In immunohistochemistry, it is well known that the majority of monoclonal antibodies to keratins work best on fresh frozen tissue specimens, yet in clinical practice most biopsies are routinely fixed in formaldehyde. This seriously limits the range of keratins that can be reliably assessed in retrospective studies (particularly where only rare archival material exists) and where subtle changes during tissue differentiation may be important. Antigen retrieval using exposure to microwave radiation is one technique that has been applied successfully to other tumour markers (e.g., p53). However, few papers have used this method when immunolabelling for keratins, in spite of the widespread use of antikeratin antibodies as markers of differentiation. The effect of keratin antigen retrieval using microwave processing was assessed on a range of oral mucosal biopsies, since the oral cavity displays a wide range of keratins. A panel of six well characterized antibodies was chosen: LP34 (Ck1, 5, 6, 18), LH1 (Ckl0), LL025 (Ck16), A53 BA2 (Ck19), AE8 (Ck13), and E3 (Ck17). For each specimen, one piece was stored in liquid nitrogen and another piece fixed in formalin. Tissue sections were cut from each and, using the peroxidase avidin biotin technique, keratin expression was recorded for a frozen section, a dewaxed section, and a microwave-heated dewaxed section. Although overall there was a 25% improvement in identification of keratins after microwaving, some antibodies performed better than others. Given that keratins have been shown to be of value in tumour diagnosis, this study suggests that microwave processing of archival material can be a valuable adjunct to such analysis.