We have recently demonstrated that the keratin K3 gene, which is active in the suprabasal human corneal epithelium, is missing in the genome of the mouse. We show that a normal K3 gene exists in a wide variety of mammals while in rodents the gene is converted to a pseudogene with a very strong sequence drift. The availability of K5(-/-) mice provides a unique opportunity to investigate type-specific keratin function during corneal differentiation in the absence of both K5 and K3. Here, we report that the deletion of K5, which in wild-type mice forms a cytoskeleton with K12, does neither cause keratin aggregation nor cytolysis in the cornea. This is due to the induction of K4 in corneal epithelial cells, normally restricted to corneal stem stem cells residing in the limbus. Using a combination of antibodies and RT-PCR, we identified additional keratins expressed in the mouse cornea including K23 which was previously thought to be specific for pancreatic carcinomas. This reflects an unexpected complexity of keratin expression in the cornea. Our data suggest that in the absence of mechanical stress, corneal differentiation does not depend on distinct keratin pairs, supporting a concept of functional redundancy, at least for certain keratins.