Neonatal transfusion practices during 1989 of 452 institutions involved in transfusing infants were surveyed by questionnaire. Most respondents (77%) transfused fresh frozen plasma appropriately (ie, primarily to treat coagulation disorders). However, 11% stated that their most frequent use of fresh frozen plasma was solely to treat hypovolemia, a practice generally not recommended. Seventy-eight percent of respondents transfused platelets to treat bleeding infants with blood platelet counts of less than 50 x 10(9)/L; 84% gave platelets to sick, premature neonates with counts of less than 50 x 10(9)/L whether or not bleeding was evident. Only 35% of respondents transfused granulocytes for neonatal sepsis; most institutions used buffy coats isolated from units of blood--a product readily available, but of questionable efficacy when compared with leukapheresis granulocytes. Ninety-three percent of respondents provided blood components with low risk of transmitting cytomegalovirus: components from seronegative donors were used by 84%, leukocyte-reduced products by 6%, and a combination by 10%. Thirteen percent of respondents gave gamma-irradiated blood components to all and 46% gave them to some neonates to prevent graft vs host disease. Forty-one percent did not routinely irradiate. Ten percent of respondents used leukocyte reduction instead of gamma irradiation to prevent graft vs host disease, a practice currently not advocated. Thus, national transfusion practices for neonates are variable, controversial, and, occasionally, other than those usually recommended. Additional research and educational efforts are needed to ensure optimal transfusion therapy.