Abstract
A search for a rare target in a complex display is important in a number of important tasks, including baggage screening at airport security, or screening for cancerous cells. Wolfe, Horowitz and Kenner (Nature, 435, 2005) demonstrated that the probability of missing a target increases as target prevalence decreases. These high error rates seem to occur because low prevalence encourages participants to set a response criterion that causes them to make rapid ‘target-absent’ responses. Wolfe et al. used complex approximations of luggage x-rays as stimuli. Last year, we demonstrated that the effect of low prevalence could be seen with less complex stimuli where the search is nonetheless relatively inefficient (a search for a T among Ls; Rich et al. VSS 2006). How far can this effect be pushed? In simple feature search tasks, the target is said to “pop-out”. Would observers miss a horizontal target among vertical distractors at low prevalence? Surprisingly, error rates were significantly greater at 2% target prevalence (∼14%) than at 50% (∼2.5%), demonstrating that participants miss even obvious targets if they are infrequent. In a second experiment, we forced participants to wait a minimum duration before responding. This effort to reduce possible speed-accuracy trade-offs reduced miss errors at low prevalence (∼5%), although this rate was still higher than in 50% prevalence minimum exposure conditions (∼2%). Thus, low prevalence can induce observers to miss highly salient targets and speed-accuracy trade-offs account for most but not all of this effect.
ANR is supported by NHMRC & R.G. Menzies Fellowships (Australia)