A population-based computer record linkage of infant births and deaths was conducted for 1978 and 1979 covering Canadian provinces. Birthweight was inversely related to risk of postneonatal death for all causes examined, including accidental deaths. Length of gestation was inversely associated with risk, but the strength of the relationship was much weaker than that noted for birthweight. A logistic regression model was used to assess the effects of variables, as reported on birth certificates, on postneonatal mortality. Maternal age less than 25 years, unmarried marital status and one or more previous births were all statistically significantly related to increased risk.