Ten commercially available antikeratin antisera were tested immunohistochemically on fresh frozen and formalin-fixed paraffin-embedded tissue. Eight of the antisera were in addition tested on protein-immunoblottings. For six of the antisera a good correspondence was found between our immunoblots and data given by the manufacturers. Two monoclonal antisera did not react with keratin proteins. On immunohistochemical testing two of the antibodies showed qualitatively identical staining on both frozen and paraffin sections without background staining. Three of the antibodies reacted weakly or not at all on paraffin sections but gave acceptable staining on frozen sections. Three of the antibodies showed acceptable staining on paraffin sections, but background staining on frozen sections and one antibody gave the reverse staining pattern. For one of the antibodies it was impossible to obtain an acceptable staining due to high non-specific binding of the secondary antibody. None of the antikeratins were true panepithelial tumour markers as all of them failed to detect keratin in at least one of the epithelial tumours. However, a combination of two or three antikeratins (Hybritech AE1 + AE3, Becton Dickinson No 7650, DAKO A622) covered most or all epithelial tumours examined. It is concluded that commercially available antisera show great variability with respect to quality and reactivity indicating that the majority need further purification, characterization and testing on tissues before they are introduced on the commercial market.