Crossover refers to a pattern of performance on the line bisection test in which short lines are bisected on the side opposite the true center of long lines. Although most patients with spatial neglect demonstrate crossover, contemporary theories of neglect cannot explain it. In contrast, we show that blending the psychophysical construct of magnitude estimation with neglect theory not only explains crossover, but also addresses a quantitative feature of neglect that is independent of spatial deficits. We report a prospective validation study of the orientation/estimation hypothesis of crossover. Forty subjects (17 patients with and without neglect following unilateral brain injury and 23 normal controls) completed four experiments that examined crossover using line bisection, line bisection with cueing, and reproducing line lengths from both memory and a standard. Replicating earlier findings, all except one subject group exhibited crossover on the standard line bisection test, all groups showed a spontaneous preference to orient attention to one end of the lines, and all groups overestimated the length of short lines and underestimated long lines. Biases in attentional orientation and magnitude estimation are exaggerated in patients with neglect. The truly novel finding of this study occurred when, after removing the line from the bisection task, the direction of crossover was completely reversed in all subject groups depending on where attention was oriented. These findings are consistent with our hypothesis of crossover: (1) crossover is a normal component of performance on line bisection; (2) crossover results from the interplay of biases in attentional orientation and magnitude estimation; and (3) attentional orientation predicts the direction of crossover, whereas a disorder of magnitude estimation, not previously emphasized in neglect, accounts for the quantitative changes in length estimation that make crossover more obvious in neglect subjects. Paradoxically, we observed that the traditional line bisection test is suboptimal for exploring crossover because lines elicit spontaneous orientation responses from subjects that confound experimental manipulations of attention. We conclude that attentional orientation and magnitude estimation are necessary and sufficient to explain crossover and that bias in magnitude estimation is a core component of neglect. © 2005 Massachusetts Institute of Technology.