Recent work on human attention and representational systems has benefited from a growing interplay between research on normal attention and neuropsychological disorders such as visual neglect. Research over the past 30 years has convincingly shown that, far from being a unitary condition, neglect is a protean disorder whose symptoms can selectively affect different sensory modalities, cognitive processes, spatial domains and coordinate systems. These clinical findings, together with those of functional neuroimaging, have increased knowledge about the anatomical and functional architecture of normal subsystems involved in spatial cognition. We provide a selective overview of how recent investigations of visual neglect are beginning to elucidate the underlying structure of spatial processes and mental representations.Once considered a unitary syndrome, it is now accepted that visuospatial neglect results from the interplay of damage to several different cognitive processes [1,2]. Even within such relatively simple tasks as copying or spontaneous drawing, patients show many qualitatively distinct patterns of omission and distortion in 'left' space [3].Deficits of attention, intention, global-local processing, spatial memory and mental representation can all contribute to the clinical picture of neglect, which accordingly cannot be traced back to the disruption of a single supramodal process [1,4]. Many of the symptoms traditionally ascribed to left neglect share little in common other than a contra-lesional gradient of increasing impairment, with comparatively well -preserved performance on the ipsilesional (right) side of space [1,5].Although characterising and remediating the clinical condition [1,2,5] are clearly crucial, the primary interest for cognitive neuroscience is how conditions such as visual neglect can inform existing cognitive theory and functional anatomy. That an abnormal bias towards one side of space can exist in the absence of contralateral peripheral sensory or motor loss (e.g. visual field losses or hemi-paralysis) suggests that impairment to higher level processes must be involved. The standard approach taken by cognitive neuropsychology characterises complex mental processes as information-processing systems with separate but interconnected components. Some of the informationprocessing stages involved in visual object recognition (see Fig. 1) illustrate the potential levels and links where lateralised deficits can arise after brain damage.This model charts the processing routes from preattentive assignment of simple structural features (i.e. without the need for focused attention and conscious awareness) to post-attentional integration of local and global processing (involving selective attention and conscious awareness) and the assignment of relevant spatial frames of reference prior to recognition and naming.