220 Opinion TRENDS in Ecology and Evolution Vol.18 No.

5 May 2003

Phylogenetic comparative methods

and the geography of speciation
Jonathan B. Losos and Richard E. Glor
Department of Biology, Campus Box 1137, Washington University, St Louis, MO 63130-4899, USA

The geography of speciation has long been contentious. that such phylogenetic approaches can make to the study
In recent years, phylogenetic approaches have been of the geography of speciation.
proposed to determine the geographical mode of speci-
ation. If reliable, these methods not only provide a Phylogenetic approaches and the geography of
means of settling the debate about the geography of speciation
speciation, but also indicate that sympatric speciation Methods and assumptions
is surprisingly common and that peripatric speciation is Most interspecific phylogenetic approaches to the geo-
relatively rare. Similar to any phylogenetic inference, graphy of speciation work by examining the distribution of
reconstructions of speciation mode are only useful if sister taxa: for example, if they are sympatric, then
the underlying assumptions of the method are met. In speciation is inferred to have been sympatric and if they
this case, the key assumption is that the geographical are allopatric, speciation is inferred to have been allopatric
range of both extant and ancestral species at the time (Fig. 1). For extant taxa, this approach leads to the
of speciation can be inferred from present-day distri- comparison of sister species; for deeper nodes in a
butions. We discuss whether, and under what circum- phylogeny, the procedure becomes more complicated
stances, such assumptions could be met. We conclude because the geographical distribution of ancestral taxa
that interspecific phylogenies are unable to test alterna- must first be inferred (Box 1).
tive hypotheses concerning the geography of speciation The Achilles heel of this approach is that it requires the
rigorously because of the lability of geographical ranges reconstruction of the geographical distribution of species
and the lack of correlation between the role of adaptive at the time of speciation. The obvious problem with this
processes and geographical mode of speciation. approach is that the current distribution of a species is
not necessarily a reliable indicator of the historical
In spite of half a century of research, resolution of the geographical range of the same species (K.A. Crandall,
controversies concerning the manner in which SPECIATION PhD thesis, Washington University, 1993, [8,9,14,24,25]).
(see Glossary) occurs does not appear to be imminent [1]. One need only consider the alteration of the ranges of
Given the great success of phylogenetic approaches over species that occurred during and after the Quaternary ‘Ice
the past 15 years in addressing a wide variety of Ages’ to recognize how radically ranges have shifted in
evolutionary questions (e.g. [2– 6]), attempts to apply recent evolutionary history, not only in glaciated areas
these methods to the study of speciation are not surprising [26 –28], but also in tropical forests and oceans [29 – 31].
(reviewed in [7]). Three lines of evidence indicate that the geographical
Spurred by Lynch’s influential paper [8], recent studies range of species can, and often does, change substantially
have attempted to use interspecific phylogenetic com- over short periods of time because of climate change,
parative methods to address one of the most contentious colonization of new areas, extinction of competitors, and a
issues in speciation: the geographical mode of speciation host of other reasons. First, the fossil record documents the
[2,9 – 24]. The results of such studies have attracted
considerable attention, both for their promise of resolving Glossary
long-standing questions and for their unexpected findings, Allopatric speciation: speciation resulting from divergent evolution of
such as the suggestions that PERIPATRIC SPECIATION is populations that are geographically isolated from each other.
Parapatric speciation: speciation resulting from divergent evolution of
relatively rare and that SYMPATRIC SPECIATION is surpris- populations that are geographically adjacent to each other.
ingly common [2,8,11,14,19]. If reliable, these methods at Peripatric speciation: a subset of allopatric speciation in which a peripherally
last provide an effective approach toward settling disputes isolated population diverges to become a new species.
Peripheral isolates: a geographically isolated population on the periphery of a
about the frequencies of different geographical modes of species’ range.
speciation. However, if the assumptions of these methods Speciation: divergent evolution resulting in two species from an initial
ancestral species.
are not met, the resulting conclusions might not be Sympatric speciation: speciation occurring within a single geographical area.
reliable. Here, we assess the extent to which these Vicariant speciation: a subset of allopatric speciation in which two populations
assumptions are realistic and evaluate the contribution become isolated by the fragmentation of an initially continuous range into two
or more allopatric populations, each of which is substantial in size (i.e. neither
is a peripheral isolate).
Corresponding author: Jonathan B. Losos ([email protected]).

http://tree.trends.com 0169-5347/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0169-5347(03)00037-5
 Opinion TRENDS in Ecology and Evolution Vol.18 No.5 May 2003 221

Fig. 1. Testing hypotheses with interspecific phylogenies. Current geographical distributions of three species (a) are used to infer geographical modes of speciation. When
species that overlap geographically are sister taxa (b), sympatric speciation is inferred. When sister species do not overlap geographically (c), sympatric speciation is
rejected and allopatric speciation is inferred. When sister species do not overlap and the range of one is smaller than the range of the other (d), peripatric speciation is
inferred.

