The identity-by-descent (IBD) matrix is the core of the variance component QTL model. The true IBD matrix comes from a distribution of IBD matrices given the marker information, but its expectation is normally used in QTL analysis. This gives incorrect likelihood values since the extra uncertainty in estimating the IBD matrix is not included. Previous studies have concentrated on small pedigrees where the correct likelihood can be derived. For large pedigrees this approach is not feasible. We therefore developed a Monte Carlo method for calculating the likelihood in- corporating the uncertainty of the estimated IBD matrix. The aim of this study is to implement the Monte Carlo Full Likelihood (MCFL) algorithm and to compare the true likelihood with the like- lihood based on the expected IBD matrix for large pedigrees. Our simulation results show that the likelihood based on the expected IBD matrix approximates the true likelihood well and may there- fore justify the use of the expected IBD matrix in empirical QTL analysis. Our MCFL method can actually be computationally more efficient than the expectation method for large pedigrees with a small founder generation, because the rank of the true IBD matrix is much lower than the rank of the expected IBD matrix, especially when the genetic markers are highly informative. Using the IBD matrices produced in our MCFL method we may also simplify the modeling of epistasis for linked QTL.