To avoid methodologic pitfalls in previous observational studies linking formula supplementation in the hospital to early discontinuation of breast-feeding, a controlled clinical trial of restricted supplementation was conducted. In a pretrial sample of 621 newborns, a comparison of two "well-baby" nurseries found no differences in either hospital supplementation practices or the proportion of infants still being breast-fed at 4 or 9 weeks postpartum. Restriction of supplementation in one of the nurseries for the trial period (n = 781) did not result in higher breast-feeding rates at 4 or 9 weeks. There was, however, a slightly greater mean percent of birth weight lost in the restricted group (6.0% v 5.1%; P less than .001). In examining the control group for evidence of an "observational" association, it was found that infants still breast-feeding at 4 or 9 weeks were far more likely to have been unsupplemented than those no longer being breast-fed. It thus appears that formula supplementation in the hospital is a marker, rather than a cause, of breast-feeding difficulty.