The probability of missing a target increases in low target prevalence search tasks. Wolfe and Van Wert (2010) propose 2 causes of this effect: reducing the quitting threshold, and conservatively shifting the decision making criterion used to evaluate each item. Reducing the quitting threshold predicts that target absent responses will be made without fully inspecting the display, increasing misses due to never inspecting the target (selection errors). The shift in decision criterion increases the likelihood of failing to recognize an inspected target (identification errors). Though there is robust evidence that target prevalence rates shift quitting thresholds, the proposed shift in decision making criterion has little support. In Experiment 1 we eye-tracked participants during searches of high, medium, and low prevalence. Eye movements were used to classify misses as selection or identification errors. Identification errors increased as prevalence decreased, supporting the claim that decision criterion becomes more conservative as prevalence decreases. In addition, as prevalence decreased, the dwell time on targets increased while dwell times on distractors decreased. We propose that the effect of prevalence on decision making for individual items is best modeled as a shift in criterion in a drift diffusion model, rather than signal detection, as drift diffusion accounts for this pattern of decision times. In Experiment 2 we replicate these findings while presenting stimuli in an rapid serial visual presentation (RSVP) stream. Experiments 1 and 2 were consistent with the conclusion that prevalence rate influences the item-by-item decision criterion, and are consistent with a drift diffusion model of this decision process.