Sympatric Speciation in the Post “Modern Synthesis” Era of Evolutionary Biology
Sympatric speciation is among the most controversial and challenging concepts in evolution. There are a multitude of definitions of speciation alone, and when combined with the biogeographic concept of sympatric range overlap, consensus on what sympatric speciation is, whether it happens, and its importance, is even more difficult to achieve. Providing the basis upon which to define and judge sympatric speciation, the Modern Evolutionary Synthesis (Huxley in Evolution: the modern synthesis. MIT Press, Cambridge, 1942) led to the conclusion that sympatric speciation is an inconsequential process in the generation of species diversity. In the post Modern Synthesis era of evolutionary biology, the PCR revolution and associated accumulation of DNA sequence data from natural populations has led to a resurgence of interest in sympatric speciation, and more importantly, the role of natural selection in lineage diversification. Much effort is currently being devoted to elucidating the processes by which the constituents of an initially panmictic population can become reproductively isolated and evolve some level of reproductive incompatibility without the complete cessation of gene flow due to geographic barriers. The evolution of reproductive isolation solely due to natural selection is perhaps the most controversial manner by which sympatric speciation occurs, and it is that which we focus upon in this review. Mathematical model simulations indicate that even strict definitions of sympatric speciation are possible to satisfy, empirical data consistent with sympatric divergence are accumulating, but irrefutable evidence of sympatric speciation in natural populations remains elusive. Genomic investigations are advancing our ability to identify genetic patterns caused by natural selection, thereby advancing our understanding of the power of natural selection relative to gene flow. Overall, sympatric lineage divergence, especially at the sub-species level, may have led to a substantial portion of biodiversity.
KeywordsLineage Population Divergence Selection Gene flow
Modern Evolutionary Synthesis and Speciation
The Modern Evolutionary Synthesis, sparked by Dobzhansky’s (1937) population genetic work with natural populations of Drosophila fruit flies, reconciled Darwin’s (1859) theory on the origin of species by natural selection, Mendelian inheritance (1866), macroevolution in paleontology (Simpson 1944), and population genetics theory for sexually reproducing organisms (Huxley 1942). The primary tenets of the Modern Evolutionary Synthesis as they relate to speciation include: (1) evolution is defined as the change in allele frequencies within and among populations (Dobzhansky 1937); (2) allele frequencies change in response to natural selection or mutation and genetic drift; (3) speciation is defined as the evolution of reproductive incompatibility (Wright 1940; Mayr 1942), (4) reproductive isolation is defined as the cessation of gene flow between two populations, occurs due to spatial isolation in allopatry, and is the required first step for reproductive incompatibility to evolve between two populations (Mayr 1963). The role of natural selection in driving reproductive isolation while populations overlap is relegated to reinforcement (Dobzhansky 1950) following an initial period of strictly allopatric isolation (Dobzhansky 1937) or possibly parapatric speciation where gene flow is reduced by incomplete spatial isolation (Mayr 1963). In rejecting the concept of sympatric speciation, where reproductive isolation and incompatibility evolves without geographic isolation, Mayr (1963), famously and ardently rebuked the possibility that natural selection could contribute to isolating sympatric populations. Maynard Smith (1966) surmises, “The crucial argument against sympatric speciation, admirably summarised by Mayr (1963), is that no mechanism consistent with the known facts of genetics can be suggested.” In short, gene flow and chromosomal recombination should oppose the relatively weak action of natural selection to divide populations and generate species by preventing the build up of linkage disequilibrium between loci under selection and those that contribute to reproductive isolation. Using a mathematical population genetic model, Maynard Smith (1966) demonstrated that sympatric speciation was possible, but the required conditions were not believed to be common in nature. It is noteworthy that the modern synthesis was primarily formulated based on studies of terrestrial species, with the great diversity of marine invertebrates primarily omitted from consideration (Love 2009), with the exception of Mayr’s (1954a) work with tropical sea urchins.
Post “Modern Synthesis” Evolutionary Biology and Speciation
The advent of the polymerase chain reaction (PCR) in 1983 (see Mullis 1990), chain-termination DNA sequencing (Sanger et al. 1977) and most recently massively parallel DNA sequencing (Ellegren 2008; Mardis 2008) combined with the proliferation of phylogenetic and phylogeographic statistical analysis (Avise and Wollenberg 1997) has ushered in a new era of evolutionary biology. There has been a resurgence in the ideas that (1) natural selection can generate reproductive isolation in the face of gene flow, that (2) parapatric speciation where reproductive isolation is driven simultaneously by geographic isolation and selection, is common (Nosil and Feder 2012) and that (3), in the extreme case, sympatric speciation via natural selection is possible, even with continuous and unrestricted gene flow (Dieckmann and Doebeli 1999; Coyne 2011). Even Mayr (2001) acknowledged likely cases of sympatric speciation in “cichlids, sticklebacks, whitefishes, etc …”, although it should be noted that the general consensus is that three spine stickleback fish (Gasterosteus aculeatus) did not undergo sympatric divergence (reviewed in Bolnick 2011). Technological advances have recently made the identification and genotyping of genome-wide data possible in any species through genome reduction (Baird et al. 2008) and transcriptome sequencing (Barbazuk et al. 2007). In the present era of evolutionary biology, researchers must reconcile these large amounts of genetic data from natural and laboratory populations with our understanding of evolutionary processes. A major challenge is resolving the nature of the relationship between natural selection, genetic/genomic processes, and geographic isolation in driving patterns of gene flow between populations. The manner in which reproductive isolation evolves without complete allopatric isolation is one of the most pressing questions in speciation today (Marie Curie SPECIATION Network 2011) and in this “golden age” of evolutionary genetics (Nadeau and Jiggins 2010) researchers are poised to make major advances in our understanding of speciation (Nosil and Feder 2012).
The Paradox and Challenge of Speciation Research
The process of speciation in sexually reproducing organisms is generally acknowledged to occur on a time scale that far exceeds that of the researchers observing it (Via 2009). The paradox of studying speciation is that when focusing on the early stages of lineage divergence, one does not know if presently observed partitioning will lead to speciation, and when focusing on sister species, the signatures of the past processes that led to speciation are lost or not interpretable. Because cases of sympatric speciation are considered to be very rare relative to allopatric and parapatric speciation (Coyne and Orr 2004), it is especially difficult to provide air-tight evidence that sympatric speciation has occurred, even by the most liberal biogeographic definitions (Coyne 2007). Coyne and Orr (2004) suggest that two presently sympatric species can be assumed to be the result of sympatric speciation when the species exhibit monophyly relative to other closely related species, reproductive isolation, and a circumstance where allopatric isolation seems particularly unlikely. Bolnick and Fitzpatrick (2007) note that these criteria are conservative and a failure to meet them does not exclude sympatric speciation. Species arising in sympatry can become allopatrically distributed, might not exhibit monophyly, and/or might not be unambiguously identifiable as separate species. Additionally, demonstrating that these criteria are met does little to convince a skeptic that sympatric speciation has occurred (see Stuessy 2006; Schliewen et al. 1994). Mathematical models are, perhaps, the only way of testing the viability of alternative scenarios of how speciation proceeded or will proceed (see Gavrilets et al. 2007; Gavrilets and Vose 2007; Sadedin et al. 2009), but such models are only viable if their assumptions are valid (Turelli et al. 2001). Consequently, studying divergence in a variety of natural populations along the continuum from a panmictic population to reproductive isolation, with and without secondary contact, and reproductive incompatibility should advance our knowledge of how speciation proceeds and improve the realism of speciation modeling.
Here we present a review of sympatric speciation, beginning with a discussion of the terminology of species concepts, sympatric speciation, ecological speciation, and divergence-with-gene-flow. Mirroring the calls for speciation research to focus on the processes generating population divergence, we review and discuss the results of empirical and theoretical research relevant to the lineage bifurcation process in sympatry in three phases: disruptive selection without reproductive isolation, the evolution of reproductive isolation, and post reproductive isolation. We conclude with a summary of points for the future of research on sympatric divergence
Complicated, Controversial Concepts and Terminology
Species Concepts and Speciation
A review of the processes that drive speciation should begin with a discussion of what species are. The controversy surrounding the definition of a species arises from (1) the large number of speciation concepts that result in disparate diagnoses of species identity (Mayden 1997; de Queiroz 2005, 2007; Hey 2006), (2) the different motivations for defining a species, (3) the view that species represent a higher level of organization than the researchers observing them (de Queiroz 2007) and (4) that species are artificial categorical classifications placed upon populations of individuals (lineages) that evolve continuously through time (Hart 2010). Non-operational species concepts are broad in scope with limited practical application for assigning species identity, as in the Evolutionary Species Concept (Simpson 1951) which defines species as groups of organisms with unique evolutionary roles, tendencies, and historical fates. Operational species concepts typically provide a criterion by which individuals can be classified by species group but typically are not applicable in all circumstances, such as the Biological Species Concept (Wright 1940; Mayr 1942, 1963) where reproductive incompatibility defines a species boundary. Differences of opinion on what the important properties of species are leads to different concepts and different conclusions about whether or not speciation has occurred, which Coyne and Orr (2004) consider to be a necessary criterion for demonstrating that (sympatric) speciation has occurred. In contrast, proponents of eliminating the species category have argued that species are at best a proxy or artificial construct: “the recognizable but indirect consequence of processes that act on variety among individual cells or multicellular organisms to produce diversity among populations” (Hart 2010). Individuals, rather than species, are the locus of action in evolutionary processes and population-level variance, the relative importance of population-level processes, and how those processes contribute to the production of diversity for different taxa, at different times, and under different ecological circumstances should be the focus of research (Hendry et al. 2000; Hart 2010; Sobel et al. 2010).
The operational Biological Species Concept is almost ubiquitously adopted by those studying speciation in sexually reproducing natural populations (see Coyne and Orr 2004), but there are alternatives. Mayden (1997, 1999) and de Queiroz (1998, 1999, 2007) have devised organizational frameworks that unify speciation concepts, and they are likely to result in similar species diagnoses (Naomi 2011). Both frameworks are based on a non-operational species concept and several operational criteria that can be used to confirm whether the non-operational definition is satisfied. Naomi (2011) combines the ideas of Mayden (1997, 1999) and de Queiroz (1998, 1999, 2007) into a third species concept. Naomi (2011) adopts the non-operational species concept of Mayden (1997, 1999), where “a species is a single lineage of ancestral descendant populations of organisms which maintains its identity from other such lineages and which has its own evolutionary tendencies and historical fate” (Wiley 1978; see Wiley and Mayden 2000a, b, c) and combines that with the contingent biological properties of species (intrinsic reproductive isolation, diagnosability, reciprocal monophyly, ecological distinctiveness, and phenetic distinguishability) from de Queiroz’s (2007) unified species concept. The power of the speciation concepts of Mayden (1999), de Queiroz (2007), and Naomi (2011) is that they are compatible with both views of species as real entities and as human constructs devised to simplify the description of nature. Adopting a species concept more inclusive of the diversity life broadens the applicability of speciation research.
