Simultaneous Quaternary Radiations of Three Damselfly Clades Across

vol. 165, no. 4 the american naturalist april 2005

E-Article Simultaneous Quaternary Radiations of Three Damselﬂy Clades across the Holarctic

Julie Turgeon,1,2,* Robby Stoks,1,3,† Ryan A. Thum,1,4,‡ Jonathan M. Brown,5,§ and Mark A. McPeek1,k

1. Department of Biological Sciences, Dartmouth College, the evolution of mate choice in generating reproductive isolation as Hanover, New Hampshire 03755; species recolonized the landscape following deglaciation. These anal- 2. De´partement de Biologie, Universite´ Laval, Que´bec, Que´bec yses suggest that recent climate ﬂuctuations resulted in radiations G1K 7P4, Canada; driven by similar combinations of speciation processes acting in dif- 3. Laboratory of Aquatic Ecology, University of Leuven, Chemin ferent lineages. de Be´riotstraat 32, B-3000 Leuven, Belgium; 4. Department of Ecology and Systematics, Cornell University, Keywords: Enallagma, speciation, radiation, ampliﬁed fragment Ithaca, New York 14850; length polymorphism (AFLP), mtDNA, phylogeny. 5. Department of Biology, Grinnell College, Grinnell, Iowa 50112

Submitted October 22, 2004; Accepted December 27, 2004; The fossil record recounts recurrent cycles of mass ex- Electronically published February 9, 2005 tinction immediately followed by rebounds in biodiversity throughout Earth’s history (Jablonski 1986, 1994; Benton 1987; Raup 1991; Sepkoski 1991). A few of these events profoundly reshaped global biodiversity (e.g., the end- abstract: If climate change during the Quaternary shaped the Permian mass extinction erased up to 96% of the world’s macroevolutionary dynamics of a taxon, we expect to see three fea- species; Raup 1979; Jablonski 1994; but see Raup 1991), tures in its history: elevated speciation or extinction rates should date but most of these have been more limited in their taxo- to this time, more northerly distributed clades should show greater nomic scope (Raup 1991). Climatic upheavals over the discontinuities in these rates, and similar signatures of those effects past 1.65 million years have had similar, though less severe, should be evident in the phylogenetic and phylodemographic his- effects on the world’s biota. Repeated glacial advances and tories of multiple clades. In accordance with the role of glacial cycles, speciation rates increased in the Holarctic Enallagma damselflies dur- retreats during the Pleistocene caused shifting and frag- ing the Quaternary, with a 4.25# greater increase in a more northerly mentation of species ranges across the globe to the extent distributed clade as compared with a more southern clade. Finer- that many fossil assemblages 10,000–20,000 years ago have scale phylogenetic analyses of three radiating clades within the north- little resemblance to extant assemblages, even though ern clade show similar, complex recent histories over the past 250,000 many of the same species still exist (Overpeck et al. 1992; years to produce 17 Nearctic and four Palearctic extant species. All Holman 1993; Coope 1995; Graham et al. 1996; see also three are marked by nearly synchronous deep splits that date to reviews by Hewitt 1996, 2004). These events drove extinct approximately 250,000 years ago, resulting in speciation in two. This was soon followed by significant demographic expansions in at least many species that presumably could not respond rapidly two of the three clades. In two, these expansions seem to have pre- enough to both the direct consequences of changes in the ceded the radiations that have given rise to most of the current physical environment and indirect consequences resulting biodiversity. Each also produced species at the periphery of the clade’s from interacting with different species (Kurteń and An- range. In spite of clear genetic support for reproductive isolation derson 1980; Martin and Klein 1984). Along with extinc- among almost all species, mtDNA signals of past asymmetric hy- tions, climatic fluctuations also produced range fragmen- bridization between species in different clades also suggest a role for tations, shifts, bottlenecks, and subsequent expansions, the consequences of which presumably generated ideal geo- * E-mail: [email protected]. graphic, demographic, ecological, genetic, and social con- † E-mail: [email protected]. ditions to promote speciation in taxa as well (e.g., Carson ‡ E-mail: [email protected]. and Templeton 1984; Vrba 1985; Coyne and Orr 2004). § E-mail: [email protected]. However, the importance of recent climatic oscillations k Corresponding author; e-mail: [email protected]. to the macroevolutionary dynamics of various taxa re- Am. Nat. 2005. Vol. 165, pp. E78–E107. ᭧ 2005 by The University of Chi- mains controversial. For example, the mammal fossil rec- cago. 0003-0147/2005/16504-40696$15.00. All rights reserved. ord shows substantial faunal turnover and recent and rapid Worldwide Radiation of Damselflies E79 speciation in many taxa (Kurteń and Anderson 1980; Lister rates, whereas taxa at lower latitudes should show less 2004), but analyses of molecular distances between sister change. Moreover, taxa whose macroevolutionary dynam- taxa in birds have failed to support a recent upturn in ics were shaped by the same climate event should all show avian speciation rates (Klicka and Zink 1997, 1999; Zink coincident signatures of those dynamics in multiple clades, et al. 2004). Likewise, fossil and subfossil collections of if more than one clade survived (e.g., altered diversification beetles show little change over the Quaternary and are rates, range fragmentation, bottlenecks, and expansions; largely composed of still extant species (Coope 2004), but Rogers and Harpending 1992; Rogers 1995). Speciation phylogenetic analyses of the Cicindela tiger beetles do show and extinction rates can be evaluated directly for taxa with a weak trend for elevated speciation rates during the last a good Quaternary fossil record (e.g., beetles: Coope 1995, million years (Barraclough and Vogler 2002; see also 2004; mammals: Kurteń and Anderson 1980; Lister 2004), Knowles 2000, 2001; Knowles and Otte 2000 for another but for most taxa that do not have adequate fossil records, insect example). No general statement can be made about indirect methods must be utilized (e.g., Nee et al. 1992, the predominant responses of plants to climatic changes 1994). during the Quaternary either; the dominant response of In this article, we examine the diversification history of arborescent species was extinction rather than speciation the Enallagma damselflies (Zygoptera: Coenagrionidae) or stasis, but speciation rates of vascular plants often in- across the Holarctic to test whether Quaternary climate creased during periods of glaciation (Willis and Niklas change shaped the macroevolutionary dynamics of this 2004). clade. First, we test the degree to which species that have Clearly, not all taxa have responded similarly to recent been shown in previous studies to share mtDNA haplo- climatic oscillations. Many different factors may influence types are reproductively isolated from one another by an- whether and how taxa are affected by such climatic events. alyzing a large amplified fragment length polymorphism First and foremost, biogeography would seem to be crucial; (AFLP) data set. Next, we add sequences from four Pa- taxa inhabiting higher latitudes should be more affected learctic Enallagma species and one more Nearctic species by glacial events than taxa at lower latitudes. Phylogeo- to the data set of Brown et al. (2000) and reanalyze this graphic studies suggest this to be the case, with northerly larger data set to test whether speciation and extinction distributed species showing strong phylogeographic signals rates in the genus have changed recently and whether the of recent range fragmentation and expansions, whereas magnitudes of discontinuities differ between more north- more southerly distributed species show phylogenetic sig- erly and more southerly distributed clades. Finally, we annals characteristic of stable long-term occupancy of their alyze a much larger mtDNA data set to reconstruct the ranges (Bernatchez and Wilson 1998; Hewitt 2004; Martin phylogenetic and phylodemographic histories of the three and McKay 2004). Vagility should also strongly influence radiating clades. If these three clades radiated in response the tempo of range shifts and the degree of isolation that to recent climatic events, we expect similar phylogenetic may be generated from range fragmentations (Coope 1995, and phylodemographic patterns across the three clades. 2004). Vagile taxa may respond to climate change by rapidly shifting their ranges, while less vagile taxa may be more susceptible to extinction and may be most likely to The Study System diversify following climatic upheavals (Dynesius and Jans- son 2000; Jansson and Dynesius 2002; but see Bouchard The Enallagma damselflies provide an amazing example and Brooks 2004). Species inhabiting relatively uncommon of a taxon that appears to have very recently radiated to or ephemeral environmental conditions may also be more produce continent-wide faunas. With 38 described species, likely to be driven extinct than species inhabiting relatively Enallagma is one of the two most speciose odonate genera common environment types. Similarly, species with broad in North America (the anisopteran genus Gomphus also ecological tolerances are expected to be less affected than has 38 North American species; Westfall and May 1996). taxa with more narrow ecological needs (e.g., Knowles Previous phylogenetic analyses of the Nearctic species sug- 2000). gest that 17 are derived from two progenitors that radiated Understanding whether taxa were influenced macro- very recently (Brown et al. 2000; fig. 1). Throughout the evolutionarily by glacial cycles during the Quaternary re- article we will identify these two radiating lineages as the quires that we undertake focused comparisons of related Nearctic “hageni” and “carunculatum” clades (fig. 1). The clades that show various degrees of response, since taxo- local species richness of this clade is maximal in the New nomically broad surveys may obscure critical differences England region of the United States (Connecticut to among groups. If glacial cycles shaped the macroevolu- Maine), which was deglaciated beginning approximately tionary dynamics of some taxa, we should first see that 15,000 years ago, but species in this clade can be found taxa at higher latitudes have greatly altered diversification across the Nearctic and into the Neotropics (Westfall and E80 The American Naturalist

Figure 1: Phylogenetic hypothesis for 38 of the 42 Enallagma species found across the Holarctic. The hypothesis was generated by a neighbor- joining analysis of the data set produced by Brown et al. (2000) plus sequences for Enallagma novaehispaniae, Enallagma circulatum, Enallagma cyathigerum, Enallagma deserti, and Enallagma risi. The basal split divides the genus into a Northern and a Southern clade. Within the Northern clade, our analyses focus on the three radiating groups—the four species in the Palearctic clade, the 10 species in the Nearctic “hageni” clade, and the seven species in the Nearctic “carunculatum” clade.

May 1996; McPeek and Brown 2000; Donnelly 2004). In These species are Enallagma cyathigerum (ranging across contrast, the other primary clade of the genus has highest northern Eurasia from Iberia to Kamchatka), Enallagma local species richness in the southeastern United States, risi (central Asia), Enallagma circulatum (Japan and ad- but representatives have ranges stretching north into jacent coastal Asian areas), and Enallagma deserti (north- southern Canada and west to the Great Plains (Saskatch- west Africa; Askew 1988; Westfall and May 1996). Our ewan to Texas), with one species (Enallagma basidens) hav- analyses below (fig. 1) support the arguments of Paulson ing recently expanded its range to cover much of the con- et al. (1998) to elevate the Palearctic Enallagma boreale tinent (Donnelly 2004). This primary clade also shows circulatum to E. circulatum. Also, up to now, E. cyathigerum several recent speciation events, but most lineages are de- was thought to be a single species with a circumpolar rived from much older speciation events (fig. 1). Previ- distribution. However, as we show below, the Nearctic and ously, we have called these the “Northern” and “Southern” Palearctic entities that share the specific epithet “cyathig- clades, respectively, for their biogeographic diversity pat- erum” are really two species—one in each region. There- terns, and we will use these monikers here (fig. 1). fore, based on precedence, we will throughout this article In addition to these Nearctic species, four Enallagma refer to the Palearctic species as E. cyathigerum and the species can be found in the Palearctic (Askew 1988), but Nearctic species as Enallagma annexum (M. L. May, per- their phylogenetic affinities up to now have been untested. sonal communication). Worldwide Radiation of Damselflies E81

