Abstract
Estimation of premorbid or expected IQ in children from IQs of family members was evaluated for precision, sensitivity, and diagnostic efficiency. Prediction of a child's IQ from familial ability has limited precision, and the sizable discrepancies from familial IQ estimates required for statistical significance reduce their sensitivity. In a Monte Carlo simulation, familial IQ discrepancies had stronger positive than negative predictive power for distinguishing severely head-injured from normal children, with acceptable predictive power being obtained over a limited range of circumstances. Thus, while a significant discrepancy may be meaningful, the absence of one is often inconclusive. Familial IQ discrepancies should be used very cautiously, if at all, and are not substantially more accurate than discrepancies from demographic IQ estimates.