Varying target prevalence reveals two dissociable decision criteria in visual search

Abstract

Target prevalence powerfully influences visual search behavior. In most visual search experiments, targets appear on at least 50% of trials [1-3]. However, when targets are rare (as in medical or airport screening), observers shift response criteria, leading to elevated miss error rates [4, 5]. Observers also speed target-absent responses and may make more motor errors [6]. This could be a speed/accuracy tradeoff with fast, frequent absent responses producing more miss errors. Disproving this hypothesis, our experiment one shows that very high target prevalence (98%) shifts response criteria in the opposite direction, leading to elevated false alarms in a simulated baggage search. However, the very frequent target-present responses are not speeded. Rather, rare target-absent responses are greatly slowed. In experiment two, prevalence was varied sinusoidally over 1000 trials as observers' accuracy and reaction times (RTs) were measured. Observers' criterion and target-absent RTs tracked prevalence. Sensitivity (d') and target-present RTs did not vary with prevalence [7-9]. These results support a model in which prevalence influences two parameters: a decision criterion governing the series of perceptual decisions about each attended item, and a quitting threshold that governs the timing of target-absent responses. Models in which target prevalence only influences an overall decision criterion are not supported.

Figure 1. Experiment One: Very high prevalence elevates false alarms and target absent RTs

Fig 1a: False alarm and miss error rates as a function of target prevalence (50% and 98%), Fig. 1b: Signal detection measures: Average sensitivity (d′) and criterion (c ) values. Fig 1c: Average reaction time (RT) for correct target present (hit) and absent

Figure 2. Experiment Two: Changing target prevalence changes the pattern of errors and target-absent RTs

Fig 2a: Miss (black, solid symbols) and false alarm errors (gray, open), trade off as prevalence (dashed line) varies over 1000 trials. Fig 2b: D_a (black, solid symbols), a signal detection measure of sensitivity does not vary systematically with prevalence but C₂ (gray, open), a criterion measure, does. Fig 2c: Hit RTs (black, solid) change very little with prevalence while True Negative responses (open, gray) vary markedly. False alarm errors (black *) do not vary with prevalence, though they appear to become faster during the experiment. Miss errors (gray *) vary with prevalence in a manner similar to true negatives. (See also Supplemental Figures S1a–c)

Figure 3. The drift diffusion model

In a standard drift-diffusion account of a two-alternative forced-choice (2AFC) task, information begins accumulating a start point generates one response (here “yes”) if it reaches an upper bound and another (“no”) if it reaches a lower bound. For a fixed drift rate, sensitivity (D′) can be varied by varying the separation of the bounds and criterion can be varied by changing the starting point. (See also Supplemental Figures S1d–e)

Figure 4. A multiple-decision model for visual search

In this model, the observer makes a 2AFC decision about each item that is selected. If an item is classified as a target, a “yes” response is generated. If not a new item will be selected unless a target-absent decision is generated when a quitting signal exceeds its threshold. The quitting signal is modeled as a diffusion process. (See also Supplemental Figures S1d–e)