Eleven patients were seen in consultation because of anemia or bleeding during chloramphenicol therapy. Four others were being treated with chloramphenicol without the physician in charge being aware of the subtle hematologic changes that were under way. Depression of erythropoiesis, manifested in a drop in reticulocyte count, was the first and most frequent warning of trouble; it was followed in order by suppression of thrombopoiesis and leukopoiesis. The reversible stage is unpredictable and sometimes quite short. In each of these cases the bone marrow recovered after chloramphenicol administration was stopped. Chloramphenicol should not be given for trivial infections. When it must be used, serial reticulocyte counts should be done, and a sudden or severe drop calls for study of the bone marrow.