The diagnosis of marrow involvement in non-Hodgkin's lymphoma (NHL) relies on morphology with support from immunophenotyping by flow cytometry (FCM). We assessed the relative sensitivity of morphology, FCM, and consensus primer polymerase chain reaction (PCR) of antigen receptor genes in the detection of marrow involvement. In 78 of 100 (78%) cases, there was concordance between FCM and PCR. FCM detected more cases of clonality in B-cell neoplasia. There were 40 cases with objective evidence of involvement by B-cell neoplasia. In this group, FCM had a sensitivity of 97.5% (39 of 40); PCR had a sensitivity of 67.5% (27 of 40). In contrast, PCR had a sensitivity of 71.4%, and FCM a sensitivity of 28.6%, in T-cell neoplasia. In all 12 cases with involvement detected by biopsy, there was objective evidence of clonality. However, clonality was detected in four of seven patients with chronic lymphocytic leukemia and in five of eight patients with T-cell neoplasia in the absence of morphologically detectable disease. Clonality was identified in only one of seven patients with B-cell lymphoma in which the biopsy was interpreted as "suspicious but not diagnostic of involvement." We conclude that morphology remains of central importance in the evaluation of marrow involvement in NHL. We show that FCM and PCR identify involvement in the absence of morphologically apparent disease. In B-cell neoplasms, FCM remains the method of choice for the detection of clonality. PCR for T-cell receptor gene rearrangements may be an important adjunct to the diagnosis of marrow involvement in patients with T-cell neoplasms.