A soy-based infant feed was compared with a humanized cow's milk product in newborn preterm babies. Forty preterm infants appropriate for gestational age with birth weights of between 1,500 and 1,800 g were randomly allocated, as they were admitted, to either a soy feed or a humanized cow's milk formula--both commercial products. The trial lasted 35 days for each patient. With a formula intake of 180 ml/kg/day, the patients on the soy feed did not progress well, and most had to be withdrawn from the trial, whereas the group fed on cow's milk fared satisfactorily. With formula intakes of 200 ml/kg/d, progress in the group on the soy feed was improved. However, the body weights and serum albumin levels of the babies receiving the soy product were significantly lower than those of babies on the cow's milk formula at the end of the trial. Until more work has been done, we suggest that soy formulas should be employed with caution in the feeding of small preterm infants.