occurrence of range shifts through time: some extant speciation, but also that entire clades do not experience
species are now found in localities in which they did not range contraction or expansion and that extinction does
occur prehistorically, and vice versa [32– 40]. Second, not occur [2,15,55,56].
observational studies in historical time have directly
documented countless changes in geographical distri- Allopatric versus sympatric speciation
butions [26,41 – 46]. Third, population genetic analyses Because of the unreliability of inferences of geographical
routinely uncover evidence for geographical range expan- ranges in the past, interspecific phylogenetic approaches
sions (e.g. [47,48]). As a result, the present-day range of a are unlikely to add much to the debate over sympatric
species will often differ greatly from the range of that versus ALLOPATRIC SPECIATION . Consider first sympatric
species when it first arose; for this reason, several authors speciation. Can a null hypothesis that speciation was not
have concluded that evolutionary inferences concerning sympatric be rejected if sympatric species are found to be
the geography of species in the past will often not be sister taxa? Given the potential for the evolutionary
reliable [14,41,49–51]. lability of geographical range, the finding that sympatric
These concerns are magnified further when attempts species are sister taxa does not strongly refute the
are made to reconstruct the geographical speciation mode alternative possibility that the species speciated in
of ancestral taxa (e.g. [8,9,19,21,22]). Recent studies have allopatry or parapatry and subsequently expanded their
revealed that, under many circumstances, phylogenetic ranges to come into sympatry. Indeed, this is the heart of
comparative methods are unable to reconstruct ancestral the classic debate about sympatric speciation: proponents
traits accurately; this is particularly true when the trait is note that sympatric species are closely related and infer
evolutionarily labile, such that the rate of trait evolution is that sympatric speciation has occurred; detractors con-
high relative to the rate of speciation [52– 54], as might be sider sympatric speciation to be unlikely on theoretical
the case for geographical range. Moreover, inferences grounds and consider allopatric speciation followed by
about the geographical distributions of ancestral taxa range expansion to be a more probable explanation [56,57].
deduced from the ranges of their descendants are based on Thus, phylogenetic approaches to the study of putative
a method that does not seem very robust (K.A. Crandall, cases of sympatric speciation really do no more than test
PhD thesis, Washington University, 1993, [25]): this the usually implicit assumption that the sympatric species
method reconstructs the geographical range of an ances- truly are sister taxa. Finding that the two species are not
tral species as the sum of the ranges of its descendants sister taxa would, of course, undermine support for a
(Fig. 2, Box 2). These reconstructions thus not only assume sympatric speciation scenario (e.g. [24,58,59]), but finding
that geographical ranges of species remain constant after that they are sister taxa would not address the primary

Box 1. Phylogenetic reconstruction of the geographical mode of speciation

Phylogenetic approaches to the geography of speciation (e.g. [2,8,56,78]) When one or both of the sister taxa are ancestral nodes in a
use present-day geographical distributions and phylogenetic relation- phylogeny, the geographical range of these ancestral taxa must be
ships to infer the geographical context of speciation. In this approach, the inferred before sister-taxon comparisons can be made. In this case, the
current distribution of an extant taxon is used as a character state and ancestral taxon is assigned a geographical distribution that encom-
comparison of the range of sister taxa enables a decision to be made passes the distribution of all of its descendants [8,12,56,78]. Then, the
about which mode of speciation was involved. For extant sister species, ranges of sister taxa are compared in the same manner as are those
present-day distributions are compared and their size and extent of the ranges of extant sister species.
overlap of the ranges is used to infer geographical mode of speciation. For For the phylogenetic method to provide accurate inferences about
example, sister species with broadly overlapping ranges would be speciation, two related assumptions must be met. First, when compar-
inferred to have originated by sympatric speciation, whereas sister ing sister species, one must assume that the current geographical
species in which the ranges were nonoverlapping and in which the range distribution correlates strongly with the geographical distribution at the
of one taxon was relatively small (,5%) compared with the range of the time of speciation. Second, to infer the geographical distributions of
other species would be considered as evidence of peripatric speciation. ancestral taxa, one must assume that distributions are so static that
Precise criteria for distinguishing geographical modes differ among ancestral geographical ranges can be inferred from the ranges of their
authors. For example, Mattern and McLennan [19] consider any overlap descendants [2,8,15,56,71]. Because of the restrictive nature of this latter
among the ranges of sister taxa to be indicative of sympatric speciation, assumption, even some proponents of these methods recommend
whereas Lynch [8] considers ,20% overlap to be trivial. focusing on more recent divergence events (e.g. [8]).

http://tree.trends.com
222 Opinion TRENDS in Ecology and Evolution Vol.18 No.5 May 2003

This is not to say, however, that phylogenetic studies

can never be used to make a case for sympatric speciation;
combined with additional evidence that argues against a
scenario of nonsympatric speciation followed by range
expansion, phylogenetic studies can make the case for
sympatric speciation more compelling [10]. The best
documented such case involves two crater lakes in
Cameroon, each of which is occupied by its own clade of
cichlid fish [60] (Fig. 3). The most parsimonious conclusion
is that, in each crater, a single ancestral species colonized
the crater and subsequently speciated repeatedly. The
topographical and ecological homogeneity of these small
(,5 km2) craters suggests few opportunities for allopatric
TRENDS in Ecology & Evolution
speciation; indeed, it is situations such as this, in which an
Fig. 2. Geographical ranges of lions (Panthera leo), tigers (P. tigris) and jaguars
allopatric phase of differentiation seems inconceivable,
(P. onca). Mattern and McLennan [19] reconstructed speciation modes in a study that have prompted many claims of sympatric speciation
of feline evolutionary diversification and concluded that sympatric speciation was in the literature. The most plausible allopatric scenarios
the most common mode of speciation in cats. For example, they concluded that
the lion is the sister taxon to a clade comprising the tiger and the jaguar. The
would require either that ancestral populations became
ranges of the lion (blue) and tiger –jaguar (red) clades overlap in western Asia geographically isolated in these tiny and apparently
(purple) and the species differ ecologically, leading to the conclusion that the split homogeneous lakes (cf. [61,62]) or that each of the 20
between these two lineages occurred through sympatric speciation (historical
range of lion and tiger are based on maps from [77]). Through this sort of reason- species in the two lakes was the result of dispersal by
ing, the authors conclude that sympatric speciation accounted for 51.8% of felid related species that lived outside of the lake and that
speciation events, even though they acknowledge that felines are ‘large, highly
mobile creatures’ and that the assumption that ‘postspeciation dispersal does not
subsequently went extinct, thus rendering the lake species
overwhelm speciations patterns’ might have ‘been violated to some extent’. Note monophyletic relative to other extant species. Such
that this is an extreme version of the phylogenetic method. Other proponents of scenarios are plausible, although less parsimonious than
this method (e.g. [8]) might not consider the range overlap substantial enough for
a claim of sympatric speciation.
a straightforward hypothesis of sympatric speciation.
Although phylogenetic studies have suggested surpris-
ingly high rates of sympatric speciation, allopatric specia-
concerns of critics, which do not hinge on the evolutionary tion is still usually identified as the predominant mode in
relationships of the taxa. such studies [8,9,14,63]. This result is not surprising
For these reasons, interspecific phylogenetic studies by because the taxa in most phylogenetic studies apparently
themselves add little support to claims of sympatric were chosen, not because they had been suggested as
speciation (or for claims of PARAPATRIC SPECIATION , to potential cases of nonallopatric speciation, but rather
which the same criticisms apply). For example, vertebrates, because data about geographical ranges and phylogenetic
particularly species that tend to be highly mobile, have relationships were available for those taxa. If allopatric
always been considered to be unlikely candidates for speciation is the predominant mode of speciation in most
sympatric speciation. Hence, phylogenetically based studies types of animal, as most workers believe, and if study taxa
that concluded that 52% of speciation events in felids [19] are chosen randomly with respect to likelihood of speciat-
(Fig. 2) and .20% of speciation events in birds [8,14,16] were ing in a particular manner, then one would expect to find a
the result of sympatric speciation were quite surprising. In predominance of cases of allopatric speciation.
light of the potential lability of geographical range discussed However, for the same reasons discussed with regard to
above, we suspect that we are not alone in finding these sympatric speciation, finding that sister taxa are allopatric
results unconvincing (see [14] for similar conclusions from a is not definitive support for allopatric speciation; sympa-
detailed analysis of 13 avian clades). tric speciation followed by range shifts leading to allopatry

