Pregnancy intervals were calculated for 54,369 later-born singletons delivered during 1969 in the State of North Carolina. Subsequent infant deaths among this cohort were categorized into probable sudden infant death syndrome (SIDSp), other postneonatal home deaths, all postneonatal hospital deaths, all neonatal deaths, and all deaths due to congenital malformations. Each death was matched with a control drawn from the population of infants surviving the first year of life. The matching variables were maternal age. race, mother's education and number of previous live births. All categories of death showed higher proportions of short intervals among cases than controls. Comparing the interval distributions by means of the χ2 test for linearity produced significant results for only SIDSp and neonatal deaths. It was concluded, however, that short-pregnancy interval probably has a similar effect on the risk of death from all five cause of death groups. The introduction of birthweight as a fifth matching variable left no significant differences in the distributions of pregnancy intervals between cases and controls. This was interpreted as evidence that short pregnancy interval exerts its influence on risk of death in infancy through its effect on birthweight. The estimated reduction in prematurity and infant mortality that might result from the elimination of short pregnancy intervals was 5% and 6%, respectively.