Sympatry and Sympatric Speciation
The definition of sympatric speciation is as contentious as that of a species (reviewed in Butlin et al. 2008; Fitzpatrick et al. 2008). Sympatric, parapatric, and allopatric are terms used to classify the relationship between the geographic distribution of two or more populations or species. All share the Greek root patra, meaning fatherland, and differ in the prefixes sym—same, para—abutting or side by side, and allo—different or other (Poulton 1904). A peripatric (peri—near or around, Mayr 1954b) distribution is a particular variety of allopatry defined by a small, geographically distinct founding population and a larger source population. These terms, by themselves, evoke little debate among biologists until they are linked with speciation.
The primary difficulty in reaching a consensus on the definition of sympatric speciation is encountered because, unlike allopatry which can cause reproductive isolation (Coyne 2007), sympatry does nothing to facilitate speciation, and thus is potentially a catchall category that provides little description of the speciation process. In contrast, when a population becomes allopatrically divided without subsequent gene flow, the lineages are free to evolve independently simply due to their spatial isolation. In essence, sympatric speciation is often defined by what it is not. In its most liberal, inclusive form, sympatric speciation is a category of speciation events in which the broad-scale geographic segregation of individuals doesnotrestrict gene flow and thereby contribute to speciation. Under this most-inclusive biogeographic definition, sympatric speciation includes cases where individuals were spatially segregated on a fine spatial scale within the range of the ancestral population. Polyploid (Otto and Whitton 2000; Wood et al. 2009) and allochronic speciation (Coyne and Orr 2004) are examples of speciation processes where the number of chromosome sets or the timing of sexual reproduction, respectively, isolates lineages. In these cases, broad-scale biogeographic distribution does not contribute to realized reproductive isolation between populations, and thus, these are examples of sympatric speciation under this broad biogeographic definition.
Researchers, however, cannot agree on one definition of sympatric speciation, leading some to suggest that we drop the term completely (Fitzpatrick et al. 2008, 2009; Coyne 2011). Definitions of sympatric speciation can be classified as biogeographic (Mallet et al. 2009) or population genetic (Fitzpatrick et al. 2009; see Bolnick and Fitzpatrick 2007). Biogeographic sympatric speciation, or speciation in sympatry, can be defined as speciation despite the realized or unrealized potential for unrestricted gene flow due to geographic overlap. Whereas, the definition of population genetic sympatric speciation is more restrictive: speciation despite unrestricted gene flow or random mating, at least at the onset of lineage divergence. Thus, a speciation event classified as population genetic sympatric speciation can typically also be classified as biogeographic sympatric speciation, but not vice versa. Population genetic definitions of sympatric speciation are further subdivided between those where random mating is considered with respect to the genotype at the loci under disruptive selection or the birth place of the mating partners (Gavrilets 2004).
While the differences in definitions of sympatric speciation are slight, they lead to conflicting classifications of speciation events (reviewed in Bolnick and Fitzpatrick 2007). We illustrate this point with a hypothetical example contrasting the use of sympatric in terms of biogeography and in terms of speciation. Consider two species that are sympatric in terms of their broad-scale biogeography. If these species are sister taxa and it is known that their lineages diverged from a single population that occupied the same biogeographic range, this case of speciation in sympatry can be classified as biogeographic sympatric speciation. If the lineages speciated by becoming specialized in different habitats within their ranges, and if genetic exchange is inherently non-random between habitats, then the definition of population genetic sympatric speciation is not met. It can be further argued that the species are not sympatrically distributed on a finer geographic scale, and thus the speciation was not truly sympatric but rather microallopatric (Getz and Kaitala 1989). On the other hand, Maynard Smith (1966) notes that cases where a population occupies two niches in a heterogeneous environment and mating takes place in the natal niche leading to automatic isolation and the evolution of separate species, might be better classified as allopatric due to “behavioral isolation”. The key factor in Maynard Smith (1966) suggesting this scenario is allopatric seems to be the fact that reproductive isolation was a byproduct of niche affiliation whether or not geography played a role. These examples indicate that there could be a consensus on what happened in the speciation event but disagreement on whether or not it should be classified as sympatric speciation, and in order to move forward, we must avoid getting mired in irresolvable semantic arguments.
In the interest of clarity, it is important for researchers to specifically define the terminology they use—here we use the term sympatric speciation in its most inclusive biogeographic sense, as defined above, where populations are not geographically isolated on a broad scale. We use the term allopatric speciation to refer to cases where only broad scale geographic barriers act to stop gene flow and confer reproductive isolation. In parapatric speciation, both geographic and non-geographic barriers contribute to reproductive isolation. We are not endorsing one definition of sympatric speciation over another, but in a review of speciation in sympatry, we do not want to exclude anything that an appreciable proportion evolutionary biologists subscribe to as sympatric speciation.
Ecological Speciation and Divergence-With-Gene-Flow
The terms ecological speciation and divergence-with-gene-flow were both coined to move beyond debating the biogeographic definitions of speciation and yet neither term quite captures and isolates the controversial aspects of the targeted underlying process. Ecological speciation is interpreted by some to describe the situation where natural selection drives reproductive isolation without the aid of geographic isolation (see Via 2009), but has been repeatedly defined to involve any speciation event where natural selection contributed to the divergence of populations (Schluter 2001; Rundle and Nosil 2005; Schluter 2009) and thus potentially includes most speciation events (Sobel et al. 2010). Sobel et al. (2010) ask “when is speciation nonecological?” Ecological speciation includes quite divergent scenarios, such as where (1) two populations become allopatrically isolated and the populations evolve to have different characteristics due to differential selective forces, or (2) in secondary contact one population undergoes character displacement due to competition for a common resource, or (3) a panmicitic population splits into two populations in sympatry due to disruptive natural selection. In a discipline where consensus is hard to come by, evolutionary biologists tend to agree that ecological adaptation has been involved in most speciation events in sexually reproducing organisms (Templeton 2008; Schluter 2009; Sobel et al. 2010), although allopolyploidy could qualify as non-ecological. Classifying the factors that promote reproductive isolation as ecological or not (Dobzhansky 1937; Nosil et al. 2009b) and determining the relative contribution and order of occurrence of each isolating factor is a promising avenue by which divergence events can be classified by ecological influence (Sobel et al. 2010).
Allopatry, reproductive isolation, reproductive incompatibility and gene flow
It is critical to distinguish between the states of allopatry, lack of gene flow, reproductive isolation, and reproductive incompatibility to prevent confusion in the discussion of lineage divergence. Allopatry, the state of two populations inhabiting non-overlapping geographic ranges, may or may not be accompanied by zero gene flow and reproductive isolation, but does not inherently confer reproductive incompatibility. Reproductive isolation is characterized by the lack of gene flow, which can be elicited by reproductive incompatibility, geographic isolation (allopatry) or a number of other factors such as allochronic reproduction, polyploidy, selection, or any combination of these factors. Reproductive isolation or lack of gene flow among populations, however, does not confirm reproductive incompatibility (Mayr 1963). The state of complete reproductive incompatibility does not inherently require present or past allopatry but does imply reproductive isolation and the absence of ongoing gene flow. It should be noted that when testing for the presence of gene flow using population genetic methods, the temporal resolution is often quite coarse and it can be difficult, but not always impossible, to determine whether gene is contemporary or recently ceased
Stages of Lineage Divergence in Sympatry
Disruptive Selection Without Substantial Reproductive Isolation
Genomics for the masses: reduced representation sequencing and selection
The primary impediment to identifying candidate loci that experience disruptive natural selection has been the cost of discovery—genomes are huge relative to affordable sequencing technology and it has been virtually impossible to align an appreciable portion of the genomes of two individuals at a price conducive to population genetic analysis where many individuals must be compared. Reduced representation genomic sequencing allows one to sequence a tractable portion of the genome at the same nucleotide positions among individuals. One effective method is restriction site associated DNA sequencing (RADseq; Miller et al. 2007; Baird et al. 2008), where only the DNA adjacent to restriction sites is sequenced using Illumina next generation sequencing technology. In an attempt to identify genome-wide signatures of natural selection using RADseq, Hohenlohe et al. (2010a) compared the genetic differentiation at ~40,000 informative SNPs among marine populations of three spine-sticklebacks (Gasterosteus aculeatus) and populations that have recently and independently colonized freshwater habitats and are phenotypically convergent. Hohenlohe et al. (2010a) found parallel patterns of genetic differentiation between marine and freshwater populations in many chromosomal locations, indicating that convergent evolution may be occurring through parallel changes at a genome-wide scale. While not necessarily an example of divergence despite continuous gene flow, the important finding was that genome scans can be used to identify patterns of disruptive selection. Importantly, RADseq does not require a reference genome and can be performed on any species at a cost comparable to that of sequencing a handful of diploid loci, thereby opening the world of genomics to all taxa and facilitating the discovery of needles of disruptive selection in a genomic haystack (Allendorf et al. 2010)
Due to the inherently heterogeneous nature of most natural settings, selectively driven divergence in allele frequencies on at least a small subset of the genome is likely to be very common in populations, but is not necessarily easily demonstrated. One must be able to differentiate between divergent selective pressures and random variation. Even statistical outliers in a genome scan might reflect random variation rather than selection. It is also quite possible that loci experiencing weak or intermediate levels of selection will not be detectable as statistical outliers (Michel et al. 2010), thus experiments and creative sampling designs might be necessary to detect all loci that are experiencing disruptive selection.
Studies of recently-diverged, broadly sympatric populations, such as in the Rhagoletis pomonella hawthorn and apple maggots (Feder 1998; Feder et al. 2003) or the marine intertidal Littorina saxatilis snails (Johannesson et al. 1995, 2010), are critical in understanding the effects of selection on population divergence, but these systems already exhibit partial reproductive isolation. In order to obtain a complete picture of the initiation of divergence-via-selection-with-continuous-gene-flow, one must study the transition from panmixia to divergence at loci under selection as it occurs. This task seems daunting in natural systems, but in the laboratory, insect-host systems lend themselves well to the study of the first steps in the divergence process (see Fry 2003). For example, the early stages of a laboratory induced host shift are being investigated in Callosobruchus maculatus beetles (Messina et al. 2009; Messina and Jones 2011). A host shift was induced by presenting the beetles with one atypical host, Lens culinaris lentil seeds. Following initially poor fitness (larval survival <1 %) and a bottleneck, larval survival rose to >80 % in <20 generations. The F1 generation and backcrosses demonstrated an effect of additive inheritance likely related to the ability to detoxify the lentil seeds. This type of system presents an excellent experimental opportunity wherein the insects can be presented with their ancestral host and a novel host in varying proportions while controlling the level of gene flow between the ancestral line and the new host line to explore the consequences of varying selection and gene flow on the initial stages of divergence. Further, the development of the genomic signature of the earliest stages of divergence can be tracked, i.e. the emergence of genomic islands of differentiation. Is there a threshold level of competition that drives the host shift, or a threshold level of gene flow that prevents host shift? How many physically unlinked loci exhibit enhanced divergence, how quickly do genomic regions of divergence evolve, and how broad are the genomic regions of divergence? Combining laboratory experiments designed to manipulate the conditions affecting divergence with genome-wide surveys of genetic differentiation would advance our understanding of the initial stages of divergence, and the experiments we suggest could identify a role for density dependence in initiating sympatric divergence.
