CAG repeat expansion in the Huntington's disease gene (HD) was examined in postmortem brains from 310 clinically diagnosed and 15 'at risk' individuals. Presence of an expanded CAG allele (>37 units) was the cause of the disorder in almost all cases (307 of 310). Despite a diversity of reporting clinicians, neurological and psychiatric onset and age at death all displayed significant inverse correlations with CAG number indicating that diagnosis of onset is reasonably accurate, and that most patients die from the disease and its complications. Neuronal changes before clinical onset are not detected by conventional microscopic examination as three out of 15 'at risk' brains had an expanded CAG allele but no neuropathology. The cause of HD-like neuropathology in three exceptional brains from clinically diagnosed individuals is unclear. The disorder in these cases could be an HD phenocopy or result from alternative mutational mechanisms at the HD locus.