While preterm birth and restricted fetal growth are strongly associated with infant mortality, the extent to which these associations are modified by placental abruption remains unknown. A retrospective cohort study was carried out to examine the risk of infant mortality among twins in relation to abruption, and explore the independent contributions of preterm birth and restricted fetal growth to these associations. The study was restricted to women who had delivered twins at 22 weeks' gestation or more and fetuses weighing 500 grams or more in the United States (1989–2000). Risks of preterm birth (less than 37 weeks' gestation), fetal growth restriction and infant mortality, in relation to placental abruption, were evaluated. All analyses were adjusted for potential sociodemographic confounding factors. The association between restricted twin fetal growth and abruption was the strongest among the most severely growth-restricted babies (i.e., less than 1 centile), with the strength of association diminishing with increasing birthweight centiles. The risk of preterm birth among pregnancies with and without abruption were 80.1% and 51.9%, respectively (relative risk [RR] 1.5, 95% confidence interval [CI] 1.4–1.6). The risk of small-for-gestational-age (SGA; birthweight of less than the 10th centile for gestational age) among abruption and nonabruption births was 11.7% and 9.2%, respectively (RR 1.3, 95% CI 1.2–1.4). Compared with twins of the appropriate growth delivered at term, the relative risks for infant mortality in the presence of abruption were 9.9 (95% CI 5.4–18.2) for term-SGA, 25.0 (95% CI 22.3–28.1) for preterm-non-SGA, and 36.2 (95% CI 28.4–46.1) for preterm-SGA births. The association between infant mortality and abruption among twins appears largely mediated through preterm birth, and to a lesser extent, through SGA. The association between fetal growth and abruption is strongest among the severely growth-restricted babies, suggesting that the origins of placental abruption may develop in early pregnancy.