Significant intra- and interobserver variability exists in diagnosing and grading oral epithelial dysplasia. Mutations in the tumor-suppressor gene p16 are common in oral cavity dysplastic lesions, but whether immunohistochemical detection of the gene product p16(INK4a) (p16) can be used as a reliable biomarker for dysplasia is unclear. In total, 119 biopsy specimens representing various oral cavity sites and degrees of dysplasia were retrieved from the pathology files of Emory University Hospital. Formalin-fixed, paraffin-embedded sections were stained with hematoxylin and eosin (H&E) and with a monoclonal antibody to p16 (LabVision Corporation, Clone JC2). A blinded review of the H&E slides and the pattern and degree of p16 expression was independently performed by two pathologists. A consensus was obtained when diagnoses differed. Morphologic diagnoses were then compared to p16 immunohistochemical expression. Overall, 61/119 (51%) cases showed no p16 immunoreactivity, including 12/33 (36%) cases of no dysplasia, 11/28 (39%) cases of mild dysplasia, and 38/58 (66%) cases of moderate/severe dysplasia. The remaining cases showed p16 expression limited to the basal and suprabasal nuclei and generally confined to the lower one-third of the epithelium. A logistic regression model showed a trend toward absent p16 expression with increasing severity of dysplasia (P=0.006). Decreased expression of p16 in dysplastic lesions, as found in this study, may reflect the biologic events involving loss of p16 gene function in the pathogenesis of oral cancer. Our findings suggest that p16 immunohistochemistry is not helpful in differentiating dysplastic from nondysplastic mucosa in oral cavity biopsies, and thus is not a reliable biomarker for use in routine clinical practice.