An ecological study of infant mortality and air pollution was conducted in the Czech Republic. Routinely collected data on infant mortality and air pollution in the period 1986-88 were analysed for the 46 of the 85 districts in the republic for which both were available. The independent effects of total suspended particulates (TSP-10), sulphur dioxide (SO2), and oxides of nitrogen (NOx) adjusted for district socioeconomic characteristics, such as income, car ownership, and abortion rate, were estimated by logistic regression. We found weak positive associations between neonatal mortality and quintile of TSP-10 and SO2. Stronger adjusted effects were seen for postneonatal mortality, with a consistent increase in risk from the lowest to the highest TSP-10 quintile (p < 0.001). Weaker and less consistent evidence of a positive association with NOx (p = 0.061) was observed. The strongest effects were seen for postneonatal respiratory mortality, which increased consistently from lowest to highest TSP-10 quintile (p = 0.013). There was also a suggestion of a positive association with SO2 (p = 0.062). The highest to lowest quintile risk ratios for postneonatal respiratory mortality were 2.41 (95% Cl 1.10-5.28) for TSP-10, 3.91 (0.90-16.9) for SO2, and 1.20 (0.37-3.91) NOx. The specificity of the association between air pollution quintile (especially TSP-10) and postneonatal respiratory mortality is consistent with the known effects of air pollution on respiratory disease morbidity in children. These ecological associations require confirmation in an individually based study.