Material and Methods ping Kit (Applied Biosystems, Foster City, CA). Three primer pairs were used for selective amplification: the Eco Damselflies were collected across the Holarctic by collab- RI-ACA primer labeled with the FAM fluorochrome was orators and ourselves (tables 1, A1; Brown et al. 2000). paired with the Mse I-CAC, Mse I-CTA, and Mse I-CAA Only adult males were included in the analyses considered primers to establish the pairs. Amplified products were in this article because they could be unambiguously as- run on an ABI 3100 sequencer, and AFLP peaks were signed to morphospecies based on the morphology of their categorized and scored using the ABI Genotyper version cerci, the clasping structures used by males to hold females 3.0 software. We identified a total of 1,170 polymorphic during mating (Westfall and May 1996). Females use the AFLPs across the 22 species. The AFLP peak profiles were tactile cues from the shapes of cerci to identify males dur- highly repeatable both in estimated size and fluorescence ing courtship (Paulson 1974; Robertson and Paterson intensity. 1982). All individuals were either preserved in ethanol or Ϫ Њ In phylogenetically distant lineages, some subsets of dried and stored at 80 C before being included in the AFLP fragments of similar length are unlikely to be iden- study. We have constructed three molecular data sets: tical by descent (Vekemans et al. 2002). Therefore, we have AFLPs for all but three species in the Northern clade (miss- analyzed the two groups of species identified in previous ing are Enallagma circulatum, Enallagma deserti, and En- mtDNA analyses (Brown et al. 2000) separately. First, we allagma novaehispaniae); the mtDNA data set of Brown et used PAUP, version 4.0b10 for Windows (Swofford 2001), al. (2000) with added sequences from E. novaehispaniae to construct 1,000 pseudoreplicate neighbor-joining trees and the four Palearctic species; and a very large data set while assuming the log-determinant model (Lockhart et of mtDNA sequences for the Northern clade species. al. 1994; Steele 1994) to test whether species formed monophyletic groups. Analyses by UPGMA and maximum par- Nuclear DNA: AFLPs simony algorithms yielded nearly identical topologies and so are not reported here. We also performed reallocation Quantifying extinction and speciation rates among clades procedures using the method and software designed for is open to error when not all defined taxa reflect true AFLP data by Duchesne and Bernatchez (2002; AFLPOP) biological species. Therefore, we first tested whether rec- to reallocate individuals to species based on the allele fre- ognized morphospecies based on cerci morphology cor- quencies measured in each morphologically defined spe- responded to mutually reproductively isolated entities cies; using species allelic frequencies, each individual is based on classifications using nuclear AFLPs. Previous reallocated to the species in which its genotype is most studies identified two groups of Nearctic species that have likely to belong. We also performed analyses of molecular a large degree of unsorted polymorphisms for mtDNA variance (Excoffier et al. 1992) to test specific hypotheses (Brown et al. 2000). This unsorted polymorphism within concerning the degree of genetic differentiation among each group could exist for at least three reasons: species species and among localities. For these analyses, we in- are recently diverged and have not had time to establish cluded only populations and species for which we had a mutually reciprocal monophyly in mtDNA; species are minimum of four individuals from at least two sites. Hy- largely reproductively isolated from one another but hypotheses involving hierarchical structures were analyzed bridize to some degree; or only two species truly exist, using Arlequin, version 2.0 (Schneider et al. 2000), and each of which is polymorphic in cerci morphology. hypotheses requiring other statistical designs were coded For AFLP analyses, we included a total of 376 individ- directly in Matlab, version 6.5 (Mathworks 2002). We em- uals from 64 populations among 17 Nearctic species and phasize that our only goal for these AFLP analyses was to 26 individuals from seven populations of two Palearctic establish whether morphologically defined species are re- species (tables 1, A2). Two or more populations were in- productively isolated. We are presently conducting a more cluded for every species except three (Enallagma davisi, detailed analysis of population structure for selected taxa Enallagma doubledayi, Enallagma praevarum), and we gen- with much greater sample sizes and geographically inten- erally genotyped four to 20 individuals per population per sive sampling. species. Our goal for this work was not to examine population structure within species but rather to determine Mitochondrial DNA whether morphospecies are distinct genetic entities. Many of these individuals were also used in the mtDNA sequence Speciation and Extinction Rates. We first extended and analyses described in the next sections. reanalyzed the data set of Brown et al. (2000) to estimate DNA was extracted from individuals using standard speciation and extinction rates and test whether speciation phenol methods. The AFLP markers were developed using rates accelerated sometime during the Quaternary in the the manufacturer’s instructions with the AFLP Plant Map- two primary Enallagma clades. The data set of Brown et Table 1: Number of populations, number of individuals, and population locations for the 22 Enallagma species included in the mtDNA or AFLP studies AFLPs mtDNA No. No. No. No. Species populations individuals Population locations populations individuals Population locations Enallagma durum 2 6 Connecticut, Rhode Island 4 19 Connecticut, Massachusetts, Rhode Island Enallagma annexum 5 18 Michigan, New Hampshire, Alaska, Colorado 14 31 New Hampshire, Vermont, Idaho, Washington, British Columbia, Alberta, Ontario, Michigan Enallagma boreale 6 20 Alaska, Colorado, California, New Brunswick, New 11 31 Vermont, Alberta, California, Michigan, Idaho, Con- Hampshire, Michigan necticut, British Columbia, New Brunswick Enallagma clausum 4 11 California, Utah, Manitoba 4 6 California, Manitoba, Utah, Nevada Enallagma davisi 1 4 North Carolina 1 2 North Carolina Enallagma ebrium 7 82 Maine, Michigan, New Hampshire, Vermont, 27 83 Idaho, Alberta, Manitoba, Prince Edward Island, British Manitoba Columbia, Saskatchewan, Michigan, Ontario, Wiscon- sin, New York, New Hampshire, Vermont, Pennsylva- nia, Maine, Quebec Enallagma hageni 10 107 Maine, Michigan, Nova Scotia, Vermont, Manitoba 30 138 Vermont, Alberta, Minnesota, Iowa, Prince Edward Is- land, Maine, Wisconsin, British Columbia, Nova Sco- tia, Michigan, Ontario, Saskatchewan, Quebec, Manitoba Enallagma laterale 2 18 Pennsylvania, Maine 7 27 Connecticut, Pennsylvania, New Jersey, Maine, Massachusetts Enallagma minusculum 3 12 Nova Scotia, Massachusetts, Maine 4 21 Massachusetts, Maine, Nova Scotia, New Brunswick Enallagma recurvatum 2 8 New Jersey, Massachusetts 4 15 Massachusetts, New Jersey Enallagma vernale 2 7 New Hampshire, Vermont 2 3 New Hampshire Enallagma anna 2 8 California, Utah 3 5 Utah, California, Nevada Enallagma aspersum 4 16 Rhode Island, Michigan, Vermont, Maine 5 36 Nova Scotia, New Hampshire, Rhode Island, Maine, Michigan Enallagma carunculatum 3 13 Idaho, California, Wisconsin 5 10 Michigan, California, Wisconsin, New Mexico, Arizona Enallagma civile 5 24 Maryland, Rhode Island, Nova Scotia, California, 9 29 Rhode Island, Massachusetts, Michigan, Vermont, New Hawaii Hampshire, Pennsylvania, Nova Scotia Enallagma doubledayi 1 2 Florida 3 9 Florida, South Carolina, Massachusetts Enallagma geminatum 4 15 Vermont, Ohio, Maine, Connecticut 9 42 Rhode Island, Connecticut, Pennsylvania, Ohio, New Jersey, New York, Michigan, Massachusetts Enallagma praevarum 1 5 New Mexico 2 2 Texas, New Mexico Enallagma cyathigerum 5 18 Spain, Belgium, Sweden, Russia (Yakutia and 7 13 Russia, Sweden, Spain, Poland, Kamchatka) Enallagma deserti 0 0 … 1 2 Algeria Enallagma risi 2 8 China, Mongolia 5 8 China, Kazakhstan, Russia Enallagma circulatum 0 0 … 1 2 Japan Total 71 402 … 158 534 … Note: More detailed lists of genetic and locality information are given in tables A1 and A2. Worldwide Radiation of Damselflies E83

2 p ! al. (2000) consisted of two to three sequences of 842-bp x1 51.8,P .001 ). To test whether violating this as- mtDNA fragments spanning the cytochrome oxidase I and sumption made any difference to our conclusions, we con- II genes and the intervening leucine tRNA from each of structed trees from the full range of possible substitution 33 Nearctic Enallagma. We have added two sequences from models (from the F81 model with no rate heterogeneity one other Nearctic species (E. novaehispaniae, a species in to the fullGTR ϩ G ϩ I model with all parameters esti- the southwestern United States and Mexico) and a number mated from the data set; Swofford et al. 1996) both as- of sequences from four Palearctic species (E. circulatum, suming and not assuming a molecular clock. We found Enallagma cyathigerum, E. deserti, Enallagma risi; table 1). strong correlations among branch lengths from all these This comprises 38 of the 42 species thought to belong to analyses (all with Pearson product moment correlations Enallagma. ofr 1 0.93 ). The correlation was alsor p 0.93 between A neighbor-joining analysis of the entire data set first the branch lengths of trees from theGTR ϩ G ϩ I models determined the placement of these five new species within assuming and not assuming a molecular clock. In addition, the genus. We used sequences from two species in the we used branch length estimates from a number of sub- sister genus Ischnura (Ischnura posita and Ischnura ram- stitution models in the methods described below and burii) as outgroups to root the tree (Brown et al. 2000). found that the substitution model had no effect on the This analysis indicated that E. novaehispaniae is basal in conclusions drawn from these analyses of speciation and the Northern clade, and the four Palearctic species form extinction rates. The only differences were that the esti- a clade within the Northern clade (fig. 1). mated times for deeper splits in the tree were estimated To estimate speciation and extinction rates across the (as expected) to be farther in the past for more complex genus using the methods of Nee et al. (1994), we needed substitution models. We, therefore, feel that our results to estimate branch lengths for the Enallagma phylogeny are robust. Unless otherwise stated, analyses reported here assuming a molecular clock. We used PAUP to construct use branch lengths derived using theGTR ϩ G ϩ I sub- maximum likelihood trees using various molecular substitution model and assuming a molecular clock (fig. 1). stitution models (Swofford 2001). To speed calculations, From the tree in figure 1, we calculated internode dis- we randomly chose one sequence to include for each spe- tances and used these to test for heterogeneity in diver- cies, and we constrained the analysis to consider only trees sification rates through time and between the Northern with branching structures found in previous analyses and Southern clades. We first calculated g (Pybus and (Brown et al. 2000; McPeek and Brown 2000). The con- Harvey 2000), which tests whether speciation and extinc- straint trees forced the branching of the major subclades tion rates remain constant through time. The g has a in the Northern clade as in figure 1 (i.e., E. novaehispaniae standard normal distribution, withg 1 0 implying that basal, followed by the Nearctic carunculatum clade, En- rates speed up near the present; a one-tailed test was as- allagma durum, Nearctic hageni clade, and the Palearctic sumed. We then applied the maximum likelihood analyses clade; see fig. 1) but left taxa unresolved within each. We of Nee et al. (1994) to estimate the speciation and ex- put no credence in the resulting branching patterns within tinction rates under various models of rate heterogeneity each clade because most species share unsorted polymor- across the tree. In the most parameter-rich model, we phisms, and in fact these analyses assigned zero lengths to assumed that the Northern and Southern clades differed many internal branches within each of these clades (fig. in speciation and extinction rates and that these rates si- 1). A few of the basal splits in the Southern clade are also multaneously changed in both clades at some point in the not well supported (Brown et al. 2000; McPeek and Brown past; this gives four speciation and four extinction rates 2000). We present results with the Southern clade con- and the time of change between the two sets. Simpler strained as in the article by Brown et al. (2000), but we models were constructed by equating various rates (e.g., repeated this analysis with different branching orders at the simplest diversification model equates all speciation the base of the Southern clade and found that the resulting rates and equates all extinction rates, giving only one spe- trees made no difference to the conclusions about diver- ciation and one extinction rate for the entire genus). The sification rates. Therefore, we present results only for the Solver algorithm of Microsoft Excel was used to estimate tree in figure 1. parameters. We used Akaike’s Information Criterion to Analyses using Modeltest, version 3.06 (Posada and choose the best-fitting model (Akaike 1974). We applied Crandall 1998), indicated that theGTR ϩ G ϩ I nucleo- Brower’s (1994) insect mtDNA molecular clock calibration tide substitution model (Swofford et al. 1996) was most of 2.3% per million years for mtDNA substitutions to appropriate for this data set. Furthermore, a molecular estimate times of molecular divergences. clock could not be statistically supported for the data (likelihood ratio test for trees assuming theGTR ϩ G ϩ I sub- Phylogenetics and Phylodemography of the Three Radiating stitution model with and without a molecular clock Northern Clades. In a previous article (Turgeon and E84 The American Naturalist

McPeek 2002), sequences were analyzed from 283 indi- mutations (Fu 1997). We used Arlequin (Schneider et al. viduals of five Enallagma species (Enallagma ebrium, En- 2000) to estimate these parameters for the mismatch dis- allagma hageni, Enallagma laterale, Enallagma minusculum, tributions and to calculate Fu’s Fs statistic. We also esti- and Enallagma recurvatum). To these data, we have added mated the average genetic distances between clades, cor- sequences from 251 individuals of 17 species for this anal- recting for the effects of bottlenecks and unequal ysis to produce a data set containing sequences from all population sizes (Gaggiotti and Excoffier 2000, imple- species in the Northern clade except E. novaehispaniae. mented in Arlequin). For most species, we have obtained sequences from populations that cover large areas of their ranges (table 1). These sequences were obtained for the same ∼1,000-bp Results mtDNA fragment analyzed in the previous section (see Nuclear DNA: AFLPs Brown et al. 2000; Turgeon and McPeek 2002 for methods). The final data set here contained unambiguous se- For all three damselfly clades, neighbor joining (NJ), quences of 868 bp. UPGMA, and maximum parsimony (MP) analyses of the Based on analyses of this phylogenetically more limited AFLP data sets using all genotyped individuals yielded data set using Modeltest (Posada and Crandall 1998), we similar topologies. In the carunculatum clade, all individ- assumed the Tamura (1992) model of nucleotide substi- uals clearly grouped by species based on their AFLP ge- tution with unequal base frequencies and a gamma shape notypes (fig. 2a; note that for presentation clarity and parameter of 0.7705. Analyses with alternative phyloge- brevity, fig. 2a and 2b show only one randomly chosen netic methods (parsimony, distance, and maximum like- individual from each population of each species. However, lihood methods) all resulted in similar trees that yielded our narrative here describes analyses including all indi- almost no supportable resolution within each of the three viduals.) Moreover, the phylogenetic hypotheses produced radiating clades. To explore the phylogenetic relationships by NJ and MP were largely consistent with the mtDNA among haplotypes within each of these clades, we con- haplotype network for this clade (see “Mitochondrial structed haplotype networks using the TCS parsimony al- DNA”); only the placement of Enallagma doubledayi gorithm (Posada et al. 2000) and minimum spanning net- (based on a single population) was inconsistent with the works using Arlequin (Schneider et al. 2000), which gave mtDNA haplotype network (cf. figs. 2a,4a). Other ana- nearly identical networks. The 95% confidence limit for lytical techniques also confirmed that individuals in the network connection in TCS was 13 steps, and so the net- carunculatum clade group by species. Analyses of molec- works for the various clades were unconnected. ular variance (AMOVA) also showed substantial differ- Given the strong support for reproductive isolation entiation among the carunculatum clade species (P ! among most species (see “Results”), the degree of shared .001; 34.5% of total variation attributable to differences haplotypes among species and the lack of phylogenetic among species). Likewise, all individuals were perfectly resolution within clades suggested that these clades radi- reallocated to their species using AFLPOP; in all cases, the ated very recently. To explore the demographic events and next choice of species for reallocation was much less likely the timing of these radiations, we computed mismatch (log-likelihood differences ranging between 63 and 193). distributions for each species separately. The frequency In contrast, only a subset of the species in the hageni distribution of pairwise differences between haplotypes clade are genetically differentiated (fig. 2b). Individuals of (i.e., the mismatch distribution) is expected to be a ragged Enallagma clausum and the four coastal plain endemics but monotonically declining function if the recent de- (Enallagma davisi, Enallagma laterale, Enallagma minus- mographic history of the species has been stable (Watter- culum, Enallagma recurvatum) all group by species in phy- son 1975; Rogers and Harpending 1992). However, recent logenetic analyses (fig. 2b), and AMOVA identifies sub- and rapid demographic expansions produce a wave in the stantial species differentiation among them (P ! .001 ). In shape of this distribution. The position of the wave crest contrast, individuals of Enallagma annexum, Enallagma bo- is estimated by the parameter t and can be used to date reale, and Enallagma vernale form a single group, indi- the expansion (Rogers and Harpending 1992). Again, we cating reproductive isolation from other species in the used Brower’s (1994) insect mtDNA molecular clock cal- clade, but these species do not form distinct sublineages. ibration to estimate these dates. Other signals for recent Rather, their populations cluster together by geography, demographic expansions include less raggedness of the not by species (AMOVA, species differences:P 1 .25 ; pop- mismatch distribution, as measured by Harpending’s rag- ulation differences within species:P ! .03 ; fig. 2b). Like- gedness index (Harpending 1994), and a significant neg- wise, individuals of Enallagma hageni and Enallagma ative value for Fu’s Fs statistic, which characterizes pop- ebrium form a single group isolated from all other species, ulations with an excess of rare haplotypes and young but populations of different species living in the same Worldwide Radiation of Damselflies E85

