I examined mitochondrial ND2 sequences in species of tits (genus Parus) to determine whether substitutions were neutral or under selection. Haplotype trees in these species are shallow and geographically unstructured, which could be a signature of recent demographic expansion or purifying selection. McDonald-Kreitman tests revealed a pattern of excess replacement polymorphisms for closely related taxa. Replacement substitutions tended to be less common in the basal parts of haplotype trees, suggesting that they are selected against over evolutionary time. Thus, the pattern of selection is consistent with a model of mildly deleterious haplotypes. To investigate the significance of this selection, the ND2 gene was subdivided into surface and transmembrane portions, and synonymous-non-synonymous sites, and FST-values (the amount of variance distributed among populations) computed for each. Among sampling localities in P.major, both partitions gave similar estimates of FST, which is consistent with demographic expansion as the cause of the shallow trees. In P.montanus, two localities contained high percentages of individuals that had a single replacement substitution in the transmembrane portion of the gene, which inflated FST-values relative to the surface sites. Thus, the interpretation of the genetic differentiation of these two sites could be either geographical isolation or selection. Given that the sole substitution causing the high FST-values was a replacement one, selection is implicated. However, this substitution occurred in only 2 of 12 localities, revealing that overall the phylogeographic pattern was not biased by selection, once the cause of the two high pairwise FST-comparisons was recognized. Investigators should consider the source of significant genetic heterogeneity within species, to determine if the variation is likely caused by demographic isolation or selection. However, it seems unlikely that most phylogeographic analyses are invalidated by the effects of natural selection.