Evolution of Reproductive Isolation
Partial reproductive isolation is necessary to elicit genome-wide allele frequency shifts in neutral loci, but it may take several generations before these allele frequency shifts occur. Migration, mutation rate and drift will combine to determine how much time must pass before a diagnostic genetic pattern can be detected. However, when a population is unevenly split by divergence, with some common and most rare alleles at various loci lacking in the smaller partition, shifts in allele frequencies could be detectable immediately. Hence, different conditions can lead to different patterns of genetic variation (Feder and Nosil 2010). Critically, when partial reproductive isolation is achieved, selection and/or any other factor controlling isolation may or may not interact to drive further reproductive isolation. It is also worth noting that the estimation of genetic differentiation for SNPs is fundamentally different than for markers with more than two alleles when using Wright’s (1943) FST and present day analogues due to a strong inverse relationship between allelic diversity and FST—high allelic diversity results in small estimates of differentiation (Hedrick 2005; Meirmans 2006; Jost 2008). As genome-wide surveys shift from the analysis of SNPs to nucleotide sequences and SNP haplotypes, estimates of genetic differentiation will have to account for this bias (reviewed in Bird et al. 2011b).
Population Genetic Models, Natural Selection and Reproductive Isolation
The perceived difficulty in the evolution of sister species in sympatry is achieving reproductive isolation despite the homogenizing forces of gene flow and chromosomal recombination. Researchers seek to elucidate the roles of fine scale spatial partitioning and natural selection in the evolution of reproductive isolation and determine if complete cessation of gene flow due to geographic isolation is required for complete reproductive isolation. In order to determine how a population progresses from panmixia to two independently evolving lineages in sympatry, one must look more closely at the factors contributing to reproductive isolation.
Modeled conditions contributing to the evolution of reproductive isolation in sympatry
Models of sympatric speciation are typically focused on the factors that contribute to reproductive isolation because the evolution of reproductive isolation via selection despite continuous gene flow is not a forgone conclusion. Additionally, once complete reproductive isolation is achieved, speciation will follow if conditions remain the same. This brief review summarizes some speciation modeling efforts to elucidate the factors affecting reproductive isolation. It is noteworthy that the vast majority of modeling efforts adopt a population genetic definition of sympatric speciation, but we present the following list with regards to the more permissive biogeographic definition of sympatric speciation. Also note that while we detail simple models in the text where one locus affects one trait, this list includes models where traits could be controlled by several loci
Disruptive selection: Soft selection, where deaths of individuals with less fit genotypes substitute for the background level of mortality rather than add to it (hard selection), increases the chances that reproductive isolation will evolve (Gavrilets 2005). Strong soft disruptive selection is more likely to lead to isolation and population divergence (Barton 2010), especially when the selective regime is spatially heterogeneous (Gavrilets 2005). Intermediate levels of hard disruptive selection leads to population divergence in most models (Gavrilets and Vose 2007; Gavrilets et al. 2007; Thibert-Plante and Hendry 2011). In contrast to competition for resources, for sperm competition, speciation is more likely in broadcast-spawners when population density and competition to fertilize eggs is high and hard selection is strong (Tomaiuolo et al. 2007). Further, if the mortality of individuals due to disruptive selection occurs prior to reproductive maturity, speciation will be more likely because of a stronger relationship between fitness, genotype, and niche affiliation (Gavrilets 2005). Indeed, Barton (2010) shows that if an extreme specialist cannot survive to reproduce in the wrong niche, then speciation occurs, otherwise an intermediate generalist phenotype evolves. Taking a genomic view, the evolution of reproductive isolation should become more likely as more physically unlinked loci simultaneously experience multifarious disruptive selective pressures because physical linkage with loci controlling reproductive isolation is more likely to occur (Feder and Nosil 2010)
Recombination: Recombination resists the evolution of reproductive isolation by breaking the association or linkage disequilibrium between a locus under divergent selection and the loci contributing to reproductive isolation (Felsenstein 1981). Complete reproductive isolation becomes more likely as the number of traits required to attain reproductive isolation decreases (Smadja and Butlin 2011), especially when the evolution of linkage disequilibrium is required for traits to become associated. Close linkage between a locus under disruptive selection and the loci controlling reproductive isolation increases the likelihood that isolation will evolve by decreasing the rate or likelihood of recombination (Gavrilets 2005). AIT facilitate the build up reproductive isolation by reducing the number of loci that have to resist the effect of recombination. If a locus under disruptive selection and loci controlling reproductive isolation occur in genomic regions with decreased recombination, the evolution of reproductive isolation is more likely (reviewed in Nosil and Feder 2012)
Number of loci controlling assoratative mating: Population divergence becomes more likely when fewer loci underlie traits under disruptive selection and traits that confer assortative mating (Gavrilets 2004; Gavrilets et al. 2007; Gavrilets and Vose 2007)
Natal niche fidelity/migration: In a haploid Maynard Smith (1966) model, low migration or a substantial level of natal niche fidelity greatly increases the chances that reproductive isolation and speciation will occur due to decreased production of heterozygotes (Gavrilets 2005). If migration occurs in juveniles before selection removes individuals then reproductive isolation is more likely because of a positive relationship between fitness and natal niche (Gavrilets 2005). Additionally, the migration rates for each sex must be different for speciation to occur in a haploid locus model
Heterozygote fitness: Heterozygotes at the locus under divergent selection must be much less fit than homozygotes in the “correct” niche for speciation by divergent selection to occur (Maynard Smith 1966). In a model of divergence for the Littorina saxatilis intertidal snail, where there is a region of hybrid superiority sandwiched between two regions alternate ecotype superiority, speciation is inhibited by hybrid viability (Sadedin et al. 2009). Pardoxically, the zone of hybrid superiority facilitated the formation of stable ecotypes (Sadedin et al. 2009), because hybrid superiority in the intermediate zone favored the generation of genetic variation required to invade new habitat. Thus, the same conditions that increased the chances of developing stable alternate ecotypes also resisted complete speciation
Automatic isolating traits: When a locus under disruptive selection contains alleles that simultaneously cause assortative mating, speciation is more likely (Maynard Smith 1966). In fact, any situation where assortative mating is automatically associated with a locus under disruptive selection, such as close linkage, increases the probability of speciation (Dieckmann and Doebeli 1999; Gavrilets 2004; Servedio et al. 2011). Speciation, however, can require more than strong assortative mating (Thibert-Plante and Hendry 2011). Additionally, assortative mating that is automatically conferred by niche affiliation, increases the probability that reproductive isolation will evolve (Gavrilets and Vose 2007; Barton 2010)
Distribution of resources or niches: Disruptive selection due to competition for a limited unimodally distributed resource can lead to speciation (Dieckmann and Doebeli 1999) but a bimodal distribution of resources, such as a bimodal distribution in seed size for granivorous birds, makes speciation more likely (Thibert-Plante and Hendry 2011). Changes over time in the environmental conditions (e.g. exhaustion of resources) might lead to concomitant changes on the intensity of selection
Genetic variance/mutation rate: Genetic variance is required for the emergence of novel genotypes able to colonize a new niche (Gavrilets et al. 2007). High levels of genetic variance also allow greater amounts of niche differentiation and reproductive isolation (Barton 2010). It follows that higher mutation rates will accelerate the evolution of reproductive isolation (Bolnick 2004; Waxman and Gavrilets 2005b) and increase the likelihood of speciation
Population size: In a model of the Nicaraguan crater lake cichlids (Amphilophus spp.), speciation is most likely if populations are not small because the low genetic variation inherent in small populations resists or delays speciation (Gavrilets et al. 2007). As modeled Amphilophus populations become large, the probability of the evolution of a generalist that exploits both niches increases, therefore very large populations might also resist speciation (Gavrilets et al. 2007). Further, Sadedin et al. (2009) suggest that when the spatial area favoring intermediate forms of Littorina saxatilis, and thus the carrying capacity, is small relative to the area favoring alternate forms, then speciation is more likely. Otherwise, when the area favoring intermediates is large, stable ecotypes may form under certain conditions but isolation remains incomplete. Furthermore, effective population sizes (coupled with low migration) favor the formation genomic islands of differentiation when selection affects one or few loci
Gene flow: Geographic structure and limited dispersal favor the build up of reproductive isolation by promoting population partitioning and divergence between ecotypes. Population connectivity was the most relevant factor found to explain the differences in progress towards ecological speciation observed in in three species of lizards in the same selective environment exhibiting phenotypic convergent evolution and different degree of neutral differentiation between ecotypes, where the habitat specialist with less suitable habitat showed the highest genetic divergence (Rosenblum and Harmon 2011). However, Barton (2010) finds that small amounts of gene flow between divergent demes in two different niches increases genetic variation and facilitates speciation. Similarly, hybridization between two divergent lineages can generate the diversity necessary to exploit new niches (Jiggins et al. 2008)
Assortative mating/mate choice: When there is little cost associated with being choosy (Bolnick 2004; Gavrilets 2004; Waxman and Gavrilets 2005b), strong mate choice makes reproductive isolation and speciation more likely (Gavrilets 2005; Gavrilets et al. 2007) and is more important than the nature of resource competition (Thibert-Plante and Hendry 2011). Increasing the penalty for being choosy, i.e. when passing up a mating opportunity results in decreased fitness, increases the amount of time for reproductive isolation to build up and may prevent speciation (Bolnick 2004)
Sexual dimorphism: If a population lacks the capacity for sexual dimorphism or disruptive selection can not lead to sexual dimorphism, then speciation is more likely (Bolnick and Doebeli 2003). If the evolution of sexual dimorphism is possible, speciation is still possible, but less likely because adaptive splitting may favor a stable sexual dimorphism where each sex partitions the resource rather than the evolution of reproductive isolation with two lineages partition the resource
Stochasticity: Due to an inability to achieve a set of parameters that consistently leads to speciation, Sadedin et al. (2009) and Thibert-Plante and Hendry (2011) find that stochasticity is an import determinant of whether reproductive isolation is achieved and speciation occurs
Traditional population genetic models that explain how differential selection between two niches can favor reproductive isolation often involve two loci (Orr 1996; Bolnick and Fitzpatrick 2007; Pinho and Hey 2010). These models typically are constructed with at least one adaptive locus (Locus 1 where alleles A and a confer fitness advantages according to niche) experiencing divergent selection between the two habitats (for the sake of simplicity we will define the niches as habitats) and a locus controlling mating fidelity (Locus 2, with alleles M and m) with respect to the Locus 1 genotype. The locus controlling mating fidelity can behave according to a one or two allele model (Felsenstein 1981). In one allele models, fixation of a single allele in Locus 2 (MM) confers assortative mating between individuals harboring the same genotype at Locus 1. For example, individuals with the AAMM genotypes will only mate with others harboring AAMM, and aaMM will only mate with aaMM. Consequently, recombination does not resist divergence because all individuals are fixed for the M allele. In contrast, with two allele models genotype MM confers mating fidelity with genotype AA and genotype mm with aa. In the absence of recombination, selection against hybrids will favor AAMM and aamm genotypes and lineage bifurcation is likely. Recombination, however, will result in non-optimal aaMM and AAmm genotypes resisting sympatric population divergence (Felsenstein 1981). Recombination rates can be reduced and linkage disequilibrium can be increased by strong selection against hybrids, tight physical linkage between loci, or by loci residing in chromosomal regions where recombination rates are decreased. Thus when the genetic architecture of assortative mating is well described by a two allele model, lineage bifurcation due to disruptive selection is less likely than for a one allele model. Dieckmann and Doebeli (1999, 2005) use an adaptive dynamic model to challenge whether recombination resists sympatric population partitioning and speciation, but the relevance of their model’s assumptions to natural populations is uncertain (Waxman and Gavrilets 2005a, b; Gourbiere and Mallet 2005).
