Can the evolutionary-rates hypothesis explain species-energy relationships?


†Author to whom correspondence should be addressed. E-mail:


  • 1There is growing consensus that much of the marked broad-scale spatial variation in species richness is associated with variation in environmental energy availability, but at least nine principal mechanisms have been proposed that may explain these patterns.
  • 2The evolutionary-rates hypothesis suggests that high environmental energy availability elevates rates of molecular evolution, promoting faster speciation, so that more species occur in high-energy areas because more evolve there. Direct tests of this hypothesis are rare and their conclusions inconsistent. Here we focus on assessing the support for its underlying assumptions.
  • 3First, the evolutionary-rates hypothesis assumes that high energy levels promote mutation. There is certainly evidence that high levels of ultraviolet radiation increase mutation rates. High temperatures may also reduce generation times and elevate metabolic rates, which may promote mutation. On balance, data support a link between rates of metabolism and mutation, but a link between the latter and generation time is more equivocal and is particularly unlikely in plants.
  • 4Second, the evolutionary-rates hypothesis assumes that mutation rates limit speciation rates. This may be true if all else was equal, but correlations between mutation and speciation are probably very noisy as many other factors may influence rates both of sympatric and allopatric speciation, including the occurrence of physical isolation barriers, the magnitude of selection and population size.
  • 5Third, the evolutionary-rates hypothesis assumes that there is a strong correlation between current and historical energy levels. Factors such as tectonic drift may weaken such relationships, but are likely to have had negligible effects over the time period during which the majority of extant species evolved.
  • 6Fourth, the evolutionary-rates hypothesis assumes that changes in species ranges following speciation do not sufficiently weaken the correlation between the rate of speciation in an area and species richness. The ranges of many species appear to alter dramatically following speciation, and this may markedly reduce the strength of the relationship, but to what extent is unclear.
  • 7In sum, the degree to which the evolutionary-rates hypothesis can explain spatial variation in species richness remains surprisingly uncertain. We suggest directions for further research.


The question of why some areas of the globe contain markedly more species than do others has fascinated ecologists for over a hundred years (Wallace 1878; Hutchinson 1959; Gaston 2000). More than 30 hypotheses have been proposed to explain this variation (Willig, Kaufman & Stevens 2003), and while many have received limited support there is increasing evidence that environmental energy availability can account for much of the pattern (Hawkins et al. 2003; Pimm & Brown 2004). Such species–energy relationships may take a variety of forms, but these almost invariably include a marked positive phase (Mittelbach et al. 2001), especially at macroecological scales where a large proportion of species–energy relationships follow a monotonically increasing curve (Waide et al. 1999; Whittaker & Heegaard 2003).

Despite numerous studies that describe the form of species–energy relationships, identification of their causation remains elusive (Currie et al. 2004; Evans, Warren & Gaston 2005a). One prominent possibility is the evolutionary-rates or climate-speciation hypothesis (Rohde 1978, 1992, 1999). This proposes that high energy availability increases mutation rates, speeding up molecular evolution, thus increasing speciation rates and elevating species richness. Theoretical and practical difficulties have limited the number of studies that have investigated how rates of molecular evolution and speciation vary with energy availability. Consequently, the extent to which the evolutionary-rates mechanism promotes species–energy relationships is unclear (Evans et al. 2005a).

Here, we first briefly summarize the currently available empirical evidence for the evolutionary-rates hypothesis. Our main aim, however, is to draw on the wider ecological and evolutionary literature to assess the validity of four principal assumptions that underlie the hypothesis. First, that high energy availability elevates mutation rates. Second, that mutation rates exert a strong influence on speciation rates. Third, that historical and current energy levels are strongly positively correlated. Fourth, that a region's current species richness is positively correlated with the rate at which species evolved in that region. We conclude by suggesting directions for future research into the evolutionary-rates hypothesis.

