Abstract: Haplotype inference and reconstruction have become an essential step in linkage analysis and association analysis. The methods for haplotype inference are all based on the complete genotype data sets. However, genotypes of some individuals in each pedigree may be partly missing or ungenotyped. In this article, we propose a new EM algorithm to perform haplotype inference incorporating ungenotyped individuals in pedigree studies. We also give the standard errors of estimated parameters and evaluate the performance of our method by simulation studies.