In cases where assortative mating is a function of niche residency rather than the locus under divergent selection among niches (see Barton 2010), an alternative three locus model for divergence may be appropriate. In addition to a locus under divergent selection conferring fitness advantages in different niches (Locus 1) and a locus for mating fidelity (Locus 2), a third locus controlling natal niche fidelity (Locus 3) is necessary for reproductive isolation to evolve. To be clear, we are talking about a scenario where the locus for assortative mating controls mating with individuals within and outside of the niche occupied, rather than with the genotype at Locus 1 that confers fitness advantage according to niche. For example, consider a hypothetical population of broadcast-spawning limpets that occupies intertidal and subtidal habitats where genotype AA in Locus 1 is more fit intertidally, aa is more fit subtidally, and Aa has intermediate fitness in both habitats. As explained above, the genotype at Locus 2 (one or two allele model) would control mating fidelity by habitat (intertidal with intertidal and subtidal with subtidal) which could decrease the probability of producing Aa heterozygotes depending upon the strength of selection (see Maynard Smith 1966; Gavrilets and Vose 2007). Random settlement of larvae, however, resists reproductive isolation by shuffling the Locus 1 genotypes among habitats in each recruitment event. In this scenario there is strong selection for natal niche fidelity—selecting the parental habitat—which, if achieved, decreases the chance of producing Locus 1 heterozygotes and contributes to reproductive isolation. Examples of cases where a locus controlling natal niche fidelity is necessary for reproductive isolation might include the most recent common ancestors of the Lord Howe Island Howea palms (Savolainen et al. 2006; Babik et al. 2009) and the Hawaiian Cellana limpets (Bird et al. 2011a), where seeds or larvae do not necessarily germinate or settle, respectively, in the habitat where they are most fit. Locus 3 could be characterized by a one allele or two allele model, and recombination will resist the evolution of reproductive isolation in the two allele models. Increasing the number of loci involved in the build up of linkage disequilibrium to achieve reproductive isolation should decrease the probability that reproductive isolation will be achieved due to recombination, but automatic isolating traits, including allelic pleitropy (Maynard Smith 1966; Smadja and Butlin 2011; Servedio et al. 2011) can effectively decrease the number of loci involved in the evolution of reproductive isolation.
Automatic Isolating Traits (AIT)
The success of Mayr’s (1963) description and defense of allopatric speciation as the primary mode of speciation among sexually reproducing organisms in nature is at least partially rooted in the fact that complete mating and natal niche fidelity is an automatic byproduct of the commonly observed condition of geographic isolation. Hence, it is easy to envision that geographically isolated populations will diverge or have diverged. It seems much more difficult for assortative mating to evolve in populations that are not geographically isolated but experience divergent selective pressures (Maynard Smith 1966; Felsenstein 1981; Gavrilets 2004). Mating fidelity (and possibly niche fidelity) would have to evolve despite gene flow and recombination.
Is the evolution of reproductive isolation in sympatry magic?
The widespread adoption of the term “magic trait” is curious. “Magic traits”, as defined by Servedio et al. (2011), are traits encoded by “magic genes” that are subject to divergent selection and that, at least partially, contribute to non-random mating either inherently or by allelic pleiotropy, and thus represent a specific class of AIT. The term “magic trait” was originally derogatory and used to refer to a trait that was thought to be exceedingly unlikely to occur in nature (Gavrilets 2004), one that is under disruptive selection and controls reproductive isolation, and was originally termed “pleitropism” by Maynard Smith (1966). Researchers have subsequently demonstrated that certain traits which automatically confer some level of reproductive isolation do occur more commonly than was appreciated. These traits are certainly not magic, but the term “magic trait” has persisted (Servedio et al. 2011). While admittedly catchy, the term “magic trait” is counterproductive to the field of evolution when reported to the general public. Given the skepticism that the theory of evolution continues to evoke in a non-trivial portion of the general public of the United States of America and other parts of the world for more than 150 years since its inception (Darwin 1859), when describing a process that actually occurs in nature, perhaps a better descriptor than “magic” is in order. It would be unfortunate to have text books where some types of speciation rely on magic. Here we suggest the term “automatic isolating traits”, which encompasses pleiotropisms, “magic traits”, and other traits that decrease the homogenizing effect of recombination on the establishment of reproductive isolation in the face of gene flow
In Fig. 3, we present a simple graphic depicting the impact of AIT, that affect assortative mating and/or natal niche fidelity, on the locus under disruptive selection and whether reproductive isolation is achieved. We further divide mating and niche fidelity with respect to either the genotype of the adaptive locus or niche residence. When both complete mating and natal niche fidelity are AIT, either as byproducts of the adaptive genotype or niche residency (Fig. 3a, b, d, e), gene flow ceases between individuals in the two niches, the adaptive locus becomes alternately fixed, and the lineages are free to diverge. Overall, in cases where there is automatic assortative mating with respect to the adaptive locus, as might be observed in fish with a color polymorphism driven by disruptive selection, then reproductive isolation among genotypes will occur. Interestingly, if there is complete mating fidelity with respect to genotype, but no natal niche fidelity (Fig. 3c), the adaptive locus becomes alternately fixed (given a low level of hybrid fitness or a high level of mating selectivity to counteract recombination) in two independent lineages with no gene flow. Even so, both lineages occur in both niches as long as selection against a non-adaptive genotype is not 100 % efficient. In the juxtaposed scenario where niche fidelity is controlled by the adaptive locus, but there is no mating fidelity (Fig. 3g), genotypes at the adaptive locus sort by niche but homozygotes are constantly produced, resisting lineage bifurcation. In the three other scenarios (Fig. 3f, g, i) genotypes at the adaptive locus do not sort by niche and heterozygotes are constantly formed, resisting lineage bifurcation. For panels f-i, secondary selection on assortative mating and potentially natal niche fidelity must evolve despite gene flow and recombination for the lineages to bifurcate sympatrically.
Genomics and the Evolution of Reproductive Isolation
The concept of divergence hitchhiking (DH), which has emerged from empirical genomic studies of Acyrthosiphon pisum pea aphids, has been reported as a solution to the primary challenge facing sympatric speciation—how does reproductive isolation evolve despite continuous gene flow (Via and West 2008; Via 2009)? Via (2009) contends that if selection drives progressively increasing reproductive isolation, then effective rates of recombination become progressively reduced around loci under strong disruptive selection, thereby allowing genomic regions of differentiation to extend or grow to encompass increasingly distant physically linked neutral loci along the chromosome. The major issue with DH, as a solution to the sympatric speciation debate, is its underlying assumption that reproductive isolation will inherently increase through time due to increased specialization. This is the controversial aspect of sympatric speciation, and it is not directly addressed by DH. Decreased effective recombination rates due to disruptive selection are not proposed to result in increased reproductive isolation via a positive feedback loop. Rather, they are proposed to result in a descriptive animation of a genome scan for differentiation, from initial divergence to speciation, where metaphorical genomic islands of differentiation first rise and then spread across the genome. Feder and Nosil (2010) model DH, mainly to determine if it is possible to replicate the sequence of events proposed in DH by Via (2009). They found that the development and enlargement of genomic islands via DH can be modeled but their formation is restricted to a certain region of the parameter space, with effective population size, number of loci affected by selection, and strength of selection controlling the extension of linkage disequilibrium on both sides of the loci under disruptive selection. In many simulations, before the islands can grow very large, genome-wide neutral variation rises and metaphorically submerges the islands as the genome-wide neutral variation increases in a pattern that Nosil and Feder (2012) term genomic hitchhiking. Feder and Nosil (2010) conclude that increasing the number of loci under disruptive selection increases the chances of speciation, mainly because neutral loci can diverge at the same rate as those under selection due to the assumed progressive increase in reproductive isolation.
If one does not assume that reproductive isolation will inherently evolve, there may be a mechanism by which stochasticity, physical linkage, and recombination combine to promote further reproductive isolation and population divergence that is consistent with allele-based quantitative population genetic models, although we acknowledge at the outset that the conditions under which this can happen are likely to be highly restrictive. Close physical linkage between a locus under disruptive selection and a locus controlling reproductive isolation can lead to at least partial splitting of a population due to reduced recombination rates (Felsenstein 1981); therefore, the number of loci that experience disruptive selection between niches in a natural setting will affect the probability that reproductive isolation will evolve automatically due to linkage with neutral loci that affect reproductive isolation. If some loci experiencing disruptive selection are physically linked, it could further increase the number of hitchhiking neutral loci (Via 2009; Feder and Nosil 2010). For more reproductive isolation to evolve, it is possible that the decrease in gene flow due to hitchhiking will allow allele frequencies to diverge at a second set of loci subject to weaker disruptive selection. If some of the loci subject to weaker selection are physically linked to loci affecting reproductive isolation, then reproductive isolation could increase. It would seem, however, that the circumstances leading to this effect are rare. Studies of hybrid zones often find that many loci, genome-wide, experience high levels of migration (for example see Yatabe et al. 2007), thus physical linkage does not seem to have an ubiquitous impact on whether partially reproductively isolated populations automatically evolve greater levels of reproductive isolation despite continued gene flow. Nonetheless, multifarious divergent selective pressures between niches could increase the probability that partial reproductive isolation will be associated with a locus under disruptive selection.
In a review of the empirical literature, Nosil et al. (2009a) find that in species experiencing partial to complete reproductive isolation (such as Lord Howe Island Howea palms, intertidal Littorina saxatilis snails, and Coregonus clupeaformis lake whitefish), on average, 5–10 % of loci (mostly AFLP data) are characterized as outliers in genome scans for genetic differentiation. It is unknown, however, how many of these loci were outliers at the onset of partial reproductive isolation. If disruptive selection is applied to a panmictic population and automatic isolating traits only result in partial reproductive isolation, the primary manner in which reproductive isolation could increase involves selection against admixture and recombination in a reinforcement-like process.
