Abstract
In the present study we investigated observers' ability to make use of saliency-information in a search display. In Experiment 1 participants were instructed to make an eye movement to either the most or the least salient singleton in a search display. The results showed that short-latency eye movements were driven by stimulus-saliency irrespective of the instruction given to the participants. In contrast, long-latency saccades were driven by instruction only and showed no effect of stimulus-salience. The goal of Experiment 2 and 3 was to investigate whether observers were aware of the salience and identity of the items selected. In Experiment 2 and 3, the search display was masked contingent to the eye movement, such that observers viewed a mask when their eyes landed on the saccadic goal. Following each trial, participants were asked to report whether they had correctly selected the most salient item (Experiment 2); or correctly selected the right identity, namely the line element tilted to the right (Experiment 3). Results showed that, regardless of saccadic latency, sensitivity as measured by d' was dramatically low when participants were asked to report on the saliency of the saccadic goal. In contrast, participants were near perfect to report on the identity of the saccadic goal. We conclude that visual selection based on saliency information occurs only when selection latencies are low, and that saliency-driven selection is impenetrable to consciousness.