OBJECTIVE The duration of exclusive breast-feeding in Bangladesh is low. Though several studies have been carried out on breast-feeding in Bangladesh, the factors influencing the duration of breast-feeding are not studied. The purpose of this study was to assess the duration of breast-feeding among children in Bangladesh and to study socio-economic and demographic factors affecting the duration. METHODS The study included 5068 mother-child pairs, selected on the basis of Bangladesh Demographic Health Survey (BDHS) of 1999-2000. The risk of giving supplementary food at different ages was estimated by life table survival method. Independent effects of total duration of breast-feeding of each of the study variables were estimated by using Cox's regression model. RESULTS The median duration of full breast-feeding was 3.67 months whereas mean and median durations of total breast-feeding was 31.3 and 30 months respectively. Life table analysis showed that 69.9 per cent women gave supplementary food to their babies before reaching six months of age. Cox's regression analysis revealed that women who had lived in rural areas were less likely to terminate breast-feeding than those living in urban areas. Women who had completed at least secondary education were more likely to stop breast-feeding than less or uneducated mothers. Children born in high economic status families had higher risk of stopping breast-feeding compared to those in low economic status families. Further, the deliveries assisted by the relatives had lower risk of terminating breastfeeding than by the health professionals. CONCLUSIONS According to the study results, women with higher education, high economic level, lower birth interval and delivery assisted by health personnel had lower duration of breastfeeding. Future breast-feeding programme in Bangladesh should give special attention to these women since they breast-feed relatively shorter periods of time.