Abstract: There is a lot of uncertainty in the process of establishing Standard penetration test(SPT)based empirical liquefaction assessment model which inevitably has model uncertainty within it. The database used in this paper contains 8 earthquakes between 1962 and 1976. Adopting the maximum likelihood method, the authors calibrate the model uncertainty of liquefaction discriminants of Code for Seismic Design of Industrial and Civil Buildings(TJ11-74),Code for Seismic Design of Buildings(GB50011-2001) and Code for Seismic Design of Buildings(GB50011-2010). Conclusions are that the three liquefaction discriminants are less accurate in the analysis of case histories before 1974, and the accuracy is higher in the analysis of case histories after 1974. The reason may be that SPT blow counts of case histories before 1974 did not come directly from SPT test. According to Bayesian information criterion, the GB50011-2001 liquefaction discriminant has the highest model probability and fits best the dataset. The result of soil liquefaction discrimination also has model uncertainty. The liquefaction probability of soil can be calculated based on the model error of liquefaction discriminants and the ratio of measured SPT blow counts to critical SPT blow counts.