Chromosomal inversions present a genomic phenomenon that may favor further strengthened reproductive isolation (Nosil and Feder 2012). Chromosomal inversions, as well as any structural or point mutations that reduce recombination for a chromosomal region (Jaarola et al. 1998; Feder et al. 2011) present another avenue by which loci under divergent selection and loci controlling reproductive isolation could evolve and maintain linkage disequilibrium, thereby promoting divergence in sympatry. Our understanding of the relevance of genomic processes to the evolution of reproductive isolation in the face of continuous gene flow is still nascent, but rapidly advancing, and could shed light new light on the topic (reviewed in Nosil and Feder 2012). However, overall, it seems that traditional population genetic principles remain the primary underpinning in models of the evolution of reproductive isolation due to selection.
Post Reproductive Isolation
When complete reproductive isolation evolves between two populations, differential selective pressures and independent patterns of genetic drift are not resisted by gene flow and the two sets of genomes are free to diverge. If the evolution of reproductive isolation is rapid, it could be difficult to decipher between the genetic signature of partial and complete reproductive isolation, but given enough time, differences between the lineages will extend beyond shifts in allele frequencies to differences in the alleles, themselves. Genome-wide, allele frequencies at loci under purifying selection should remain the most similar, but allele frequencies at other loci will diverge dependent upon divergent selective pressures, mutation rate and drift. It follows that deciphering loci under selection from neutral loci is difficult and may require specific sampling designs or experiments (Hohenlohe et al. 2010b; Michel et al. 2010; Nosil and Feder 2012). As allelic sequence divergence progresses, measures of genetic differentiation that incorporate the evolutionary relationship between alleles will be desirable, such as ΦST (Excoffier et al. 1992; Bird et al. 2011b), the coalescent (Kingman 2000), or phylogenetic reconstructions (Stamatakis 2006).
Resumption of Gene Flow Following Isolation and Reproductive Incompatibility
Reproductive isolation can be reversible (Ribeiro and Caticha 2009). While discussions of resumed gene flow typically focus on secondary contact following a period of allopatry, it is conceivable, and possibly likely, that gene flow will resume after the evolution of reproductive isolation despite continuous gene flow. Since the basic unit of evolutionary time in sexually reproducing species is the generation (the average age of breeding adults), as the migration rate (NeM—number of individuals per generation) decreases below 1, gene flow becomes discontinuous (Slatkin 1987) with increasing periods of time without gene flow. It is likely that, in many cases, complete reproductive isolation is achieved not in a singular cessation of gene flow, but after several cycles of gene flow cessation and resumption, like a dripping faucet. Fundamentally, this is a minor departure from divergence via selection with continuous gene flow but is genetically indistinguishable from scenarios where spatial isolation drive gene flow restrictions. It is also possible that, after several generations of selectively driven isolation, gene flow could resume, thereby leaving a genetic/genomic pattern that may be confounded with secondary contact following a period of allopatric isolation. Selective pressures are variable, and as biological communities and the environment changes, it is at least theoretically possible that selectively driven reproductive isolation could decrease when selective pressures change (Ribeiro and Caticha 2009). This highlights the fact that while allopatric and sympatric speciation are often described as residing at opposite ends of a continuum of possibilities for how speciation proceeds, both can be fundamentally similar and differ mainly by the manner in which reproductive isolation is achieved.
Determining Whether Lineage Bifurcation and Speciation was Sympatric
If complete reproductive isolation is not achieved, then lineages are not evolving in complete independence and the divergence process may become arrested, leading Coyne and Orr (2004) to identify reproductive isolation as one of several criteria that must be satisfied to demonstrate that sympatric speciation, or any type of speciation, has occurred. Studies that focus on sympatric lineages or species that are reproductively isolated typically seek to determine what processes are most likely to have contributed to the achievement of reproductive isolation. Coyne and Orr (2004) outline a method of determining whether sympatric speciation is likely to have resulted in presently observed sympatric sister species that entails treating allopatric speciation as the null hypothesis and attempting to disprove this null hypothesis by demonstrating that the lineages are unlikely to have undergone an allopatric phase. Bolnick and Fitzpatrick (2007) provide a supplementary table listing 12 species groups (mainly insects diverging between hosts and fish in isolated crater lakes) that are generally accepted to satisfy Coyne and Orr’s (2004) criteria and 27 more possible examples of sympatric speciation. Since 2007, there have been at least two more sets of sibling species for which Coyne and Orr’s (2004) criteria for sympatric speciation are claimed to have been met, Japanese Hexagrammos spp reef fish (Crow et al. 2010) and endemic Hawaiian Cellana spp. limpets (Bird et al. 2007, 2011a; Bird 2011). In both cases, a skeptic will likely remain unconvinced that sympatric speciation has occurred, thereby identifying the major issue with determining how reproductive isolation was achieved—the process cannot be observed and signatures of what took place fade.
A variety of researchers have built on the initial efforts of Coyne and Orr (2004) to outline a general method for identifying whether sympatric speciation is likely to have led to presently observed sister species. Crow et al. (2010), take the process one step farther by testing the level of pre- and post-zygotic isolation in sympatric sister species and an allopatric sibling species—Coyne and Orr (2004) proposed that the presence of only prezygotic isolation in sympatric sister species could be good evidence that sympatric speciation occurred. Papadopulos et al. (2011) point out that even if sympatric species are not monophyletic, it does not preclude that speciation occurred in sympatry with continuous gene flow. For example, a paraphyletic pattern could be generated by the colonization of a remote habitat by one of the populations diverging in sympatry. Bolnick and Fitzpatrick (2007) take issue with the method of treating allopatric speciation as the null hypothesis in classifying the biogeographic mode of speciation because (1) allopatric speciation is difficult to falsify, leading to an unknown Type II error rate; (2) ad hoc scenarios of how allopatry occurred become more credible than sympatric scenarios without statistical testing; and (3) parapatric speciation is not considered but might be the most common manner in which speciation occurs. Johannesson (2009, 2010) argues that assuming a null hypothesis of allopatric divergence is a biased approach in cases where non-allopatric divergence is suspected, and rather, efforts should be made to reject both allopatric and non-allopatric scenarios leading to reproductive isolation. Bird et al. (2011a) take a slightly different approach by asking which mode of speciation has the most support and attempt to confirm both allopatric and non-allopatric divergence scenarios with the available data, but also note the inherently subjective nature of such efforts. Ultimately, it is extremely difficult (potentially impossible in the majority of cases) to be certain whether reproductive isolation was achieved in allopatry, parapatry, sympatry, or some combination of the three when studying well established sister species. Consequently, we are left to test the likelihood of alternative hypotheses about the scenarios that led to presently observed isolation, report the most well supported hypotheses given the available data, model the system based on available data, and acknowledge the inherent uncertainty in determining whether reproductive isolation was a byproduct of geographic isolation, selectively driven isolation, or some combination of the two.
Model of sympatric speciation in broadcast-spawning invertebrates
Models of sympatric speciation and automatic isolating traits have been reviewed extensively by terrestrial biologists in the past 5 years, but Tomaiuolo et al.’s (2007) model of sympatric speciation in broadcast-spawning marine invertebrates is rarely mentioned. Broadcast-spawners release their gametes into the water column where fertilization occurs. Fertilization success in depends upon the timing of gamete release—somewhat similar to certain flowering plants. Polyspermy, the case of an egg being fertilized by multiple sperm, results in unviable larvae. Polyspermy is density dependent, and thus if population densities are high, strong disruptive selection is applied to spawning timing. Assuming an initially normal distribution of spawning times in the ancestral population, the individuals with mean spawning timing are more likely to experience polyspermy and hence individuals with early or late spawning timing will be more fit. Mathematical simulations show that the model is viable, and may explain population divergence in the Montastraea annularis complex of Caribbean corals (Tomaiuolo et al. 2007). In this model, spawning timing is an automatic isolating trait because it defines the character under disruptive selection, and reproductive isolation is a byproduct of the timing of gamete release
Divergence Hindcasting with Genome Surveys
A major challenge in speciation research is the limited window of time in which observations of a divergence event can be made relative to the, typically, much greater amount of time required for speciation to proceed. The analysis of genetic variation and genomic patterns promise to expand our power of inference into historical events that contributed to the presently observed levels of divergence among sympatric populations, thereby facilitating the discernment between secondary contact and divergence without any period of complete geographic isolation. Theoretically, identifying cases of secondary contact is a problem that can be reduced to determining the timing of gene flow relative to divergence (Niemiller et al. 2008). While presently implemented methods of inferring the timing of gene flow events are suspect (IMa2, Migrate; Sousa et al. 2011; Strasburg and Rieseberg 2011; Gaggiotti 2011), Sousa et al. (2011) suggest that coalescent isolation with migration simulations (Wang and Hey 2010) can be modified to include migration parameters for discrete time periods and additional parameters for the time at which migration rates change, allowing one to test the probability of gene flow timing.
Complex model simulation such as that suggested by Sousa et al. (2011) can be addressed by Approximate Bayesian Computation (ABC, Beaumont et al. 2002). In ABC, the likelihood functions do not need to be specified. Instead, a set of models are treated as categorical variables and posterior probabilities are estimated based on the coalescent using parameter values that are randomly drawn from the prior distribution. The success of this approach relies on choosing the summary statistics that are informative about the parameters of interest. In other words, the summary statistics chosen need to be informative of the time of divergence, the time of migration and/or their correlation. Work in this arena is ongoing, and seems a promising avenue to eventually be able to test the biogeographic history of divergence within and among species (Sousa et al. 2009; Hamilton et al. 2005; Hickerson et al. 2010). ABC allows simultaneous hypothesis testing and parameter estimation even assuming very complex biogeographic models, as well as incorporation of prior information, e.g. dating of a fossil ancestor, timing of relevant geological events or ancient DNA data (Beaumont 2008; Bertorelle et al. 2010; Hickerson et al. 2010; Ilves et al. 2010).
It may be possible to infer patterns of isolation and secondary contact by examining the distribution of genetic divergence times among chromosomes. Patterson et al. (2006) inferred a pattern of isolation and secondary contact followed by isolation after finding that the X chromosomes of chimpanzees and humans diverged 1 my more recently than the autosomes, a bimodal distribution of divergence times. Patterson et al. (2006) reason that a strict allopatric divergence event without secondary contact should result in a unimodal distribution of chromosomal divergence times and thus hominin and chimpanzee lineages initially separated but then exchanged genes before finally separating. Dobzhansky (1937)–Muller (1942) genetic incompatibilities (DMI) which result in intrinsic hybrid inviability are most densely concentrated in the X chromosome (Coyne and Orr 1989; Masly and Presgraves 2007). Patterson et al. (2008) hypothesize that intense selection against hybridization between to the two X chromosome lineages due to DMI led to the extinction of one lineage (also see Tucker et al. 1992). Given that all ancestral autosome lineages were retained, it must also be true that the surviving X lineage was intensely selected for. While the interpretation of Patterson et al. (2006, 2008) has been challenged (Wakeley 2008; Presgraves and Yi 2009), the underlying concept that differences in divergence times of genes under selection and neutral genes could, in at least some circumstances, help to decipher between secondary contact and divergence despite gene flow.