Figure 2: Neighbor-joining trees derived from 1,170 amplified fragment length polymorphism (AFLP) loci for the (a) Nearctic “carunculatum” clade, the (b) Nearctic “hageni” clade, and (c) Enallagma cyathigerum and Enallagma risi of the Palearctic clade. Analyses that include only one randomly chosen individual from each sampled population are shown in a and b; analyses including all individuals in the data set give identical results (see table A2 for locality and sample size information). Panel c shows all genotyped individuals. Locations for Nearctic individuals are given by U.S. state and Canadian provincial postal codes. Bootstrap values (1,000 replicates) 150% are shown. region are less genetically differentiated than conspecific Enallagma vernale individuals were more often reallocated populations from different areas (AMOVA, species differ- to E. boreale (67%) than to itself. Again, these results par- ences:P 1 .13 ; population differences within species: P ! allel those of figure 2b. .004; fig. 2b). Our AFLP analysis also indicates that pop- We were only able to obtain AFLPs for Enallagma cy- ulations of the Atlantic and Continental mtDNA haplotype athigerum and Enallagma risi populations in the Palearctic clades (identified in Turgeon and McPeek 2002; see “Mi- clade. Although E. cyathigerum and E. risi are not recip- tochondrial DNA”) of E. hageni are not differentiated rocally monophyletic, E. risi does form a group nested 1 ().P .25 within the paraphyletic E. cyathigerum in the NJ analyses Reallocation analyses with ALFPOP for the hageni clade (fig. 2c). The two E. risi populations from Inner Mongolia yielded results that agree with those obtained by phylo- are most similar to the Yakutia population of E. cyathig- genetic clustering methods. Individuals of E. clausum and erum, which is to their northeast. European E. cyathigerum the four coastal endemics were all reallocated to their spe- populations form another well-supported clade. The Kam- cies without ambiguity (100%; range of log-likelihood dif- chatka population of E. cyathigerum forms the basal clade ferences for the next most probable species: 33–175). Real- location, however, was imperfect between E. hageni and in this analysis (Alaskan E. annexum population was used E. ebrium as well as among E. boreale, E. vernale, and E. toroottheNJtreeinfig.2c). Reallocation analyses with annexum. While most individuals of E. hageni were cor- AFLPOP also correctly reassigned all E. risi and E. cy- rectly reallocated (90%), individuals of E. ebrium were athigerum individuals to species (range of log-likelihood allocated to E. hageni nearly as often (42%) as they were differences: two to 20 for E. risi and three to 51 for E. to E. ebrium, a result that is in agreement with the clus- cyathigerum). Given that Enallagma deserti is only found tering in figure 2b. Among the other three species, E. bo- in north Africa, and Enallagma circulatum is similarly al- reale and E. annexum were either correctly reallocated lopatric from E. cyathigerum and E. risi and well differ- (68% and 50%, respectively) or allocated to one another. entiated from the other three species in mtDNA (see “Mi- E86 The American Naturalist

Figure 2: (Continued) tochondrial DNA”), we consider the four Palearctic species arately. The shapes of these taxon accumulation curves to be reproductively isolated from one another. suggest a rapid increase in diversification rate in the very recent past (Kubo and Iwasa 1995). Analysis of internode distances using the methods of Pybus and Harvey (2000) Mitochondrial DNA yielded a value ofg p 5.16 (P ! .0001 ), which strongly Speciation and Extinction Rates. Due to the ambiguous supports this conclusion. species status identified in the AFLP analyses, we have Estimates of speciation and extinction rates also support pruned E. boreale, E. vernale, and E. ebrium from the phy- a rapid and recent increase in speciation rate, particularly logeny in figure 1 to estimate speciation and extinction in the Northern clade. Because of the statistical difficulty rates; that is, for this analysis we assume that E. hageni in testing a particular date for diversification rate heter- and E. ebrium are one “species” and that E. annexum, E. ogeneity (e.g., Barraclough and Vogler 2002), we defined boreale, and E. vernale are one “species.” Applying Brower’s a priori the speciation and extinction rate discontinuity to (1994) molecular clock calibration to the tree given in occur sometime in the period between the common an- figure 1 provides an estimate of the basal split in the genus cestor of the Palearctic and the Nearctic hageni clades and at about 15 million years ago (the comparable estimate the deepest split within any of the three radiating Northern derived from the tree assuming the simplest F81 substi- clades (see fig. 1). This places the rate discontinuity some- tution model gives about 9 million years ago; we infer that where between 1.26 and 0.51 million years ago, assuming the true value is bracketed by these numbers). Figure 3 theGTR ϩ G ϩ I substitution model (or between 1.14 and presents the accumulation of taxa through time for the 0.46 million years ago, assuming the simplest F81 substi- entire tree and for the Northern and Southern clades sep- tution model). Using Akaike’s Information Criterion, the Worldwide Radiation of Damselflies E87

Figure 2: (Continued) best-fit model of diversification has all extinction rates species form the fourth well-supported clade placed equal to 0, the same speciation rates in the two clades among the Nearctic clades. Phylogenetic resolution within before the discontinuity, and different speciation rates in the three multispecies clades was very poor, as documented the two clades after the discontinuity (table 2). Moreover, in previous studies (Brown et al. 2000; Turgeon and speciation rates in both clades increase after the discon- McPeek 2002). The TCS and minimum spanning networks tinuity, with a 4.25# greater increase for the Northern of the three multispecies haplotype clades show similar clade (table 2, parameter combination e; the same con- phylogenetic patterns: each is subdivided into at least two clusions are obtained if we fix the discontinuity time any- subclades, with one of these subclades subsequently ra- where between 2.50 and 0.25 million years ago.) diating (fig. 4). The Palearctic clade has Enallagma circulatum separated Phylogenetics and Phylodemography of the Three Radiating from the rest of the network by nine mutational steps, Northern Clades. To examine the phylogenetic and phy- and within the other subclade, haplotypes are wide ranging lodemographic patterns within each of the radiating and do not cluster by species (fig. 4a). In the Nearctic Northern clades, we obtained a total of 139 unique se- carunculatum clade, two subclades are separated by six quences among 534 individuals collected from 158 differ- mutational steps. One subclade contains only Enallagma ent populations among 22 of the 23 Northern clade species geminatum haplotypes, and the other contains haplotypes (excepting Enallagma novaehispaniae; table 1). Represen- of the remaining species (fig. 4b). Within this second sub- tative sequences have previously been deposited in GenBank clade, Enallagma aspersum haplotypes cluster together, but (accession numbers: AF064985–AF065040, AF512684). As the remaining haplotypes do not cluster by species, and in previous analyses (Brown et al. 2000; McPeek and two haplotypes were found in multiple species. Enallagma Brown 2000), these mtDNA sequences form four related praevarum haplotypes may also segregate within the net- haplotype clades (fig. 1). Haplotypes of Enallagma durum, work, forming another subclade, but limited sampling an endemic of the eastern North American coast, form makes this conclusion uncertain. one clade. The two other Nearctic clades (fig. 1) contain The Nearctic hageni clade also shows genetic subdivision haplotypes from multiple species, and representatives of into two subclades separated by six mutational events. each clade can be found across almost the entire North Previously, we characterized the two hageni subclades as American continent. Haplotypes from the four Palearctic “Continental” and “Atlantic” based on their geographic E88 The American Naturalist

Figure 3: Taxon accumulation curves for the Northern and Southern clades and the entire phylogeny shown in figure 1. The abscissa gives the age of nodes, after converting genetic distances into time, based on Brower’s (1994) molecular clock calibration. The ordinate is given on a loge scale. As references, the long-dashed vertical line identifies the date of the beginning of the Quaternary period (1.8 million years ago), and the short- dashed vertical line identifies the date of the beginning of the Wisconsinan glaciation (0.24 million years ago). distributions (fig. 4c; Turgeon and McPeek 2002). We had whereas E. geminatum populations from outside this area found Atlantic haplotypes only in E. hageni within a nar- have only carunculatum haplotypes (fig. 4b,4c). Likewise, row geographic range (Connecticut to Saco River water- E. aspersum populations from Rhode Island to Nova Scotia sheds of New England). Here, more extensive sampling have mixtures of carunculatum and both Continental and has identified two Atlantic haplotypes in four E. ebrium Atlantic hageni haplotypes but only carunculatum haplo- from Saskatchewan (all other E. ebrium haplotypes, from types outside this area (fig. 4b,4c). In contrast, we have Prince Edward Island to British Columbia, have been found no carunculatum haplotypes in hageni clade species members of the Continental clade), whereas E. hageni from in this geographic area despite intensive sampling. In the the same area of Saskatchewan have Continental haplo- western part of our sampling range, we also found hageni types (fig. 4c). Unlike the other two clades, genetic sub- haplotypes in two Enallagma carunculatum and two En- division did not result in speciation, because E. hageni and allagma anna from northern California (fig. 4c), suggesting E. ebrium are represented in both subclades. However, another local area of apparent asymmetric hybridization. given the extent to which all other species share haplotypes, Haplotype mismatch distributions also revealed signals the Continental subclade did radiate very recently. of recent rapid population expansions in both the radiating Another striking feature of the Nearctic hageni clade and nonradiating Nearctic subclades (table 3). Because of network is that several haplotypes from this clade were their ambiguous species status, table 3 presents mismatch found in individuals of species typically belonging to the analyses assuming that E. annexum,E.boreale, and E. ver- carunculatum clade (haplotypes shown in gray boxes in nale, and E. hageni and E. ebrium represent five biological fig. 4c). This apparent asymmetric hybridization was ex- species and assuming that these are only two biological tensive for populations of E. aspersum and E. geminatum species. For the hageni clade, E. annexum, E. boreale, and from northeastern North America. Enallagma geminatum both the Continental and Atlantic haplotype groups within populations from New Jersey to Rhode Island have E. hageni show haplotype mismatch patterns consistent mixtures of carunculatum and Atlantic hageni haplotypes, with recent population expansions: all show raggedness Worldwide Radiation of Damselflies E89

Table 2: Parameters and statistics for the rates of accumulation of taxa in the Northern and Southern clades Parameters No. free Ϫ parameters North l1 North l2 North m1 North m2 South l1 South l2 South m1 South m2 log(LIK) AIC p p p p p p p a, 1 24.4 Nl1 0 Nm1 Nl1 Sl1 Nm1 Sm1 161.0 320.0 p p p p p p b, 2 133.7 Nl1 135.5 Nm1 Nl1 Sl1 Nm1 Sm1 177.1 350.2 p p p p p p c, 2 33.6 Nl1 0 Nm1 17.8 Sl1 Nm1 Sm1 162.7 321.3 p p p p p p d, 2 10.7 141.87 0 Nm1 Nl1 Sl1 Nm1 Sm1 185.5 367.1 p p p p p e, 3 10.7 255.8 0 Nm1 Nl1 60.7 Nm1 Sm1 190.1 374.3 p p p p f, 4 10.7 255.8 .0 Nm1 Nl1 60.7 Nm1 Sm1 190.1 372.3 p p p p g, 4 7.9 255.8 0 Nm1 12.8 60.7 Nm1 Sm1 190.5 372.9 h, 8 8.9 490.5 3.1 379.4 12.8 91.8 .0 247.8 191.0 366.0 Note: These analyses are based on the internodes in the tree given in figure 1 and were calculated using branch lengths expressed in molecular distances. 6 Speciation rates before (l1) and after (l2) and extinction rates before (m1) and after (m2)10 years ago were estimated for each clade using the maximum likelihood techniques derived by Nee et al. (1994). Parameter estimates, likelihood scores, and Akaike’s Information Criterion (AIC) for the best-fit model for various parameter combinations are given. For each model, free parameters are given by their estimated numbers, and fixed parameters are identified p p p p by “ constraint” (e.g., “ 0” means constrained to equal zero, “ Nl1” means constrained to equal North l1).LIK likelihood .

