Abstract
BACKGROUND: Visual search behavior is influenced by target prevalence. Wolfe & Van Wert (2010) found that, whereas target-absent response times (RT) and decision criterion varied with prevalence, target-present RT and overall sensitivity (d') did not. These results support a model that assumes target prevalence to influence two relatively distinct search processes. The first process governs the acceptance / rejection of each attended item. The second process governs the timing for terminating search with a target-absent response. PRESENT STUDY: The present study examines two tenets of this model. First, as the model predicts that target prevalence independently exerts its effects on decision criterion and target-absent RTs via different paths, the sizes of the two effects should be independent. In Experiment 1, I asked participants to search for an oriented line among vertical lines, manipulated target prevalence (20% vs. 80%), and measured decision criterion and target-absent RT. Contrary to the model's prediction, prevalence effects on the two measures correlated across participants (r=0.545, p=.013), indicating that the two processes are not entirely independent. Second, the model predicts that prevalence has no effect on target-present RT, because prevalence affects only the per-item decisions, but not the inspection duration for each search item. Indeed, most studies did not find this effect (with some exceptions), possibly because the effect was small and went undetected. In Experiment 2, I attempted to magnify this effect by adding orientation variations to the distractors, so that they became more confusable with the target. The result now showed a significant target-prevalence effect on target-present RT (p=.019). CONCLUSION: Taken together, the present results indicate that target prevalence affects both target-present and target-absent decisions, suggesting that both decisions may be governed by a common decision process. In my presentation, I will outline a decision mechanism that supports both response types and accommodates existing findings.
Meeting abstract presented at VSS 2017