Incidence studies of Parkinson's disease (PD) are important for both health-care planning and epidemiological research. This report reviews the methods and results of previous incidence studies of PD and makes recommendations for future studies. Original articles that described the incidence of PD were located using several strategies. The methods were summarised, and the results of studies with similar methodologies were compared on a standardised population. Twenty-five incidence studies were included. Each used different methods to identify incident patients, although most screened both primary care and hospital records. Only eight studies were prospective, and only two of these had any follow-up. The diagnostic criteria for PD varied (11 studies used two or more cardinal motor features, four used the UK Brain Bank criteria), as did the exclusion criteria and the definition of an incident case. In 16 studies, attempts were made to confirm the diagnosis by examination of patients by a specialist as part of the study. None of the studies used identical methods, but five were sufficiently similar to merit comparison. Four of these gave a similar incidence (16-19/100000/year), but one from Italy had a much lower incidence (8.4/100000), the reason for which was unclear. Five studies found significantly greater incidence in men. This review highlights the difficulties in performing good quality incidence studies of PD. Further incidence studies using standardised methods are required. A set of minimal scientific criteria has been devised to improve the quality and consistency of future studies.