Estimates of speciation times are subject to a number of potential errors. One source of bias is that effective population size (Ne) has been shown to influence substitution rates. This issue is of particular interest for phylogeographic studies because population sizes can vary dramatically among genetically structured populations across species' ranges. In this study, we used multilocus data to examine temporal phylogeographic patterns in a widespread North American songbird, the Northern Cardinal (Cardinalis cardinalis). Species tree estimation indicated that the phylogeographic structure of C. cardinalis was comprised of four well-supported mainland lineages with large population sizes (large Ne) and two island lineages comprised of much smaller populations (small Ne). We inferred speciation times from mtDNA and multilocus data and found there was discordance between events that represented island-mainland divergences, whereas both estimates were similar for divergences among mainland lineages. We performed coalescent simulations and found that the difference in speciation times could be attributed to stochasticity for a recently diverged island lineage. However, the magnitude of the change between speciation times estimated from mtDNA and multilocus data of an older island lineage was substantially greater than predicted by coalescent simulations. For this divergence, we found the discordance in time estimates was due to a substantial increase in the mtDNA substitution rate in the small island population. These findings indicate that in phylogeographic studies the relative tempo of evolution between mtDNA and nuclear DNA can become highly discordant in small populations.