Combination of evidence in a probabilistic visual search and detection task

Abstract Subjects performing a visual search and detection task on an inherently noisy display were provided with unreliable categorical cues as a secondary source of information, in a study of combination of evidence in psychophysical judgment. Seventy-six subjects served in a separate-groups design with five cue modes and an uncued control. The secondary information produced reliably better detection efficiency only when it included signal location information, as without this the gain in detections was offset by a higher false alarm rate. The reliance on the cues was conservative, probably due to underestimation of their reliability, and negative cues did not appear to affect subjective signal probability relative to performance without cues. Performance with the secondary information is discussed in relation to a model for cue utilization based on signal detection theory and Bayes' theorem.