Faculty Proceedings, Presentations, Speeches and Lectures


Estimation of WAIS-III Premorbid IQ in Neurologic and Psychiatric Patients

Event Location / Date(s)

Chicago, IL / February 14 - 17, 2001

Document Type

Conference Proceeding

Presentation Date


Conference Name / Publication Title

29th Annual Meeting of the International Neuropsychological Society



Premorbid intellectual level is typically estimated in clinical settings, but the accuracy of these IQ estimates has not been evaluated using the WAIS–3. This study examined the accuracy of the Barona et al. (1984) demographic equation, the WRAT–3 reading subtest, and the WAIS–3 Vocabulary subtest as estimates of WAIS–3 FSIQ in 68 neurologically normal psychiatric patients and 101 patients with central nervous system disorders. Psychiatric patients with DSM–4 diagnoses of mood disorders (N=40), anxiety disorders (N =22), and somatization disorder (N=6) obtained a mean FSIQ of 99 (SD=13.4). IQ estimates based on the Barona formula (M=103, SD=8.3), WRAT–3 (M=99, SD=11.7), and Vocabulary subtest (M5101, SD514.5) correlated significantly with obtained WAIS–3 FSIQ (rs5.59, 71, and 78, respectively). Neurologic patients with diagnosis of dementia (N=33), head trauma (N=28), CVA (N=19) and other documented disorders obtained a mean FSIQ of 89 (SD514.5), significantly lower than estimates based on the Barona formula (M=101, SD=9.5), WRAT–3 (M=95, SD=15.1), and Vocabulary subtest (M=94, SD=16.4). Specificity (percentage of psychiatric patients accurately classified) and sensitivity (percentage of neurologic patients accurately identified) were calculated using 10, 12, and 15 point declines from estimated premorbid IQ as evidence of CNS dysfunction. Estimates using the Barona formula significantly discriminated groups, with specificities of 72%, 75%, and 88% for each cut-off score. Respective sensitivities were 57%, 49%, and 43%. The WRAT–3 also discriminated groups significantly, with respective specificities of 79%, 87%, and 97% for each cut-off score. Corresponding sensitivity values were 43%, 33%, and 23%. Estimates using the Vocabulary subtest did not significantly discriminate groups.

This document is currently not available here.