A growing literature has documented group differences between boys with and without attention deficit-hyperactivity disorder (ADHD) on neuropsychological tests; however, whether or not such tests can discriminate individuals with ADHD from non-ADHD controls remains unclear. This study used conditional probability and receiver operating characteristic analyses to examine the efficiency of test-based diagnostic discriminations in a large sample of referred boys with and without ADHD. Single neuropsychological tests had limited discriminating ability at various cutoff scores. When multiple tests were used together, prediction of ADHD status improved but overall diagnostic efficiency remained limited. Diagnostic efficiency did not differ when medicated and nonmedicated index children were considered separately. Results suggest that children with ADHD show variable deficits on neuropsychological tests of attention and executive functions. Impairments on multiple neuropsychological tests are predictive of ADHD, but normal scores do not rule out the diagnosis. The prognostic implications of variable neuropsychological deficits in children with ADHD require further investigation.