Energy availability

Environmental energy availability can be measured using numerous metrics, but these fall into two main categories. First, solar energy metrics record the amount of solar radiation reaching the earth's surface. Both temperature and ultraviolet (UV) radiation are suitable solar energy metrics, and exhibit marked spatial variation (Fig. 1). However, in vegetated areas individuals that live beneath the canopy may be exposed to lower levels of UV radiation than indicated, owing to the absorption and reflection of UV light by vegetation. Other terrestrial species may reduce exposure to UV radiation by becoming crepuscular or nocturnal, or develop protective measures such as increasing investment in DNA repair mechanisms. Spatial variation in UV and temperature exhibit broadly similar patterns, with both variables reaching maximum values within tropical regions, although there are also marked differences, e.g. high-altitude regions typically have low temperatures but high levels of UV radiation (Fig. 1). Some of these high-altitude regions are also associated with high species richness, such as parts of the Andes, and other areas with high UV radiation are also known for the high species richness of some taxa, such as the plants of the Cape Floristic region. The Sahara and other desert regions, however, although having high temperatures and exposure to UV radiation, typically have low species richness.

Figure 1.

Energy availability, measured as mean ground level air temperature (°C), images on the right, and UV radiation (J m−2) reaching the earth's surface, images on the left, for (a) January, (b) April, (c) July and (d) October. UV images are from a single representative year (2000), while temperature data are averages between 1961 and 1990. In UV maps yellow represents high values, red intermediate and purple low values, dark bands are those for which data are unavailable, often because of high cloud cover that would reduce UV radiation. In temperature maps yellow represents high values, green intermediate and dark blue the lowest values. Temperature data are from New et al. (2002), UV data are from the NASA Earth Observatory (

Second, productive energy metrics record the amount of plant resources available to consumers; examples include net primary productivity (NPP) and the normalized difference vegetation index (NDVI). Productive energy metrics respond positively to a combination of solar radiation and water, while solar energy metrics respond only to the former. The two types of energy metrics may thus exhibit very different spatial patterns. The evolutionary-rates hypothesis is typically concerned with solar energy metrics, and while productive energy availability may also influence evolutionary rates (e.g. biotic interactions may be stronger in areas with more productive energy and thus increase speciation rates; Schemske 2002), discussion of this hypothesis is largely beyond the scope of this paper.

Is molecular evolution faster in high-energy areas?

The one study that explicitly investigates how evolutionary rates vary with environmental energy availability found that these rates for flowering plants correlate positively with both temperature and UV radiation (Davies et al. 2004). The evolutionary-rates hypothesis predicts that evolutionary rates should generally decline with latitude, given that energy availability typically does so. This conflicts with data on avian evolutionary rates, measured across the complete nuclear genome and two mitochondrial genes, which exhibit no correlation with latitude (Bromham & Cardillo 2003). Clearly, additional data, from more species and a broader range of taxa, are required before the general nature of relationships between evolutionary rates and energy availability can be discerned.

Other studies do not directly report spatial variation in evolutionary rates but investigate spatial patterns in genetic diversity and the factors, such as environmental energy, associated with high levels of diversity. These may be of limited value with regard to the evolutionary-rates hypothesis if, once a certain level of genetic diversity has accumulated, populations diverge and much of the new diversity is lost from the ancestral population and passed to the novel population, i.e. if the rate at which mutations increase with energy is close to the rate at which speciation increases with energy. In reality, however, population divergence is likely to be influenced by many factors other than the amount of genetic diversity within a population, as suggested by the marked variation in levels of genetic diversity within species, and the mutation-energy relationship is likely to have a steeper gradient than the speciation–energy relationship. Studies reporting how levels of genetic diversity vary along energy gradients are thus likely to provide insights into how energy influences the rate of molecular evolution.

Populations with relatively fast rates of molecular evolution will, other things being equal, contain more genetic diversity than more slowly evolving populations. Intensive studies across a range of taxa in Israel indicate that genetic diversity is greatest in regions with high temperatures (Nevo 1998). Such patterns are compatible with faster rates of molecular evolution in high-energy areas, but the underlying determinants are uncertain as genetic diversity also increases with aridity and pollution, suggesting that environmental stress, rather than temperature per se, may be the causal factor (Nevo 1998; Imasheva 1999).

