Autoimmune hemolytic anemia (AHA) may be either primary (ie, "idiopathic," one third of all patients) or secondary (ie, associated with underlying illness, two thirds of all patients). A positive Coombs antiglobulin test is the most important criterion for diagnosis of AHA, and characterization of RBC coating (as to whether it is by IgG alone, by IgG plus complement, or by complement alone) may be valuable in ruling out certain underlying illnesses as causative in selected patients. Many limitations to the antiglobulin test must be kept in mind. As routinely performed, a positive result requires greater than 500 molecules of IgG per RBC. Red blood cells from normal subjects have less than 35 molecules of IgG per RBC. Up to 2% to 5% of patients with AHA will have RBC coating in the 50 to 500 molecules per cell range and may therefore have "false"-negative antiglobulin tests. Apparent failures of strength of antiglobulin reactions to correlate with severity of in vivo RBC destruction may result from technical factors or may reflect intrinsic differences among autoantibodies (eg, IgG subclass, affinity). A likely mechanism of in vivo RBC destruction in AHA has been demonstrated in immune-mediated in vitro RBC "rosette" formation about macrophages and lymphocytes bearing specific receptors for subclasses of IgG and for subcomponents of complement. Increased avidity of macrophages for coated RBCs in response to such stimulus as infection may play an additional role. Treatment involves control of underlying disease and the judicious use of adrenal steroids, splenectomy, and immunosuppressive agents. Transfusions, while life-saving in life-threatening anemia, carry substantially increased risks in AHA patients. Their use should be strictly limited.