Resolution of the phylogenetic relationships among organisms, at or above the species level, has usually been performed using mitochondrial DNA (mtDNA) genes 
. There are several reasons to justify that mtDNA is the most commonly used marker in animal phylogenetics including its elevated substitution rate, which ensures a strong phylogenetic signal, and its reduced effective population size, Ne
, which increases the probability of reciprocal monophyly of the gene among different species. In spite of the advantages of mtDNA, the use of a single marker in phylogenetics has a critical limitation: it allows obtaining only one particular genealogy among all possible ones compatible with the true populations or species tree 
. Thus, making inferences from a single gene could lead, in particular instances, to erroneous biological conclusions.
Indeed, different unlinked loci can have conflicting genealogical histories as a result of the stochastic process of coalescence of alleles due to random genetic drift and incomplete lineage sorting 
. The probability of observing discordant gene trees depends on Ne
and the branch lengths of the species tree. When these branch lengths (measured in generations) are of the same magnitude as Ne
, a situation typically observed in phylogenies of closely related species, deep coalescences in ancestral populations become highly probable and, consequently, the gene tree topology may differ from that of the species tree 
. The discordance among gene trees can occur not only in topology but also in branch lengths 
. Noticeably, this source of heterogeneity is always present since different genes are likely to coalesce at different times. This will produce different gene trees (with different branch lengths) even if the topologies are identical. In fact, branch lengths and tree topology heterogeneity can be considered two sides of the same coin and conclusions based on the former can also be applied to the latter. In the last few years, phylogenetics focused on the species tree, rather than on individual gene trees, has received considerable attention since these new methods take into account the coalescence of genes to construct species trees and to estimate reliable split times and demographic parameters 
The mutational process is also an important source of variation among gene trees and can significantly affect both classical phylogenetic and species tree reconstruction processes. Short genes and genes with low substitution rates are highly affected by this randomness due to their low information content. Actually, in animals, most nuclear genes have very low substitution rates compared to mitochondrial genes. In this sense, we may raise the question of whether the incorporation of a highly informative marker (e.g., a mitochondrial gene) into a species tree inference can improve the efficiency of the estimates. That is, the addition of this marker may help to reduce the quantity of data (i.e., the number of independent markers) necessary to achieve an adequate level of accuracy in estimating species trees. On the other hand, the incorporation of a mitochondrial gene into the inference might compromise the performance of these methods due to the introduction of a large variance in the substitution rates among loci and a more complex mutational model. It is thus important to determine which properties of this marker predominate in the construction of species trees.
Here, we analyze the influence of deep coalescence and mutational processes in the estimation of divergence times under conditions likely to be encountered in natural populations, mainly focusing on recent mammalian species diversification. For this purpose, we conducted a multilocus simulation study using empirical values of nucleotide diversity and substitution rates obtained from a wide range of mammals. In addition, we examined, under different conditions, the performance of different multilocus strategies (with or without mitochondrial genes) in estimating divergence times in a species tree framework. Finally, we also investigated the power of the different multilocus sets in estimating the species tree topology in the particular situation of a species radiation. The results of these analyses can be useful to make informed choices about the optimal number and types of markers necessary to obtain reliable estimates of species tree parameters in empirical studies of mammals and other groups with similar population sizes.