The Parkinson Psychosis Rating Scale (PPRS) was designed to assess the severity of specific symptoms of levodopa-induced psychosis in patients with Parkinson's disease. The scale is administered by the clinician and consists of six items, each rated from 1 (no symptoms) to 4 (extreme symptoms); total score ranges from 6-24 points. To test its reliability, validity, and stability, the authors administered the PPRS to 29 patients with parkinsonian psychosis before and after a 6-week course of treatment with the 5HT3 antagonist ondansetron. Each patient and their caregivers were interviewed by three raters. Results were compared with two other available psychotic symptom scales and one cognitive scale given to the same patients. It was found that the interrater reliability for the total PPRS scores and for each of the individual items was high (Spearman's rho = 0.80-0.99). A significant reduction in PPRS scores was obtained after ondansetron administration, concomitant with a decrease in the Global Functional Impairment score, which was assessed by the family. The changes in PPRS scores occurred irrespective of cognitive dysfunction, as measured by the Minimental State Examination. It appears that the PPRS is a relevant, reliable, valid, and easily-administered instrument for measuring the severity of parkinsonian psychosis. Further studies are needed to evaluate its utility in other organic psychoses and in the assessment of additional pharmacologic treatments for levodopa-induced psychosis.