The weight and height data of 9261 children from birth onwards were gathered on their admission to primary school in 1989 in the area of Lyon, and compared to French references. In comparison with the latter data, the weight and height for age were found to be higher, and there was an excess of children with a high weight for height. The weight for height was higher than the mean of the reference distribution plus two standard deviations in 6.2% of the children. Birth hypertrophy increased the risk of obesity at age six. A new criterion of neonatal hypertrophy, taking various constitutional characteristics into account, was found to be associated to a twofold risk of subsequent obesity, and was shown to be more appropriate in predicting subsequent obesity than classical definitions based on the weight or on the weight for gestational age or sex.