Random-dot autostereograms (RDASs) were used to investigate attention shifts along the sagittal plane in distractor-free tasks of high perceptual load. In three experiments using a same/different comparison task, the shape of the gradient over five different depths was examined and the conditions under which the gradient is and is not observed were compared. When the target set consisted of five similar objects, a robust asynunetric depth gradient was observed. When the target set consisted of two dissimilar objects, no gradient was observed. The results support a hypothesis of a viewer-centered asymmetric attention gradient in the depth plane that is dependent on perceptual or attentionalload defmed by target-set discriminability.It is well established that attention can be allocated to different positions in a two-dimensional (2-D) visual field (Eriksen & St. James, 1986;LaBerge, 1983;Posner, 1980;Posner & Cohen, 1984). However, maneuvering through a three-dimensional (3-D) world also requires shifts ofattention in depth. Moving toward a distant object, for example, may involve constant shifts of visual attention between objects in the immediate foreground and those in the distance.Studies using real 3-D scenes have shown distance effects as well as directional asymmetries. These effects are measured as costs and benefits to validly and invalidly cued targets in a variation ofthe classic Posner (1980) task and suggest that attention can be allocated to different depth planes. For example, in experiments that present curved rows of lights at different depths (Downing & Pinker, 1985) or single lights aligned along the sagittal axis of the observer (Gawryszewski, Riggio, Rizzolatti, & Umilta, 1987), responses are faster to validly cued targets and slower to invalidly cued targets. Moreover, there is an asymmetry: A faster reallocation ofattention is found when observers switch attention from a far to a near (F-N) location than when they switch attention from a near to a far (N-F) location, suggesting that attention in depth operates from a body-centered awareness.Criticisms ofthese studies have addressed the fact that target intensities may vary in depth in a real 3-D scene This research was supported by Natural Sciences and Engineering Research Council of Canada Grant OGPOl70353 to 1.M.S. We thank Paul Atchley, Claude Alain, Jan Theeuwes, and two anonymous reviewers for their comments on earlier drafts of this manuscript. Thanks to Shanthy Ratnasingari for collecting the data for Experiment 2 and to Stephanie Hevenor for collecting the data for Experiment 3. Correspondence concerning this article should be directed to 1. M. Shedden, 406 Psychology, McMaster University, Hamilton, ON, L8S 4K I Canada (e-mail: shedden@mcmaster.caorstephena@psych.utoronto.ca). (Andersen & Kramer, 1993;Atchley, Kramer, Andersen, & Theeuwes, 1997). This may lead to behavioral differences that are based on discrimination owing to intensity rather than on differences in depth. Furthermore, the use oflong cue-to-target stimulus onset asy...