In this paper, a limited modified BFGS method for solving large-scale unconstrained optimization problems is proposed. The proposed algorithm generates quasi-Newton directions using a modified BFGS method suggested by Biggs (1973). The modified BFGS method is then extended to the limited memory version. In order to use only minimum storage for the modification, the modification is only applied to the last BFGS corrections. Numerical results indicate that an improvement is achieved.