This study examined the interrater reliability of neuropsychological diagnoses produced by clinical neuropsychologists across 4 medical centers. These diagnoses were based on evaluations using a comprehensive battery of commonly used neuropsychological test instruments, interview, history and medical chart review. The diagnoses of individual neuropsychologists were compared to those made by members of an external review panel for each patient evaluated. Patients were first diagnosed as showing cognitive impairment versus no cognitive impairment. If a patient was diagnosed as impaired, a specific neuropsychological diagnosis was assigned. The diagnostic classification for cognitive impairment was moderately reliable [κ = .48 ± s.e.(κ) = .062]. The interrater reliability for specific diagnoses was in the fair to good range [κ = .44 ± s.e.(κ) = .029]. These levels of reliability are comparable to those found for other psychiatric and neurologic specialties and for medical diagnoses made by other health care disciplines. (JINS, 2002, 8, 555–565.)