The purpose of this study is to examine whether the co-residence of spouses and children affects self-reported general health among older men and women in a rural area of Bangladesh. Binary logistic regression has been used to explore the impact of spouses and children on self-reported health, with particular attention to the gender of children and interactions with chronic disease. The data are from the Matlab Health and Socio-Economic Survey. A sample of 765 women and 979 men aged 60 or more years with at least one surviving child was available. The principal result is that for an older woman, optimum self-reported health is most likely when a spouse and at least one son and one daughter are present. Any deviation from this family pattern (either no spouse or children of only one sex) leads to a significantly increased risk of poor self-reported health. On the other hand, among older men there were no differences in self-reported health among the various spouse-child combinations. The relationship between a balanced gender distribution of children and optimum self-reported health among older women may explain the levelling out of fertility at roughly three children per women despite intensive family planning promotion in the area. Further reductions in fertility (an important policy concern) may depend on improving the substitutability of sons and daughters in the support of their elderly mothers.