The impact of birth weight on cognitive performance in late adolescence was assessed in a study of 20,567 male infants born in Jerusalem between 1964-1970. The 17-year follow-up was performed by matching computerized data base records. Stepwise multiple regression analysis was used to estimate the effect of birth weight on intelligence test scores, adjusting for the influence of ethnic origin, maternal and paternal education, social class (determined by area of residence), maternal age, and birth order. These confounders explained 22% of the variance in intelligence test scores. The adjusted differences in intelligence test scores were significantly lower for groups with birth weight less than 2000 g (-6.5 points), 2000-2499 g (-3.6 points), and 2500-2999 g (-1.6 points) compared with the group weighing 3000-3499 g. Most low birth weight infants achieved intellectual performance within the normal range. However, despite this reassuring finding, low weight at birth in male subjects was found to have a statistically significant independent association with inferior intelligence test achievements in late adolescence.