indices close to 0; all have large negative values for Fu’s Discussion

Fs statistic; and all except the Atlantic hageni subclade have their modal mismatch class 10 (table 3; fig. 5a). The peaks Our analyses of diversification rates across the entire genus in mismatch distributions estimate the time of the most clearly show that both primary clades of Enallagma were rapid expansion (t). The relative positions of peaks for influenced by climatic upheaval during the Quaternary. these species suggest that E. annexum, E. boreale, and the For most of their histories, the Northern and Southern Continental subclade in E. hageni all expanded at about clades had very similar diversification rates, but speciation the same time (table 3). Applying Brower’s (1994) mo- rates in both clades have greatly increased sometime in lecular clock estimate suggests that these expansions all the last million years (fig. 3; table 2). This increase was date well before 15,000 years ago but at or after the clades’ much more pronounced in the Northern clade, where the ∼ # subdivisions, and so probably the date of most rapid ex- estimated speciation rate increased by 24 , as compared ∼ # pansion is following subdivision (table 3). The Atlantic E. with an increase of 6 in the Southern clade. This dif- hageni subclade shows signatures of a much more recent ference is due primarily to the disparity in the number of expansion (table 3); the confidence interval for its expan- new species created per progenitor in each clade. In the sion time is consistent with expansion as the glacier re- Northern clade, an average of seven extant species was derived from each progenitor that diversified (i.e., 21 ex- treated 15,000 years ago. The three coastal plain endemics tant species from three progenitors). In contrast, for the in this clade show no genetic signals of population ex- speciation events occurring during the same time in the pansion (table 3). Southern clade, an average of only 2.25 extant species was In the carunculatum clade, E. geminatum, E. aspersum, produced for each progenitor (nine extant species from and E. doubledayi have significant negative values for Fu’s four progenitors). Three of these four Southern clade spe- F and raggedness indices near 0, suggesting that these s ciation events resulted in a widespread species that now three species also expanded recently (table 2; fig. 5b). Con- ranges across much of eastern North America and a species fidence intervals for their expansion times are also con- that is restricted to the Atlantic coastal plain (McPeek and sistent with expansion as the glacier retreated (table 3). Brown 2000). The fourth was the splitting of a progenitor We saw no signature of expansions in the Palearctic taxa, into three species that all now inhabit this same coastal but our sample sizes are too small to conclude this with plain. This comparison first highlights the importance of confidence. the Atlantic coastal plain to the Nearctic diversification of In contrast, E. durum, the one major nonradiating clade Enallagma; in addition to the new Southern clade species, that is sister to the three radiating clades, shows the classic four derived hageni clade species and one derived carun- pattern of a lineage with long-term demographic stability: culatum clade species are Atlantic coastal plain endemics. highest mismatch frequency class at 0 (fig. 5c), a ragged- More importantly, these gross differences between the ness index substantially larger than 0, and a small value Northern and Southern clades suggest biogeography as for Fu’s Fs statistic that is not different from 0 (table 3). critically mediating the effects of climate change on the E90 The American Naturalist macroevolutionary dynamics of clades (reviewed in Jans- (Hagen 1861; M. L. May, personal communication) and son and Dynesius 2002). Both the Northern and Southern “Enallagma boreale” should be divided into the Palearctic clades showed a recent elevation in speciation rate, and as Enallagma circulatum Selys 1883 and the Nearctic E. bo- expected, the Northern clade showed a substantially reale Selys 1875 (Paulson et al. 1998). greater increase than the Southern clade. Many of the While analyses of mtDNA haplotypes revealed extensive Southern clade species have ranges that extend from the shared polymorphisms among many species within each Gulf of Mexico to southern Canada, including the three Nearctic clade, our analyses of nuclear polymorphisms widespread species in the singleton speciation events (En- (AFLP) established that genome-wide differentiation has allagma traviatum, Enallagma signatum, Enallagma ves- accumulated among almost all species. This was particu- perum). Presumably, these species ranges were contracted larly true among species of the carunculatum clade, in south during periods of glaciation, and they have simply which all individuals cluster according to morphological extended their ranges north as ecological conditions al- taxonomy. In the hageni clade, many species were also well lowed (cf. Edmands 2001). Evidence for such range con- differentiated from one another; however, Enallagma hag- traction and expansion has been found over the same area eni and Enallagma ebrium did not form distinct genetic in the salamander Gyrinophylus porphyriticus, which ranges clusters within the sublineage that they jointly form, nor along the Appalachian mountains from Georgia to Maine did E. annexum, E. boreale, and Enallagma vernale. Hy- (W. H. Lowe, in preparation). Gyrinophylus populations bridization among incompletely reproductively isolated from West Virginia south show high mtDNA haplotype species may explain this lack of genetic differentiation. diversity, whereas populations north of West Virginia are Indeed, our AFLP data provide evidence that conspecific almost monotypic for one haplotype, suggesting a very populations from different geographic areas are more ge- rapid recolonization of this area following deglaciation. In netically differentiated from one another than from het- contrast, our analyses of the radiating Northern clades erospecific populations in the same region. Enallagma hag- show strong signals of range fragmentation, followed by eni and E. ebrium co-occur in lakes from Nova Scotia to demographic expansions involving the entire ranges of British Columbia (Walker 1953; McPeek 1989, 1990b), and species (Turgeon and McPeek 2002; this study), which is a putative hybrid between these two species, based on characteristic of species at higher latitudes that would have sexual morphology, was recently reported (Catling 2001). been severely affected by glacial advances and retreats (Ber- Likewise, E. annexum and E. boreale also co-occur in lakes natchez and Wilson 1998; Hewitt 2004). Moreover, our across much of the continent (Walker 1953; McPeek 1989, results decipher three complex yet similar and nearly si- 1990b). However, these data are also consistent with other multaneous evolutionary scenarios all involving past ge- interpretations, given our limited sampling thus far for netic divergence, demographic expansions, and recent ra- both groups (four to 20 individuals per population) in the diation of three progenitors to produce the extant AFLP data set. At one extreme, species within each group Enallagma assemblage across the Holarctic. may have diverged very recently in multiple locales, and the morphospecies we see today are completely reproductively isolated (cf. Schluter and Nagel 1995). At the Genetic Relationships and Reproductive other extreme, each group may be only one species, with Isolation among Species a polymorphism of cerci and mesostigmal plate. More de- Our genetic analyses confirm the existence of two inde- tailed genetic studies are currently under way to more pendently radiating Nearctic clades (Brown et al. 2000) thoroughly evaluate these hypotheses. Nevertheless, con- and a third radiating clade in the Palearctic that is sister clusions about the phylogenetics and phylodemographics to the Nearctic hageni clade. Because the Palearctic clade of these radiations do not depend on the species status of is paraphyletic within a much larger Nearctic clade, Pa- these taxa. learctic Enallagma appear to have originated from a relatively recent invasion of the Old World by an ancestor Northern Clade Radiations across the Holarctic of the Nearctic hageni clade. The split between these two clades dates to approximately 1.0–1.3 million years ago, Comparisons of haplotype network structures of the three depending on the substitution model assumed. Moreover, radiating clades also revealed striking similarities in their the segregation of Holarctic taxa currently sharing specific phylogenetic and phylodemographic histories. Although epithets into long-separated Nearctic and Palearctic dates for genetic subdivisions and demographic expansions mtDNA clades suggests that they should be split into Ne- are only approximate, all three radiating clades were arctic and Palearctic species; that is, “Enallagma cyathig- marked by a major genetic split occurring approximately erum” should be divided into the Palearctic E. cyathigerum 250,000 years ago, a date that corresponds roughly to a (Charpentier 1740), and the Nearctic Enallagma annexum glacial period in North America (Martinson et al. 1987;

E92 The American Naturalist

Bennett 1997). Then, each of the radiating subclades un- this clade is either E. hageni or E. ebrium. The greater derwent significant demographic expansions clearly pre- pairwise nucleotide diversity in E. hageni/E. ebrium than dating the expansions experienced by nonradiating sub- in other clade members also suggests one of the two as clades (table 3). the progenitor of the hageni clade. Given the extent to Within the carunculatum clade, relationships among which all 10 species share haplotypes, the Continental sub- haplotypes suggest that speciation events first cleaved En- clade must have explosively radiated very recently. We have allagma geminatum in eastern North America (based on also identified phylogeographic signals of recent range ex- its current range) and probably Enallagma praevarum in pansion in both E. hageni and E. ebrium, presumably as- the southwest from a much more widely distributed pro- sociated with colonization of deglaciated areas because genitor about 250,000 years ago (based on the number of their current ranges would have been almost completely fixed differences between the clades and correcting for covered with ice 15,000 years ago (Turgeon and McPeek genetic diversity and unequal population sizes; Gaggiotti 2002). and Excoffier 2000). This was followed by speciation pro- The recent rise in speciation rates (fig. 3) and nearly ducing Enallagma aspersum about 110,000 years ago (again synchronous events of genetic subdivisions and demo- in eastern North America based on current range) and graphic expansions within the radiating clades strongly finally the radiation of this progenitor very recently to give suggest the importance of environmental changes asso- the remaining four species. Enallagma carunculatum and ciated with Pleistocene glaciation in sparking the radiations Enallagma civile both now have distributions spanning of these clades. Moreover, the phylodemographic similar- much of North America. Enallagma anna is a western ities among these radiating clades stand in stark contrast species, and Enallagma doubledayi is found only on the to the fourth clade that did not radiate—Enallagma durum, Atlantic coastal plain. Furthermore, E. carunculatum’s which displays the classic genetic signatures of long-term greater average pairwise differences among haplotypes as demographic stability. We hypothesize that this contrast compared to E. civile suggests E. carunculatum as the pu- may be due to the unique association of E. durum with tative ancestral species of this clade. brackish water and thus its coastal distribution (Westfall Likewise, the Palearctic clade has E. circulatum separated and May 1996). The range of E. durum probably shifted from the rest of the network by nine mutational steps. to follow the coastline as sea levels rose and fell over the Within the other subclade, haplotypes are wide ranging last 250,000 years, while the other clades were fragmented and do not cluster by species (fig. 2b). Again, one species and bottlenecked. The demography of E. durum would first separated from a widely distributed progenitor, fol- have been relatively stable, and its range would have re- lowed by a much more recent splitting of one subclade mained relatively intact, in spite of its distribution shifting into multiple species. on the landscape because of habitat tracking. This contrast The Nearctic hageni clade also showed genetic subdi- indicates that ecological characteristics have important vision and subsequent radiation of only one subclade (see biogeographic consequences, which in turn determine the also Turgeon and McPeek 2002), but unlike the other two effects of climatic changes on diversification. clades, this initial range fragmentation of E. hageni/E. ebrium into Continental and Atlantic subclades (dating to Mechanisms of Speciation the same time as the E. geminatum speciation event) did not result in a completed speciation event. This range These results put into temporal context the proposed fragmentation, which must predate the radiation of the mechanisms generating new species in these radiations. In Continental clade, suggests that the putative progenitor of eastern North America, where centrarchid species domi-

Figure 4: Minimum spanning networks for the three radiating clades of Enallagma damselflies: (a) Palearctic clade, (b) Nearctic “carunculatum” clade, and (c) Nearctic “hageni” clade. Boxes represent haplotypes identified in the study, open circles are haplotypes not identified but inferred for the network, and lines connect haplotypes differing by one nucleotide substitution (solid lines for unambiguous connections and dashed lines when more than one connection is possible for a haplotype). Species are identified by letter codes (ann p Enallagmaanna, asp p Enallagma aspersum, car p Enallagmacarunculatum,, civ p Enallagma civile dou p Enallagmadoubledayi, gem p Enallagmageminatum, pra p Enallagma praevarum, anx p Enallagmaannexum, bor p Enallagmaboreale, cla p Enallagmaclausum,, dav p Enallagma davisi ebr p Enallagmaebrium, hag p Enallagmahageni, lat p Enallagmalaterale, min p Enallagmaminusculum, rec p Enallagma recurvatum,,ver p Enallagma vernale cya p Enallagmacyathigerum, risi p Enallagmarisi, des p Enallagmadeserti, cir p Enallagma circulatum), and Nearctic locations where haplotypes were collected are identified by U.S. state and Canadian provincial postal codes (see table A1 for more detailed locality and sample size information; fig. 4c here is an expansion of fig. 4 in Turgeon and McPeek 2002; note that the labels for E. hageni and E. ebrium were inadvertently reversed in fig. 4 in Turgeon and McPeek 2002). Haplotypes of the carunculatum clade species are shaded gray to help identify these haplotypes in the hageni clade network. Figure 4: (Continued) E. hageni (years ago) Expansion time mean (95% CI) 1.67) 51,913 (3,245–82,196) enced mtDNA fragment to the expansion Ϫ ) t ( Expansion time mean (95% CI) s .69 … … .61.52 … … … … .34 … … .15 … … F 8.13** 3.30 (1.30–4.83) 162,082 (64,253–237,445) 4.34* 4.36 (1.69–7.52) 214,290 (83,130–369,735) 8.29*** 1.06 (.07 6.22*** 1.14 (.00–3.25) 56,239 (0–160,115) 2.36* 1.20 (.00, 2.60) 59,091 (0, 127,866) 1.38 … … 1.23 … … 7.53*** 1.63 (.04,1.75* 2.13) 80,181 (1,770, .50 104,466) (.04, 1.04) 24,728 (1,819, 51,078) 1.08 … … Ϫ Ϫ Ϫ Ϫ Ϫ 18.95*** 3.97 (2.13–6.29) 186,072 (104,515–309,268) 25.74*** 4.93 (2.19–7.98) 242,508 (107,415–392,742) 22.55*** 4.49 (2.11–6.37) 220,828 (104,122–313,201) Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ have been excluded from these analyses. index raggedness E. anna Harpending’s values. Also, because the AFLP analyses suggested extensive hybridization between , and s F .54.89.44 .16 .50 .17 2.11 … … .68 .06 .83 .75 1.10 … … .62 .05 .97 .25 .77 .11 .67 … .20 … … .39 .18 .34 .49 1 SD) 1.51 … 1.721.73 .51 1.67 … … .69 … … … 1.69 .02 1.58 .04 1.82 .01 1.83 .04 1.90 .03 1.59 .06 1.641.01 .13 .07 .70 … … 2.02 .07 1.11 .07 ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע ע E. carunculatum , (mean Pairwise differences E. aspersum , Unique haplotypes , these statistics were also calculated when combining haplotypes for these groups. All metrics are derived from Arlequin, version E. geminatum 3 3 2.00 6 32 2.81 2 2.00 2121 …21 … … … … … … … … … … … … … … 9 5 1.06 8 3 1.50 38 2 6 .67 3.55 65 28 3.23 31 16272115 2.93 4 3 3 .68 1.38 .48 31 14 3.47 29 6 1.59 13 6 2.93 2927 13 4 1.89 .41 19 3 .32 No. E. vernale individuals , and E. boreale ” clade haplotypes in , clade haplotypes found in these species. hageni E. vernale a , Atlantic clade 34 11 .98 , Continental clade 188 43 3.58 E. annexum hageni , and a a , Continental clade 76 13 3.02 , Continental clade 112 36 3.74 a E. ebrium E. ebrium E. boreale and between , and and , Atlantic clade 4 2 1.00 , Atlantic clade 30 9 .85 .001 Genetic and demographic statistics derived from the mismatch distributions for the seven haplotype clades and subclades shown in figure 3 .01 ! .05 ! ) estimates. Expansion times are not estimated for species that did not have significantly negative P ! t P P E. ebrium These analyses exclude ** . Note: The expansion time (years ago) was calculated by applying Brower’s (1994) 2.3% divergence/million years molecular clock estimate for the sequ ***a . *. Enallagma vernale E. annexum Enallagma clausum Enallagma recurvatum Enallagma boreale Enallagma laterale Enallagma minusculum Enallagma davisi Enallagma geminatum Enallagma annexum E. hageni E. ebrium E. hageni Enallagma civile Enallagma carunculatum Enallagma ebrium E. hageni Enallagma circulatum Enallagma deserti Enallagma cyathigerum Enallagma praevarum Enallagma doubledayi and time ( Table 3: Species and haplotypes Enallagma hageni Enallagma aspersum Enallagma anna 2.0 (Schneider et al. 2000). Also, “ Enallagma risi Enallagma durum

