The determination of statistical significance in genetic linkage studies is complicated by many factors, such as missing individuals or uninformative markers, and the validity of theoretical results is often questionable. Although many simulation-based methods have been proposed to determine empirically the statistical significance, they are either not generally applicable to complex pedigree structures, or not able to preserve the observed genetic information content at each locus in the pedigrees. We have developed and implemented a general and computationally efficient randomization procedure in GENEHUNTER that applies to arbitrary pedigree structure and preserves the observed information content at each locus. We applied this method to the Problem 1 data set of the Genetic Analysis Workshop 11. The performance of this new method was similar to the method implemented in GENEHUNTER-PLUS, and both outperformed the conservative approach in GENEHUNTER.