This paper investigates the high correlation in infant mortality across siblings using micro-data for each of the fifteen major states of India. The main finding is that, in thirteen of the fifteen states, there is evidence of a causal effect of a child death on the risk of death of the subsequent child in the same family (a scarring effect), which is identified after controlling for observed and unobserved heterogeneity at the family level. The two states in which evidence of scarring is weak are Punjab, the richest, and Kerala, the state that is most advanced in socio-economic terms. In the other states, scarring effects are large. Indeed, the only other covariate that has a marginal effect on mortality that is as big, or bigger, than the survival status of a preceding sibling is an indicator for mothers having attained secondary or higher levels of education. These results show that policies targeted at reducing infant mortality will have social multiplier effects through helping avoid the death of subsequent siblings. The size of the scarring effect depends upon the gender of the previous child in three states, in a direction consistent with son-preference. Comparison of other covariate effects across the states offers some new insights, there being no previous research that has compared the determinants of infant mortality across the Indian states.