UNDERSTANDING THE ORIGINS OF AREAS OF ENDEMISM IN PHYLOGEOGRAPHIC ANALYSES: A REPLY TO BRIDLE ET AL.

Abstract

One of the chief strategies used to protect biodiversity isthe identification of areas of endemism—regions in whichunique species or populations are concentrated. It often hasbeen assumed that such areas result from nonrandom histor-ical processes that affect a biota, such as vicariance of dis-tributional ranges by geological or climatological events.This might produce concordant patterns of distribution, aswell as genetic divergence, among clades of distantly relatedspecies, such as monkeys, toads, and grasshoppers. However,explanations due to vicariance and fragmentation have ap-propriately come under closer scrutiny, because independentprocesses, such as adaptation to local ecological conditionsor lineage sorting, might produce similar biogeographic pat-terns. Hence the use of hypothesis testing in biogeography,as opposed to narrative scenarios, has grown (Knowles 2001;Templeton 2002; Lessa et al. 2003).We recently identified concordant areas of endemism(AOEs) in monkeys and toads on the biodiverse Indonesianisland of Sulawesi that we attributed to habitat fragmentation(Evans et al. 2003a). Bridle et al. (2004) proposed that theidentification of these AOEs is premature. They presented analternative analysis and concluded that a hypothesis of iso-lation by distance would explain the distributional patternsin the Celebes toad,Bufo celebensis. Thus, the main concernsof Bridle et al. (2004) are: (1) other processes in addition toor instead of fragmentation, such as isolation by distance,may explain the distribution of diversity on Sulawesi, (2) notall Sulawesi species or populations have AOEs that corre-spond with Sulawesi macaques and toads, and (3) a hypoth-esis of isolation by distance rather than fragmentation is suf-ficient to explain our data fromB. celebensis mitochondrialDNA (mtDNA). We agree with the first two of these asser-tions, as we have discussed (Evans et al. 2003a), but disagreewith the third.In this response we review evidence that supports frag-mentation as the cause of concordant AOEs in Sulawesi mon-keys and toads. We describe how assumptions made in thepartial correlation test by Bridle et al. (2004) are not satisfiedby the data, and how these inappropriate assumptions delivera misleading result. We further demonstrate that when theseassumptions are not made, our data onB. celebensis stronglysupport a significant correlation between genetic distance andSulawesi AOEs that is independent of geographical distance,a result that is also supported by our other analyses (Evanset al. 2003a). Finally, we point out that although exceptionsto a more common pattern of endemism are expected, theseareas provide useful biological information for conservationdecisions.Null hypotheses, nested clade analysisEvans et al. (2003a) tested the null hypothesis of panmixiaof B. celebensis mtDNA with nested clade analysis (NCA);deviations from the null were explored with alternative ex-planations including isolation by distance and fragmentation(Templeton 1998). Bridle et al. (2004) criticized the use ofNCA, and cited a critique of NCA that we also cited byKnowles and Maddison (2002). However, they did not dem-onstrate how NCA might have caused error in our study. Oneconclusion of Knowles and Maddison (2002) is that NCAhas a tendency to assign deterministic processes such as frag-mentation to patterns that may be realized by stochastic pro-cesses such as lineage sorting. Although this seems plausiblefor a small number of congruent patterns, the ‘‘mirrorimage’’of seven separate AOEs for toads and monkeys is supportedby phylogenetic analyses and by parametric bootstrap tests,and argues against stochastic lineage sorting as the sole causeof the phylogeographic pattern of B. celebensis mtDNA.If abiotic processes influenced the distribution of diversityon Sulawesi, multiple sympatrically occurring species withdifferent ecological adaptations might share a common dis-tribution of diversity. Bridle et al. (2004) suggested that weincorrectly tested the null hypothesis that they embrace—thatgenetic variation is due to isolation by distance (a hypothesiswe tested with NCA). However, they missed the point thatthe null hypotheses of the parametric bootstrap tests (Huel-senbeck et al. 1996; Goldman et al. 2000) explored anotheralternative: that B. celebensis mtDNA sequences form a cladein each AOE.Isolation by distance is expected to result in older lineagestypically being widespread, whereas younger ones have amore limited distribution (Crandall and Templeton 1993; Po-sada and Crandall 2001). Although stochastic processes orpoor sampling can yield no overlap without any barrier togene flow (Irwin 2002), in the absence of barriers to dispersalother than distance some geographic overlap in clades canresult. That we detected no geographic overlap of the twooldest mtDNA clades of B. celebensis and also no overlap ofseven other divergent clades suggests the role of factors inaddition to geographic distance in influencing AOEs. Ofcourse, as Bridle et al. (2004) pointed out, multiple molecularmarkers and other sources of information are preferable to a