This paper presents a Bayesian method for species delimitation using multilocus sequence data. The method uses a Bayesian framework to estimate the posterior probabilities of species assignments, taking into account uncertainties due to unknown gene trees and the ancestral coalescent process. The method relies on a user-specified guide tree to avoid integrating over all possible species delimitations. The statistical performance of the method is examined using simulations, and the method is illustrated by analyzing sequence data from rotifers, fence lizards, and human populations. Species delimitation is a critical task in biology, important for conservation, epidemiology, and evolutionary biology. Traditional methods rely on morphological traits, which may not accurately reflect species diversity. Molecular genetic data can provide additional information about species identification, including population identities, levels of gene flow, hybridization, and phylogenetic relationships. However, traditional methods may fail to identify cryptic species. Multilocus sequence data can provide support for different species delimitations using theoretical models that combine species phylogenies and gene genealogies via ancestral coalescent processes. The method uses a Bayesian approach to generate posterior probabilities of species assignments, incorporating information on plausible species membership from morphology, paleontology, and other sources. The method is implemented in the C program bpp, which replaces MCMCcoal. The computation is proportional to the number of loci, and is affected more by the number of sequences in the alignments than by the number of potential species on the guide tree. The method is tested on simulated data and real datasets, showing good performance in identifying species delimitations. The method is also compared to other species delimitation methods, such as structure, and is found to be particularly useful for identifying cryptic species. The method is also shown to be robust to different prior assumptions and to handle large datasets with many loci and sequences. The method is expected to be especially useful for identifying cryptic species that are in sympathy. The impact of alternative models of speciation allowing migration etc. on the statistical performance of our method and the similarities and differences between our algorithm and population assignment algorithms, such as structure, merit further study. At a minimum, species delimitation should rely on many kinds of data, such as morphological, behavioral, and geographical evidence. Studies of behavior, estimation of the frequency and fitness of hybrids, and so on are essential in defining species, although coalescent analysis of genomic data provides valuable information.This paper presents a Bayesian method for species delimitation using multilocus sequence data. The method uses a Bayesian framework to estimate the posterior probabilities of species assignments, taking into account uncertainties due to unknown gene trees and the ancestral coalescent process. The method relies on a user-specified guide tree to avoid integrating over all possible species delimitations. The statistical performance of the method is examined using simulations, and the method is illustrated by analyzing sequence data from rotifers, fence lizards, and human populations. Species delimitation is a critical task in biology, important for conservation, epidemiology, and evolutionary biology. Traditional methods rely on morphological traits, which may not accurately reflect species diversity. Molecular genetic data can provide additional information about species identification, including population identities, levels of gene flow, hybridization, and phylogenetic relationships. However, traditional methods may fail to identify cryptic species. Multilocus sequence data can provide support for different species delimitations using theoretical models that combine species phylogenies and gene genealogies via ancestral coalescent processes. The method uses a Bayesian approach to generate posterior probabilities of species assignments, incorporating information on plausible species membership from morphology, paleontology, and other sources. The method is implemented in the C program bpp, which replaces MCMCcoal. The computation is proportional to the number of loci, and is affected more by the number of sequences in the alignments than by the number of potential species on the guide tree. The method is tested on simulated data and real datasets, showing good performance in identifying species delimitations. The method is also compared to other species delimitation methods, such as structure, and is found to be particularly useful for identifying cryptic species. The method is also shown to be robust to different prior assumptions and to handle large datasets with many loci and sequences. The method is expected to be especially useful for identifying cryptic species that are in sympathy. The impact of alternative models of speciation allowing migration etc. on the statistical performance of our method and the similarities and differences between our algorithm and population assignment algorithms, such as structure, merit further study. At a minimum, species delimitation should rely on many kinds of data, such as morphological, behavioral, and geographical evidence. Studies of behavior, estimation of the frequency and fitness of hybrids, and so on are essential in defining species, although coalescent analysis of genomic data provides valuable information.