Numerous studies report greater genetic diversity in lower latitude populations of many Palearctic and Nearctic taxa (e.g. Nevo & Beiles 1991; Bernatchez & Wilson 1998; Edmands 2002; Grivet & Petit 2002; Ödeen & Björklund 2003; Chek, Austin & Lougheed 2003). Although such patterns suggest that molecular evolution may be faster at lower latitudes, they must be interpreted cautiously for three reasons. First, negative correlations between genetic diversity and latitude are not universal (Gaston 2003). Second, genetic diversity tends to correlate positively with population size, and most studies do not take spatial variation in the latter into account, although such variation is known to be marked. Third, at least within the extent of temperate regions, glaciation events may generate greater diversity in relatively low-latitude populations by exterminating populations at the highest latitudes and promoting diversification of lower latitude populations by subdividing them into refugia, from which only a small proportion of lineages subsequently expand northwards (Hewitt 2000; Grivet & Petit 2002). While latitude and genetic diversity remain negatively correlated when analyses are confined to regions not covered by ice (freshwater fish, Bernatchez & Wilson 1998; vertebrates in general, Martin & McKay 2004), the effects of glaciation on evolutionary events can extend well beyond the glacial limit (Jackson & Sheldon 1994).

In summary, evidence that molecular evolution is faster in high-energy regions is somewhat limited, but this is largely a consequence of the paucity of appropriate studies and the difficulty of controlling for confounding factors. It remains plausible that energy availability is positively correlated with evolutionary rates.

Is speciation faster in high-energy areas?

The fossil record has the potential to provide much useful information on evolutionary processes, including spatial variation in their rates. This potential is reduced by the lower probability of fossil preservation in the tropics and the poorer exploration of the tropical fossil record (Jablonski 1993; Smith 1994) and, with regard to the evolutionary-rates hypothesis, by inadequate knowledge of paleoclimates, although it is likely that for long periods the tropics have had higher levels of environmental energy than more temperate regions. Despite these limitations it is noteworthy that first appearances of 26 marine orders in the fossil record are more numerous in the tropics than expected by chance (Jablonski 1993). In addition, the Foraminiferan fossil record indicates that temporal increases in species richness have been faster in tropical regions than temperate ones (Buzas, Collins & Culver 2002), and there is some evidence for such patterns in marine bivalves (Crame 2002). The fossil record thus indicates that speciation rates may have been higher in tropical areas.

Regions with fast speciation rates should contain a disproportionate number of recently evolved endemics. While many studies document patterns of spatial variation in the number of endemic species, very few investigate which areas contain more endemics than is expected given the total number of species in an area (but see Jetz, Rahbek & Colwell 2004). Moreover, none takes species richness into account while also distinguishing between relict endemics and those that have recently evolved. Such analyses are vital if studies of endemic distributions are to inform the debate on spatial variation in speciation rates.

All else being equal, faster speciation in high-energy areas should result in such regions containing younger taxa than areas where speciation is slower. Obtaining reliable estimates of species’ ages can, however, be problematic. Studies based on phylogenies suffer from three problems. First, age estimates are generally based on a molecular clock, but there is debate concerning its reliability (e.g. Simon et al. 1996; Rodriguez-Trelles, Tarrio & Ayala 2001; Bromham & Woolfit 2004; Chaw et al. 2004; Gillooly et al. 2005). Second, is the ‘missing branches problem’, in which the omission of extinct taxa from a phylogeny results in ages being overestimated (Chown & Gaston 2000). Third, molecular phylogenies provide estimates of the time back to the common ancestor of two DNA sequences, which is typically longer than the time back to the common ancestor of the two species, again leading to overestimates of taxon age, although methods are available that attempt to correct for this (Nichols 2001). Studies of speciation rates that use the fossil record (Stehli & Wells 1971; Hecht & Agan 1972; Durazzi & Stehli 1973; Jablonski 1993) are limited by the aforementioned problems of relative exploration and taphonomic setting. Moreover, both approaches are problematic because estimates of the mean age of an area's taxa may be influenced by dispersal of recently evolved species and variation in extinction rates (Gaston & Blackburn 1996; Kerr & Currie 1999). The nature of spatial variation in taxonomic age thus remains uncertain (Chown & Gaston 2000), although it may improve in the future through increased availability of molecular phylogenies and advances in phylogenetic analyses.

