This study examines the effects of biological, demographic and social variables, as reported on birth certificates, on neonatal and postneonatal mortalities. Examination of both birth and infant death certificates was performed for infants born in 1989 to residents of Tohoku, Tokai and Kyushu regions (n = 409,679), i.e. about a third of the total number of infants born in Japan. Record linkage rates in the neonatal and postneonatal periods were 99% and 92% respectively. The neonatal and postneonatal mortality rate of the cohort including unlinked deaths were 2.65/1,000 live-born infants and 2.02/1,000 neonatal survivors respectively. Both neonatal mortality rate and postneonatal mortality rate increased exponentially as birth weight decreased. Infant's sex and gestational period exerted significant influences on neonatal mortality after controlling for birth-weight by logistic regression. However, no overall effect on neonatal mortality was seen for maternal age, live-birth order, previous stillbirths and marital status, when birth-weight was taken into account. This suggests that the effect of social-demographic disadvantage on neonatal mortality is predominantly mediated through a downward shift in the birth-weight distribution. On the other hand, postneonatal mortality was influenced substantially by socio-demographic factors. Maternal age of less than 20 years and more than 35 years, one or more previous live-birth and unmarried marital status were all significantly related to increased risk of postneonatal death after controlling for birth-weight. Occupational type classification of the household, also, had a significant effect on postneonatal mortality, with agriculture, self-employed, or blue collar employee being related to increased risk.