Using data from 123 children (aged 6-12 years) referred consecutively to a pediatric neuropsychiatry clinic by community physicians for assessment of Attention-Deficit/Hyperactivity Disorder (ADHD) and related problems, we investigated the effects of informant (parent, teacher), tool (interview, rating scale), and method for combining symptom reports ("and," "or" algorithms), on the diagnosis of ADHD and its subtypes. Results indicated that as many as 50% of cases were reclassified from one subtype to another, depending on whether information was derived from one or two informants, a semistructured clinical interview and/or rating scale, and the algorithm used to combine informant reports. We conclude that the diagnosis of "DSM-IV" ADHD subtypes is capricious in that it is influenced by clinicians' decisions regarding informants, instrumentation, and method for aggregating information across informants and instruments. (Contains 6 tables.)