Depression in Parkinson's disease (dPD) is difficult to diagnose because depressive symptoms can overlap with symptoms of Parkinson's disease (PD). Subject-rated scales such as the 30-item Geriatric Depression Scale (GDS) may be useful in screening for dPD. There were 57 patients (33 men, 24 women; mean age, 58.6 years [SD +/- 8.4]) enrolled in a study of pallidotomy for intractable PD who were evaluated for depression before and after surgery. Subjects were evaluated using the 17-item Hamilton Depression Rating Scale (HDRS), Structured Clinical Interview for Diagnostic and Statistical Manual-III (SCID), and the GDS. SCID was used to diagnose major depression with confirmation by an expert geropsychiatrist. Receiver-operating curves (ROC) were used to identify cutoff points with maximal discriminant validity for diagnosing dPD. A total of 213 evaluation time points were included for the 52 patients with time points that included a valid SCID diagnosis, GDS, and HDRS. A ROC established points of maximum specificity/sensitivity for the GDS at a cutoff of 9/10 (sensitivity = 0.809, specificity = 0.837, positive predictive value [PPV] = 0.584, negative predictive value [NPV] = 0.939) and for the HDRS at a cutoff of 12/13 (sensitivity = 0.810, specificity = 0.821, PPV = 0.580, NPV = 0.934). The GDS was moderately correlated with the HDRS (Pearson's r = 0.54; P < 0.001). The GDS is useful in screening for dPD. A cutoff score of 9/10 has acceptable discriminant validity for dPD, and the GDS has a moderate correlation with the HDRS in PD patients.