We exploit the occurrence of postpartum depression (PPD), which has a random component according to the medical community, to estimate causal effects of a salient form of mental illness on couples' relationship status. We estimate single-equation models as well as bivariate probit models that address the potential endogeneity of PPD. We find that this relatively prevalent mental illness reduces the probability the couples are married (by 22–24 %) as well the probability that they are living together (married or cohabiting) (by 24–26 %) 3 years after the birth of the child. Models stratified by relationship status at the time of the birth indicate that PPD makes it more likely that unions dissolve (particularly among baseline cohabitors) and less likely that unions are formed (particularly among baseline non-cohabitors). The findings contribute to the literature on the effects of mental illness on relationships, to the broader literature on socioeconomic status and health, and to the more specific scientific and policy literatures on the effects of PPD on families.