We studied infant mortality rates in Canada within specific gestational age and birthweight categories after using probabilistic techniques to link information in Statistics Canada's live births data base (1985-94) with that in the death data base (1985-95). Gestational age- and birthweight-specific mortality rates in 1992-94 were contrasted with those in 1985-87 with changes expressed in terms of relative risks with 95% confidence intervals [CI]. Statistically significant reductions in infant mortality were observed beginning at 24-25 weeks of gestation and extended across the gestational age range to post-term births. Crude infant mortality rates, infant mortality rates among those > or = 500 g and among those > or = 1000 g decreased by 22%, 25% and 26%, respectively, from 1985-87 to 1992-94. The magnitude of the reductions in infant mortality rates ranged from 14% [95% CI 7, 21%] at 24-25 weeks of gestation to 40% [95% CI 31, 47%] at 28-31 weeks. Almost all reductions in gestational age- and birthweight-specific infant mortality between 1985-87 and 1992-94 were due to approximately equal reductions in neonatal and post-neonatal mortality. Live births > or = 42 weeks of gestation did not follow this rule; post-neonatal mortality rates among such live births decreased significantly by 51% [95% CI 26, 68%], although neonatal mortality rates showed no significant change. The mortality reductions observed across the gestational age and birthweight range are probably a consequence of specific clinical interventions complementing improvements in fetal growth. Temporal changes in the outcome of post-term pregnancies need to be carefully examined, especially in relation to recent changes in the obstetric management of such pregnancies.