Social class differences in height have been recognised for many centuries. However, few studies have examined the extent to which these differences are made up of differences in leg length or trunk length. This paper reanalyses cross-sectional information on children examined in Britain in the 1930s. We assess associations between socio-economic status and diet and the components of childhood stature. The analyses were based on the records of 2990 children aged 2 years to 14 years 9 months who were examined in the Carnegie (Boyd Orr) survey of diet and health (1937-39). z-Scores for the measures of childhood stature were calculated using polynomial regression techniques with the study population as the standard. Univariable and multivariable statistical techniques were used to assess the relationships between childhood height, leg length and trunk length, and dietary and socio-economic factors measured at the level of the household. Leg length was the component of stature most strongly associated with measures of childhood diet and socio-economic status. A greater part of the difference in stature between socio-economic groups was caused by differences in leg length rather than trunk length. In multiple regression analyses, district of residence and family food expenditure were generally the two factors most strongly related to stature. In a subsample of the surveyed children, for whom birthweight information was available, trunk length and leg length were equally strongly related to birthweight. Leg length appears to be a particularly sensitive indicator of childhood socio-economic circumstances. Although contemporary studies highlight the importance of biological factors in determining childhood height, the data analysed in this study suggest that socio-economic circumstances were also important in explaining height differentials in prewar Britain.