Two in vitro tests were applied to 54 consecutive patients with severe aplastic anemia who were treated in Seattle with antithymocyte globulin. In the first test, peripheral blood mononuclear cells were collected from each patient before antithymocyte globulin therapy and then treated with a panel of monoclonal antibodies and complement. The treated peripheral blood mononuclear cells were assayed for erythroid burst-forming units (BFU-E). This test was designed to determine whether removing various subpopulations of peripheral blood mononuclear cells would increase the number of detectable BFU-E. In the second test, peripheral blood was collected within 48 hr after completion of antithymocyte globulin therapy, and cells were immediately assayed for BFU-E without any further treatment. Data from both tests were analyzed to determine whether the in vitro results correlated with patient response to therapy. Binary logistic regression analyses indicate that a modest correlation (p = 0.04) exists between test 1 in vitro results and patient response to therapy. However, the strength of this association appears to decrease as the interval between diagnosis and treatment increases. In contrast, test 2 had a very significant correlation (p = 0.001) with response to therapy among patients diagnosed more than 1 mo prior to treatment, whereas such an association was not apparent among patients treated within 1 mo of diagnosis.