The predictive validity of developmental testing was investigated in term, preterm and retarded children. Verbal, performance and locomotor development were assessed at various ages and individual development curves constructed. Inter-age correlations between development tests at nine to 24 months and intellectual assessment at seven years revealed a strong effect on prediction for age at testing and level of mental performance. Categorisation by level of mental performance demonstrated that at seven years 98.6 per cent of the children with developmental quotients (DQs) greater than 85 at 24 months achieved IQs greater than 85, while 98.7 per cent of the seven- to eight-year-old children with DQs greater than 85 at nine to 24 months achieved IQs greater than 85. The Griffiths language and performance scores and their combination were the strongest predictors of later intellectual functioning. The social score was of moderate significance, while the locomotor score had no predictive value. Analysis of individual development curves revealed that prediction was hampered in some children by factors not detectable by statistical analysis, such as dissociations in development, organic impairment and major life events. Among the prenatal, perinatal and postnatal variables studied, only socio-economic status was of predictive significance; its effect depended on the level of mental performance and was most marked in term children, but it had no effect on retarded children.