Abstract
According to the saliency theories of visual search, basic features like luminance, orientation, and size are extracted in parallel across the visual field. Activation in a master saliency map is determined by local differences between these features, and this activation determines the allocation of attention (Itti & Koch, 2001). In a previous experiment, luminance was found to follow the saliency hypothesis that attention will visit items in order of their salience (Tsai & Peterson, 2006). In this study, participants were tested using a visual search task involving a target and distractor items arranged in a circle around the periphery of the display. The feature dimensions tested were luminance, orientation, and size. Salient objects were brighter, darker, larger, smaller, or more tilted than the other objects in the display. In addition, each display contained two salient objects, with one more salient than the other (psychophysically matched). The target coincided with any of the items at chance level. Subjects were faster to report the target in the brightest, darkest, or largest trials when it occurred in the most salient item. However, in contrast to the predictions of the saliency model, this speed advantage disappeared when the target appeared in the second-most salient item. That is, attention was biased to visit the most salient item first, but this bias did not extend to the visiting the second-most salient item next. In the conditions where the most salient item was the smallest or defined by orientation, there were no significant effects on reaction times.