AbstractUnderstanding of visual hallucinations is developing rapidly. Single‐factor explanations based on specific pathologies have given way to complex multifactor models with wide potential applicability. Clinical studies of disorders with frequent hallucinations—dementia, delirium, eye disease and psychosis—show that dysfunction within many parts of the distributed ventral object perception system is associated with a range of perceptions from simple flashes and dots to complex formed figures and landscapes. Dissociations between these simple and complex hallucinations indicate at least two hallucinatory syndromes, though exact boundaries need clarification. Neural models of hallucinations variably emphasize the importance of constraints from top down dorsolateral frontal systems, bottom up occipital systems, interconnecting tracts, and thalamic and brainstem regulatory systems. No model has yet gained general acceptance. Both qualitative (a small number of necessary and sufficient constraints) and quantitative explanations (an accumulation of many nonspecific factors) fit existing data. Variable associations of hallucinations with emotional distress and thought disorders across and within pathologies may reflect the roles of cognitive and regulatory systems outside of the purely perceptual. Functional imaging demonstrates that hallucinations and veridical perceptions occur in the same brain areas, intimating a key role for the negotiating interface between top down and bottom up processes. Thus, hallucinations occur when a perception that incorporates a hallucinatory element can provide a better match between predicted and actual sensory input than does a purely veridical experience. Translational research that integrates understandings from clinical hallucinations and basic vision science is likely to be the key to better treatments. WIREs Cogn Sci 2010 1 781–786This article is categorized under: Psychology > Perception and Psychophysics Neuroscience > Cognition