The recent publication of the Wechsler Adult Intelligence Scale (WAIS-III), the most widely used standard test of intelligence, requires the development of a new short form for use with patients with schizophrenia for many clinical and research purposes. We used regression analyses of complete WAIS-III data on 41 outpatients with schizophrenia and 41 education-, and age-matched healthy subjects to determine the best combination of subtests to use as a short form. Excluding three subtests that are time-consuming to administer, and requiring that the solution includes one subtest from each of the four WAIS index scores, the combination that most fully accounted for the variance in full-scale IQ (FSIQ) for both participants with schizophrenia (R(2)=0.90) and healthy controls (R(2)=0.86) included the information, block design, arithmetic, and digit symbol subtests. When the restrictions regarding which subtests could enter were relaxed, the best four-subtest solution included information, block design, comprehension, and similarities. Although the latter explained 95% of the variance in FSIQ for schizophrenia participants and 90% of the variance for healthy controls, it consistently overestimated FSIQ for the schizophrenia group. We recommend the four-factor short form for use in future research and clinical practice in which a quick, accurate IQ estimate is desired.