The clinical utility of the DSM-IV eating disorder (ED) diagnostic criteria among practicing clinicians has not been formally evaluated, despite the considerable diagnostic challenges these disorders present. This study evaluated inter-rater reliability between research and clinical diagnoses, identified discrepantly rated diagnostic criteria, and evaluated ED subtype use in a naturalistic treatment setting. Seventy-six adolescent and young adult female patients consecutively admitted to a residential ED program were evaluated independently by clinicians (unstructured clinical interview) and research assessors (Structured Clinical Interview for DSM-IV). Clinicians and research assessors conferred concordant ED diagnoses in 80.3% of cases ( κ = 0.70), thus highlighting the clinical utility of the extant DSM-IV diagnostic scheme in this specialty ED treatment setting. All but two discordant cases included a diagnosis of ED not otherwise specified (EDNOS). Clinicians applied ED subtypes in just 20.4% of eligible cases, and were significantly more likely to apply subtypes to major depressive disorder. Although clinical and research interviews yielded substantial reliability, EDNOS had the lowest reliability among the ED diagnoses. Moreover, infrequent subtype application on this specialty unit raises questions about the clinical utility of DSM-IV anorexia nervosa and bulimia nervosa subtypes, even in the context of clinically useful overarching categories.