In Pelotas, Brazil, 400 newborns from low income families were followed-up until 26 wk of life to study the relationship between their feeding patterns and growth as modified by access to water and by diarrhea. Effects of access to water were the strongest among non-breastfed infants. In houses without indoor water taps, the weight gain of non-breastfed infants during the first 3 mo was approximately half that of partially or predominantly breastfed infants (P < 0.001). In houses with indoor water taps, non-breastfed infants' growth was similar to or exceeded that of predominantly breastfed infants from 2 mo. Predominantly breastfed infants' growth was similar in houses with and without water taps. Breastfed infants had less weight loss per day of diarrhea than non-breastfed infants during the first 4 mo and less diarrhea through 6 mo of life, particularly in houses without taps, in which diarrhea was most prevalent. The existence of a "weanling's dilemma" was approached by comparing the duration of the detrimental effects of not breastfeeding (i.e., 0-3 mo in this study) with the age at which breast milk alone becomes less than optimal for growth (i.e., at 5 mo). Because these two points did not coincide, we conclude that there is no "weaning's dilemma" in this population.