Abstract: Background: Healthy breastfeeding practice in the United States depends decisively on high rates of initiation at the delivery hospital. We sought to estimate the component of hospital variation in rates of exclusive breastfeeding at discharge that was dependent on demographic composition. Isolating that component can help to illuminate the potential independent contribution of hospital policies, practices, and staff behaviors. Methods: Electronic birth certificate data in New Jersey from 1996 to 2001 (n = 545,837) were used to measure variations in hospital-level rates of breastfeeding initiation. The method of infant feeding within 24 hours before hospital discharge was reported as exclusive breastfeeding, formula feeding, combination feedings, other methods, and unknown. Rates of exclusive breastfeeding by hospital were standardized to remove sociodemographic differences in hospitals’ service populations that influence initiation rates. Results: Sociodemographic variables predicted about 60 percent of the variation in hospital-specific rates of exclusive breastfeeding at discharge. Hospitals that were designated intensive or regional perinatal centers, delivered higher volumes of infants, and served more breastfeeding-prone populations were only slightly more likely to have higher adjusted rates compared with other hospitals; considerable unexplained variation remained. Conclusions: Standardized exclusive breastfeeding rates pointed to the contribution of population demographics to breastfeeding initiation, and other contributions, including hospital practices, are also important. To protect, promote, and support breastfeeding, a more detailed evidence base on hospital policies and practices should be developed, and hospitals should review their policies and practices in light of documented best breastfeeding practice. (BIRTH 32:2 June 2005)