A variety of stimulus factors have been shown to influence the degree of leftward displacement of perceived line midpoint (i.e., pseudoneglect), which typifies the performance of normal subjects in line bisection tasks [M.E. McCourt & G. Jewell: Neuropsychologia 37, 843-855 (1999); G. Jewell & M.E. McCourt: Neuropsychologia 38, 93-110 (2000)]. One such factor is the position of lines within the visual field, where two conflicting patterns of bisection error have been reported. Some authors report a centrifugal pattern of error, where perceived line midpoint shifts away from the vertical midline, regardless of line position, i.e., relatively leftward for leftward displaced lines and vice versa. Others have reported a centripetal pattern of bisection error, where perceived line midpoint is always displaced centrally, toward the vertical midline, regardless of line position. There is no satisfactory explanation for these discrepant findings. An experiment using a tachistoscopic forced-choice line bisection protocol is described which discloses that neurologically normal right-handed subjects (N=82) typically display a centrifugal pattern of bisection error when lines are azimuthally displaced over a relatively small range, whereas a centripetal pattern is observed when lines are displaced over a wider range. Results from ancillary control experiments, in which eye position was measured during testing, confirm that systematic differences in gaze direction do not occur as a function of line position, and thus cannot account for the different patterns of bisection error. We conclude that stimulus context significantly modulates the strategy with which observers deploy spatial attention. When line position is constant, or varies over a narrow range, observers hold attention steady and widen its aperture to accommodate the relevant range of spatial location. Centrifugal bisection error is thus produced by the asymmetric cueing effect of laterally displaced lines, according to the activation-orientation theory [M. Kinsbourne: Acta Psychologica 33, 193-201 (1970)]. When the range of line position exceeds the aperture of focal attention, we hypothesize that observers adopt a strategy in which attention is dynamically scanned in the direction of azimuthally displaced lines. The effects of attentional scanning on line bisection performance are quite robust. The centripetal scanning proposed to occur for widely displaced lines is consistent with the centripetal pattern of bisection error in this condition.