E94 Figure 5: Mismatch distributions among mtDNA haplotypes for species in the (a)“hageni,” (b)“carunculatum,” and (c) Palearctic clades. The distribution for Enallagma durum, the Nearctic lineage that did not radiate, is also shown in c. Statistics derived from these distributions are given in table 3. E96 The American Naturalist nate the fish fauna, most Enallagma species are found as aspersum suggests that this hybridization occurred some larvae only in water bodies with fish as the top predators, time ago (i.e., hageni clade alleles in E. aspersum and E. but four (E. annexum, E. aspersum, E. boreale, E. double- geminatum have had time to accumulate unique muta- dayi) are found as larvae only in water bodies where in- tions). Our results also identify northern California as an- sectivorous fish are absent and large dragonfly larvae are other potential area of past asymmetric hybridization be- the top predators (Johnson and Crowley 1980; McPeek tween these two clades, since we also found hageni 1990b, 1998). Species segregate because phenotypic dif- haplotypes in E. carunculatum and E. anna. ferences between the two groups make them differentially Unidirectional mitochondrial hybridization is fre- vulnerable to fish and dragonfly predators (McPeek 1990a, quently seen when females of one species are locally rare 1995, 1999, 2000, 2004; Stoks et al. 2003). If E. hageni or and cannot find suitable (i.e., conspecific) males and even- E. ebrium was the progenitor of the hageni clade and if E. tually acquiesce to mating with males of the locally com- carunculatum was the progenitor of the carcunculatum mon species (Wirtz 1999; Randler 2002). This hybridi- clade, three independent evolutionary habitat shifts oc- zation must have occurred as E. hageni recolonized areas curred to produce today’s dragonfly lake species: E. as- of northeastern North America already occupied by E. persum and E. doubledayi in the carunculatum clade, and geminatum and E. aspersum, two species that arose earlier the lineage with E. annexum and E. boreale in the hageni in the radiation. Previous analyses of these data have also clade. These habitat shifts were accomplished by parallel shown a strong phylogeographic signal of range expansion adaptive changes in morphology, physiology, and behavior for E. hageni from the center of the continent toward this to increase survival in the face of dragonfly predation area (Turgeon and McPeek 2002). This hybridization im- (McPeek 1995, 1999, 2000, 2004; McPeek et al. 1996; Stoks plies that small founder populations of E. hageni were et al. 2003). being created in this area and that mate choice was vul- However, most new species from these radiations in- nerable to change in these populations, a scenario that is habit only fish lakes, which also appear to be the ancestral consistent with the importance of mate choice evolution habitat for the genus (all species in the Southern clade are in founder populations to speciation (Lande 1981; Gav- found only in fish lakes; McPeek and Brown 2000). More- rilets and Boake 1998; Meffert and Regan 2002). Inter- over, many species in the hageni and carunculatum clades estingly, this is also the area where three of the four en- now co-occur in fish lakes across much of North America demic hageni clade species (Enallagma laterale, Enallagma (Johnson and Crowley 1980; McPeek 1990b, 1998), and minusculum, Enallagma recurvatum) presumably arose and these species are phenotypically and ecologically very sim- now co-occur. Sexual selection has generated new species ilar to one another (McPeek 1990a, 1998, 1999, 2000; in many diverse and widespread clades (Barraclough et al. McPeek et al. 1996; McPeek and Brown 2000; Stoks et al. 1995; Cuervo and Møller 1999; Arnqvist et al. 2000; Stuart- 2003). We postulate that these fish lake species differen- Fox and Owen 2003). Sexual conflict (Holland and Rice tiated primarily in characters important to sexual selection 1998; Parker and Partridge 1998; Gavrilets 2000; Gavrilets and mate recognition, that is, the male cerci, which are and Waxman 2002) may also promote diversification by morphological structures of males used to grasp females rapidly evolving differences among populations in isolated during mating, and the female mesostigmal plates, which fragments, which may be further enhanced during sec- are the structures used by females to evaluate the shapes ondary contact (Parker and Partridge 1998; Kawata and of the cerci of grasping males in order to discriminate Yoshimura 2000). among mates (Paulson 1974; Robertson and Paterson The literature is replete with molecular studies that iden- 1982). tify recently radiating clades (Hodges and Arnold 1994; Colonization of new areas during times of range ex- Knox and Palmer 1995; Johns and Avise 1998; Roderick pansion also may have promoted conditions conducive to and Gillespie 1998; Lovette and Bermingham 1999; Sher- diversification (Lande 1981; Carson and Templeton 1984; bakov 1999; Knowles and Otte 2000; Sturmbauer et al. Gavrilets and Boake 1998; Regan et al. 2003; but see Coyne 2003; Verheyen et al. 2003; Wilder and Hollocher 2003; and Orr 2004). In fact, our analysis has identified genetic Gillespie 2004), some with spectacular increases in species signatures of the mating systems of some species briefly richness (Albertson et al. 1999; Allender et al. 2004; and breaking down recently. Enallagma hageni was apparently see overviews in Schluter 2000; Hewitt 2004). However, hybridizing with both E. aspersum and E. geminatum in conflict over the general influence of Pleistocene climatic northeastern North America in the recent past (fig. 2c). changes on species diversification may arise from both The directionality of introgression suggests that hybridi- methodological and taxonomic biases. Ribera and Vogler zation was primarily by E. hageni females mating with E. (2004) suggest that, for beetles, the conflict between phy- aspersum and E. geminatum males. In addition, the prev- logenetic studies, which support hypothesis of rapid Pleis- alence of private hageni alleles in E. geminatum and E. tocene radiations, and studies of the fossil record (e.g., Worldwide Radiation of Damselflies E97

Coope 2004), which support stasis, may be a reflection of cichlid lineages in Lake Malawi that have established sim- the geographic bias of fossils toward higher latitudes; their ilar diversities of nuptial coloration across multiple clades studies of endemic Iberian diving beetles indicate that cli- (Allender et al. 2003). mate fluctuations produced the vicariance processes re- Clearly, different clades responded differently to the cli- sponsible for speciation. (Note that in this case, they pro- matic upheavals during the Quaternary. Many clade-level posed that higher latitude populations are less likely to differences may account for why some were affected and speciate because range contraction does not isolate north- others were not. Obviously, the distributions of clades ern taxa in refugia.) Use of genitalic structures alone to across the landscape played a crucial role; clades at higher delineate species may also miss significant diversification; latitudes would have been more impacted simply because for example, Carisio et al. (2004) demonstrated that sig- they would have experienced greater climatic fluctua- nificant “subspecific” diversification occurred in Mediter- tions—as we saw in the Enallagma. However, how and ranean dung beetles during the Pleistocene due to isolation why clades responded to such changes may be more par- in patchily distributed habitats, and geographic differen- ticular to the specific features of different clades. Changing tiation among subspecies persists. climates may be more likely to drive members of some We believe that comparisons of closely related clades clades with particular features extinct, simply shift the can help resolve this issue, as well as considerations of the ranges of others, and spark diversification in still others. roles played by biogeography, ecology, and different spe- Focused studies comparing clades with different responses ciation modes in mediating these effects. Few, however, will allow us to determine how biogeography may interact have compared related clades in order to compare their with the ecological and phenotypic properties of various macroevolutionary dynamics, particularly in the case of clade members to determine these responses. Such studies how Quaternary climate change may have driven diver- will be invaluable as guides to conservation efforts as we sification. One such study by Price and colleagues ex- try to anticipate species’ responses to climate change in amined the diversification of warblers in the Himalayan the future. mountains of south Asia and in the White Mountains of northeastern North America (Price et al. 1998, 2000). As Acknowledgments in this study, they found that the clade at higher latitudes (i.e., the North America clade) radiated recently, whereas We owe a tremendous debt of gratitude to all the people the clade farther from the influences of recent glaciation (listed in tables A1 and A2) who sent us damselflies from contained only older species (Price et al. 1998). They also around the Holarctic. We would also like to thank G. suggest that sexual selection played a significant role in Hewitt, W. Lowe, and three anonymous reviewers for their the radiation of the North American clade. Chapple and valuable comments that helped to greatly clarify our pre- Keogh (2004) have also examined how sister clades of sentation. J.T. was supported by a postdoctoral fellowship skinks (Squamata) radiated to fill arid and temperate hab- and a research grant from the Natural Sciences and En- itats in Australia, and clade comparisons have shown how gineering Research Council of Canada. R.S. was supported Caribbean Anolis lizards have recapitulated similar eco- by a postdoctoral fellowship and grants from the Fund for logical types on multiple islands (Losos et al. 1998). One Scientific Research Flanders (Belgium). M.A.M. was sup- spectacular example illustrating the importance of mating ported by National Science Foundation grant IBN- signal evolution is the parallel diversification of multiple 0130021 and funds from Dartmouth College.

APPENDIX

Sampling Localities

Table A1: Sampling localities for mtDNA analyses giving collection year and collector (if not an author) for each species included in the mtDNA study Year Haplotypes Species collected Locality information (number) Collector Enallagma durum 1996 Worden Pond, Route 110, South Kingston, Rhode 545 (3) Island, USA E. durum 2001 Sachem Pond, Block Island, Rhode Island, USA 545 (5), 546 (1) E. durum 2001 Lake Quannapowitt, Wakeﬁeld, Massachusetts, USA 545 (5) E98 The American Naturalist

Table A1 (Continued) Year Haplotypes Species collected Locality information (number) Collector E. durum 2001 Canal on Ferry Road, Old Sambrook, Connecticut, 545 (5), 547 (1) USA Enallagma annexum 1995 Ferson Road Marsh, Hanover, New Hampshire, 001 (1) USA E. annexum 1999 Beaver Pond, Bear Brook State Park, Suncook, 001 (2) New Hampshire, USA E. annexum 2000 Beaver Pond, Moose Mountain, Etna, New 001 (1) Hampshire, USA E. annexum 2000 Beaver Pond, Killington, Vermont, USA 022 (1) E. annexum 2000 Brush Lake, Boundary County, Idaho, USA 022 (1), 063 (3), 066 (1) D. Paulson E. annexum 1999 Langendorfer Lake, King County, Washington, USA 022 (1), 034 (1) D. Paulson E. annexum 2000 Tabor Lake, Prince George, British Columbia, 022 (2) L. Ramsay and Canada S. Cannings E. annexum 2000 Chickakoo Lake, Stony Plain, Alberta, Canada 034 (1), 069 (1), 071 (1), J. Acorn 072 (1), 073 (1) E. annexum 2000 Ketmark no. 1, Brooks, Alberta, Canada 034 (1) C. Rice E. annexum 1999 Davis Lake, Ferry County, Washington, USA 034 (2) D. Paulson E. annexum 2000 Minor Lake, Algonquin Provincial Park, Ontario, 070 (1) Canada E. annexum 2001 Auke Lake, Juneau, Alaska, USA 034 (2), 083 (1) J. Hudson E. annexum 2000 Crescent Lake, Pend Oreille County, Washington, 067 (1) USA E. annexum 2000 Pond Laboratory, Kellogg Biological Station, 032 (3), 068 (1) Hickory Corners, Michigan, USA Enallagma boreale 2000 Beaver Pond, Killington, Vermont, USA 001 (1), 007 (1), 064 (1) E. boreale 2000 Basford Basin 1, Brooks, Alberta, Canada 022 (1) J. Acorn E. boreale 2003 Honey Lake, Susanville, California, USA 022 (1) E. boreale 2000 Gull Lake, Aspen Beach, Alberta, Canada 022 (2), 080 (1), 081 (1) J. Acorn E. boreale 2000 Little Brevort Lake, Brevort, Michigan, USA 065 (3), 075 (1) E. boreale 2000 Pond Laboratory, Kellogg Biological Station, 074 (1), 032 (2), 064 (1) Hickory Corners, Michigan, USA E. boreale 2000 Brush Lake, Boundary County, Idaho, USA 077 (1), 034 (2), 078 (1), D. Paulson 076 (1) E. boreale 2000 Wood Creek Pond, Norfolk, Connecticut, USA 065 (2) E. boreale 2000 Little Cranberry Lake, Valemount, British 034 (3), 059 (1) L. Ramsay and Columbia, Canada S. Cannings E. boreale 2000 Rock Lake, Brooks, Alberta, Canada 079 (1) C. Rice E. boreale 2000 Gaspisie Site 1, Cloridorme, New Brunswick, 064 (2) Canada Enallagma clausum 2003 Eagle Lake, Susanville, California, USA 034 (2) E. clausum 2003 Lake Winnipegosis, Winnipegosis, Manitoba, 034 (1) M. Hughes Canada E. clausum 2003 Bear River Migratory Bird Refuge, Box Elder 034 (1) E. Pilgrim County, Utah, USA E. clausum 1996 Walker Lake, Mineral, Nevada, USA 022 (2) J. Simpkin Enallagma davisi 2001 Pine Lake, Marston Road, Marston, North Carolina, USA Enallagma ebrium 2000 Lake Benewah, Benewah County, Idaho, USA 001 (2), 057 (2) D. Paulson E. ebrium 2000 Ketmark 1, Brooks, Alberta, Canada 001 (2) C. Rice E. ebrium 2000 Chickakoo Lake, Stony Plain, Alberta, Canada 001 (3), 061 (1), 062 (1) J. Acorn E. ebrium 2003 Ochre River, Lake Dauphin, Manitoba, Canada 001 (1), 084 (1) M. Hughes E. ebrium 2000 Save Easy, Alberton, Prince Edward Island, Canada 009 (1) Worldwide Radiation of Damselﬂies E99