Patterns of divergence timing between regions of chromosomal inversion and non-inversion are one avenue by which differences in divergence timing among genome regions can be diagnostic for a history of allopatry or sympatry in presently sympatric populations. Feder et al. (2011) performed computer simulations to model the establishment of chromosomal rearrangements in “sympatric” and “mixed-model” (allopatry followed by secondary contact) scenarios, finding that each model generates unique predictions. The models rely on inversions reducing recombination and facilitating linkage disequilibrium in selectively favored alleles among loci. In the face of gene flow, inversions that maintain linkage disequilibrium among favored alleles are, themselves, selectively favored and go to fixation. In strictly allopatric populations, there is little selective advantage of non-deleterious inversion and thus inversions do not often go to fixation, which is supported by the observation in Drosophila spp. that eight of nine allopatric sister species pairs do not exhibit fixed inversions, while five of six sympatric pairs do (Noor et al. 2001). Thus, it is theoretically possible to distinguish between divergence with continuous gene flow and allopatric divergence followed by secondary contact. The secondary contact model predicts that inversions in different regions of the genome will share divergence times that are older than other genomic regions because the inversions occurred and became established prior to the allopatric period, do not readily admix and thus reflect the age of original allopatric divergence (Feder et al. 2003, 2011). In contrast, the sympatric model predicts that divergence times for inverted regions should be randomly distributed because the divergence times mark the inversion events as they occur.
Our ability to infer past evolutionary events will increase with our knowledge of the genomic make-up and allelic variation of species within lineages. The study of the evolution of the human lineage is such an example, where the availability of highly accurate whole genome sequence information of modern humans and a growing body of information on the human allelic variation (Altshuler et al. 2010) provide indirect venues to infer the evolutionary history of the lineage. A recent investigation of the allelic variation in isolated groups in sub-Saharan African populations identified three genomic segments in autosomal non-coding regions exhibiting very deep haplotype divergence and very high levels of linkage disequilibrium in hunter-gatherer groups. This pattern of polymorphism is consistent with a history of ancient admixture, the introgressions having originated via relatively recent interbreeding (approximately 35 kya) with hominim forms that diverged from the ancestors of modern humans approximately 700 kya (Hammer et al. 2011). Whole-genome sequencing of the extant populations and/or, in the hypothetical case that remnants of the ancestors of these lineages were discovered, ancient DNA sequence analysis (Chan et al. 2006) could provide additional data to assist in interpreting present day patterns of DNA variation.
Systems that Facilitate the Identification of Sympatric Speciation Due to Selection
The confluence of a variety of circumstances can increase the probability of correctly concluding that sympatric speciation has occurred. Sibling lineages that are geographically isolated in a common, restricted location, suggest that there is a bound on the geographic range under which the divergence occurred and reduces the probability that range changes can explain the presence of sympatric sibling lineages. For example, endemic sympatric sister lineages residing in small, isolated oceanic archipelagos or crater lakes are cases in which non-sympatric scenarios seem unlikely (Coyne and Orr 2004). Sibling species being allopatrically distributed among isolated regions make it easier to identify, barring extinction, when allopatrically derived isolation drives speciation among isolated regions. Small, isolated locations often exhibit a depauperate fauna and flora, potentially increasing ecological opportunity for natural selection to drive divergence in taxa that might not otherwise diverge, as is proposed for adaptive radiation (Schluter 2000). It is also important to recognize that the scale of geographic isolation is relative to the spatial scale of gene flow (Kisel and Barraclough 2010).
In the absence of isolation, Crow et al. (2010) propose that sister species exhibiting an extensive sympatric distribution combined with little restriction to gene flow within the species’ ranges are also fruitful scenarios in which to investigate sympatric speciation. Given the broad geographic extent of gene flow in many marine species of invertebrates and teleost fishes with planktonic larvae (Weersing and Toonen 2010) and the relative paucity of marine speciation studies when compared to terrestrial and freshwater habitats, the oceans may harbor numerous undiscovered examples of putative sympatric divergence and speciation and at the very least, divergence via natural selection despite continuous gene flow.
Particular taxa seem to be more predisposed to the evolution of population partitioning and speciation in sympatry. Specialist phytophagous insects seem to be a hotbed of sympatric speciation (Bolnick and Fitzpatrick 2007), possibly due to the discrete nature of the host resource and a tendency to mate on the host plant (potentially due to AIT). Many contend that intense sexual selection in Lake Victoria Haplochromis has led to speciation within a single lake (Seehausen 1997). Krug (2011) identifies various groups of marine gastropods that have an overabundance of young sympatric sister species, some of which may have evolved in sympatry. Investigations of the similarities among taxa where evolution in sympatry is a likely explanation of currently observed diversity could be a fruitful line of inquiry.
Parallel patterns of divergence are championed as desirable traits for studying sympatric divergence (Johannesson et al. 2010). It is also desirable to study sympatric lineage divergence in set of sympatric sibling lineages exhibiting a variety of divergence levels (Marie Curie SPECIATION Network 2011; Merrill et al. 2011), such as in the Nicaraguan crater lake Amphilophus cichlid fish (Barluenga et al. 2006; Barluenga and Meyer 2010) and Hawaiian Cellana limpets (Bird et al. 2011a). The combination of varying stages of divergence, potential sexual selection, and isolation make species like the cichlid fish Amphilophus one the most promising systems in which to study sympatric speciation.
How Common is Sympatric Population Divergence and Speciation?
Sympatric speciation, like adaptive radiation (Schluter 2000), is a biological phenomenon that is considered to be a rare or special case of evolution (Coyne and Orr 2004; Schluter 2009). Despite the inherent uncertainties in determining the biogeographic mode of speciation, it is widely accepted that geographic isolation is likely to play a role in the majority of speciation events (Mayr 1963; Coyne and Orr 2004). It is also well appreciated that selection, whether natural or sexual, also plays a prominent role in speciation (Darwin 1859; Dobzhansky 1937; Mayr 1963; Coyne and Orr 2004; Schluter 2009; Sobel et al. 2010). Despite the challenges in determining if sympatric sibling species evolved in sympatry, researchers have attempted to roughly estimate the frequency with which sympatric speciation may have contributed to contemporary biodiversity relative to allopatric and parapatric speciation. It should be noted that most of these methods rely on observed species ranges, but present day range may not reflect past distributions (sympatry may occur after divergence in geographic isolation, or allopatry may occur after divergence without geographic isolation) so there is a degree of uncertainty involved with such efforts. Further, while sympatric sister species receive large amounts of scrutiny on whether they were always sympatric, allopatric and parapatric sister species typically receive no scrutiny as to whether they may have once been sympatrically distributed that could lead to overestimation of prevalence.
Noting the caveats and pitfalls of estimating the prevalence of a mode of speciation in nature, there seems to be growing divide in the perceived prevalence of sympatric speciation in terrestrial and marine habitats. Phylogenetic investigations of the prevalence of sympatric sister species resulting from a single colonization in island birds found little evidence for sympatric sister species, and thus no evidence supporting sympatric speciation (Coyne and Price 2000). Losos and Schluter (2000) investigated the relationship between island size and number of Caribbean Anolis lizard species originating within each island and found that sister species do not originate on small islands, indicating that geographic area is important for speciation to occur. Barraclough and Vogler (2000) predict that speciation allopatry, parapatry, and sympatry will result in different patterns of relatedness and range overlap based on phylogenetic simulations of lineage divergence followed by geographic range shifts. When comparing real phylogenetic and range data from eight clades (birds, fish, and insects) composed of ~20 species each, only Rhagoletis pomonella was consistent with model predictions for speciation in sympatry. Following the same model, Krug (2011) compares existing phylogenetic and range data for a variety of marine gastropods and, in stark contrast to results from terrestrial and freshwater species, finds that many recently diverged sister species are sympatric. At the very least, speciation and range expansion in marine gastropods follows a substantially different pattern than was observed in the birds, insects, and fish investigated by Barraclough and Vogler (2000). Krug (2011) highlights that in addition to sympatric speciation, allopatry followed by secondary contact (transient allopatry), rapid evolution of sperm and egg recognition proteins, or a shift in life history from planktonic to direct development could explain the relative abundance of sympatric sister gastropod species. Briggs (2007) suggests that sympatric speciation instigated by intense competition for food resources may explain a “considerable” portion of tropical reef fish diversity in the Indo-Pacific and Atlantic centers of marine diversity, which are located in the East Indes Triangle and the southern Caribbean Sea, respectively. It has been noted that the Modern Synthesis largely ignores marine species (Love 2009), and the phylogenetic work of Krug (2011) and others (such as Hellberg 1998, Tegula snails; Munday et al. 2004, Gobiodon fish; Rocha et al. 2005, Halichoeres fish; Rocha and Bowen 2008, review of reef fishes; Kelly and Eernisse 2008, Mopalia chitons; Crow et al. 2010, Hexagrammos fish, Johannesson et al. 2010, Littorina saxatilis snails; Bird et al. 2011a, Cellana limpets) that identify several sympatric sibling lineages indicates that much remains to be learned about speciation and the prevalence of divergence despite continuous gene flow in the ocean relative to more well-studied habitats.
In perhaps the most extensive investigation of the prevalence of sympatric speciation to date, Papadopulos et al. (2011) reconstructed chloroplast DNA phylogenies for all of the plant genera on Lord Howe Island (~16 km2), off the coast of Australia in the Pacific Ocean. These phylogenies were used to infer the manner in which each plant species on Lord Howe Island arose. In all, 55 % of the species were determined to be colonists that exist elsewhere, 25 % are endemic and allopatrically derived, 8 % are putative endemic sympatric sister or sibling species, 5 % of species may have originated in allopatry or sympatry, 1 % are the result of hybrid speciation, and there is not enough data to assess the origins of 6 % of the species. All 20 of the species classified as arising in sympatry were found to occupy different habitats. Additionally, Papadopulos et al. (2011) find a significantly higher frequency of endemic congeners (57 %) than non-endemic congeners (34 %), consistent with expectations where at least some of the endemic species were the result of sympatric speciation. Coyne (2011) points out that the phylogenetic data presented by Papadopulos et al. (2011) is based on chloroplast DNA sequence, and confirmation with nuclear sequence is necessary to bolster their findings. Additionally, approximately half of the sympatric sister species determinations (4 % of all species) are based on phylogenies where more than 40 % of potential sister species are missing, and thus are more questionable assignments, but Papadopulos et al. (2011) contend that the preponderance of data indicate that sympatric speciation has occurred multiple times on Lord Howe Island. Sympatric speciation is not the dominant mode of speciation in Lord Howe Island plants but it certainly may be more prevalent than expected.
It would be quite interesting to apply some of the techniques listed here at the subspecies level. While Coyne and Orr (2004) accentuate the importance of studying good species pairs, Via (2009) contends that studying speciation at earlier stages of divergence is relevent to speciation research. We agree with both, and advocate studies of divergence at every stage. The prevalence of sympatric disruptively selected alleles, ecotypes, and species could illuminate our understanding of the prevalence of sympatric divergence in nature. What is even more certain is that studying evolution across the early stages of lineage divergence will often reveal more details of the mechanisms that generate reproductive isolation than studying good species that are already reproductively incompatible.