Box 2. The relationship between geographical modes of speciation and adaptive evolution
As a corollary to phylogenetic approaches to the geography of indicate that allopatric populations are more likely to speciate when
speciation, several workers [12,13,16] have suggested that not only selection causes their adaptive divergence [80 –83]. Consequently,
the geographical mode of speciation, but also the underlying evidence favoring the vicariant mode of geographical speciation does
process driving speciation, can be determined from examination of not constitute evidence that adaptation was not involved in the
phylogeny. In particular, they argue that some modes of speciation, speciation process; quite the contrary, these studies indicate that
such as peripatric speciation, involve adaptive evolution, whereas vicariant speciation is more likely when adaptation is involved.
other modes, such as vicariant speciation, do not. However, this Furthermore, founder-induced or peripatric speciation does not require
assumption is unwarranted – no relationship necessarily exists that the daughter species undergo adaptive divergence [84]. For
between the geographical mode of speciation and underlying example, Powell [85] used experimental studies to argue that founder
evolutionary process [79]. effects might lead to pre-mating isolation among populations of
Although parapatric and sympatric speciation almost certainly Drosophila pseudoobscura without accompanying adaptive diver-
involve adaptive processes, vicariant speciation and peripatric specia- gence. Hence, at least in allopatric speciation, no relationship exists
tion have no necessary relationship with adaptive versus nonadaptive between the mode of speciation and the extent of adaptive divergence
evolutionary processes. For example, field and laboratory studies during the speciation process.

http://tree.trends.com
 Opinion TRENDS in Ecology and Evolution Vol.18 No.5 May 2003 223

For example, when sister taxa occur on different islands

Pungu maclareni [64,65], are separated by physical barriers (e.g. [66]), or
share a common pattern of geographical distribution with
Konia eisentrauti
other sets of sister taxa [23], then allopatric speciation
Konia dikume would seem to be the most probable interpretation.
Sarotherodon linnellii

Barombi Mbo Crater

Vicariant versus peripatric speciation
Sarotherodon caroli Several studies have extended the approach outlined
Sarotherodon steinbachi above to distinguish between different types of allopatric
speciation (e.g. [8,9,16]). In this case, it is the relative size
Myaka myaka
of the ranges of species at the time of speciation that
Sarotherodon lohbergeri distinguishes between peripatric (or PERIPHERAL ISOLATES )
and VICARIANT SPECIATION ; similarly sized ranges indicate
Stomatepia mariae
vicariant speciation, whereas asymmetric range sizes
Stomatepia pindu indicate peripatric speciation (Fig. 1). For the same
Stomatepia mongo reasons as discussed above, however, inferences about
ancestral geographical range size are unlikely to be
Sarotherodon g. gailiaeus 'Meme' accurate in many circumstances, and hence phylogenetic
Sarotherodon gailiaeus 'Ejagham'
comparative methods might provide little insight [14].
Moreover, if Chesser and Zink [14] are correct that only
Sarotherodon g. gailiaeus 'Cross'
those peripatrically derived species that rapidly expand in
Sarotherodon galilaeus sanagaensis range size are likely to survive, then this method will not
only be inaccurate, but also biased against detecting
peripatric speciation.
Tilapia sp. 'Ejagham 1' Another approach for distinguishing peripatric from
vicariant speciation was proposed recently by Chan and
Tilapia sp. 'Ejagham 2' Moore [13], who suggested that measures of tree balance
Tilapia sp. 'Ejagham 3' could be used to distinguish between the two. In a nutshell,
Chan and Moore [13] argue that vicariant speciation will
Tilapia sp. 'Ejagham 4' produce more balanced tree topologies than will peripatric
speciation because, in peripatric speciation, the processes
Tilapia snyderae/ T. bythobates
leading to speciation occur solely in the peripheral
Tilapia bakossi population. Consequently, one daughter species (the
Bermin Crater

