This study assesses the extent to which children under five years of age attending a growth-monitoring program are representative of the population as a whole. Bias in the prevalence of underweight estimated through growth- monitoring is assessed by comparing data from the program with prevalence estimates from the community- based Demographic and Health Surveys of 1988 and 1994. Geographic patterns of attendance at growth- monitoring are also examined through a comparison with census data, and trends in growth-monitoring data are also assessed. Provincial and national estimates of the prevalence of underweight from the two sources were not significantly different in 1988, but significant differences in prevalence estimates were identified in 1994. This suggests that growth-monitoring attendees were less representative of the general population in 1994 than at the start of the study period. The methodology used is transferable elsewhere, since the same data sets exist for many other African countries.