This study investigates the possible effects of pre-term births and low birth weight on infant mortality rates (IMRs) over a 15-year period in Ribeirão Preto, Brazil, based on surveys carried out in 1978/79 and 1994. The 1978/79 survey included 6750 births over a 12-month period and the 1994 survey 2846 births over a 4-month period. Infant deaths were retrieved monthly from the city register. Infant mortality rate decreased from 36·6 to 16·9 deaths per 1000 over 15 years. The decrease in IMR was larger in the 2500–2999 g group than in any other group. The observed falls in IMR were attributable to decreases in birth-weight-specific mortality rates. Likewise, there was a general decrease in IMR in mild, moderate and severe pre-term births. The incidence rate ratio of infant mortality between surveys was 0·46 (95% CI 0·34–0·63); it increased to 0·57 (95% CI 0·35–0·75) when adjusted for birth weight and other factors in the model and rose to 0·69 (95% CI 0·49–0·97) when adjusted for length of gestation and other variables. The increase in pre-term births and low birth weight may have had, at most, a marginal effect on the IMR. Progress in the care of newborns may have decreased the mortality risk, but even mild pre-term birth still has an impact on infant mortality. There is room for further improvement in IMR by tackling the high rates of pre-term birth.