The accuracy of the clinical diagnosis of corticobasal degeneration (CBD) is unknown. To determine its diagnostic accuracy, we presented 105 cases with known neuropathologic diagnoses, including CBD (n = 10), progressive supranuclear palsy (PSP, n = 24), Parkinson's disease (n = 15), diffuse Lewy body disease (n = 14), multiple system atrophy (n = 16), postencephalitic parkinsonism (n = 7), Pick's disease (n = 7), Creutzfeldt-Jakob disease (n = 4), Alzheimer's disease (n = 4), vascular parkinsonism (n = 3), and Whipple's disease (n = 1), as clinical vignettes to six neurologists unaware of the autopsy findings. Reliability was measured with the kappa statistics. The neurologists' clinical diagnoses were compared with clinicopathologic diagnoses for sensitivity, specificity, and positive predictive values at first and last clinic visits. The group reliability for the diagnosis of CBD significantly improved from moderate for the first visit (mean = 34 months after onset) to substantial for the last (68 months after onset). For the first visit, mean sensitivity for CBD was low (35%), but specificity was near-perfect (99.6%). For the last visit, mean sensitivity minimally increased (48.3%), and specificity remained stable. False-negative misdiagnoses mainly occurred with PSP. False-positive diagnoses were rare. The extremely low sensitivity of the clinical diagnosis of CBD suggests that this disorder is markedly underdiagnosed. Although the validity of the clinical diagnosis might have been improved if neurologists could have examined these patients, more important is that this disorder was misdiagnosed by the primary neurologists. In our data set, the best predictors for the diagnosis of CBD included limb dystonia, ideomotor apraxia, myoclonus, and asymmetric akinetic-rigid syndrome with late onset of gait or balance disturbances.