149 previously untreated patients with Parkinson's disease were recruited over a three year period and randomly allocated to either low dose levodopa-carbidopa (< or = 600/150 mg/day) or low dose bromocriptine (< or = 30 mg/day). A five year follow up is reported on the 126 patients who completed the dose titration and who have not developed features of atypical Parkinsonism. Levodopa-carbidopa in low dosage adequately controlled symptoms in most patients and delayed the appearance of dyskinesia and end of dose failure for about two years longer than conventional doses. Only a few patients could be managed for more than one year on low dose bromocriptine alone; these patients had mild disease and asymmetric signs. Patients randomised to bromocriptine did not develop dyskinesia or troublesome end of dose failure until levodopa-carbidopa was added. The prevalence of dyskinesia in this group was lower than in patients given levodopa-carbidopa alone. The prevalence of end of dose failure was similar in the two randomisation groups once levodopa was introduced.