In a prospective randomized trial, asymptomatic very low-birth-weight infants in a neonatal intensive care unit were either electively transfused with red cells to maintain capillary haematocrit greater than 0.35 l/1 (group 1; n = 9), or were not transfused (group 2; n = 10). Individuals from both groups were excluded if they subsequently received non-elective transfusions, necessitated by their clinical condition. Electively transfused infants gained weight more rapidly than their non-transfused counterparts, but the duration of hospitalization was not shortened. Criteria of morbidity, such as patent ductus murmurs, apnoea and failure to thrive, were similar in both groups. We conclude that in the absence of clinical indications, transfusion to achieve a hematocrit greater than 0.35 l/1 as an empiric procedure, improves weight gain but the risks of transfusion are likely to outweigh the benefits.