We have used multiple sources to identify a population-representative cohort of newly diagnosed patients with parkinsonism and Parkinson's disease in the UK over a 2-year period. All patients have been invited to participate in a detailed clinical assessment either at home or in an outpatient clinic. These assessments have been used to refine clinical diagnoses of parkinsonism using established criteria, and describe some of the phenotypic variability of Parkinson's disease at the time of diagnosis. The crude incidence of Parkinson's disease was 13.6/10[superscript 5yr-1] [confidence interval (CI) 11.8-15.6 and of parkinsonism was 20.9/10[superscript 5yr-1] (CI 18.7-23.3). Age-standardized to the 1991 European population, the incidence figures become 10.8/10[superscript 5yr-1] (CI 9.4-12.4) for Parkinson's disease and 16.6/10[superscript 5yr-1](CI 14.8-18.6) for parkinsonism. Thirty-six per cent of the Parkinson's disease patients had evidence of cognitive impairment based on their performance in the Mini-Mental State Examination, a pattern recognition task, and the Tower of London task. The pattern of cognitive deficits seen among these patients using these and further cognitive tasks suggests that sub-groups of patients based on cognitive ability might be identifiable even in the early stages of disease, which may reflect regional differences in the underlying neuropathological processes.