The soil-water characteristic curve(SWCC)could be used to describe the relationship between the water content and the matrix suction in unsaturated soil and played important roles in unsaturated soil mechanics. Several parameters in fitting models and significant uncertainty of SWCC were obtained by experiment. The uncertainty of SWCC fitting parameters was analyzed using the Bayesian theory. The curve-fitting parameters were regarded as the random vectors. And the Van Genuchten model was used to fit the experimental data of SWCC for the granite residual soils. The posterior distributions of model fitting parameters were obtained by the Markov Chain Monte Carlo simulation delayed rejection adaptive metropolis. Confidence intervals of uncertain model parameters of SWCC were obtained by proposed approach.