peripheral populations) must complete the speciation

Tilapia gutturosa
process before being able to speciate again, whereas the
Tilapia imbriferna other (the main population) can speciate again immedi-
ately; thus, the latter clade will tend to accumulate more
Tilapia flava
speciation events, leading to an unbalanced phylogenetic
Tilapia bemini/ T. spongotroktis/ T. thysi topology. By contrast, in vicariant speciation, the specia-
tion process will affect both populations equally, and thus
Tilapia guineensis 'Cross' no consistent difference should exist in the time at which
Tilapia kottae
either is able to speciate again; as a result, more sym-
metrical phylogenetic topologies would be expected.
Tilapia guineensis 'Ivory Coast' However, this method assumes allopatric speciation from
the outset and cannot distinguish between allopatric,
Tilapia zillii
parapatric and sympatric speciation; similar to vicariant
speciation, parapatric and sympatric speciation should
TRENDS in Ecology & Evolution
also produce balanced tree topologies because these
speciation processes should affect both daughter species
Fig. 3. Phylogenies of cichlid species from two small African crater lakes. The
monophyly of species in each crater and the lack of obvious barriers to gene flow in the same way. Moreover, many other explanations, such
within them makes sympatric speciation the most plausible interpretation. Repro- as the evolution of characters that promote speciation,
duced, with permission, from [60]. have been put forward to explain tree imbalance [67 – 70];
thus, the existence of imbalance, even in clades in which
could produce the same present-day pattern. Indeed, fossil allopatric speciation seems to have occurred, would not
data reveal cases in which currently allopatric species necessarily be indicative of peripatric speciation.
were sympatric in the late Pleistocene [37]; moreover,
some species of plants that are likely to have arisen Patterns of character evolution
sympatrically by allopolyploidy now have allopatric Patterns of character evolution have also been suggested
distributions separated by hundreds of kilometers [10]. as a line of evidence that can support geographical
As for sympatric speciation, the inclusion of additional interpretations (e.g. [15,16,71]), but such approaches are
data can make a case for allopatric speciation compelling. fraught with assumptions and are unlikely to be generally
http://tree.trends.com
224 Opinion TRENDS in Ecology and Evolution Vol.18 No.5 May 2003

Fig. 4. Plots of geographical range overlap through time. In three out of ten actual phylogenies [fairy wrens (Malurus), cranes (Gruidae) and swordtail fish (Xiphophorus)],
most comparisons among recently divergent taxa are allopatric [as indicated by the percent range overlap (x-axis) versus evolutionary age (y-axis)], a signature of allopatric
speciation with infrequent range shifts (although the swordtail fish do not show the predicted steady increase in overlap with age). However, the remaining seven groups
closely resemble the results of null models that assume frequent range shifts, conditions under which allopatric and sympatric null models are indistinguishable. The
inability to distinguish the observed patterns from those produced by the null models suggests that inferences concerning the geographical mode of speciation cannot be
made reliably. Reproduced, with permission, from [9,21].

applicable. For example, Friesen and Anderson [16] ancestral geographical ranges. However, it is based on the
suggest that vicariant speciation can be ruled out for assumption that, because geographical ranges shift over
taxa that exhibit relatively high rates of character change time, the geographical signal of speciation decays with age
because such high rates are the signature of adaptive such that the deeper one looks in a phylogeny, the less
differentiation, which is supposed to not be involved in sister taxa will exhibit a geographical pattern that
vicariant speciation. However, other processes can pro- conforms to the geographical mode of speciation they
duce high rates of character change and adaptive experienced.
differentiation might play an important role in vicariant Barraclough and Vogler [9] conducted an important test
speciation (Box 2). We suspect that no general relationship of this approach by using simulations to assess what
exists between patterns of character evolution and the patterns would be expected for a phylogeny given different
geographical mode of speciation. modes of speciation and different rates of geographical
range change. These models suggest that, when a
Shifts in range overlap through time particular mode of speciation predominates in a particular
Several studies have used phylogenies to test geographical group and ranges shift stochastically following speciation,
modes of speciation by examining changes in geo- modes of speciation can be inferred by plotting range
graphical range size and overlap of sister taxa through overlap versus evolutionary age (Fig. 4). However, these
time [8 – 10,21,24] (Fig. 4). If, for example, allopatric null models also indicate that, when rates of geographical
speciation is predominant, then recently diverged sister range change are high, it will be impossible to distinguish
taxa will tend to be geographically nonoverlapping, and among different geographical modes of speciation because
the degree of overlap is likely to increase between deeper range shifts will obscure the geographical pattern of
phylogenetic clades as a result of geographical range speciation for even the most recent events (Fig. 4).
shifts. Conversely, if sympatric speciation is the norm, In an examination of actual phylogenies, Barraclough
then recent sister taxa will be entirely overlapping, but and Vogler [9] found that, although many sister species are
sister clades deeper in the tree will be more likely to have allopatric and range overlap increases with age in some
shifted their ranges and thus overlap less. Similar groups, suggesting allopatric speciation, this correlation is
reasoning is used to distinguish peripatric from vicariant always weak and confidence limits around the y-intercept,
speciation. This test is a significant advance over previous which represents geographical overlap at time of specia-
methods because it does not require the reconstruction of tion, are often very large. Unfortunately, in most groups,
http://tree.trends.com
 Opinion TRENDS in Ecology and Evolution Vol.18 No.5 May 2003 225

