A major development in recent research on human visual attention has been the increasing interplay between research on normal attentional mechanisms, and accounts of unilateral neglect and extinction after brain-damage in terms of damage to these mechanisms. Although there are potential pitfalls in this approach, it has already proved useful. This is illustrated for the debate over whether segmentation processes precede spatial attention in vision. This debate began in the normal literature, but has since motivated several studies of visual neglect and extinction. These reveal that various segmentation processes can influence which region of a scene will be neglected or extinguished, implying that grouping may precede the abnormal bias in spatial attention, and showing that considerable residual processing can take place in the neglected or extinguished visual field. The extent of this residual processing is tentatively related to emerging anatomical data.