In a double-blind prospective trial, 64 children, 3 to 36 months of age, who had diarrhea for at least 14 days were randomly assigned to receive either a milk-based diet containing 6 g/kg of body weight per day of lactose or the same diet in which the lactose was > 95% prehydrolyzed with β-galactosidase. Clinical and nutritional outcomes were compared. The groups were similar at the start of the study. Four of 33 patients (12.1%) in the lactose group were considered to have treatment failure because of excessive purging with or without refusal to accept the diet, compared with 1 of 31 patients (3.2%) in the hydrolyzed lactose group (P = .20). Among successfully treated boys, fecal excretion was initially similar, but on days 3 to 5 of the trial the lactose group purged a mean 74.4 g/kg per day (95% confidence limits 17.8, 131.0) compared with 42.0 g/kg per day (95% confidence limits 11.4, 72.6) in the hydrolyzed lactose group (P < .01). Diarrhea stopped within 30 hours of hospital admission in 11 children in the hydrolyzed lactose group (35.5%) compared with 1 child in the lactose group (3.3%) (P < .001). Fecal excretion of carbohydrate, nitrogen, and energy was significantly greater in the lactose group (P < .01), but there were no significant differences in fat excretion or in incremental weight change during hospitalization. Feeding lactose-containing nonhuman milk as the sole nutrient source to children with persistent diarrhea resulted in substantially greater purging which was sufficiently severe to increase the risk of dehydration in these children.