The diagnostic accuracy of the Child Behavior Checklist (CBCL) scales for attention-deficit hyperactivity disorder (ADHD) were examined using receiver-operating characteristic (ROC) analysis. Three logistic regression models were estimated in an initial sample of 121 children with and without ADHD. These models were then tested in a cross-validation sample (N = 122) and among the 111 brothers and 108 sisters of the initial and cross-validation samples. In all four groups, the CBCL Attention Problems scale had the highest discriminating power for ADHD. Adding other scales did not increase the area under the curve of the ROC significantly. These findings suggest that the CBCL Attention Problems scale could serve as a rapid and useful screening instrument not only to help identify cases likely to meet criteria for ADHD in clinical settings but also to identify cases of ADHD among the siblings of children with ADHD.