Reliability estimates for psychological tests are almost always reported for nonclinical populations (e.g., the normative samples). Such practice will no longer be sufficient as the new standards for testing call for an adequate assessment of psychometric properties within the specific population being tested. The purpose of this study was to provide internal consistency reliability estimates for clinical groups on the Wechsler Adult Intelligence Scale–Third Edition. The study included data from 403 clinical participants composed of 10 groups of adults recruited as part of the WAIS–III clinical validity studies. Split-half reliability coefficients were obtained for these groups replicating the procedure used in the WAIS–III. With 8 of the clinical groups, the split-half reliability coefficients were comparable to, or even higher than, those reported for the WAIS–III standardization sample. In general, the split-half coefficients for the Verbal subtests tended to be higher than the coefficients for the Performance subtests. The high magnitude and general pattern of these coefficients demonstrate that the WAIS–III scales do not include additional error variance above and beyond what is reported in the WAIS–III–WMS–III Technical Manual when it was used to assess certain clinical groups. For the ADHD/ADD and learning disabilities groups, however, the internal consistencies coefficients of some subtests were relatively lower, although not statistically significant, than the normative sample. These findings may reflect more heterogeneity within the groups. The implications for assessment and for using alternate methods of determining the psychometric properties in these populations are discussed. (JINS, 2001, 7, 862–866.)