Event-related brain potentials (ERPs) were recorded in response to unilateral arrays of letters flashed in rapid, randomized sequences to left and right visual field locations. Subjects were required to focus attention exclusively on either left or right field stimuli, or to divide attention in different proportions between the two fields, with the aim of detecting infrequent target letters. Both d' and percent hits for target detections increased significantly as attentional allocation to a stimulus location increased. Attention operating characteristic (AOC) curves for the target detection scores were highly similar in form to those for the amplitudes of the long-latency, endogenous ERP components-N350-650 and P400-800 (P300). All of these measures showed gradual, nearly rectangular tradeoff functions. In contrast, the AOC curves for the early sensoryevoked components displayed steep, nearly linear amplitude tradeoffs as attention was increasingly allocated to one visual field at the expense of the other. The early and late ERP components were considered as indices of separate but interacting levels of attentional selection having different operating principles.The ability to direct attention voluntarily to selected regions of visual space plays a central role in human perception. When attention is focused on a particular location, stimuli presented at that location can be responded to with greater speed (see, e.g., Eriksen & Yeh, 1985; Hawkins, Shafto, & Richardson, 1988;Posner, Snyder, & Davidson, 1980) and discriminated with greater accuracy (Downing, 1988; Prinz metal , Presti, & Posner, 1986). Yet despite the growing evidence that spatial attention influences perceptual processing, considerable controversy exists concerning the mechanisms that underlie the facilitated performance observed to occur with spatial attention. It has been argued, for example, that changes in decision and response bias factors rather than enhanced perceptual processing mediate the performance benefits of spatial attention (e.g., Milller & Findlay, 1987;Muller & Rabbitt, 1989;Shaw, 1984;Sperling, 1984).Mechanisms of visual selective attention have also been studied extensively, using physiological measures of sensory and perceptual processing. In particular, eventrelated brain potentials (ERPs) provide measures of the transmission of sensory information through successive levels of the afferent pathways and cortical receiving areas. ERP studies of visual-spatial attention in humans have identified attention-related amplitude modulations of the early, sensory-evoked PI and Nl peaks as well as longer latency ERP components related to selective stimulus processing (Eason,