Pathologists are now asked frequently to determine the primary site for metastatic carcinomas of unknown origin (MCUO), using adjunctive morphological techniques such as electron microscopy, immunohistology, and other modalities. The authors present an algorithmic immunohistochemical approach to this problem that is based on their experience with over 2,800 routinely-processed epithelial malignancies of various types. These have been studied with antibodies to keratins, vimentin, epithelial membrane antigen, MOC-31, tumor associated-glycoprotein-72 (recognized by monoclonal antibody B72.3), prostate-specific antigen, thyroglobulin, gross cystic disease fluid protein-15, carcinoembryonic antigen, CA-125, CA19-9, placental alkaline phosphatase, S100 protein, and estrogen receptor protein. The algorithm that is structured around these 14 analytes is based on the relative predictive value of each marker, which in turn, determines its place in the sequence of interpretation. The authors' experience with this approach shows 67% accuracy with regard to the ultimately determined site of origin for MCUO, a figure which is similar to that reported by other investigators.