those who were testable as adults), there was a significant association between child and adult IQ (r = .45; p < .01). Within the total cohort, adult best estimate IQ was not related to a history of language regression (rho = .04) or ADI diagnostic total (r = 0.26). Age at initial diagnosis (r = .39), childhood language level (rho = .41), epilepsy (rho = .37) and gender (rho = .32) were all significantly (p < .01) associated with IQ at follow-up, but a multiple regression analysis including these variables failed to identify any overall significant predictor.