. Here we investigate the extent to which this probe benefit generalizes to targets that are probed infrequently or not at all. In theory, if probes increase quitting thresholds, the benefits might generalize. In one paradigm, Ss searched for Ts and Os among L and Q distractors. A control block consisted of 240 trials and each target appeared 24 times. The experimental block had the same 240 trials plus 50 randomly dispersed probe trials, and we varied the percentage of probe trials that matched each target. Results suggested that the probe benefit was proportional to the number of probes that matched a specific target. Eye-tracking data suggest that probes increased quitting thresholds, but this failed to produce a generalized benefit because it was coupled with an attentional bias toward features of the commonly probed targets. In a second approach, Ss searched for weapons among an array of photorealistic objects. Targets were exemplars drawn from three basic-level categories (guns, knives, axes). In a control block, each basic-level target appeared with a 6.67% prevalence rate. In the probe block, 80% of probe targets matched one basic-level category, 20% matched a second, and 0% matched the third. We hoped that probing two of the basic-level categories would produce a benefit that generalized to the superordinate category, however results suggest that the probe benefit did not generalize to the category. Thus, the LPE continues to be a stubborn problem.