A series of premature infants was studied for the presence of bacterial infection. On the basis of clinical evidence and bacteriological studies, they were divided into three groups in which sepsis was considered to be proven, possible, or unlikely. Band neutrophil counts were elevated most frequently in the "sepsis-proven" group and the elevation occurred usually within 24 hours of onset of signs of disease. Qualitative changes in neutrophils (Döhle bodies, toxic granulation, and vacuolization) were more frequent in the sepsis-proven group and, together with the band count, provided valuable techniques for the diagnosis of bacterial infections. Thrombocytopenia occurred frequently in the sepsis-proven group and seemed to result from increased utilization or destruction of platelets rather than failure of production. In such cases, evidence of intravascular coagulation was minimal and it was concluded that thrombocytopenia had resulted from a direct effect of the bacteria or its products on platelets and/or endothelium.