Although household partition is generally thought to be an event with significant behavioural implications in rural South Asia, little is known about the underlying determinants of household partition, or the extent to which partitioned households tend to operate as a single economic and social unit. In this paper the author uses longitudinal data collected in rural Bangladesh to provide a number of new insights into the process of household partition. There are three main parts to the paper. The first consists of a descriptive analysis of household structure which indicates that partition is an important determinant of household structure in this population, particularly for young couples in the early stages of family formation. Secondly, a procedure is developed for the analysis of household partition, which makes use of data on relationship to head of household. A multivariate analysis of household partition is then used to evaluate a number of new and existing hypotheses about the relationship between economic and demographic characteristics and the probability of household partition. One result that is of particular interest is that presence of daughters of a non-head increased the probability of partition significantly, but that of sons does not. Thirdly, data on the educational attainment of children are used to provide an indirect measure of the extent to which recently partitioned households continued to operate as a single economic and social unit. Although partitioned households remained in close proximity, they exhibited significant independence with regard to decisions about the educational attainment of children, something that is not apparent in jointly-resident sub-households. www.jstor.org/stable/2175228