In the design of a large liquid-metal fast breeder reactor (LMFBR), the bias-factor method is usually applied to reduce the error of predicted values of neutronics parameters. These bias factors are obtained through the analysis of mock-up experiments. When there exist some differences between the reactor to be designed and its mock-up experimental system, it is impossible to be free from extrapolation errors even after the application of the bias factor. This paper presents an evaluation model for the above kind of extrapolation error, which still remains after the biasing, due to cross-section uncertainties. As an example of an application of this model, the extrapolation error of the design parameters of a 1000-MW(electric) fast breeder reactor was evaluated for the case where bias factors from the large LMFBR mock-up critical experiment, ZPPR-10D, were available. As a result, the error in keff was found to range 0.3 to 1.1% depending on how precisely the reactivity effect of higher plutonium isotopes (especially 241Pu) was predicted. The extrapolation error was predicted to be <2.5% for the control rod worth and also for the fission rate distributions of 239Pu and 238U. It was also shown that the extrapolation error for the control rod worth was reduced by use of a bias factor constructed from some different rod patterns.