Halligan and Marshall [Cortex 27 (1991) 623] devised a new test to evaluate the hypothesis that in visual neglect, left space is systematically compressed rightwards. In the critical condition of the original study, rows of horizontally arranged numbers with a target arrow pointing to one of them from the opposite margin of the display were presented. When asked to verbally identify the number indicated by the arrow, a right brain-damaged patient with left neglect and hemianopia often indicated a number to the right of the target. The more the target was located on the left, the greater the response shift rightward, as if rightward compression were linearly proportional to the co-ordinates of Euclidian space. However, a possible alternative account could be that the patient’s attention was attracted by the numbers located to the right of the target digit, thus biasing her responses toward numbers on the right. To explore this hypothesis, we asked normal participants and patients with right hemisphere lesions, with and without neglect or hemianopia, to mark on the margin of a sheet the approximate location indicated by an arrow situated on the opposite margin. In three different conditions, the arrow indicated either one of several numbers or lines in a row, or a blank location on the sheet margin. Only patients with left neglect, and especially those with associated hemianopia, deviated rightward, and then crucially only on those conditions where visible targets were present, consistent with the attentional bias account.