Table A1 (Continued) Year Haplotypes Species collected Locality information (number) Collector E. ebrium 2000 Pelican Lake, Prince George, British Columbia, 034 (2), 059 (1) L. Ramsay and Canada S. Cannings E. ebrium 2000 Murdoch Lake, Prince George, British Columbia, 034 (3), 058 (1), 060 (1) L. Ramsay and Canada S. Cannings E. ebrium 2001 Trapper’s River, Prince Albert National Park, 054 (3), 055 (1) G. Hutchings Waskesiu, Saskatchewan, Canada E. ebrium 2000 Pond Laboratory, Kellogg Biological Station, 008 (2) Hickory Corners, Michigan, USA E. ebrium 2001 George Reserve, Pinckney, Michigan, USA 008 (2) E. ebrium 2000 Lake Katcheqanooka, Lakefield, Ontario, Canada 009 (1) C. Jones E. ebrium 2000 Sawyer Creek, Lakefield, Ontario, Canada 008 (2), 009 (1) C. Jones E. ebrium 2001 Highway 2 Pond, Brule, Wisconsin, USA 008 (1) R. duBois E. ebrium 2001 Gilbert Lake, Brule State Forest, Solon Springs, 008 (1) R. duBois Wisconsin, USA E. ebrium 2000 Mosher Pond, Route 104, Oswego County, Lake 008 (1), 009 (3) Oneida, New York, USA E. ebrium 2000 Moss Lake, Houghton, New York, USA 008 (1), 009 (4) E. ebrium 1999 Hadlock Lake, West Fort Ann, New York, USA 008 (1), 009 (4) E. ebrium 1999 Beaver Pond, Moose Mountain, Etna, New 008 (1), 009 (2), 056 (1) Hampshire, USA E. ebrium 1999 McDaniel’s Marsh, Enfield, New Hampshire, USA 009 (4) E. ebrium 1999 Ninevah Lake, Grahamville, Vermont, USA 009 (2) E. ebrium 2000 Bresee Pond, Hubbardton, Vermont, USA 009 (4) E. ebrium 2000 Elfin Lake, Wallingford, Vermont, USA 009 (1) E. ebrium 2000 Hough Lake, Sudbury, Vermont, USA 009 (4) E. ebrium 2000 State Game Area 18, south of Erie, Pennsylvania, 009 (4) USA E. ebrium 2001 Dolby Pond, East Millisocket, Maine, USA 008 (2), 019 (2) E. ebrium 2000 Drew Pond, Highway 110, Meecors, Maine, USA 009 (2) P. Brunelle E. ebrium 2000 Lac Romulus, ZEC de la Blanche, Quebec, Canada 019 (4) Enallagma hageni 2000 Spectacle Pond, Island Pond, Vermont, USA 001 (1), 011 (1), 015 (1), 016 (1) E. hageni 2000 Chickakoo Lake, Stony Plain, Alberta, Canada 001 (4), 034 (1) J. Acorn E. hageni 2000 Low Lake, Ely, Minnesota, USA 022 (4), 027 (1) J. M. Brown E. hageni 2000 Save Easy, Alberton, Prince Edward Island, Canada 022 (2), 023 (1), 025 (1) E. hageni 2000 East Lake at Route 16, Prince Edward Island, 021 (1), 022 (2), 024 (1) Canada E. hageni 1995 Lovewell Pond, Fryeburg, Maine, USA 036 (15), 037, 038 (2), 039, 040, 04, 044 E. hageni 2000 Ticklenaked Lake, Boltonville, Vermont, USA 036 (2), 043 (2) E. hageni 2000 Amhearst Lake, Plymouth Vermont, USA 036 (1), 042 (2) E. hageni 2000 Martin Pond, Green Bay, Vermont, USA 009 (1), 011 (3), 036 (2) E. hageni 2000 Little Salem Pond, Newport, Vermont, USA 008 (1), 017 (1), 011 (2) E. hageni 2001 Minnesuing Lake, Brule State Forest, Solon Springs, 003 (1) R. duBois Wisconsin, USA E. hageni 2001 Gilbert Lake, Brule State Forest, Solon Springs, 002 (2) R. duBois Wisconsin, USA E. hageni 2000 Murdoch Lake, Prince George, British Columbia 034 (1) L. Ramsay and S. Cannings E. hageni 2000 Dolby Pond, East Millisocket, Maine, USA 009 (2) E. hageni 2001 Nyanza, Nova Scotia, Canada 007 (3) E. hageni 2001 Little Mushamush Lake, Middle North Cornwall, 012 (1), 013 (2), 014 (1) Nova Scotia, Canada E100 The American Naturalist

Table A1 (Continued) Year Haplotypes Species collected Locality information (number) Collector E. hageni 2001 Swan Creek, New Brunswick, Canada 011 (3) E. hageni 2000 Kennedy Lake, Manistique, Michigan, USA 007 (3), 035 (1), 027 (2), 032 (1) E. hageni 2000 Little Brevort Lake, Brevort, Michigan, USA 032 (1), 034 (1) E. hageni 2000 Strouble Lake, Epoutette, Michigan, USA 031 (2), 032 (1), 033 (1) E. hageni 2000 3 Lakes II, Richland, Michigan, USA 027 (2), 028 (1), 030 (1), 031 (4) E. hageni 2000 Bab Lake, Algonquin Provincial Park, Ontario, 009 (2), 010 (1) C. Jones Canada E. hageni 2000 Crane Creek Bog, Buckhorn, Ontario, Canada 009 (2) C. Jones E. hageni 2000 Crystal Lake, Starbuckville, Horicon, New York, 009 (15), 026 (3) USA E. hageni 2001 Trapper’s River, Prince Albert National Park, 004 (2), 005 (1), 006 (1) G. Hutchings Waskesiu, Saskatchewan, Canada E. hageni 2000 Lac Romulus, ZEC de la Blanche, Quebec, Canada 011 (1), 019 (2), 018 (1), 020 (1) E. hageni 2001 DLL Pond, Pont-Rouge, Quebec, Canada 008 (1), 009 (1) E. hageni 2000 Lac a la Truite, Daveluyville, Quebec, Canada 008 (1) E. hageni 2003 Sclater River, at PTH20, Manitoba, Canada 035 (1), 084 (1) Enallagma laterale 2000 Wood Creek Pond, Norfolk, Connecticut, USA 049 (2) E. laterale 2000 Peck’s Pond, Milford, Pennsylvania, USA 048 (7) E. laterale 1995 Lake Nummy, Belleplain State Forest, Cape May 048 (4) County, New Jersey, USA E. laterale 1995 Perley Pond, Sandy Beach Road, Sebago, Maine, 048 (2), 049 (1) USA E. laterale 1995 Otter Pond, SR302 Bridgton, Maine, USA 048 (3) E. laterale 1995 Lovell’s Pond, Barnstable Township, Marston Mills, 051 (2), 049 (1), 050 (1) Massachusetts, USA E. laterale 1995 Herring Pond, Wellfleet Township, Massachusetts, 049 (4) USA Enallagma minusculum 1995 Horseleech Pond, Truro Township, Cape Cod, 007 (4) Massachusetts, USA E. minusculum 1995 Perley Pond, Sandy Beach Road, Sebago, Maine, 007 (9) USA E. minusculum 2000 Little Mushamush Lake, Middle North Cornwall, 007 (1), 047 (1), 022 (2) Nova Scotia, Canada E. minusculum 2000 Lake George, New Brunswick, Canada 022 (4) Enallagma recurvatum 1995 Horseleech Pond, Truro Township, Cape Cod, 007 (1), 046 (3) Massachusetts, USA E. recurvatum 1995 Little Cliff Pond, Nickerson State Park, Brewster, 007 (3), 045 (1) Massachusetts, USA E. recurvatum 1995 Batsto Lake, SR 542, Burlington County, New 007 (4) Jersey, USA E. recurvatum 1995 Paper Mill Lake, SR 347, Belleplain State Forest, 007 (3) Cape May County, New Jersey, USA Enallagma vernale 1999 Lower Baker Pond, Gilmer’s Corner, New 002 (1), 022 (1) Hampshire, USA E. vernale 2000 McDaniel’s Marsh, Enfield, New Hampshire, USA 082 (1) Enallagma anna 2003 Legrand Creek, Logan, Coche County, Utah, USA 535 (1) E. Pilgrim E. anna 2003 Big Pine, Inyo County, California, USA 022 (2) R. Caesar E. anna 1996 Taylor Ditch, Pershing, Nevada, USA 528 (2) J. Simpkin Worldwide Radiation of Damselflies E101

Table A1 (Continued) Year Haplotypes Species collected Locality information (number) Collector Enallagma aspersum 2001 Route 10 E, Albany Cross, Nova Scotia, Canada 001 (1), 509 (1), 508 (1), 504 (1) E. aspersum 1999 Rabinowicz Pond, Etna, New Hampshire, USA 007 (1), 035 (1), 036 (1), 505 (1), 502 (2), 503 (1) E. aspersum 2001 Long Pond, Block Island, Rhode Island, USA 509 (1), 036 (1), 501 (1), 502 (2) E. aspersum 2000 Webber Pond, Plummer Hill Road, Lovell, Maine, 036 (3), 510 (1), 507 (1), USA 506 (1) E. aspersum 2000 Pond Laboratory, Kellogg Biological Station, 502 (6) Hickory Corners, Michigan, USA Enallagma carunculatum 2001 Prairie Road, Springport, Calhoun County, 532 (1) Michigan, USA E. carunculatum 2001 Chileno Valley Road, Petaluma, California, USA 022 (1) K. Biggs E. carunculatum 2001 Upper St. Croix, Solon Springs, Wisconsin, USA 535 (1), 544 (1), 542 (1) R. duBois E. carunculatum 2001 Apache Creek, Catron County, New Mexico, USA 540 (1), 541 (1) S. Dunkle E. carunculatum 2001 Luna Lake, Apache County, Arizona, USA 535 (2), 543 (1) S. Dunkle Enallagma civile 2001 Sachem Pond, Block Island, Rhode Island, USA 535 (5) E. civile 1995 Hinkley’s Pond, Brewster, Massachusetts, USA 537 (1) E. civile 1995 Lovell’s Pond Marston, Massachusetts, USA 535 (1), 538 (1) E. civile 1995 Horseleech Pond, Truro Township, Cape Cod, 535 (1), 536 (1) Massachusetts, USA E. civile 2000 Pond Laboratory, Kellogg Biological Station, 535 (2), 534 (2) Hickory Corners, Michigan, USA E. civile 2000 Ticklenaked Lake, Boltonville, Vermont, USA 535 (1), 534 (1) E. civile 1999 Beaver Pond, Moose Mountain, Etna, New 539 (1) Hampshire, USA E. civile 2000 Raubenstein Reservoir, Hanover, Pennsylvania, USA 535 (1), 534 (5) E. civile 2001 Digby Campground Pond, Digby, Nova Scotia, 535 (5) Canada Enallagma doubledayi 1995 Sun-N-Lakes Golf Course, Lake Placid, Florida, 529 (3), 533 (1) USA E. doubledayi 1995 Flamingo Bay, Savannah River Ecology Lab, Aiken, 529 (3), 530 (1), 531 (1) South Carolina, USA E. doubledayi 1995 Waquoit Pond, Mashpee, Massachusetts, USA 532 (1) Enallagma geminatum 2001 Hourglass Pond, Block Island, Rhode Island, USA 036 (1), 512 (3), 513 (1), 511 (1) E. geminatum 2000 Oncas Lake, Connecticut, USA 036 (2), 513 (2), 524 (1), 525 (1) E. geminatum 2000 Minis Lake, Roseto, Pennsylvania, USA 036 (3), 512 (1), 514 (1) E. geminatum 2000 East Branch Reservoir, Chardon, Ohio, USA 513 (1), 515 (1) E. geminatum 2000 While Lake, Blairstown, New Jersey, USA 036 (3), 526 (1), 517 (1) E. geminatum 1995 Paper Mill Pond, Belleplain State Forest, Cape May 516 (1) County, New Jersey, USA E. geminatum 1999 Burnt Pond, Starbuckville, New York, USA 512 (1), 513 (4) E. geminatum 2001 Palmatier Lake, Hastings, Michigan, USA 518 (1), 519 (1), 521 (1), 520 (2) E. geminatum 1995 Stony Pond, Barnstable Township, Cape Cod, 512 (3), 513 (1), 523 (1), Massachusetts, USA 522 (1) Enallagma praevarum 1996 Unknown location in Texas, USA 527 (1) S. Dunkle E. praevarum 2001 Apache Creek, Catron County, New Mexico, USA 527 (1) S. Dunkle E102 The American Naturalist

Table A1 (Continued) Year Haplotypes Species collected Locality information (number) Collector Enallagma circulatum 1999 Niigata Prefecture, Itsigawa, Japan 900 (2) Y. Hayashi Enallagma cyathigerum 1999 Bay between Capes Helthi and Liya, Kamchatka, 907 (2) K. Kurowski Russia and N. Minakawaa E. cyathigerum 2000 O Porrinõ, Centeańs, Spain 905 (2) A. Cordero Rivera E. cyathigerum 1999 Kvarnfors, Umea, Sweden 901 (1) F. Johannson E. cyathigerum 1999 Holmsund Hamn, Holmsund, Sweden 909 (1) F. Johannson E. cyathigerum 1999 Groot Schietveld, Brasschaat, Belgium 901 (1), 905 (1) M. De Block E. cyathigerum 1999 Jasne Lake, Torzym, Poland 901 (2) R. Bernard E. cyathigerum 2001 Aldan River, Tommot, Yakutia, Russia 906 (2), 908 (1) O. Kosterin Enallagma deserti 2000 Tasura Dam, Souk-Ahras, Tebessa, Algeria 904 (2) B. Samraoui Enallagma risi 1999 Manzhouli, Inner Mongolia, China 903 (2) H. Dumont E. risi 1999 Hohhut, Inner Mongolia, China 903 (1) H. Dumont E. risi 1999 Chingistay, Kazakhstan 901 (1), 902 (1) K. Reinhardt E. risi 1999 Novosibirsk Province, Shradikha Rivult Estuary at 903 (2) O. Kosterin Novosibirsk Water Reserve, Russia E. risi 1999 Tyva Republic, Kyzyl, Russia 901 (1) O. Kosterin Note: Haplotype identification numbers correspond to those in figures 1 and 2 (number of individuals with that haplotype are given in parentheses). a Supplied by D. Paulson.

