A sizeable literature shows that intergenerational coresidence remains widespread in much of the developing world. However, this literature mostly relies on cross-sectional data and thereby overlooks key changes affecting longitudinal patterns of household formation and dissolution. Addressing this question, the present study examines premortem household separations – sons forming their own household before their father’s death – in rural India. These separations violate traditional norms and offer a pivotal measure of household change in patrilocal contexts. The 1999 round of the REDS provides a unique opportunity to study these separations by applying survival analysis models on a national sample of rural households. Contrary to results from cross-sectional studies, preliminary findings show a sizeable increase in premortem household separations during the second half of the 20^th century. Further analyses reveal that landlessness and having multiple brothers predict higher rates of household separation, whereas education has no significant association with household separation.