This paper aims to overcome the problem “fragility of knowledge” with which conversational contextualism is confronted by presenting Wittgensteinian contextualism model which replaces “conversational context” dependency of knowledge with “language game” dependency of knowledge. Wittgensteinian contextualism can be distinguished into two standpoints. One is “contextualism of justification” which insists on the variability of justification rules with language games, the other is “contextualism of hinge” that insists on variability of error possibilities to be excluded with language games. Wittgensteinian contextualism which combine these two contextualism can make knowledge stable by fixing and structurizing the context.