Diversification rates measure the net gain in species richness that occurs through speciation and extinction, and can provide some useful information on how energy availability influences these processes. Crude estimates of spatial variation in diversification rates, measured as the number of species within higher taxonomic groups, indicate that they are faster in tropical regions (Attrill, Stafford & Rowden 2001). A more reliable estimate of diversification rates can be provided by using DNA sequence data to group species into monophyletic sister clades of equal age. Such approaches have seldom been applied to questions concerning spatial variation in diversification rates, but reveal a negative correlation between latitude and diversification rates in swallowtail butterflies and passerine birds (Cardillo 1999). Conversely, a more extensive analysis of avian diversification rates demonstrates that tribes with low rates are not disproportionately located in the high-latitude regions with low energy availability, suggesting that energy availability may not be the primary factor influencing speciation rates (Ricklefs 2003). Similarly, a comparison of the diversification rates in two clades of Enallagma damselflies in the USA reveals faster diversification in the more northerly clade (McPeek & Brown 2000). The continuing rapid reduction in the costs of DNA sequencing (Kocher 2004) should facilitate the construction of robust phylogenies, enabling similar tests to be conducted on a much wider range of taxa. Until this is achieved it would be unwise to draw conclusions on the general nature of the association between diversification rates and energy availability and it is currently uncertain if speciation proceeds faster in high-energy areas. The lack of consensus on the relationship between energy and speciation rates means that it is crucial to evaluate the evolutionary-rates hypothesis on the basis of its underlying assumptions.

Assumption 1: high energy availability increases mutation rates

Solar energy metrics such as heat and UV radiation may act directly as mutagenic agents. In addition, solar and productive energy metrics may influence secondary factors, such as generation time, metabolic rate and population size, which are associated with mutation rates.

direct pathways

UV radiation can spontaneously change DNA and, when absorbed by cell contents, may generate free radicals that cause oxidative damage. These mutational effects are not confined to terrestrial systems, but can occur in water at depths of up to 30 m (Karentz & Lutze 1990). They have been conclusively demonstrated, under laboratory conditions, in a range of taxa (Berg et al. 2000; Shin, Mellon & Turker 2002). Indeed, experiments that seek to induce mutations routinely do so by exposing cells to UV radiation (e.g. Smith & Drake 1998). Evidence for the mutational effects of UV radiation in the field is more limited, but positive correlations occur between UV exposure and mutation rates in fungi (Lutzoni & Pagel 1997), higher plants (Tuteja et al. 2001) and humans (Ziegler et al. 1993). These relationships appear to be causal, as the nature of the base transitions are of the type commonly induced by UV exposure. Studies demonstrating that high temperatures increase mutation rates are less numerous, but include experiments on bacteria and Drosophila (references in Rohde 1978; Grogan 1998; Castan et al. 2003). Field studies that investigate such relationships are yet more scarce, although Wright, Gray & Gardner (2003) report a positive interspecific relationship between ambient temperature and mutation rate in Mearnsia plants growing in the Pacific basin (but see Brown & Pauly 2005; Pawar 2005).

There is strong evidence that high levels of UV radiation, and potentially high temperatures, directly increase mutation rates. However, if species–energy relationships predominantly arise because high energy availability directly increases mutation rates and thus promotes speciation, species richness should be related to solar energy metrics more strongly than productive energy metrics. The converse appears to be the case, with many macroecological studies reporting that species richness is more closely associated with productive energy measures such as NPP, than with temperature alone (83 out of 85 studies reviewed by Hawkins et al. 2003; although Kaspari, Ward & Yuan (2004) found that temperature was a better predictor of ant species richness than NPP). However, while evidence for mutational effects is stronger for UV radiation than temperature, few studies use UV radiation as a predictor of species richness.

indirect pathways

Generation time

Shorter generation times give rise to a greater number of germ-line cell divisions and thus replication induced mutations per unit time; generation times may thus be negatively correlated with mutation rates (Laird, McConaughty & McCarthy 1969; Kohne 1970). Taxa in high-energy areas may have shorter generation times for two reasons. First, increased resource abundance in areas with high levels of productive energy may reduce generation times through a number of pathways including reductions in the cost of reproduction that lower the refractory period between reproductive attempts, and increasing growth rates. Second, while the mechanisms are poorly understood, low temperatures are generally associated with an increase in body size, James’ and Bergmann's rules (Atkinson & Sibly 1997; Gaston & Blackburn 2000; Fischer & Fiedler 2002; Blackburn & Hawkins 2004). As body size is positively correlated, both intra- and interspecifically, with generation time (Peters 1983; Calder 1984), the latter may be increased in cold environments with larger taxa.

