To evaluate the possibility that parkinsonian signs may be the only presenting feature of multiple system atrophy (MSA), parkinsonian patients were studied who had no atypical clinical signs and had no symptoms of autonomic dysfunction, but who reported that they had not experienced the anticipated good response to dopaminergic treatment. These stringent criteria identified 20 patients from a series of 298 consecutive parkinsonian outpatients. The following clinical pointers were analysed: (a) rate of disease progression; (b) symmetry of parkinsonian symptoms and signs; (c) occurrence of resting tremor during the first three years from onset. In addition, all patients underwent (d) acute and chronic challenge with dopaminergic drugs; (e) cardiovascular reflex autonomic function tests; (f) high field MRI. Rapid progression of disease was seen in 45% of patients, onset was symmetric in 25%, tremor was absent at onset in 70%, response to dopaminergic drug challenges was inadequate in 40%, abnormal cardiovascular reflexes occurred in 50%, and some abnormal MRI finding occurred in 35% of cases. Each of these features was equally weighted by giving to each patient a 0 to 6 point score corresponding to the number of abnormal findings. Fifteen patients scoring higher than 1 were considered at risk for having MSA: five of them were classified as clinically possible (score 2), six as clinically probable (score 3-4), and four patients were classified as clinically definite multiple system atrophy (score 5). The six pointers considered were variably combined in each patient, none of them being universally abnormal in patients with high scores. The patients were followed up for a mean 2.1 (SEM 0.65) years. All but one of the 10 patients prospectively classified as probable or definite MSA developed unequivocal clinical signs of fully symptomatic MSA. A receiver operator characteristic cure was plotted for the prospective score based on follow up diagnosis. The best compromise for trade off between sensitivity and specificity was a cut off value at a score of 3. The sensitivity and specificity of the individual pointers considered to predict fully symptomatic MSA varied considerably, and no single item could predict whether patients presenting with just parkinsonian signs went on during the two year follow up period to develop fully symptomatic MSA. Instead, the number of abnormalities offered a predictive value for the clinical prognosis of these parkinsonian patients.