Existence of clinical subgroups among cases of Wilson's disease has long been postulated and various classifications suggested but none statistically tested. This study analyses, by means of pattern recognition techniques, 400 cases from 4 series including our own (n = 195). Factor analysis (to summarise variables) and cluster analysis (to derive groups of patients) were performed. From each series 4 clusters were derived, each containing a "hepatic" and at least one "neurological" cluster, other clusters being "mixed hepatic & neurological" (3 series) and "asymptomatic" (3 series). Two series contained 2 "neurological" clusters; one dividing in terms of age and the other in terms of presence or absence of psychopathology. At least 2 factors, "hepatic" and "neurological", emerged from each series. Reasons for differences are discussed. In conclusion, Wilson's disease is not homogenous: clinical groupings are supported by statistical classification.