There is substantial debate over the criteria that should be used to group populations of a species into distinct units for conservation (e.g. evolutionarily significant units, management units, distinct population segments). However, in practice molecular genetic differentiation is often the only or main criterion used to identify such units. Most genetic studies attempting to define conservation units in animals use a single molecular marker, most often mitochondrial, and use samples from a limited number of populations throughout the species’ range. Although there are many benefits to using mtDNA, certain features can cause it to show patterns of differentiation among populations that do not reflect the history of differentiation at the nuclear genome where loci controlling traits of adaptive significance presumably occur. Here we illustrate an example of such mitochondrial–nuclear discordance in a ranid frog, and show how using mtDNA or nuclear loci alone could have led to very different conservation recommendations. We also found very high genetic differentiation among populations on a local scale, and discuss the conservation implications of our results.