Identification of biomarkers in archival tissues using immunochemistry is becoming increasingly important for determining the diagnosis and prognosis of tumors, for characterizing preinvasive neoplastic changes in glandular tissues such as prostate, for evaluating the response of tumors and preinvasive neoplastic changes to certain therapies (i.e., as a surrogate intermediate end point), for selecting patients who are candidates for specific therapies (e.g., immunotherapy) and for retrospective studies. For detecting specific biomarkers it is important to understand the limitations imposed by the fixation methods and processing of the tissues. This study was designed to determine the effects of fixation on the detection in archival paraffin blocks of selected antigens postulated to be important in tumor biology. We evaluated the antigens TGF alpha, p185erbB-2, broad spectrum keratins, p53, and TAG-72 (B72.3). Fixatives evaluated included standard preparations of neutral buffered formalin, acid formalin, zinc formalin, alcoholic formalin, ethanol, methanol, and Bouin's fixative. We found that in general neutral buffered formalin is the poorest fixative for maintaining antigen recognition by immunohistochemistry and that no single fixative was best for all antigens. The dehydrating (coagulant) fixatives (e.g., ethanol and methanol) preserved immunorecognition of p53 and broad spectrum keratins best while the slow cross-linking fixatives (e.g., unbuffered zinc formalin) were best for demonstrating TGF alpha and p185erbB-2. Fixatives other than neutral buffered formalin produced equivalent recognition of the epitope of TAG-72 by B72.3. In formalin fixed archival tissues, only a portion of the antigen signal can be detected by routine immunohistologic methods.