the correlation between evolutionary age and geographical The future of phylogenetics in the study of the
overlap is indistinguishable from simulations that incor- geography of speciation
porate high rates of geographical range change, regardless Because speciation occurs at the population/species inter-
of whether the geographical mode of speciation in the face, methods are needed that focus at that level [25,72].
simulations is allopatric or sympatric [9,21] (Fig. 4). In One recent approach has been to use phylogenetic
other words, the pattern frequently observed in real approaches to focus solely on recent divergence events,
phylogenies cannot be used to make inferences about the either between populations within a species or within
geographical mode of speciation. Given that the true rate closely related species (reviewed in [73]). These
of change in geographical range will usually be unknown approaches propose that the phylogenetic structure of
and, in many cases, might be large, drawing firm genetic variation among populations and very recently
conclusions from these studies will be difficult. The one diverged species will differ depending on geographical
exception is groups in which almost all sister taxa are mode of speciation. For example, the finding that a small
allopatric and in which only sister taxa deep in the tree and geographically isolated population was closely related
exhibit geographical overlap [24] (Fig. 4). In such cases, a to a population of a geographically widespread species,
conclusion that allopatric speciation has occurred is rendering that species paraphyletic, might indicate the
reasonable, although it seems unlikely that, even in the occurrence of recent peripatric speciation [72]. More work
absence of these methods, anyone would have considered is needed to validate these types of prediction, although
such geographical patterns to be the result of any other the more general concern still remains of whether, even
mode of speciation. among such closely related taxa, one can assume that post-
In summary, Barraclough and Vogler find that, under speciation range shifts have not obscured patterns of
the conditions of their null model, allopatric speciation and speciation.
sympatric speciation can be distinguished by plotting Other recent approaches have moved even closer
range overlap versus age, but only when range shifts have toward the population– species interface (Box 3). These
been relatively infrequent. In practice, few phylogenies historical population genetic methods can estimate several
present the clear-cut pattern exhibited by the null models, parameters that might be useful in attempts to distinguish
suggesting either that range shifts are too frequent to among alternative geographical models of speciation, such
leave a signature in most groups or that a single mode of as historical population size and rates of gene flow, and
speciation rarely dominates in a given group. could also be used to detect the signature of some

Box 3. Historical population genetics and the geography of speciation

Coalescent approaches to population genetics have been developed among alternative geographical modes at a finer scale. For example, if
over the past 20 years to examine the historical processes responsible the data fit a model of strict isolation, estimation of historical effective
for patterns of genetic variation that exist within and among population sizes of sister species at the time of speciation might be
populations (reviewed in [86]). Some coalescent methods, which possible [91]. These estimates could then be used to test the hypothesis
have been developed primarily to test demographic, genetic and of peripatric speciation, which predicts greatly different population sizes
ecological mechanisms of speciation, might also be useful for testing in sister species at the time of speciation. Hare et al. [92] used this
geographical modes of speciation. approach to find that historical effective populations sizes of two
Several methods test the simple null hypothesis of complete isolation dolphin species with antitropical distributions have been very large
of incipient species versus more complex models that involve throughout their history, which suggests that their divergence did not
divergence-with-gene-flow, natural selection, or both [87 –90]. Kliman involve small peripatric populations.
et al. [88] applied such an approach to a study of the Drosophila Other methods, such as nested clade analysis [93], directly incorpor-
simulans complex, finding that a simple isolation model provides a ate haplotype trees with geographical information to test whether
good fit to the divergence between the cosmopolitan D. simulans and phylogeographical associations are due to recurrent gene flow or
two island endemic forms, D. mauritiana and D. sechellia, suggesting historical events, such as fragmentation, colonization and range
the occurrence of allopatric speciation (perhaps not a surprising result expansion. Such analyses of the dynamical history of the geographical
given the geographical distribution of these species). By contrast, range of a population might sometimes be crucial for discriminating
Machado et al. [87] rejected the strict isolation model and found between alternative geographical modes of divergence [93,94]. For
evidence that some gene flow has occurred between D. persimilis and example, this method has been used to show that partially overlapping
D. pseudoobscura since these species began diverging, but not recently. distributions of phylogenetically distinct groups could be due to
These methods appear to be quite promising for distinguishing divergence in allopatry followed by range expansion [93].
among models of isolation, divergence-with-gene-flow, and natural Of course, all of the population-level analyses discussed above have
selection when data from multiple independent loci are available. assumptions of their own that could limit their ability to answer
However, this does not necessarily directly translate into distinguishing questions about the geography of speciation [87,92,95]. For example,
among alternative geographical modes of speciation (a purpose for methods for estimating ancestral effective population sizes assume
which these methods were not originally intended). In the case of panmixia in the ancestor and both descendants, an assumption that,
D. persimilis and D. pseudoobscura, support for the divergence-with- when violated, could result in rejection of the isolation model or
gene-flow model might be the result of limited gene flow throughout the unrealistic estimates of historical effective population sizes [87,92].
speciation process, as might be expected under sympatric or parapatric However, historical population genetic approaches will undoubtedly
speciation, or limited gene flow following secondary contact between provide fresh new insights into the process of speciation, particularly
species whose initial divergence occurred in allopatry. Similarly, when coalescent analyses of multiple independent loci are combined
selection at some loci, but not others, during species divergence with standard phylogenetic or phylogeographical analyses. Advances
might occur in sympatry or parapatry, or might be the result of such as the ongoing application of likelihood and Markov Chain Monte
secondary contact and reinforcement following speciation in allopatry Carlo methods are likely to improve the complexity and utility of these
[89]. methods even further [87]. Whether these approaches will provide
In some cases, additional analyses could be used to distinguish insight into the geography of speciation remains to be seen.