Evidence for positive correlations between the number of replication events or generation times and rates of molecular evolution is fairly strong. DNA sequences that replicate more frequently, such as the male specific sex chromosomes in birds and mammals, have faster evolutionary rates than their homologous chromosomes that undergo fewer replications (Chang et al. 1994; Li et al. 1996; Ellegren & Fridolfsson 1997; Lawson & Hewitt 2002). While such sexual bias in mutation rates may not relate solely to the number of replications (Hurst & Ellegren 1998) other factors that may covary with generation time (e.g. exposure to UV light and temperature) are held constant. Similarly, DNA evolution is faster in bacteria than their more slowly reproducing insect hosts (Moran, van Dohlen & Paumann 1995). Negative correlations between generation time and rates of molecular evolution have also been demonstrated in birds (Mooers & Harvey 1994) and mammals (Bromham, Rambaut & Harvey 1996; Li et al. 1996). Conversely, generation time and evolutionary rates were not associated in studies of 35 complete mitochondrial genomes from 14 mammalian orders (Gissi et al. 2000) and of rates of cytochrome-b evolution across 21 rodent species (Spradling, Hafner & Demastes 2002). The reasons for these differences are currently unclear.

Botanical studies have also reported negative correlations between generation time and rates of molecular evolution (Gaut et al. 1992; Laroche & Bosquet 1999; Charlesworth & Wright 2001). However, these often focus on taxa that differ in many factors other than their generation time, or fail to take phylogenetic non-independence into account, and a broad-scale review concludes that generation time does not influence rates of molecular evolution in plants (Whittle & Johnston 2003). Correlations between generation time and rates of molecular evolution may be weakened in plants because, unlike in animals, germ-line cells are produced from cells that have undergone as many divisions as somatic tissue. In some species, the inheritance of somatic mutations may also be increased by vegetative reproductive events. As the rate at which somatic mutations arise is independent of generation time, their inheritance may reduce the strength of correlations between evolutionary rates in plants and generation time (Gaut et al. 1996; Whittle & Johnston 2003).

Turning to the nature of the relationship between generation time and energy availability, numerous laboratory experiments have demonstrated that generation times are closely negatively correlated with temperature (e.g. zooplankton, Gillooly 2002; the polychaete T. heterouncinata, Finley, Mulligan & Friedman 2001; the mite Tetranychus urticae, Kasap 2004; the coleopteran Gastrophysa viridula, Honek, Jarosik & Martinkova 2003) and food availability (e.g. the polychaete Eupolymnia nebulosa, Martin et al. 1998; the crustacean Streptocephalus mackini, Anaya-Soto, Sarma & Nandini 2003). Field studies are scarcer, but also report negative correlations between generation time and energy availability (e.g. temperature – the ostracod Darwinula stevensoni, van Doninck et al. 2003; food availability – the shrimp Mysis relicta, Chess & Stanford 1998). A study of Harp Seals, Pagophilus groenlandicus, found that following a rapid decline in their food supplies the age taken to reach sexual maturity increased by approximately 50% (Frie et al. 2003). Although a few studies find no relationship between generation time and surrogate measures of energy availability, such as altitude, for example in some Lycaena butterflies, such patterns probably arise as a consequence of developmental constraints arising from a trade-off between investment in growth and investment in reproduction (Fischer & Fiedler 2002).

Support for the suggestion that temperature may influence generation times through its influence on body size is equivocal. Numerous studies report positive intra- and interspecific correlations between body size and latitude (ants, Cushman, Lawton & Manly 1993; amphibians, Ashton 2002a; birds, Ashton 2002b; Meiri & Dayan 2003; mammals, Freckleton, Harvey & Pagel 2003; Meiri & Dayan 2003). While some studies fail to find significant correlations (freshwater fish, Belk & Houston 2002; snakes, Reed 2003), most taxa appear to follow Bergmann's rule (Gaston & Blackburn 2000). However, there is only limited evidence that generation times increase with latitude. Although they do so in swallowtail butterflies (Tyler, Brown & Wilson 1994), most insect taxa do not exhibit this pattern (Chown, Addo-Bediako & Gaston 2002). Similarly, in birds, age at first breeding is predominantly a function of body size and taxon rather than geographical location – almost all passerines breed in their first year of life and tropical and high-latitude species have similar generation times (Mayr 1976).

