The height, weight and left triceps skinfold thickness of 2190 schoolchildren were measured and the heights and weights of their parents obtained from a postal questionnaire. Power-type obesity indices were used to correct weight for height in parents and children. Preliminary analysis of the distribution of obesity using arbitrarily chosen cut-off points to define the obese, normal and lean parent indicated the presence of a group of 'super-obese' families. The spline technique was used to analyse these data as continuous variables, and showed that the 'super-obese' group of families was too small to be of any practical importance. The relationship between obesity in parents and their children has also been investigated using the spline regression technique. The results show that the relationship is linear but that only 6% of the total variation of observed obesity in children is attributable to variation in the parents' reported obesity. However, a structural analysis indicates a direct relationship between obesity of parents and children. This is obscured in the regression analysis by the large within-subject variability of observed obesity in both parents and children.