We prospectively examined the predictive value of the apomorphine test for the therapeutic efficacy of sustained oral levodopa treatment in 62 patients with de novo Parkinson syndrome (no additional neurological deficit) who had not previously been treated with dopaminergic medication. Patients received 2 to 5 mg of apomorphine hydrochloride subcutaneously and a subsequent trial of oral levodopa of at least 3 months' duration. In three patients, response to apomorphine could not be evaluated owing to side effects experienced during the test. In the remaining 59 patients, the best predictor of response to oral levodopa was the apomorphine-induced relative decrease in the scores on the motor examination part of the Unified Parkinson Disease Rating Scale (UPDRS). At a cutoff value of 20% improvement in UPDRS scores, the test predicted the response to levodopa correctly in 50 patients (85%). The sensitivity of the test was 90%, specificity 88%. The positive predictive value was 95%. However, seven of 19 apomorphine test-negative patients experienced a good (n = 4) or partial (n = 3) improvement with levodopa therapy. Thus, the negative predictive value was only 63%. We conclude that response to apomorphine has a high predictive value for response to sustained oral levodopa treatment in most previously untreated patients, but a negative test should not preclude an adequate trial of oral levodopa.