A one-year field-study has been carried out in a diarrhoea-endemic area in West Africa to determine the relationship between the bacteriostatic activity of fresh human milk for Escherichia coli in vitro and freedom from diarrhoea of the infant recipients of the milk. The specific contribution of E. coli gastroenteritis to gastrointestinal diseases of infants in general is not known, nor is its particular role in the Gambian infants studied. During the study period, however, both enteropathogenic and toxigenic strains of E. coli were isolated. The incidence of diarrhoea in Gambian infants of seven age-groups from 2 days to 12 months was not significantly correlated with the bacteriostatic activity of milk. This was due rather to absence of diarrhoea in babies fed on low-activity milk than illness in those receiving highly bacteriostatic milk. Indeed, very active milk appeared to protect recipients almost completely, including seven babies of over 3 months of age, five of them during the rainy season, when the risk was high. Babies receiving lower-activity milk experienced more diarrhoea. In a situation where diarrhoeal disease is multifactorial, field evaluation of the protective action by one antibacterial property of milk is difficult. A better understanding of in vivo protection is important, and the factors which have to be taken into account are discussed.