Coalescent theory provides powerful models for population genetic inference and is now increasingly important in estimates of divergence times and speciation research. We use molecular data and methods based on coalescent theory to investigate whether genetic evidence supports the hypothesis of A. pretrei and A. tucumana as separate species and whether genetic data allow us to assess which allopatric model seems to better explain the diversification process in these taxa. We sampled 13 A. tucumana from two provinces in northern Argentina and 28 A. pretrei from nine localities of Rio Grande do Sul, Brazil. A 491 bp segment of the mitochondrial gene cytochrome c oxidase I was evaluated using the haplotype network and phylogenetic methods. The divergence time and other demographic quantities were estimated using the isolation and migration model based on coalescent theory. The network and phylogenetic reconstructions showed similar results, supporting reciprocal monophyly for these two taxa. The divergence time of lineage separation was estimated to be approximately 1.3 million years ago, which corresponds to the lower Pleistocene. Our results enforce the current taxonomic status for these two Amazon species. They also support that A. pretrei and A. tucumana diverged with little or no gene flow approximately 1.3 million years ago, most likely after the establishment of a small population in the Southern Yungas forest by dispersion of a few founders from the A. pretrei ancestral population. This process may have been favored by habitat corridors formed in hot and humid periods of the Quaternary. Considering that these two species are considered threatened, the results were evaluated for their implications for the conservation of these two species.