The evaluation of antibodies for diagnostic purposes on routinely processed sections of histologically undiagnosable neoplasms presents special problems. The authors have addressed this problem by constructing a panel of putatively mutually exclusive antibodies and testing it on sections of anaplastic neoplasms. Tissue sections of 120 routinely processed histologically undifferentiated large cell human neoplasms with a histologic differential diagnosis of carcinoma versus lymphoma and/or melanoma were stained with a panel of antibodies composed of monoclonal antikeratin AE1, monoclonal antileukocyte common antigens PD7/26 and 2B11, and rabbit anti-S-100 protein. Only cases not diagnosable by routine morphologic examination were included. PD7/26 and/or 2B11 were positive in 61 cases (supporting lymphoma), AE1 was positive in 17 cases (supporting carcinoma), and anti-S-100 was positive in 25 cases (supporting melanoma). Seventeen neoplasms failed to react with any of the panel antibodies. None of the neoplasms reacted with antibodies directed against two or more different antigens. These results indicate excellent specificity (100%) and good sensitivity (86%) of the panel antibodies on histologically undifferentiated neoplasms. These results are significant on two levels: first, as a test of the panel approach to evaluate antibodies on anaplastic neoplasms, and, second, as a demonstration of the diagnostic utility of the specific panel the authors have employed in such cases.