To examine the association between individual lifetime measures of mean exposure to air pollution and postneonatal respiratory deaths, we have conducted a matched population-based case-control study covering all births registered in the Czech Republic from 1989 to 1991 that were linked to death records. For each case of infant death, we have randomly selected 20 controls from infants of the same sex born on the same day and alive when the case died. Exposure was assigned as the arithmetic mean of all 24-hour air pollution measurements in the district of residence of each case and control for the period between the birth and death of the index case. We used conditional logistic regression to estimate the effects of suspended particles, sulfur dioxide, and nitrogen oxides on risk of death in the neonatal and postneonatal period, controlling for maternal socioeconomic status and birth weight, birth length, and gestational age. There were 2,494 infant deaths with exposure data on at least one pollutant, 133 of them from respiratory causes. The effects of all pollutants were strongest in the postneonatal period and were specific for respiratory causes. For these, rate ratios for a 50 microg/m3 increase in particles, sulfur dioxide, and nitrogen oxides were 1.95 [95% confidence interval (CI) = 1.09-3.50], 1.74 (95% CI = 1.01-2.98), and 1.66 (95% CI = 0.98-2.81), respectively, after controlling for all covariates. Only particles showed a consistent association when all pollutants were entered in one model. We found no evidence of a relation between any pollutant and mortality from other causes. These results indicate that the effects of air pollution on infant mortality are specific for respiratory causes in the postneonatal period, are independent of socioeconomic factors, and are not mediated by birth weight or gestational age.