A headwater or river capture is a phenomenon commonly invoked to explain the absence of reciprocal monophyly of genetic lineages among isolated hydrographic basins in freshwater fish. Under the assumption of river capture, a secondary contact between populations previously isolated in different basins explains the observed genetic pattern. However, the absence of reciprocal monophyly could also arise under population isolation through the retention of ancestral of polymorphisms. Here, we applied an approximate Bayesian computation (ABC) framework for estimating the relative probability of scenarios with and without secondary contact. We used Cnesterodon decemmaculatus as a study model because of the multiple possible cases of river capture and the demographic parameters estimated in a previous mitochondrial DNA study that are useful for simulating scenarios to test both hypotheses using the ABC framework. Our results showed that, in general, mitochondrial DNA is useful for distinguishing between these alternative demographic scenarios with reasonable confidence, but in extreme cases (e.g. recent divergence or large population size) there is no power to discriminate between scenarios. Testing hypotheses of drainage rearrangement under a statistically rigorous framework is fundamental for understanding the evolution of freshwater fish fauna as a complement to, or in the absence of, geological evidence.