Patients with left unilateral neglect bisect long horizontal lines to the righ t of the true centre. However, when given short lines, many of the same patients mark the midpoint to the left of the true centre, towards the otherwise neglect ed space. This paradoxical phenomenon has been termed ‘cross-over’and is diff icult to explain based on current accounts of the neglect syndrome. To explore t he causes of cross-over, in a first study we evaluated bisection of 20, 100 and 200 mm horizontal lines in groups of unilateral brain-damaged patients with ne glect and hemianopia, with neglect and no hemianopia, with hemianopia and no neg lect and without neglect or hemianopia. Cross-over of 20 mm lines was found onl y in neglect patients with hemianopia. To ascertain further the influence of vis ual field defects on cross-over, in a second study we compared the performance of two rightbrain damaged patients with contralesional neglect and inferior quad rantanopia with that of a patient with inferior quadrantanopia and no neglect. P atients bisected lines oriented so as to cross or uncross the blind quadrant of the visual field. When short 20 mm lines crossed the blind quadrant, neglect pat ients showed cross-over; when the same lines crossed the seeing quadrants cross -over was absent. These findings were confirmed by the examination of a neglect patient with sparing of the central 5°of the contralesional left visual hemifi eld in the right eye and no sparing in the left eye. In monocular viewing, cross over was present when 20 mm lines were bisected with the left eye and ab sent when bisected with the right eye. Recording of eye movements showed that at the moment of bisection left eye fixations shifted towards the contralesional line endpoint whereas right eye fixations remained anchored to the centre of th e line. With long lines, both eyes deviated ipsilesionally. These results show t hat in neglect patients ipsilesional deviation in the bisection of long lines tu rns into apparently paradoxical contralesional bisection of short ones only when these cross a retinotopically blind sector of the neglected space. Cross-over seems to depend on the small spatial effects produced by reflexive contralesiona l gaze shifts allowing eccentric fixations with the seeing hemifield. During the bisection of long lines, these effects are cancelled out by the strong attentio nal deviation induced by the marked extension of the ipsilesional line segment. This explanation establishes coherence between cross-over and current accounts of the neglect syndrome. Patients with left unilateral neglect bisect long horizontal lines to the righ t of the true center. However, when given short lines, many of the same patients mark the midpoint to the left of the true center, towards the otherwise neglect ed space. This paradoxical phenomenon has been termed ’cross-over’and is diff icult to explain based on current accounts of the neglect syndrome. To explore t he causes of cross-over, in a first study we evaluated bisection of 20, 100 and 200 mm horizontal lines in groups of unilateral brain-damaged patients with ne glect and hemianopia, with neglect and no hemianopia, with hemianopia and no neg lect and without neglect or hemianopia. Cross-over of 20 mm lines was found onl y in neglect patients with hemianopia. To ascertain further the influence of vis ual field defects on cross-over, in a second study we compared the performance of two rightbrain damaged patients with contralesional neglect and inferior quad rantanopia with that of a patient wi tharse quadrantanopia and no neglect. P atients bisected lines oriented so as to cross or uncross the blind quadrant of the visual field. When short 20 mm lines crossed the blind quadrant, neglect pat ients showed cross-over; when the same lines crossed the seeing quadrants cross -over was absent. These findings were confirmed by the examination of a neglect patient with sparing of the central 5 ° of the contralesional left visual hemifi eld in the right eye and no sparing in the left eye. In monocular viewing, cross over was present when 20 mm lines were bisected with the left eye and ab sent when bisected with the right eye. Recording of eye movements showed that at the moment of bisection left eye fixations shifted towards the contralesional line endpoint rather right eye fixations remained anchored to the center of th e line. With long lines, both eyes deviated ipsilesionally. These results show t hat in neglect patients ipsilesional deviation in the bisection of long lines tu rnsinto apparently paradoxical contralesional bisection of short ones only when the small spatial effects produced by reflexive contralesiona l gaze shifting allows eccentric fixations with the seeing hemifield. During the bisection of long lines, these effects are canceled out by the strong attentio nal deviation induced by the marked extension of the ipsilesional line segment. This explanation establishes coherence between cross-over and current accounts of the neglect syndrome.