Sympatric lineage divergence due to selection has come of age in the wake of the PCR revolution and broad DNA sequencing of populations. Discussion has turned from whether or not sympatric speciation can happen to how often and under what conditions is population and lineage divergence in sympatry most likely. Lineage divergence and speciation are driven by population level processes acting on individuals and their genes, and recognition that operational species concepts based on single diagnostic factors, such as reproductive incompatibility, are not sufficient to describe species is a step forward. Speciation occurs along a continuum of divergence and the body of research on speciation must necessarily span this continuum. Because research experiments and the researchers observing evolution exist on a smaller time scale than that of the lineage formation and divergence process, empiricists are forced to study snap-shots of the process. Here, we divided the divergence continuum into four potential sections (panmixia, divergent selection without reproductive isolation, divergent selection with evolving reproductive isolation, and divergence after complete reproductive isolation), but other strategies are possible. We suggest that all stages of the continuum require improved understanding, but mechanisms will be elucidated by focusing on taxa where reproductive isolation is incomplete. In a similar sense, as many have noted, relegating ones observations to only sympatric scenarios where selection solely drives lineage divergence is limiting. Selection and gene flow are antagonistic forces in the context of lineage divergence, and in nature, any imaginable combination of these forces is likely to occur. A primary question is, under what combinations and varieties of selection and gene flow does divergence occur, and to what extent does it advance toward reproductive incompatibility?
In the view of the Modern Synthesis (Huxley 1942), selection and gene flow collide at the level of alleles and allele frequencies, which is a view that continues today. Under strict models of evolutionary divergence based on loci under disruptive selection and loci controlling reproductive isolation, strong selection can resist otherwise unrestricted gene flow, but for selection to drive reproductive isolation, selection typically must also overcome the inter-locus genotype shuffling of recombination. Recombination can be partially or completely circumvented if (1) mate choice depends upon the fixation of an allele in both populations, (2) loci controlling reproductive isolation either are or are physically linked to the loci experiencing divergent selection, (3) divergent conditions affect the expression of characters affecting reproductive isolation, such as flowering time in plants, or (4) gene flow is reduced by factors other than selection, as is the case in Mayr’s (1963) concepts of allopatric and parapatric speciation. Biological organisms and nature are diverse, and the study of divergence, selection, and gene flow are required to fully comprehend the diversity of manners in which these processes interact. For example, Maynard Smith (1966) presciently notes that factors such as allelic pleiotropy could give selection the upper hand against gene flow, but considered it to be biologically unlikely. However, pleiotropy since has been confirmed to occur in nature (reviewed in Servedio et al. 2011; Smadja and Butlin 2011). Studying monophyletic lineages exhibiting a range of divergences and differing degrees of geographic isolation is one avenue that is likely to advance our understanding of lineage divergence and the generation of biodiversity. But, as has been the case thus far, numerous specific studies of a single pair of diverging lineages or polymorphic populations are also likely to contribute to our understanding of lineage divergence. Simultaneous modeling of the genetic and ecological processes is a necessity in order to test hypotheses based on snap shot observations of diverging lineages, and determining how varying conditions affect lineage divergence. Most prominently, the new deluge of data from genome-wide studies of DNA variation and transcriptome RNA expression is presently impacting how we view lineage divergence. Looking forward, genomics is likely to result in the most significant advances in our understanding of speciation.
So what is to come of sympatric speciation? The popularity of sympatric speciation is at a peak (Fig. 1) and if the past is any indicator of the future, it will become less popular. There are some who urge that we abandon the term ‘sympatric speciation’, while others are determining how prevalent sympatric speciation might be. Arguments over the definition of sympatric speciation have not been fruitful, and it is best for those who hold population genetic or biogeographic views of sympatric speciation to agree to disagree. As long as researchers are clear in what it means when they refer to sympatric speciation, then there is probably not a reason to question it.
Why is sympatric speciation important? Like adaptive radiation, sympatric speciation is a special case of evolution that seems to rarely occur under a restricted set of circumstances. Studying biogeographically sympatric lineage divergence advances our knowledge of how genes, individuals, species, and the environment interact to generate biodiversity. Attempts to prove that sympatric speciation, due to selection, has occurred by obtaining “air tight” (Coyne 2007) empirical evidence is likely to be a misguided endeavor. A more promising approach, should one want to assess the biogeographic mode of speciation, is to generate competing hypotheses and formally test their likelihood and sensitivity based on the available evidence. Hybrid spatial, ecological and population genetic simulations and approximate Bayesian computation (ABC) are one way to accomplish this. Efforts to determine the prevalence of sympatric speciation in nature will, necessarily, have to sacrifice confidence in any one diagnosis in order to assess broad patterns among taxa. The nature of observational and experimental research on speciation necessitates that there will always be a certain level of uncertainty. The major challenge is determining what that level of uncertainty is in an unbiased fashion.
As time passes and technology advances, the perceived prevalence of sympatric speciation in natural systems is expanding. Overall, efforts should be heavily focused on the processes generating lineage divergence in nature, which will allow an estimation of how divergence occurred, sympatric or not. To be clear, this means that increased effort should be focused at the sub-species, population level (especially in the ocean) where we can directly observe the mechanisms driving lineage divergence. As an example, rather than asking, “Can selection drive speciation without geographic isolation and a complete cessation of gene flow?”, one could ask “How much lineage divergence is possible given a certain level of selection and gene flow reduction?”. Sympatric lineage divergence could be quite common in nature, with gene flow, recombination, and selection striking a balance that makes sympatric speciation an occasional, special phenomenon. Clearly, much remains to be discovered and our understanding of speciation will continue to evolve during the post Modern Evolutionary Synthesis era of evolutionary biology.
For intellectual discussions that motivated and significantly improved this manuscript, we would like to thank Stephen Karl, Brian Bowen, Richard Grosberg, Luiz Rocha, Matthew Craig, Jonathan Whitney, Maria Pia Miglietta, Anuschka Faucci, Francesco Santini, Giacomo Bernardi, Michael Hart, Bernard Crespi, Stephen Palumbi, John Geller, Steven Morgan, Rosemary Gillespi, George Roderick, Nina Yasuda, Gustav Paulay, Christopher Meyer, Harilaos Lessios, the SICB marine speciation group, and audiences at U. C. Davis, U. C. Berkeley, U. C. Santa Cruz, U. C. Davis’ Bodega Bay Marine Laboratory, Cal. State’s Moss Landing Marine Laboratory, Stanford’s Hopkins Marine Laboratory, Simon Frasier University, University of Connecticut, Texas A&M University-Corpus Christi, Florida International University, and the University of Hawai’i. We also thank the efforts of two anonymous reviewers that helped to substantially improved this manuscript. CEB was funded by a grant from the Seaver Institute, the Hawai’i Sea Grant College Program, and the Papahanaumokuakea Marine National Monument. This is publication number 1491 from the Hawai’i Institute of Marine Biology, 8604 from the School of Ocean, Earth Sciences and Technology at the University of Hawai’i, and XXXX from the Marine Biology Program at Texas A&M University-Corpus Christi.
- Altshuler, D. L., Durbin, R. M., Abecasis, G. R., Bentley, D. R., Chakravarti, A., 1000 Genomes Project Consortium, et al. (2010). A map of human genome variation from population-scale sequencing. Nature, 467, 1061–1073.Google Scholar
- Avise, J. C., & Wollenberg, K. (1997). Phylogenetics and the origin of species. Procedings of the National Academy of Sciences, USA, 94, 7748–7755.Google Scholar
- Barton, N. H. (2010). What role does natural selection play in speciation? Philosophical Transactions of the Royal Society B-Biological Sciences, 365, 1825–1840.Google Scholar
- Beaumont, M. A. (2008). Joint determination of topology, divergence time and immigration in population trees. In S. Matsumura, P. Forster, & C. Renfrew (Eds.), Simulations, genetics and human prehistory (pp. 135–154). Cambridge: McDonald Institute for Archaeological Research.Google Scholar
- Bird, C. E. (2011). Morphological and behavioral evidence for adaptive diversification of sympatric Hawaiian limpets. Journal of Integrative and Comparative Biology, 51, 466–473.Google Scholar
- Bird, C. E., Smouse, P. E., Karl, S. A., & Toonen, R. J. (2011b). Detecting and measuring genetic differentiation. In S. Koenemann, C. Schubart, & C. Held (Eds.), Crustacean issues: Phylogeography and population genetics in crustacea (pp. 31–73). Boca Raton, FL: CRC Press.Google Scholar
- Bolnick, D. I., & Fitzpatrick, B. M. (2007). Sympatric speciation: Models and empirical evidence. Annual Review of Ecology Evolution and Systematics, 38, 459–487.Google Scholar
- Briggs, J. C. (2007). Marine longitudinal biodiversity: Causes and conservation. Diversity and Distributions, 13, 544–555.Google Scholar
- Butlin, R. K., Galindo, J., & Grahame, J. W. (2008). Sympatric, parapatric or allopatric: The most important way to classify speciation? Philosophical Transactions of the Royal Society B-Biological Sciences, 363, 2997–3007.Google Scholar
- Coyne, J. A. (2011). Speciation in a small space. Proceedings of the National Academy of Sciences USA, 108, 12975–12976.Google Scholar
- Coyne, J. A., & Orr, H. A. (1989). Two rules of speciation. In D. Otte & J. A. Endler (Eds.), Speciation and its consequences (pp. 180–207). Sunderland, MA: Sinauer Associates.Google Scholar
- Coyne, J. A., & Orr, H. A. (2004). Speciation. Sunderland, MA: Sinauer Associates.Google Scholar
- Darwin, C. (1859). On the origin of species by means of natural selection or the preservation of favored races in the struggle for life. London: J. Murray.Google Scholar
