Clinicopathological series indicate that the clinical diagnosis of Parkinson's disease is correct in only 80% of cases. Multiple system atrophy (MSA) and Steele-Richardson-Olszewski syndrome (SRO) comprise most of the misdiagnoses. By means of 18F-dopa PET the pattern of nigrostriatal dopaminergic dysfunction in 28 patients with clinically probable Parkinson's disease, 25 with MSA, and 10 patients with SRO, was assessed and compared with the pattern in 27 normal subjects. Discriminant function analysis was used to assess the ability of 18F-dopa PET to categorize individual parkinsonian patients on the basis of their caudate and putamen tracer uptake. Discriminant function analysis assigned all control subjects a normal category. One Parkinsonian patient out of 63 was classified as "normal" on the basis of PET findings, although this patient had significantly reduced putamen 18F-dopa uptake. Discriminant function analysis was less effective at distinguishing different categories of akinetic-rigid syndrome on the basis of their striatal 18F-dopa uptake, as judged against clinical criteria. Patients clinically labelled as having typical or atypical Parkinsonian syndromes were assigned the same category on PET criteria 64% and 69% of the time, respectively. When all three categories of Parkinson's disease, MSA, and SRO were considered together, clinical and 18F-dopa PET findings correlated in 64% of patients assigned a diagnosis of Parkinson's disease and 70% of those given a diagnosis of SRO; MSA was less readily discriminated, patients with MSA being assigned to MSA, Parkinson's disease, and SRO groups with equal frequency. The correlation between clinical and discriminant function analysis assignment improved when separate comparisons were made between Parkinson's disease and MSA, or Parkinson's disease and SRO groups. In these analyses, clinical and PET categorisation of MSA and Parkinson's disease agreed in 60% of cases, and of SRO and Parkinson's disease in 90% of cases. In summary, (18)F-dopa PET successfully discriminates normal subjects from parkinsonian patients, and patients with Parkinson's disease from patients with SRO, but is less reliable in distinguishing Parkinson's disease from MSA. The concomitant assessment of striatal neuronal function with additional PET tracers may be necessary to reliably differentiate typical and atypical parkinsonian syndromes.