In general and other things being equal, shorter generation times probably increase the speed of molecular evolution. However, correlations between solar or productive energy and generation times are equivocal. It would be premature to conclude that high energy availability generally reduces generation times, thus promoting faster molecular evolution.

Metabolic rate

Metabolism releases by-products, such as free radicals, which can oxidize DNA and thus induce mutations (Tuteja et al. 2001; Epe 2002). Positive correlations may therefore arise between rates of metabolism and mutation (Richter, Park & Ames 1988). High temperatures may promote faster metabolic rates for two reasons. First, they are generally associated with elevated metabolic rates in ectotherms (Gillooly et al. 2001). Intraspecific relationships of this type may, however, be stronger than interspecific ones as the latter may be weakened by cold adaptation; insect species from cold regions have a higher metabolic rate at a given ambient temperature than ones from warmer regions (Addo-Bediako, Chown & Gaston 2002). Second, high temperatures favour smaller-bodied taxa (see Generation time section) that, while having lower total energy expenditure than larger species, have higher metabolic rates per gram of body mass (McNab 2002). This second pathway assumes that there are no latitudinal gradients in cell size which may not be true (French, Feast & Partridge 1998), but evidence for such a pattern is equivocal (Litzgus, duRant & Mousseau 2004).

If fast metabolism promotes mutation, endotherms are predicted to have more rapid mutation rates than ectotherms. Such patterns are frequently reported (Thomas & Beckenbach 1989; Avise et al. 1992; Adachi, Cao & Hasegawa 1993; Bowen, Nelson & Avise 1993; Martin & Palumbi 1993). However, a few studies have provided apparently anomalous results. Avian metabolic rates are at least as high as those of mammals, but they appear to have slower rates of molecular evolution (Prager et al. 1974), even at silent sites that are not phenotypically expressed and thus experience much reduced selection (Mindell et al. 1996). Similarly, warm-water microcrustaceans do not have a faster rate of molecular evolution than those occupying colder environments, even when taking into account generation time, differential exposure to UV and selective constraints (Hebert et al. 2002). Positive interspecific associations between rates of metabolism and molecular evolution are, however, reported in comparisons of closely related taxa in which the influence of confounding variables is reduced (hummingbirds, Bleiweiss 1998).

Moreover, a model has recently been proposed (Gillooly et al. 2005) that predicts molecular evolution rates by combining the neutral theory of evolution (Kimura 1983) with the effects of body size and temperature on metabolic rate. This can explain much of the observed rate heterogeneity across different genes and thermal environments in numerous invertebrate and vertebrate taxa, and reconcile differences between fossil and molecular-based estimates of when lineages diverged. As Gillooly et al. (2005) note, however, the model does not distinguish between the metabolic rate and generation time hypotheses.

The energetic equivalence rule (Damuth 1981, 1987) combines two allometric scaling relationships, that density (D) scales with body mass (M) in the manner D ∝ M−0·75 and that metabolic rate (R) scales with body mass in the manner R ∝ M0·75, to conclude that species of different body sizes use approximately equal amounts of energy. Allen, Brown & Gillooly (2002) extend this to include temperature by incorporating the biochemical kinetics of metabolism. They suggest that the total energy flux of populations is independent of temperature, as well as body size, and present empirical data that support this, and that in ectotherms temperature positively affects metabolic rate and thus energy use. Therefore, in order to satisfy the energetic equivalence rule, species’ population abundances must be lower in areas with higher temperatures. If the total number of individuals in an assemblage is also independent of temperature, then the species richness of an assemblage must be higher in areas with high temperatures. Allen et al. (2002) conclude that, according to the extended energetic equivalence rule, the relationship between ectotherm species richness, expressed as the natural logarithm, and temperature (1000/K) should have a slope of −9. They present data for a range of ectothermic taxa that appear to match this prediction, and Kaspari et al. (2004) do so for ant assemblages, and suggest that as temperature is related to mutation rates and generation time in the same manner as it is related to metabolic rate then such relationships also provide support for the evolutionary-rates hypothesis. Such an explanation for spatial patterns in biodiversity cannot be used to explain species richness patterns in endotherms, and it is unclear why population energy flux should be independent of temperature (Storch 2003).

