One of the leading health mysteries in the U. S. is why the infant mortality rate is one of the highest among industrialized countries. Although the relationships of both maternal health and socioeconomic characteristics to infant mortality rates are well documented, little is known about whether the social environment is associated with infant health. Moreover, few ecological studies of infant mortality in the literature have adopted a spatial approach to handle the spatial dependence embedded in the data. This study is designed to fill these gaps by connecting the concept of social capital with infant mortality in the U. S. at the county level. In so doing, it employs a spatial Poisson methodology to manage spatial dependence. The major conclusions are: (1) Though structural social capital index alone is predictive of infant health, this effect disappears when residential stability and neighborhood safety (cognitive social capital) are considered. Both could better explain why infant mortality varies by county. (2) The adverse effect of low birth weight on infant health could be attenuated by social capital and other socioeconomic conditions. (3) Our study illustrates that a spatial approach is necessary, especially for ecological studies; otherwise, spatial dependence would lead to biased estimates and incorrect conclusions.