BACKGROUND: Infant mortality rates have fallen considerably in the past decades. However, because of sparseness of historical data, it is unclear how social differences in infant mortality have changed over time. METHODS: The aim of this study was to assess the association between prenatal social determinants and infant mortality in a Swedish cohort of 13,741 singletons born between 1915 and 1929. Associations and potential mediation of these determinants through other maternal and birth factors were studied using Cox regression. Analyses were repeated for deaths within the neonatal and post-neonatal periods. RESULTS: Mother's single marital status, high parity, low birthweight (LBW), low gestational age and male sex were associated with infant mortality in fully adjusted models, whereas the associations of maternal social class, region of residence and maternal age appeared to be largely mediated through other variables. Contrary to studies reporting a 'LBW paradox', no interactions between gestational age or birthweight and maternal social characteristics on infant mortality were observed in this study. High maternal age, LBW and low/high gestational age were associated with neonatal mortality, whereas single marital status, high parity, LBW, low gestational age and male sex were associated with post-neonatal mortality. An association of family social class with post-neonatal mortality was largely mediated by other maternal variables. CONCLUSIONS: Our findings of determinants of neonatal and post-neonatal mortality in 1915-1929 are strikingly consistent with results from contemporary cohorts. On the other hand, contrary to most recent findings, there is no evidence of a LBW paradox in this historical cohort.