This study aimed to evaluate the prevalence of birth-weight discordance among twins, to determine the risk cut-off point for birthweight discordance, and to clarify whether perinatal deaths are aggravated by birthweight discordance or by low birthweight itself. A population-based analysis of all twins born between 1995 and 1999 in Japan was conducted using data collected from national birth, death and stillbirth certificates. Birthweight discordance was determined as: higher birthweight minus lower birthweight divided by higher birthweight multiplied by 100. Among twins with a birthweight discordance of more than 15%, the fetal and perinatal mortality rate was higher than that of twins with a discordance of less than 15%. Ninety per cent of the relative cumulative frequency of twin live-births were within 25% of the birthweight discordance. Logistic regression analysis showed that the stillbirth of at least one of the twins is associated with the birthweight of the larger twin and birthweight discordance. It also showed that the stillbirth of both twins in the twin pair is associated with the birth-weight of larger twins and the gestational weeks. It was clarified that birthweight discordance was associated with the stillbirth of only one twin in the twin pair, and the stillbirth of both twins in the twin pair was associated with low birthweight itself.