The factorial structure of the WISC-IV for 859 Spanish children diagnosed with ADHD was examined. A bifactor model with the four factors first identified by Wechsler (2003a) was the best fit to the data. The Coding and Symbol Search subtests were particularly poor measures of "g" but relatively strong measures of the Processing Speed factor. In contrast, the Block Design (BD) and Picture Concepts (PC) subtests were relatively strong measures of "g" but weak measures of the Perceptual Reasoning factor. In fact, 80% of the BD variance and 97.8% of the PC variance and was due to the general factor. Additionally, the Wechsler bifactor model was invariant across ADHD-Combined and ADHD-Inattentive groups, permitting a direct comparison of WISC-IV across children diagnosed with these subtypes of ADHD. Only the FSIQ exhibited robust estimates of reliability ([omega] = 0.85 and [omega][subscript h] = 0.70). In contrast, the group factor scores were unreliable measures of their proposed underlying factors ([omega][subscript hs] coefficients ranging from 0.14 to 0.50). It is unlikely that WISC-IV index score profiles can validly contribute to ADHD assessments. Consequently, clinicians must produce psychometric evidence to justify the interpretation of Wechsler score profiles for children with ADHD.