http://tree.trends.com
226 Opinion TRENDS in Ecology and Evolution Vol.18 No.5 May 2003

important historical processes, such as population frag- Sulidae (Aves: Pelecaniformes): a test of alternative modes of
mentation and range shifting. These new approaches hold speciation. Mol. Phylog. Evol. 7, 252 – 260
17 Grady, J.M. and LeGrand, W.H. (1992) Phylogenetic relationships,
great promise for addressing key questions concerning modes of speciation, and historical biogeography of the madtom
speciation; whether they can resolve debate about alterna- catfishes, genus Noturus Rafinesque (Siluriformes: Ictaluridae). In
tive geographical modes of speciation, and avoid the Systematics, Historical Ecology, and North American Freshwater
pitfalls described above, remains to be seen. Fishes (Mayden, R.L., ed.), pp. 747 – 777, Stanford University Press
18 Grant, W.S. and Leslie, R.W. (2001) Inter-ocean dispersal is an
important mechanism in the zoogeography of hakes (Pisces: Merluc-
Conclusions cius spp.). J. Biogeogr. 28, 699 – 721
The past 15 years have seen a remarkable revolution in 19 Mattern, M.Y. and McLennan, D.A. (2000) Phylogeny and speciation of
comparative biology: phylogenetic historical perspectives felids. Cladistics 16, 232 – 253
have gone from being completely ignored to being 20 Patton, J.L. and da Silva, M.N.F. (1998) Rivers, refuges, and ridges:
ubiquitous. As with many conceptual breakthroughs, the geography of speciation of Amazonian mammals. In Endless
Forms: Species and Speciation (Howard, D.J. and Berlocher, S.H., eds)
however, the pendulum has swung too far the other way pp. 202– 213, Oxford University Press
and re-examination of the utility of these methods has 21 Ribera, I. et al. (2001) The effect of habitat type on speciation rates and
begun (e.g. [74 – 76]). Phylogenetic approaches, similar to range movements in aquatic beetles: inferences from species-level
any other analytical method, have their own particular phylogenies. Mol. Ecol. 10, 721 – 735
22 Taylor, C.M. and Gotelli, N.J. (1994) The macroecology of Cyprinella –
assumptions. When those assumptions are met, phylo-
correlates of phylogeny, body- size, and geographical range. Am. Nat.
genetic approaches represent powerful approaches for 144, 549 – 569
addressing many important questions in evolutionary 23 Green, M.D. et al. (2002) Assessing modes of speciation: range
biology. Conversely, when they are not met, as appears asymmetry and biogeographical congruence. Cladistics 18, 112 – 124
often to be the case here, then interspecific phylogenetic 24 Johnson, N.K. and Cicero, C. (2002) The role of ecologic diversification
in sibling speciation of Empidonax flycatchers (Tyrannidae): multi-
methods generally will not be useful.
gene evidence from mtDNA. Mol. Ecol. 11, 2065– 2081
25 Templeton, A.R. (1998) The role of molecular genetics in speciation
Acknowledgements studies. In Molecular Approaches to Ecology and Evolution (DeSalle,
We thank M. Francis Benard, L. Harmon, J. Kolbe, J. Wakeley, R. and Shierwater, B., eds) pp. 131– 156, Birkhäuser
D. Ziolkowski and three anonymous reviewers for comments. Peter 26 Burton, J.F. (1995) Birds and Climate Change, Christopher Helm
Jackson shared his expertise on the geographical range of wild felids. This 27 Hewitt, G.M. (1999) Post-glacial re-colonization of European biota.
work was supported by the National Science Foundation (DEB 9982736). Biol. J. Linn. Soc. Lond. 68, 87 – 112
28 Hewitt, G.M. (1996) Some genetic consequences of ice ages, and their
References role in divergence and speciation. Biol. J. Linn. Soc. Lond. 58, 247 – 276
1 Barton, N.H. (2001) Speciation. Trends Ecol. Evol. 16, 325– 413 29 Anderson, D.M. and Webb, R.S. (1994) Paleoclimatology – Ice-Age
2 Brooks, D.R. and McLennan, D.A. (1991) Phylogeny, Ecology, and tropics revisited. Nature 367, 23 – 24
Behavior: A Research Program in Comparative Biology, University of 30 Colinvaux, P.A. et al. (1996) A long pollen record from lowland
Chicago Press Amazonia: Forest and cooling in glacial times. Science 274, 85 – 88
