We used a rotation-sensitive movement monitor (RoMM) to quantify and characterize dyskinesia in Parkinson's disease (PD). Both upper limbs of 22 patients with dyskinetic PD were recorded and videotaped simultaneously. Three neurologists reviewed the video segments and rated severity of dyskinesia on a four-point scale; they also assessed any asymmetry of dyskinesia between the right and left side as well as the dyskinesia type (choreic, dystonic, or mixed). Mean and median clinical ratings for severity, asymmetry, and type of dyskinesia were compared with (1) the total power of the frequency power spectrum (FPS, degrees/second), (2) the percent difference of FPS values between the right and left side, and (3) the frequency (Hz) of the predominant peak, respectively. Intra- and interrater reliability was determined and a test-retest analysis was performed. FPS values showed a statistically significant correlation with the clinical ratings for dyskinesia severity. FPS difference between both sides was more sensitive than raters in detecting dyskinesia asymmetry. A predominant frequency peak of dyskinesia was obtained in all cases and ranged from 0.25-3.25 Hz. There was a significant trend for high-frequency dyskinesia to correlate with choreic type and for low-frequency dyskinesia to correlate with dystonic type. Test-retest analysis indicated a high reliability. We conclude that the RoMM is a valid, reliable, and sensitive method to quantify and characterize dyskinesia. Examples are provided suggesting that this instrument may prove useful for long-term assessment of dyskinetic patients and as a standardized tool for assessing dyskinesia in pharmaceutical or surgical trials for PD.