Several theoretical studies have demonstrated the importance of accounting for coalescent stochasticity in phylogeographical studies, however, there are few empirical examples that do so in the context of explicit hypothesis testing. Here, we provide an example from the Idaho giant salamander (Dicamptodon aterrimus) using 118 mtDNA sequences, nearly 2 kb in length. This species is endemic to mesic forests in northern and central Idaho, and several a priori hypotheses have been erected based both on palaeoclimatic grounds and from phylogeographical studies of codistributed amphibians. Phylogenetic analysis of the D. aterrimus data suggests an expansion from a single refugium south of the Salmon River, whereas the inference from nested clade analysis is one of expansion from a single refugium in the Clearwater drainage. Explicit testing of these hypotheses, using geographically structured coalescent simulations to erect null distributions, indicates we can reject expansion from the Clearwater drainage (p(CLW) = 0.089), but not expansion from the South Fork of the Salmon drainage (p(SAL) = 0.329). Furthermore, data from codistributed amphibians suggest that there may have been two refugia, and an AMOVA shows that most of the molecular variance partitioned between the Clearwater and the Salmon drainages (54.40%; P < 0.001) and within drainages (43.61%; P < 0.001). As a result, we also tested three a priori hypotheses which predicted that both the Clearwater and Salmon drainages functioned as refugia during the late Pleistocene; we could reject (P-CORD = 0.019) divergence dates during the Cordilleran glacial maxima [c. 20 000 years before present (ybp)], during the Sangamon interglacial (c. 35 000 ybp; p(SANG) = 0.032), as well as pre-Pleistocene divergence (c. 1.7 Ma; p(pP) < 0.001). Mismatch distributions and Tajima's D within the individual drainages provide further support to recent population expansion. This work demonstrates coalescent stochasticity is an important phenomenon to consider in testing phylogeographical hypotheses, and suggests that analytical methods which fail to sufficiently quantify this uncertainty can lead to false confidence in the conclusions drawn from these methods.
