A self-selected sample of 417 urban and 379 rural preschool children attending the under-5s clinics was weighed in Sierra Leone. The individual weights were related to the weight-for-age Harvard 3rd centiles and 80% levels of the Harvard medians. In this analysis there were considerable discrepancies between the sexes. Further investigation showed that the distribution of the weight-for-age Harvard centiles for girls is much wider than the distribution of the Harvard centiles for boys. These distributions appear to be unusual when compared with those from London and Hong Kong studies. Accordingly, use of the weight-for-age Harvard centiles may lead to inaccuracy in the assessment of the nutritional status of preschool children. The "percentage method" as suggested by Jelliffe (1966) and Gomez et al. (1956) does not take account of the normal range of distribution of the standards. This study has shown that the 80% levels of sex-specific median weight-for-age occupy positions as much as 600 g below the sex-specific 3rd centiles for Hong Kong Chinese children, while the 80% levels closely follow the 3rd centiles for London children in the preschool age group. Thus the Hong Kong 80% levels are not comparable with the London 80% levels. Therefore, the "percentage method" appears to be misleading in the assessment of nutritional status using weight measurements.