The effect of selective spatial attention on peripheral discrimination thresholds

Abstract

Experiments were conducted to investigate the role of attention in peripheral detection and discrimination. Advance spatial cues informed subjects about likely target positions; the task required to detect/discriminate plus localise a target briefly presented at cued or uncued locations, with accuracy as the dependent variable ("cost-benefit" analysis).Spatial cueing produced reliable advantages for cued over uncued locations, in single and in multiple element displays. However, costs plus benefits were less marked for single displays. Thus, advance knowledge of the likely target location enhances performance also when there are no competing stimuli present in the visual field. But costs plus benefits are smaller because single target onsets at uncued locations summon attention in the same "automatic" fashion as peripheral cues. Peripheral cues trigger a rapid facilitatory component (automatic), fading out within 300 msec after cue onset. Facilitation is then maintained by a less effective mechanism (controlled). Central cues initiate only this second component. Sustained, controlled, orienting in response to central cues is interruptable by automatic orienting in response to uninformative peripheral flashes. Interruption also occurs when irrelevant flashes compete with peripheral cues. However, interference is less marked for the early automatic than for the following controlled orienting component. Indication of a second position (four-location display) to be most likely resulted in a marked sensitivity gain for this position, relative to uncued locations in a single cue condition. That is, attention could be simultaneously shared between two cued positions. For a luminance detection task (single target), cued locations showed no advantage in sensitivity; but for letter detection tasks (target plus distractors), there was a marked priming effect. That is, letter detection is capacity limited, whereas luminance detection is not. In all tasks, decision criteria are largely preset according to a-priori target probabilities assigned to particular locations, i.e. more liberal for cued and more conservative for uncued locations.