Table A2: Sampling localities for AFLP analyses: populations, sample sizes, the year samples were collected, and the collector (if not an author) for each species included in the AFLP study Year Sample Species collected size Locality information Collector Enallagma durum 2001 2 Sachem Pond, Block Island, Rhode Island, USA E. durum 2001 4 Canal on Ferry Road, Old Sambrook, Connecticut, USA Enallagma annexum 2000 4 KBS Pond Laboratory, Richland, Michigan, USA E. annexum 1995 4 Lily Pond, Kankamagas Pass, Livermore Township, New Hampshire, USA E. annexum 2001 4 Auke Lake, Juneau, Alaska, USA J. Hudson E. annexum 2001 4 Mexican Cut, Rocky Mountain Biological Station, Gunnison, S. Wissinger Colorado, USA E. annexum 2001 2 Fairbanks Airport, Fairbanks, Alaska, USA T. Sformo Enallagma boreale 2000 4 KBS Pond Laboratory, Richland, Michigan, USA E. boreale 1995 4 Greeley Pond, Livermore Township, Kancamagus Pass, New Hampshire, USA E. boreale 2002 4 Gaspisie Pond, Route 132, Lanse-a-valleau, New Brunswick, Canada E. boreale 2003 2 Honey Lake Wildlife Area, Dakin Unit, Susanville, California, USA E. boreale 2001 4 Mexican Cut, Rocky Mountain Biological Station, Gunnison, S. Wissinger Colorado, USA E. boreale 1998 2 Smith Lake, Fairbanks, Alaska, USA D. Paulson Enallagma clausum 2003 4 Eagle Lake, Susanville, California, USA E. clausum 2003 4 Bear River Migratory Bird Refuge, Brigham City, Utah, USA E. Pilgrim E. clausum 2003 1 Lake Winnipegosis, Winnipegosis, Manitoba, Canada M. Hughes E. clausum 2003 2 Lake Dauphin, Manitoba, Canada M. Hughes Enallagma davisi 2001 4 Pine Lake, Marston Road, Marston, North Carolina, USA Enallagma ebrium 2000 15 Drew Pond, Highway 110, York County, Meecors, Maine, USA Worldwide Radiation of Damselﬂies E103

Table A2 (Continued) Year Sample Species collected size Locality information Collector E. ebrium 2001 19 George Reserve, Pinckney, Michigan, USA E. ebrium 2001 24 McDaniel’s Marsh, Enfield, New Hampshire, USA E. ebrium 1999 4 Lake Ninevah, Grahamville, Vermont, USA E. ebrium 2001 13 Molly Pond, Groten, Vermont, USA E. ebrium 2003 4 Ochre River, L. Dauphin, Manitoba, Canada M. Hughes E. ebrium 2003 3 Sclater River at PTH20, Manitoba, Canada M. Hughes Enallagma hageni 1999 19 Lovewell Pond, Fryeburg, Maine, USA E. hageni 2000 19 3 Lakes II, Richland, Michigan, USA E. hageni 2001 5 Little Mushamush Lake, Middle New Cornwall, Nova Scotia, Canada E. hageni 1999 4 Amhearst Lake, Plymouth, Vermont, USA E. hageni 2000 3 Spectacle Pond, Island Pond, Vermont, USA E. hageni 2000 4 Martins Pond, Green Bay, Vermont, USA E. hageni 2000 20 Ticklenaked Lake, Boltonville, Vermont, USA E. hageni 2001 25 Molly Falls Pond, Junction 232 and 2, Groton, Vermont, USA E. hageni 2003 4 Waterhen, Manitoba, Canada M. Hughes E. hageni 2003 4 Sclater River at PTH20, Manitoba, Canada M. Hughes Enallagma laterale 1995 8 Otter Pond, State Route 302, Bridgton, Maine, USA E. laterale 2000 10 Peck’s Pond, Milford, Pennsylvania, USA Enallagma minusculum 1995 4 Horseleech Pond, Truro Township, Cape Cod, Massachusetts, USA E. minusculum 2000 4 Pearly Pond, Sebago, Maine, USA E. minusculum 2001 4 Little Mushamush Lake, Middle New Cornwall, Nova Scotia, Canada Enallagma recurvatum 1995 4 Snow Pond, Truro Township, Cape Cod, Massachusetts, USA E. recurvatum 1995 4 Makepeace Lake, Makepeace Wildlife Management Area, Atlantic County, New Jersey, USA Enallagma vernale 1995 3 Lower Baker Pond, Gilmer’s Corner, New Hampshire, USA E. vernale 2001 4 Lower Symes Pond, East Ryegate, Vermont, USA Enallagma anna 2003 4 South of Big Pine, Inyo County, California, USA R. Caesar E. anna 2003 4 Legrand Creek, Logan, Utah, USA E. Pilgrim Enallagma aspersum 2001 4 South Long Pond, Block Island, Rhode Island, USA E. aspersum 2000 4 KBS Pond Laboratory, Richland, Michigan, USA E. aspersum 2000 4 Johnson’s Pond, Norwich, Vermont, USA E. aspersum 2000 4 Webber Road Pond, Plummer Hill Road, Lovell, Maine, USA Enallagma carunculatum 2000 5 Lake Chatcolet, Benewah County, Idaho, USA D. Paulson E. carunculatum 2001 4 Lake Raphine, Santa Rosa, California, USA K. Biggs E. carunculatum 2001 4 Upper St. Croix, Solon Springs, Wisconsin, USA R. duBois Enallagma civile 2000 5 Pond at Exit 76, I-70 in Maryland, USA E. civile 2001 5 Wright’s Pond, Block Island, Rhode Island, USA E. civile 2001 5 Lake O’Lawns, Nova Scotia, Canada E. civile 2001 5 Sculpture Park, Eureka Road, Roseville, California, USA K. Biggs E. civile 2003 4 Hawaii I. Cooper Enallagma doubledayi 1993 2 Sinkhole Pond, Archbold Biological Station, Lake Placid, Florida, USA Enallagma geminatum 2000 2 Lake Hortonia, Hubbardton, Vermont, USA E. geminatum 2000 4 East Branch Reservoir, SE of Chardon, Ohio, USA E. geminatum 2000 4 Perley Pond, Sandy Beach Road, North Sebago, Maine, USA E. geminatum 2001 5 Oncas Lake, Sambrook, Connecticut, USA Enallagma praevarum 2001 5 Silver Lake, Cloudcroft, New Mexico, USA S. Dunkle Enallagma cyathigerum 2000 4 O Porrinõ, Centeańs, Spain A. Cordero Rivera E. cyathigerum 1999 4 Groot Schietveld, Brasschaat, Belgium M. De Block E104 The American Naturalist

Table A2 (Continued) Year Sample Species collected size Locality information Collector E. cyathigerum 1999 4 Kvarnfors, Umea, Sweden F. Johannson E. cyathigerum 2001 3 Aldan River, Tommot, Yakutia, Russia O. Kosterin E. cyathigerum 1999 3 Bay between Capes Helthi and Liya, Kamchatka, Russia K. Kurowski and N. Minakawaa Enallagma risi 1999 4 Manzhouli, Inner Mongolia, China H. Dumont E. risi 1999 4 Hohhut, Inner Mongolia, China H. Dumont a Supplied by D. Paulson.

Literature Cited relation to speciation phenomena: the founding of new populations. Annual Review of Ecology and Systematics 15:97–131. Catling, P. M. 2001. Morphological evidence for the hybrid Akaike, H. 1974. A new look at the statistical model identification. Enallagma ebrium # hageni (Zygoptera: Coenagrionidae) from IEEE Transactions on Automatic Control AC-19:716–723. Ontario. Proceedings of the Entomological Society of Ontario 132: Albertson, R. C., J. A. Markert, P. D. Danley, and T. D. Kocher. 1999. 99–100. Phylogeny of a rapidly evolving clade: the cichlid fishes of Lake Chapple, D. G., and J. S. Keogh. 2004. Parallel adaptive radiations Malawi, East Africa. Proceedings of the National Academy of Sci- in arid and temperate Australia: molecular phylogeography and ences of the USA 96:5107–5110. systematics of the Egernia whitii (Lacertilia: Scincidae) species Allender, C. J., O. Seehausen, M. E. Knight, G. F. Turner, and N. group. Biological Journal of the Linnean Society 83:157–173. Maclean. 2003. Divergent selection during speciation of Lake Ma- Coope, G. R. 1995. Insect faunas in Ice Age environments: why so lawi cichlid fishes inferred from parallel radiations in nuptial col- little extinction? Pages 55–74 in J. H. Lawton and R. M. May, eds. oration. Proceedings of the National Academy of Sciences of the Extinction rates. Oxford University Press, Oxford. USA 100:14074–14079. ———. 2004. Several million years of stability among insect species Arnqvist, G., M. Edvardsson, U. Friberg, and T. Nilsson. 2000. Sexual because of, or in spite of, Ice Age climatic instability? Philosophical conflict promotes speciation in insects. Proceedings of the National Transactions of the Royal Society of London B 359:209–214. Academy of Sciences of the USA 97:10460–10464. Coyne, J. A., and H. A. Orr. 2004. Speciation. Sinauer, Sunderland, Askew, R. R. 1988. The dragonflies of Europe. Harley, London. MA. Barraclough, T. G., and A. P. Vogler. 2002. Recent diversification rates Cuervo, J. J., and A. P. Møller. 1999. Evolutionary rates of secondary in North American tiger beetles estimated from a dated mtDNA sexual and non-sexual characters in birds. Evolutionary Ecology phylogenetic tree. Molecular Biology and Evolution 19:1706–1716. 13:283–303. Barraclough, T. G., P. H. Harvey, and S. Nee. 1995. Sexual selection Donnelly, T. W. 2004. Distribution of North American Odonata. Part and taxonomic diversity in passerine birds. Proceedings of the III. Calopterygidae, Lestidae, Coenagrionidae, Protoneuridae, Pla- Royal Society of London B 259:211–215. tystictidae. Bulletin of American Odonatology 8:33–99. Bennett, K. D. 1997. Evolution and ecology: the pace of life. Cam- Duchesne, P., and L. Bernatchez. 2002. AFLPOP: a computer program bridge University Press, Cambridge. for simulated and real population allocation based on AFLP data. Benton, M. J. 1987. Progress and competition in macroevolution. Biological Review 62:305–338. Molecular Ecology Notes 3:380–383. Bernatchez, L., and C. C. Wilson. 1998. Comparative phylogeography Dynesius, M., and R. Jansson. 2000. Evolutionary consequences of of Nearctic and Palearctic fishes. Molecular Ecology 7:431–452. changes in species’ geographical distributions driven by Milan- Bouchard, P., and D. R. Brooks. 2004. Effect of vagility potential on kovitch climate oscillations. Proceedings of the National Academy dispersal and speciation in rainforest insects. Journal of Evolu- of Sciences of the USA 97:9115–9120. tionary Biology 17:994–1006. Edmands, S. 2001. Phylogeography of the intertidal copepod Tigrio- Brower, A. V. Z. 1994. Rapid morphological radiation and conver- pus californicus reveals substantially reduced population differen- gence among races of the butterfly Heliconius erato, inferred from tiation at northern latitudes. Molecular Ecology 10:1743–1750. patterns of mitochondrial DNA evolution. Proceedings of the Na- Excoffier, L., P. E. Smouse, and J. M. Quattro. 1992. Analysis of tional Academy of Sciences of the USA 91:6491–6495. molecular variance inferred from metric distances among DNA Brown, J. M., M. A. McPeek, and M. L. May. 2000. A phylogenetic haplotypes: application to human mitochondrial DNA restriction perspective on habitat shifts and diversity in the North American data. Genetics 131:479–491. Enallagma damselflies. Systematic Biology 49:697–712. Fu, Y.-X. 1997. Statistical tests of neutrality of mutations against Carisio, L., P. Cervella, C. Palestrini, M. DelPero, and A. Rolando. population growth, hitchhiking and background selection. Ge- 2004. Biogeographical patterns of genetic differentiation in dung netics 147:915–925. beetles of the genus Trypocopris (Coleoptera, Geotrupidae)inferred Gaggiotti, O. E., and L. Excoffier. 2000. A simple method of removing from mtDNA and AFLP analyses. Journal of Biogeography 31: the effect of a bottleneck and unequal population sizes on pairwise 1149–1162. genetic distances. Proceedings of the Royal Society of London B Carson, H. L., and A. R. Templeton. 1984. Genetic revolutions in 267:81–87. Worldwide Radiation of Damselflies E105