3 Grandcolas, P. ed. (1997) The Origin of Biodiversity in Insects: 31 Haffer, J. (1987) Quaternary history of tropical America. In Biogeo-
Phylogenetic Tests of Evolutionary Scenarios Muséum National graphy and Quaternary History in Tropical America (Whitmore, T.C.
d’Histoire Naturelle and Prance, G.T., eds) pp. 1 – 18, Oxford University Press
4 Harvey, P.H. and Pagel, M.D. (1991) The Comparative Method in 32 Delcourt, H.R. and Delcourt, P.A. (1991) Quaternary Ecology: A
Evolutionary Biology, Oxford University Press Paleoecological Perspective, Chapman & Hall
5 Maddison, D.R. and Maddison, W.P. (2000) MacClade 4: Analysis of 33 Coope, G.R. and Wilkins, A.S. (1994) The response of insect faunas to
Phylogeny and Character Evolution Version 4.0, Sinauer Associates glacial-interglacial climatic fluctuations. Philos. Trans. R. Soc. Lond.
6 Eggleton, P., Vane-Wright, R.I. eds (1994) Phylogenetics and Ecology Ser. B 344, 19– 26
Academic Press 34 Davis, M.B. (1986) Climatic instability, time lags, and community
7 Barraclough, T.G. and Nee, S. (2001) Phylogenetics and speciation. disequilibrium. In Community Ecology (Diamond, J. and Case, T.J.,
Trends Ecol. Evol. 16, 391– 399 eds) pp. 269 – 284, Harper & Row
8 Lynch, J.D. (1989) The gauge of speciation: on the frequencies of modes 35 Elias, S.A. (1992) Late Quaternary zoogeography of the Chihuahuan
of speciation. In Speciation and Its Consequences (Otte, D. and Endler, Desert insect fauna, based on fossil records from pack-rat middens.
J.A., eds) pp. 527 – 553, Sinauer Associates J. Biogeogr. 19, 285– 297
9 Barraclough, T.G. and Vogler, A.P. (2000) Detecting the geographical 36 Graham, R.W. (1988) The role of climatic-change in the design of
pattern of speciation from species- level phylogenies. Am. Nat. 155, biological reserves – the paleoecological perspective for conservation
419 – 434 biology. Conserv. Biol. 2, 391 – 394
10 Berlocher, S.H. (1998) Can sympatric speciation via host or habitat 37 Graham, R.W. and Lundelius, E.L. Jr (1984) Coevolutionary disequili-
shift be proven from phylogenetic and biogeographic evidence. In brium and Pleistocene extinctions. In Quaternary Extinctions: a
Endless Forms: Species and Speciation (Howard, D.J. and Berlocher, Prehistoric Revolution (Martin, P.S. and Klein, R.G., eds) pp. 223 – 249,
S.H., eds) pp. 99– 113, Oxford University Press University of Arizona Press
11 Briggs, J.C. (1999) Modes of speciation: Marine Indo-West Pacific. 38 Graham, R.W. et al. (1996) Spatial response of mammals to late
Bull. Mar. Sci. 65, 645 – 656 quaternary environmental fluctuations. Science 272, 1601– 1606
12 Brooks, D.R. and McLennan, D.A. (1993) Comparative-study of 39 Miller, A.I. (1997) Dissecting global diversity patterns: examples from
adaptive radiations with an example using parasitic flatworms the Ordovician Radiation. Annu. Rev. Ecol. Syst. 28, 85 – 104
(Platyhelminthes, Cercomeria). Am. Nat. 142, 755 – 778 40 Roy, K. et al. (2001) Climate change, species range limits and body size
13 Chan, K.M.A. and Moore, B.R. (1999) Accounting for mode of in marine bivalves. Ecol. Lett. 4, 366 – 370
speciation increases power and realism of tests of phylogenetic 41 Brown, J.H. (1995) Macroecology, University of Chicago Press
asymmetry. Am. Nat. 153, 332 – 346 42 Brown, J.H. and Lomolino, M.V. (1998) Biogeography, Sinauer
14 Chesser, R.T. and Zink, R.M. (1994) Modes of speciation in birds – a Associates
test of Lynch’s method. Evolution 48, 490 – 497 43 Frey, J.K. (1992) Response of a mammalian faunal element to climatic
15 Frey, J.K. (1993) Modes of peripheral isolate formation and speciation. changes. J. Mammal. 73, 43 – 50
Syst. Biol. 42, 373 – 381 44 Parmesan, C. (1996) Climate and species’ range. Nature 382, 765 – 766
16 Friesen, V.L. and Anderson, D.J. (1997) Phylogeny and evolution of the 45 Parmesan, C. et al. (1999) Poleward shifts in geographical ranges of
http://tree.trends.com
 Opinion TRENDS in Ecology and Evolution Vol.18 No.5 May 2003 227

