The acute effects of air pollution on preterm delivery were examined in a prospective cohort in Beijing, China. From early pregnancy until delivery in 1988, we followed all registered pregnant women who lived in four residential areas of Beijing. Information for both mothers and infants was collected. Daily air pollution and meteorological data were obtained independently. The sample for analysis included 25,370 resident women who gave first live births in 1988. Multiple linear regression and logistic regression were used to estimate the effects of air pollution on gestational age and preterm delivery (i.e., < 37 wk), with adjustment for outdoor temperature and humidity, day of the week, season, maternal age, gender of child, and residential area. Very high concentrations of ambient sulfur dioxide (mean = 102 micrograms/m3, maximum = 630 micrograms/m3) and total suspended particulates (mean = 375 micrograms/m3, maximum = 1 0003 micrograms/m3) were observed in these areas. There was a significant dose-dependent association between gestational age and sulfur dioxide and total suspended particulate concentrations. The estimated reduced duration of gestation was .075 wk (12.6h) and .042 wk (7.1 h) for each 100-micrograms/m3 increase in sulfur dioxide and total suspended particulates 7-d lagged moving average, respectively. The adjusted odds ratio for preterm delivery was 1.21 (95% CI = 1.01-1.46) for each in micrograms/m3 increase in sulfur dioxide, and was 1.10 (95% CI = 1.01-1.20) for each 100-micrograms/m3 increase in total suspended particulates. In addition, the gestational age distribution of high-pollution days was more skewed toward the left tail (i.e., very preterm and preterm), compared with low-pollution days. We concluded that high levels of total suspended particulates and sulfur dioxide, or of a more complex pollution mixture associated with these pollutants, appear to contribute to excess risk of preterm delivery in this population. Further work need to be carried out, with more detailed information on personal exposure and effect modifiers.