Objectives-To assess whether early treatment of congenital hypothyroidism fully prevents intellectual impairment. Design-A national register of children with congenital hypothyroidism who were compared with unaffected children from the same school classes and matched for age, sex, social class, and first language. Setting-First three years (1982-4) of a neonatal screening programme in England, Wales, and Northern Ireland. Subjects-361 children with congenital hypothyroidism given early treatment and 315 control children. Main outcome measures-Intelligence quotient (IQ) measured at school entry at 5 years of age with the Wechsler preschool and primary scale of intelligence. Results-There was a discontinuous relation between IQ and plasma thyroxine concentration at diagnosis, with a threshold at 42.8 nmol/l (95% confidence interval 35.2 to 47.1 nmol/l). Hypothyroid children with thyroxine values below 42.8 nmol/l had a mean IQ 10.3 points (6.9 to 13.7 points) lower than those with higher values and than controls. None of the measures of quality of treatment (age at start of treatment (range 1-173 days), average thyoxine dose (12-76 mu g in the first year), average thyroxine concentration during treatment (79-234 nmol/l in the first year), and thyroxine concentration 103 nmol/l less than at least once during the first year) influenced IQ at age 5. Conclusions-Despite early treatment in congenital hypothyroidism the disease severity has a threshold effect on brain development, probably determined prenatally. The 55% of infants with more severe disease continue to show clinically significant intellectual impairment; infants with milder disease show no such impairment. The findings predict that