butterfly species associated with regional warming. Nature 399, 69 Heard, S.B. (1996) Patterns in phylogenetic tree balance with variable
579 – 583 and evolving speciation rates. Evolution 50, 2141– 2148
46 Warren, M.S. et al. (2001) Rapid responses of British butterflies to 70 Mooers, A.Ø. (1997) Inferring evolutionary process from phylogenetic
opposing forces of climate and habitat change. Nature 414, 65 – 69 tree shape. Q. Rev. Biol. 72, 31 – 72
47 Branco, M. et al. (2002) Postglacial dispersal of the European rabbit 71 Wiley, E.O. (1981) Phylogenetics: The Theory and Practice of
(Oryctolagus cuniculus) on the Iberian Peninsula reconstructed from Phylogenetic Systematics, John Wiley & Sons
nested glade and mismatch analyses of mitochondrial DNA genetic 72 Harrison, R.G. (1991) Molecular-changes at speciation. Annu. Rev.
variation. Evolution 56, 792 – 803 Ecol. Syst. 22, 281 – 308
48 Matocq, M.D. (2002) Phylogeographical structure and regional history 73 Avise, J.C. (2000) Phylogeography: The History and Formation of
of the dusky- footed woodrat, Neotoma fuscipes. Mol. Ecol. 11, 229– 242 Species, Harvard University Press
49 Gaston, K.J. (1998) Species-range size distributions: products of 74 Frumhoff, P.C. and Reeve, H.K. (1994) Using phylogenies to test
speciation, extinction and transformation. Philos. Trans. R. Soc. Lond. hypotheses of adaptation – a critique of some current proposals.
Ser. B 353, 219– 230 Evolution 48, 172 – 180
50 Webb, T.J. and Gaston, K.J. (2000) Geographic range size and 75 Ricklefs, R.E. and Starck, J.M. (1996) Applications of phylogenetically
evolutionary age in birds. Proc. R. Soc. Lond. Ser. B 267, 1843 – 1850 independent contrasts: a mixed progress report. Oikos 77, 167 – 172
51 Knowles, L.L. (2001) Did the Pleistocene glaciations promote 76 Price, T.D. et al. (1997) Evolution of breeding distributions in the old
divergence? Tests of explicit refugial models in montane grasshoppers. world leaf warblers (genus Phylloscopus). Evolution 51, 552– 561
Mol. Ecol. 10, 691 – 701 77 Nowell, K. and Jackson, P. (1996) Wild Cats: Status Survey and
52 Cunningham, C.W. et al. (1998) Reconstructing ancestral character Conservation Action Plan, IUCN
states: a critical reappraisal. Trends Ecol. Evol. 13, 361– 366 78 Brooks, D.R. and McLennan, D.A. (1993) Historical ecology: phylo-
53 Schluter, D. et al. (1997) Likelihood of ancestor states in adaptive genetic components of community evolution. In Species Diversity
radiation. Evolution 51, 1699 – 1711 in Ecological Communities (Ricklefs, R.E. and Schluter, D., eds)
54 Webster, A.J. and Purvis, A. (2002) Testing the accuracy of methods for pp. 267– 280, University of Chicago Press
reconstructing ancestral states of continuous characters. Proc. R. Soc. 79 Templeton, A.R. (1981) Mechanisms of speciation: a population genetic
Lond. Ser. B 269, 143– 149 approach. Annu. Rev. Ecol. Syst. 12, 23 – 48
80 Endler, J.A. (1992) Signals, signal conditions, and the direction of
55 Butler, M.A. and Losos, J.B. (1997) Testing for unequal amounts of
evolution. Am. Nat. 139, S125 – S153
evolution in a continuous character on different branches of a
81 Marchetti, K. (1993) Dark habitats and bright birds illustrate the role
phylogenetic tree using linear and squared-change parsimony: an
of the environment in species divergence. Nature 362, 149 – 152
example using lesser Antillean Anolis lizards. Evolution 51,
82 Rice, W.R. and Hostert, E.E. (1993) Laboratory experiments on
1623 – 1635
speciation – what have we learned in 40 years. Evolution 47,
56 Wiley, E.O. and Mayden, R.L. (1985) Species and speciation in
1637– 1653
phylogenetic systematics, with examples from the North American
83 Schluter, D. (2000) The Ecology of Adaptive Radiation, Oxford
fish fauna. Ann. Mo. Bot. Gard. 72, 569 – 635
University Press
57 Mayr, E. (1963) Animal Species and Evolution, Harvard University
84 Carson, H.L. and Templeton, A.R. (1984) Genetic revolutions in
Press
relation to speciation phenomena: the founding of new populations.
58 Mayden, R.L. et al. (1992) Speciation in the North American genus
Annu. Rev. Ecol. Syst., 15
Dionda (Teleostei: Cypriniformes). In Systematics, Historical Ecology,
85 Powell, J.R. (1978) The founder-flush speciation theory: an experi-
and North American Freshwater Fishes (Mayden, R.L. et al., eds),
mental approach. Evolution 32, 465 – 474
pp. 710– 746, Stanford University Press
86 Nordborg, M. (2001) Coalescent theory. In Handbook of Statistical
59 Bernatchez, L. and Dodson, J.J. (1991) Allopatric origin of sympatric
Genetics (Balding, D.J., ed.), pp. 179 – 212, John Wiley & Sons
populations of lake whitefish (Coregonus clupeaformis) as revealed by 87 Machado, C.A. et al. (2002) Inferring the history of speciation from
mitochondrial-DNA restriction analysis. Evolution 44, 1263– 1271 multilocus DNA sequence data: the case of Drosophila pseudoobscura
60 Schliewen, U.K. et al. (1994) Sympatric speciation suggested by and close relatives. Mol. Biol. Evol. 19, 472 – 488
monophyly of crater lake cichlids. Nature 368, 629 – 632 88 Kliman, R.M. et al. (2000) The population genetics of the origin and
61 Markert, J.A. et al. (1999) Biogeography and population genetics of the divergence of the Drosophila simulans complex species. Genetics 156,
Lake Malawi cichlid Melanochromis auratus: habitat transience, 1913– 1931
philopatry and speciation. Mol. Ecol. 8, 1013 – 1026 89 Wang, R.L. et al. (1997) Gene flow and natural selection in the origin
62 Van Oppen, M.J.H. et al. (1997) Unusually fine-scale genetic of Drosophila pseudoobscura and close relatives. Genetics 147,
structuring found in rapidly speciating Malawi cichlid fishes. Proc. 1091– 1106
R. Soc. Lond. Ser. B 264, 1803 – 1812 90 Wakeley, J. and Hey, J. (1998) Testing speciation models with DNA
63 Green, D.M. et al. (1997) Cryptic species of spotted frogs, Rana pretiosa sequence data. In Molecular Approaches to Ecology and Evolution
complex, in western North America. Copeia, 1 – 8 (DeSalle, R. and Shierwater, B., eds) pp. 157 – 175, Birkhäuser
64 Coyne, J.A. and Price, T.D. (2000) Little evidence for sympatric 91 Wakeley, J. and Hey, J. (1997) Estimating ancestral population
speciation in island birds. Evolution 54, 2166– 2171 parameters. Genetics 145, 847– 855
65 Losos, J.B. and Schluter, D. (2000) Analysis of an evolutionary species- 92 Hare, M.P. et al. (2002) Genetic evidence on the demography of
area relationship. Nature 408, 847 – 850 speciation in allopatric dolphin species. Evolution 56, 804– 816
66 Macey, J.R. et al. (1998) Phylogenetic relationships among agamid 93 Templeton, A.R. (2001) Using phylogeographic analyses of gene trees
lizards of the Laudakia caucasia species group: testing hypotheses of to test species status and processes. Mol. Ecol. 10, 779 – 791
biogeographic fragmentation and an area cladogram for the Iranian 94 Berlocher, S.H. (1998) Can sympatric speciation via host or habitat
Plateau. Mol. Phylogenet. Evol. 10, 118 – 131 shift be proven from phylogenetic and biogeographic evidence. In
67 Guyer, C. and Slowinski, J.B. (1993) Adaptive radiation and the Endless Forms: Species and Speciation (Howard, D.J. and Berlocher,
topology of large phylogenies. Evolution 47, 253 – 263 S.H., eds) pp. 99 – 113, Oxford University Press
68 Heard, S.B. (1992) Patterns in tree balance among cladistic, phenetic, 95 Knowles, L.L. and Maddison, W.P. (2002) Statistical phylogeography.
and randomly generated phylogenetic trees. Evolution 46, 1818– 1826 Mol. Ecol. 11, 2623 – 2635

http://tree.trends.com