Gavrilets, S. 2000. Rapid evolution of reproductive barriers driven Knox, E. B., and J. D. Palmer. 1995. Chloroplast DNA variation and by sexual conflict. Nature 403:886–889. the recent radiation of the giant Senecios (Asteraceae) on the tall Gavrilets, S., and C. R. B. Boake. 1998. On the evolution of premating mountains of eastern Africa. Proceedings of the National Academy isolation after a founder event. American Naturalist 152:706–716. of Sciences of the USA 92:10349–10353. Gavrilets, S., and D. Waxman. 2002. Sympatric speciation by sexual Kubo, T., and Y. Iwasa. 1995. Inferring the rates of branching and conflict. Proceedings of the National Academy of Sciences of the extinction from molecular phylogenies. Evolution 49:694–704. USA 99:10533–10538. Kurteń, B., and E. Anderson. 1980. Pleistocene mammals of North Gillespie, R. 2004. Community assembly through adaptive radiation America. Columbia University Press, New York. in Hawaiian spiders. Science 303:356–359. Lande, R. S. 1981. Models of speciation by sexual selection on poly- Graham, R. W., E. L. Lundelius, M. A. Graham, E. K. Schroeder, R. genic traits. Proceedings of the National Academy of Sciences of S. Toomey, E. Anderson, A. D. Barnosky, et al. 1996. Spatial re- the USA 78:3721–3725. sponse of mammals to late Quaternary environmental fluctuations. Lister, A. M. 2004. The impact of Quaternary ice ages on mammalian Science 272:1601–1606. evolution. Philosophical Transactions of the Royal Society of Lon- Harpending, R. C. 1994. Signature of ancient population growth in don B 359:221–241. a low-resolution mitochondrial DNA mismatch distribution. Hu- Lockhart, P. J., M. A. Steele, M. D. Hendy, and D. Penny. 1994. man Biology 66:591–600. Recovering evolutionary trees under a more realistic model of Hewitt, G. M. 1996. Some genetic consequence of ice ages, and their sequence evolution. Molecular Biology and Evolution 11:605–612. role in divergence and speciation. Biological Journal of the Linnean Losos, J. B., T. R. Jackman, A. Larson, K. de Queiroz, and L. Society 58:247–276. Rodriguez-Schettino. 1998. Contingency and determinism in rep- ———. 2004. Genetic consequences of climatic oscillations in the licated adaptive radiations of island lizards. Science 279:2115–2118. Quaternary. Philosophical Transactions of the Royal Society of Lovette, I. J., and E. Bermingham. 1999. Explosive speciation in the London B 359:183–195. New World Dendroica warblers. Proceedings of the Royal Society Hodges, S. A., and M. L. Arnold. 1994. Columbines: a geographically of London B 266:1629–1636. widespread species flock. Proceedings of the National Academy of Martin, P. R., and J. K. McKay. 2004. Latitudinal variation in genetic Sciences of the USA 91:5129–5132. divergence of populations and the potential for future speciation. Holland, B., and W. R. Rice. 1998. Chase-away sexual selection: an- Evolution 58:938–945. tagonistic seduction versus resistance. Evolution 52:1–7. Martin, P. S., and R. G. Klein. 1984. Quaternary extinctions: a pre- Holman, J. A. 1993. British Quaternary herpetofaunas: a history of historic revolution. University of Arizona Press, Tucson. adaptations to Pleistocene disruptions. Herpetological Journal 3: Martinson, D. G., N. G. Pisias, J. D. Hays, J. Imbrie, T. C. Moore, 1–7. and N. J. Shackleton. 1987. Age dating and the orbital theory of Jablonski, D. 1986. Background and mass extinctions: the alternation the Ice Ages: development of a high-resolution 0 to 300,000-year of macroevolutionary regimes. Science 231:129–133. chronostratigraphy. Quaternary Research 27:1–29. ———. 1994. Extinctions in the fossil record. Philosophical Trans- Mathworks. 2002. Matlab, version 7. Natick, MA. actions of the Royal Society of London B 344:11–17. Jansson, R., and M. Dynesius. 2002. The fate of clades in a world of McPeek, M. A. 1989. Differential dispersal tendencies among Enal- recurrent climatic change: Milankovitch oscillations and evolution. lagma damselflies (Odonata: Coenagrionidae) inhabiting different Annual Review of Ecology and Systematics 33:741–777. habitats. Oikos 56:187–195. Johns, G. C., and J. C. Avise. 1998. Tests for ancient species flocks ———. 1990a. Behavioral differences between Enallagma species based on molecular phylogenetic appraisals of Sebastes rockfishes (Odonata) influencing differential vulnerability to predators. Ecol- and other marine fishes. Evolution 52:1135–1146. ogy 71:1714–1726. Johnson, D. M., and P. H. Crowley. 1980. Habitat and seasonal seg- ———. 1990b. Determination of species composition in the Enal- regation among coexisting odonate larvae. Odonatologica 9:297– lagma damselfly assemblage of permanent lakes. Ecology 71:83– 308. 98. Kawata, M., and J. Yoshimura. 2000. Speciation by sexual selection ———. 1995. Morphological evolution mediated by behavior in the in hybridizing populations without viability selection. Evolution- damselflies of two communities. Evolution 49:749–769. ary and Ecological Research 2:897–909. ———. 1998. The consequences of changing the top predator in a Klicka, J., and R. M. Zink. 1997. The importance of recent Ice Ages food web: a comparative experimental approach. Ecological Mon- in speciation: a failed paradigm. Science 277:1666–1669. ographs 68:1–23. ———. 1999. Pleistocene phylogeographic effects on avian evolu- ———. 1999. Biochemical evolution associated with antipredator tion. Proceedings of the Royal Society of London B 266:695–700. adaptation in damselflies. Evolution 53:1835–1845. Knowles, L. L. 2000. Tests of Pleistocene speciation in montane grass- ———. 2000. Predisposed to adapt? clade-level differences in char- hoppers from the sky islands of western North America (genus acters affecting swimming performance in damselflies. Evolution Melanoplus). Evolution 54:1337–1348. 54:2072–2080. ———. 2001. Did the Pleistocene glaciations promote divergence? ———. 2004. The growth/predation-risk trade-off: so what is the mech- tests of explicit refugial models in montane grasshoppers. Molec- anism? American Naturalist 163:E88–E111. http://www.journals ular Ecology 10:691–701. .uchicago.edu/AN/journal/issues/v163n5/40010/40010.html. Knowles, L. L., and D. Otte. 2000. Phylogenetic analysis of montane McPeek, M. A., and J. M. Brown. 2000. Building a regional species grasshoppers from western North America (genus Melanoplus, Ac- pool: diversification of the Enallagma damselflies in eastern North rididae: Melanoplinae). Annals of the Entomology Society of America. Ecology 81:904–920. America 93:421–431. McPeek, M. A., A. K. Schrot, and J. M. Brown. 1996. Adaptation to E106 The American Naturalist

predators in a new community: swimming performance and pred- Schluter, D., and L. M. Nagel. 1995. Parallel speciation by natural ator avoidance in damselflies. Ecology 77:617–629. selection. American Naturalist 146:292–301. Meffert, L. M., and J. L. Regan. 2002. A test of speciation via sexual Schneider, S., D. Roessli, and L. Excoffier. 2000. Arlequin: a software selection on female preferences. Animal Behaviour 64:955–965. for population genetic analysis. Version 2.0. University of Geneva, Nee, S., A. Ø. Mooers, and P. H. Harvey. 1992. Tempo and mode of Switzerland. evolution revealed from molecular phylogenies. Proceedings of the Sepkoski, J. J., Jr. 1991. Diversity in the Phanerozoic oceans: a partisan National Academy of Sciences of the USA 89:8322–8326. review. Pages 210–236 in E. Dudley, ed. Fourth International Con- Nee, S., R. M. May, and P. H. Harvey. 1994. The reconstructed gress of Systematics and Evolutionary Biology, Proceedings. Dios- evolutionary process. Philosophical Transactions of the Royal So- corides, Portland, OR. ciety of London B 344:305–311. Sherbakov, D. Y. 1999. Molecular phylogenetic studies on the origin Overpeck, J. T., R. S. Webb, and T. Webb. 1992. Mapping eastern of biodiversity in Lake Baikal. Trends in Evolution & Ecology 14: North American vegetation change of the past 18 KA: no analogs 92–95. and the future. Geology 20:1071–1074. Steele, M. 1994. Recovering a tree from the Markov model. Applied Parker, G. A., and L. Partridge. 1998. Sexual conflict and speciation. Mathematics Letters 7:19–23. Proceedings of the Royal Society of London B 353:261–274. Stoks, R., M. A. McPeek, and J. L. Mitchell. 2003. Evolution of prey Paulson, D. R. 1974. Reproductive isolation in damselflies. Systematic behavior in response to changes in predation regime: damselflies Zoology 23:40–49. in fish and dragonfly lakes. Evolution 57:574–585. Stuart-Fox, D., and I. P. F. Owen. 2003. Species richness in agamid Paulson, D. R., N. Minakawa, and R. I. Gara. 1998. Recent collections lizards: chance, body size, sexual selection or ecology? Journal of of Odonata from the Kuril Islands. Species Diversity 3:75–80. Evolutionary Biology 16:659–669. Posada, D., and K. A. Crandall. 1998. MODELTEST: testing the model Sturmbauer, C., U. Hainz, S. Baric, E. Verheyen, and W. Salzburger. of DNA substitution. Bioinformatics 14:817–818. 2003. Evolution of the tribe Tropheini from Lake Tanganyika: syn- Posada, D., K. A. Crandall, and A. R. Templeton. 2000. GeoDis: a chronized explosive speciation producing multiple evolutionary program for the cladistic nested analysis of the geographical dis- parallelism. Hydrobiologia 500:51–64. tribution of genetic haplotypes. Molecular Ecology 9:487–488. Swofford, D. L. 2001. PAUP∗: phylogenetic analysis using parsimony Price, T., H. L. Gibbs, L. de Sousa, and A. D. Richman. 1998. Different ∗ ( and other methods). Version 4.1b10 for Windows. Sinauer, Sun- timing of the adaptive radiations of North American and Asian derland, MA. warblers. Proceedings of the Royal Society of London B 265:1969– Swofford, D. L., G. J. Olsen, P. J. Waddell, and D. M. Hillis. 1996. 1975. Phylogenetic inference. Pages 407–514 in D. M. Hillis, C. Moritz, Price, T., I. J. Lovette, E. Bermingham, H. L. Gibbs, and A. D. Rich- and B. K. Mable, eds. Molecular systematics. 2nd ed. Sinauer, man. 2000. The imprint of history on communities of North Sunderland, MA. American and Asian warblers. American Naturalist 156:354–367. Tamura, K. 1992. Estimation of the number of nucleotide substi- Pybus, O. G., and P. H. Harvey. 2000. Testing macro-evolutionary tutions when there are strong transition-transversion and G ϩ C models using incomplete molecular phylogenies. Proceedings of content biases. Molecular Biology and Evolution 9:678–687. the Royal Society of London B 267:2267–2272. Turgeon, J., and M. A. McPeek. 2002. Phylogeographic analysis of a Randler, C. 2002. Avian hybridization, mixed pairing and female recent radiation of Enallagma damselflies (Odonata: Coenagrion- choice. Animal Behaviour 63:103–119. idae). Molecular Ecology 11:1989–2001. Raup, D. M. 1979. Size of the Permo-Triassic bottleneck and its Vekemans, X., T. Beauwens, M. Lemaire, and I. Roldan-Ruiz. 2002. evolutionary implications. Science 206:217–218. Data from amplified fragment length polymorphism (AFLP) ———. 1991. Extinction: bad genes or bad luck? Norton, New York. markers show indication of size homoplasy and of a relationship Regan, J. L., L. M. Meffert, and E. H. Bryant. 2003. A direct exper- between degree of homoplasy and fragment size. Molecular Ecol- imental test of founder-flush effects on the evolutionary potential ogy 11:139–151. for assortative mating. Journal of Evolutionary Biology 16:302– Verheyen, E., W. Salzburger, J. Snoeks, and A. Meyer. 2003. Origin 312. of the superflock of cichlid fishes from Lake Victoria, East Africa. Ribera, I., and A. P. Vogler. 2004. Speciation of Iberian diving beetles Science 300:325–329. in Pleistocene refugia (Coleoptera, Dytiscidae). Molecular Ecology Vrba, E. A. 1985. Environment and evolution: alternative causes of 13:179–193. the temporal distribution of evolutionary events. South African Robertson, H. M., and H. E. H. Paterson. 1982. Mate recognition Journal of Science 81:229–236. and mechanical isolation in Enallagma damselflies (Odonata: Walker, E. M. 1953. The Odonata of Canada and Alaska. Vol. 1. Coenagrionidae). Evolution 36:243–250. University of Toronto Press, Toronto. Roderick, G. K., and R. G. Gillespie. 1998. Speciation and phylo- Watterson, G. 1975. On the number of segregating sites in genetical geography of Hawaiian terrestrial arthropods. Molecular Ecology models without recombination. Theoretical Population Biology 7: 7:519–531. 256–276. Rogers, A. R. 1995. Genetic-evidence for a Pleistocene population Westfall, M. J., and M. L. May. 1996. Damselflies of North America. explosion. Evolution 49:608–615. Scientific, Gainesville, FL. Rogers, A. R., and H. Harpending. 1992. Population-growth makes Wilder, J. A., and H. Hollocher. 2003. Recent radiation of endemic waves in the distribution of pairwise genetic-differences.Molecular Caribbean Drosophila of the dunni subgroup inferred from mul- Biology and Evolution 9:552–569. tilocus DNA sequence variation. Evolution 57:2566–2579. Schluter, D. 2000. The ecology of adaptive radiation. Oxford Uni- Willis, K. J., and K. J. Niklas. 2004. The role of Quaternary envi- versity Press, Oxford. ronmental change in plant macroevolution: the exception or the Worldwide Radiation of Damselflies E107

rule? Philosophical Transactions of the Royal Society of London diversiﬁcation during the Quaternary. Philosophical Transactions B 359:159–172. of the Royal Society of London B 359:215–220. Wirtz, P. 1999. Mother species-father species: unidirectional hybridization in animals with female choice. Animal Behaviour 58:1–12. Editor: Jonathan B. Losos Zink, R. M., J. Klicka, and B. R. Barber. 2004. The tempo of avian Associate Editor: Michael E. Hellberg