We examine the impact of air pollution on infant death in California over the 1990s. Our work offers several innovations: First, many previous studies examine populations subject to far greater levels of pollution. In contrast, the experience of California in the 1990s is clearly relevant to current debates over the regulation of pollution. Second, many studies examine a few routinely monitored pollutants in isolation, generally because of data limitations. We examine four "criteria" pollutants in a common framework. Third, we develop an identification strategy based on within zip code variation in pollution levels that controls for potentially important unobserved characteristics of high pollution areas. Fourth, we use rich individual-level data to investigate effects of pollution on infant mortality, fetal deaths, low birth weight and prematurity in a common framework. We find that the reductions in carbon monoxide (CO) and particulates (PM10) over the 1990s in California saved over 1,000 infant lives. However, we find little consistent evidence of pollution effects on fetal deaths, low birth weight or short gestation.