Structural magnetic resonance imaging (MRI) data have provided much evidence in support of our current view that schizophrenia is a brain disorder with altered brain structure, and consequently involving more than a simple disturbance in neurotransmission. This review surveys 118 peer-reviewed studies with control group from 1987 to May 1998. Most studies (81%) do not find abnormalities of whole brain/intracranial contents, while lateral ventricle enlargement is reported in 77%, and third ventricle enlargement in 67%. The temporal lobe was the brain parenchymal region with the most consistently documented abnormalities. Volume decreases were found in 62% of 37 studies of whole temporal lobe, and in 81% of 16 studies of the superior temporal gyrus (and in 100% with gray matter separately evaluated). Fully 77% of the 30 studies of the medial temporal lobe reported volume reduction in one or more of its constituent structures (hippocampus, amygdala, parahippocampal gyrus). Despite evidence for frontal lobe functional abnormalities, structural MRI investigations less consistently found abnormalities, with 55% describing volume reduction. It may be that frontal lobe volume changes are small, and near the threshold for MRI detection. The parietal and occipital lobes were much less studied; about half of the studies showed positive findings. Most studies of cortical gray matter (86%) found volume reductions were not diffuse, but more pronounced in certain areas. About two thirds of the studies of subcortical structures of thalamus, corpus callosum and basal ganglia (which tend to increase volume with typical neuroleptics), show positive findings, as do almost all (91%) studies of cavum septi pellucidi (CSP). Most data were consistent with a developmental model, but growing evidence was compatible also with progressive, neurodegenerative features, suggesting a "two-hit" model of schizophrenia, for which a cellular hypothesis is discussed. The relationship of clinical symptoms to MRI findings is reviewed, as is the growing evidence suggesting structural abnormalities differ in affective (bipolar) psychosis and schizophrenia.