Objective: The objective of this study was to examine the intrarater reliability of "DSM-IV" ADHD symptoms. Method: Two-hundred-two children referred for attention problems and 49 comparison children (all 7-12 years) were rated by parents and teachers on the identical "DSM-IV" items presented in two different formats, the SNAP-IV and Conners' Revised Questionnaires, at two closely spaced points in time. Results: For the combined sample, weighted kappa scores for intrarater agreement ranged from 0.30 ("fair") to 0.77 ("good") across symptoms. Kappa scores were good with respect to agreement on the "DSM-IV" criterion of endorsement of at least six symptoms in a given cluster for Inattention (0.60 and 0.76, for parents and teachers, respectively) and Hyperactivity-Impulsivity (0.72 and 0.75, respectively). Kappas for identification of cases as AD/HD or not AD/HD were good to excellent (0.67 and 0.79 for parents and teachers, respectively). Classification as AD/HD or not AD/HD changed from the first to the second rating in 12% and 10% of cases rated parents and teachers, respectively. Conclusion: Reliability of individual ADHD symptoms appears to be suboptimal for clinical and research use and is improved, although less than ideal, at the levels of cluster endorsement and case classification. (Contains 7 tables.)