- de Queiroz, K. (1998). The general lineage concept of species, species criteria, and the process of speciation: A conceptual unification and terminological recommendations. In D. J. Howard & S. H. Berlocher (Eds.), Endless forms: Species and speciation (pp. 57–75). New York: Oxford University Press.Google Scholar
- de Queiroz, K. (1999). The general lineage concept of species and the defining properties of the species category. In R. A. Wilson (Ed.), Species: New interdisciplinary essays (pp. 49–89). Cambridge, MA: MIT Press.Google Scholar
- de Queiroz, K. (2005). A unified concept of species and its consequences for the future of taxonomy. Proceedings of the California Academy of Sciences, 56, 196–215.Google Scholar
- Dobzhansky, T. (1937). Genetics and the origin of species. New York: Columbia University Press.Google Scholar
- Dobzhansky, T. (1950). Mendelian populations and their evolution. American Naturalist, 84, 401–418.Google Scholar
- Feder, J. L. (1998). The apple maggot fly, Rhagoletis pomonella: flies in the face of conventional wisdom about speciation? In D. Howard & S. H. Berlocher (Eds.), Endless forms: Species and speciation (pp. 130–144). London: Oxford University Press.Google Scholar
- Feder, J. L., Berlocher, S. H., Roethele, J. B., Dambroski, H., Smith, J. J., Perry, W. L., et al. (2003). Allopatric genetic origins for sympatric host-plant shifts and race formation in Rhagoletis. Proceedings of the National Academy of Sciences, USA, 100, 10314–10319.Google Scholar
- Felsenstein, J. (1981). Skepticism towards Santa Rosalia, or why are there so few kinds of animals? Evolution, 35, 124–138.Google Scholar
- Fry, J. D. (2003). Detecting ecological trade-offs using selection experiments. Ecology, 84, 1672–1678.Google Scholar
- Gavrilets, S. (2004). Fitness landscapes and the origin of species. Princeton, NJ: Princeton University Press.Google Scholar
- Getz, W. M., & Kaitala, V. (1989). Ecogenetic models, competition, and heteropatry. Theoretical Population Biology, 36, 34–58.Google Scholar
- Hammer, M. F., Woerner, A. E., Mendez, F. L., Watkins, J. C., & Wall, J. D. (2011). Genetic evidence for archaic admixture in Africa. Proceedings of the National Academy of Sciences, USA, 108, 15123–15128.Google Scholar
- Hellberg, M. E. (1998). Sympatric sea shells along the sea’s shore: the geography of speciation in the marine gastropod Tegula. Evolution, 52, 1311–1324.Google Scholar
- Hendry, A. P., Vamosi, S. M., Latham, S. J., Heilbuth, J. C., & Day, T. (2000). Questioning species realities. Conservation Genetics, 1, 67–76.Google Scholar
- Hey, J. (2006). On the failure of modern species concepts. Trends in Ecology & Evolution, 21, 447–450.Google Scholar
- Huxley, J. S. (1942). Evolution: The modern synthesis. Cambridge, MA: MIT Press.Google Scholar
- Jiggins, C. D., Salazar, C., Linares, M., & Mavarez, J. (2008). Review. Hybrid trait speciation and Heliconius butterflies. Philosophical Transactions of the Royal Society B: Biological Sciences, 363, 3047–3054.Google Scholar
- Johannesson, K. (2009). Inverting the null-hypothesis of speciation: a marine snail perspective. Evolutionary Ecology, 23, 5–16.Google Scholar
- Johannesson, K., Panova, M., Kemppainen, P., Andre, C., Rolan-Alvarez, E., & Butlin, R. K. (2010). Repeated evolution of reproductive isolation in a marine snail: unveiling mechanisms of speciation. Philosophical Transactions of the Royal Society B-Biological Sciences, 365, 1735–1747.Google Scholar
- Johannesson, K., Rolan-Alvarez, E., & Ekendahl, A. (1995). Incipient reproductive isolation between two sympatric morphs of the intertidal snail Littorina saxatilis. Evolution, 49, 1180–1190.Google Scholar
- Kelly, R. P., & Eernisse, D. J. (2008). Reconstructing a radiation: the chiton genus Mopalia in the north Pacific. Invertebrate Systematics, 22, 1–12.Google Scholar
- Krug, P. J. (2011). Patterns of speciation in marine gastropods: A review of the phylogenetic evidence for localized radiations in the sea. American Malacological Bulletin, 29, 169–186.Google Scholar
- Levene, H. (1953). Genetic equilibrium when more than one ecological niche is available. American Naturalist, 87, 331–333.Google Scholar
- Marie Curie SPECIATION Network. (2011). What do we need to know about speciation? Trends in Ecology and Evolution, 27, 27–39.Google Scholar
- Mayden, R. L. (1997). A hierarchy of species concepts: The denouement in the saga of the species problem. In M. F. Claridge, H. A. Dawah, & M. R. Wilson (Eds.), Species: The units of biodiversity (pp. 381–424). London: Chapman and Hall.Google Scholar
- Maynard Smith, J. (1966). Sympatric speciation. American Naturalist, 100, 637–650.Google Scholar
- Mayr, E. (1942). Systematics and the origin of species. New York: Columbia University Press.Google Scholar
- Mayr, E. (1954a). Geographic speciation in tropical echinoids. Evolution, 8, 1–18.Google Scholar
- Mayr, E. (1954b). Change of genetic environment and evolution. In J. Huxley, A. C. Hardy, & E. B. Ford (Eds.), Evolution as a process (pp. 157–180). London: Unwin Brothers.Google Scholar
- Mayr, E. (1963). Animal species and evolution. Cambridge, MA: Belknap Press of Harvard University Press.Google Scholar
- Mayr, E. (2001). Wu’s genic view of speciation. Journal of Evolutionary Biology, 14, 866–867.Google Scholar
- Mendel, G. (1866). Versuche über Pflanzen-Hybriden. Verh. Naturforsch. Ver. Brünn, 4, 3–47 (in English in 1901, Journal of the Royal Horticulture Society, 26, 1–32).Google Scholar
- Messina, F. J., & Jones, J. C. (2011). Inheritance of traits mediating a major host shift by a seed beetle, Callosobruchus maculatus (Coleoptera: Chrysomelidae: Bruchinae). Annals of the Entomological Society of America, 104, 808–815.Google Scholar
- Messina, F. J., Mendenhall, M., & Jones, J. C. (2009). An experimentally induced host shift in a seed beetle. Entomologia Experimentalis et Applicata, 132, 39–49.Google Scholar
- Michel, A. P., Sim, S., Powell, T. H. Q., Taylor, M. S., Nosil, P., & Feder, J. L. (2010). Widespread genomic divergence during sympatric speciation. Proceedings of the Natural Academy of Science, USA, 107, 9724–9729.Google Scholar
- Muller, H. J. (1942). Isolation mechanisms, evolution and temperature. Biological Symposia, 6, 71–125.Google Scholar
- Naomi, S.-I. (2011). On the integrated frameworks of species concepts: Mayden’s hierarchy of species concepts and de Queiroz’s unified concept of species. Journal of Zoological Systematics and Evolutionary Research, 49, 177–184.Google Scholar
- Noor, M. A. F., Grams, K. L., Bertucci, L. A., & Reiland, J. (2001). Chromosomal inversions and the persistence of species. Proceedings of the Natural Academy of Science, USA, 98, 12084–12088.Google Scholar
- Nosil, P., & Feder, J. L. (2012). Genomic divergence during speciation: causes and consequences. Philosophical Transactions of the Royal Society B-Biological Sciences, 367, 332–342.Google Scholar
- Nosil, P., Harmon, L. J., & Seehausen, O. (2009b). Ecological explanations for (incomplete) speciation. Trends in Ecology & Evolution, 24, 145–156.Google Scholar
- Papadopulos, A. S. T., Baker, W. J., Crayn, D., Butlin, R. K., Kynast, R. G., Hutton, I., et al. (2011). Speciation with gene flow on Lord Howe Island. Proceedings of the National Academy of Sciences, USA, 108, 13188–13193.Google Scholar
- Patterson, N., Richter, D. J., Gnerre, S., Lander, E., & Reich, D. (2008). Reply: Complex speciation of humans and chimpanzees. Nature, 452, E4.Google Scholar
- Pinho, C., & Hey, J. (2010). Divergence with gene flow: Models and data. Annual Review of Ecology Evolution and Systematics, 41, 215–230.Google Scholar
- Poulton, E. B. (1904). What is a species? (Presidential address to the Entomological Society of London) Proceedings of the Entomological Society London (revised version in Poulton E. B. Essays on Evolution. 1889–1907. (1908) Clarendon Press, Oxford. pp. 46–94).Google Scholar
- Presgraves, D. C., & Yi, S. V. (2009). Doubts about complex speciation between humans and chimpanzees. Trends in Ecology & Evolution, 24, 533–540.Google Scholar
- Rocha, L. A., & Bowen, B. W. (2008). Speciation in coral reef fishes. Journal of Fish Biology, 72, 1101–1121.Google Scholar
- Rundle, H. D., & Nosil, P. (2005). Ecological speciation. Ecology Letters, 8, 336–352.Google Scholar
- Sanger, F., Nicklen, S., & Coulson, A. R. (1977). DNA sequencing with chain-terminating inhibitors. Proceedings of the National Academy of Sciences, USA, 74, 5463–5467.Google Scholar
- Schluter, D. (2000). The ecology of adaptive radiation. New York: Oxford University Press Inc.Google Scholar
- Schluter, D. (2001). Ecology and the origin of species. Trends in Ecology & Evolution, 16, 372–380.Google Scholar
- Seehausen, O. (1997). Distribution of and reproductive isolation among color morphs of a rock-dwelling Lake Victoria cichlid (Haplochromis nyererei). Ecology of Freshwater Fish, 6, 59–66.Google Scholar
- Servedio, M. R., Van Doorn, G. S., Kopp, M., Frame, A. M., & Nosil, P. (2011). Magic traits in speciation: ‘Magic’ but not rare? Trends in Ecology & Evolution, 26, 389–397.Google Scholar
- Simpson, G. G. (1944). Tempo and mode in evolution. New York: Columbia University Press.Google Scholar
- Simpson, G. G. (1951). The species concept. Evolution, 5, 285–298.Google Scholar
- Stam, P. (1983). The evolution of reproductive isolation in closely adjacent populations through differential flowering time. Heredity, 50, 105–118.Google Scholar
- Tucker, P. K., Sage, R. D., Wilson, A. C., & Eichler, E. M. (1992). Abrupt cline for sex chromosomes in a hybrid zone between two species of mice. Evolution, 46, 1146–1163.Google Scholar
- Turelli, M., Barton, N. H., & Coyne, J. A. (2001). Theory and speciation. Trends in Ecology & Evolution, 16, 330–342.Google Scholar
- Via, S. (2009). Natural selection in action during speciation. Proceedings of the National Academy of Sciences, USA, 106, 9939–9946.Google Scholar
- Weersing, K., & Toonen, R. J. (2010). Population genetics, larval dispersal, and connectivity in marine systems. Marine Ecology Progress Series, 393, 1–12.Google Scholar
- Wiley, E. O. (1978). The evolutionary species concept reconsidered. Systematic Zoology, 27, 17–26.Google Scholar
- Wiley, E. O., & Mayden, R. L. (2000a). The evolutionary species concept. In Q. D. Wheeler & R. Meiner (Eds.), Species concept and phylogenetic theory: A debate (pp. 70–89). New York: Columbia University Press.Google Scholar
- Wiley, E. O., & Mayden, R. L. (2000b). A critique from the evolutionary species concept perspective. In Q. D. Wheeler & R. Meiner (Eds.), Species concept and phylogenetic theory: A debate (pp. 146–158). New York: Columbia University Press.Google Scholar
- Wiley, E. O., & Mayden, R. L. (2000c). A defense of the evolutionary species concept. In Q. D. Wheeler & R. Meiner (Eds.), Species concept and phylogenetic theory: A debate (pp. 198–208). New York: Columbia University Press.Google Scholar
- Wood, T. E., Takebayashi, N., Barker, M. S., Mayrose, I., Greenspoon, P. B., & Rieseberg, L. H. (2009). The frequency of polyploid speciation in vascular plants. Proceedings of the National Academy of Sciences USA, 106, 13875–13879.Google Scholar
- Wright, S. (1940). The statistical consequences of Mendelian heredity in relation to speciation. In J. Huxley (Ed.), The new systematics (pp. 161–183). London: Oxford University Press.Google Scholar