Population size

Theory has long predicted that genetic diversity and population size are positively correlated (Wright 1931). Potential reasons include the increased number of reproductive events, leading to a greater number of replication errors; the larger number of sites, i.e. individuals, at which mutations may occur; and the positive correlation between population size and range size (Brown 1984; Gaston, Blackburn & Lawton 1997; Gaston & Blackburn 2000) which may increase the variety of conditions to which local populations are adapted and thus total genetic diversity. The neutral theory of biodiversity assumes that the probability of mutation is constant across all individuals; thus more mutations, and greater genetic variation, will occur in larger populations (Hubbell 2001). Indeed, levels of genetic variation within populations are much higher in abundant and widespread species compared with closely related rare and localized species (Karron 1987; Gitzendanner & Soltis 2000; Cole 2003) and, intraspecifically, a population's size can be estimated relatively accurately given data on the amount of genetic variation that it contains (Kuhner, Yamato & Felsenstein 1998; Schwartz, Tallmon & Luikart 1998). The greater genetic variation that occurs in large populations, which tend to have large ranges, may provide more material upon which natural selection can act and thus promote faster speciation rates. There is much evidence, however, for a unimodal relationship between geographical range size and the probability of speciation, with species with intermediate range sizes having the greatest probability of speciation (Gaston 1998, 2003; Gaston & Chown 1999; Chown & Gaston 2000; but see Rosenzweig 1995).

Theory predicts that greater resource abundance in high-energy areas may lead to larger populations (Wright 1983) and the total abundance of individuals in an assemblage often correlates positively with productive energy (Kaspari, Yuan & Alonso 2003; Hurlbert 2004; Meehan, Jetz & Brown 2004; Pautasso & Gaston 2005; but see Srivastava & Lawton 1998). However, there is rather little evidence that mean population sizes, per species, are higher in areas with more productive energy (Currie et al. 2004; Pautasso & Gaston 2005). On balance, a positive relationship between productive energy and population size, leading to greater genetic diversity and faster speciation thus seems unlikely to drive a relationship between productive energy and speciation rate.

Assumption 2: mutation rates control speciation rates

At its core, the evolutionary-rates hypothesis assumes that greater genetic diversity in high-energy areas promotes speciation as there is greater variation upon which natural selection can act. That is, speciation is mutation limited. Additional triggers are generally required to promote species divergence, and spatial variation in the occurrence of these triggers may confound and weaken relationships between genetic diversity and speciation rates. The above discussion on the relationship between population size and speciation rate provides an example of how factors other than mutation rate can influence speciation rates. Here we consider whether vicariance or dispersal events, environmental heterogeneity and environmental instability may also reduce the extent to which speciation rates are controlled by mutation. In addition, genetic variation generates the phenotypic variation that is sorted by natural selection and may ultimately lead to speciation. Speciation rates are thus more likely to be limited by mutation rates if genetic and phenotypic variation are strongly positively correlated.

Allopatric speciation requires a vicariance event, or dispersal to a previously unoccupied area, that spatially isolates populations and prevents gene flow. Geographical variation in the presence of physical isolation barriers, such as mountain ranges or water-bodies, may thus exert a large influence on spatial variation in speciation rates. While tropical regions contain a disproportionate area of oceanic islands (data from ESRI 2003), there is little evidence that other barriers are more frequent in low-latitude or high-energy regions. The effectiveness of mountains as dispersal barriers may, however, be more pronounced in areas with high levels of environmental energy. Two sites along an altitudinal gradient are less likely to overlap in their climatic conditions in the tropics than in temperate regions. Mountains may thus be effectively higher in the tropics because dispersal up or down a tropical mountain is more likely to take a species into a climatic regime to which it is not adapted (Janzen 1967; Huey 1978). Vicariance events arising from mountain barriers may thus be more frequent in warm regions, leading to positive correlations between speciation rates and energy availability. Conversely, high mutation rates may reduce the probability of allopatric speciation. If unsuitable habitat forms a dispersal barrier between two populations then high mutation rates may increase the probability that novel adaptations will arise that enable individuals to migrate through the unsuitable habitat, leading to gene flow and thus reducing the probability of speciation (Wiens 2004).

Both sympatric and allopatric speciation may be promoted by environmental heterogeneity, which may generate a diverse array of selective forces that favour specialization. For example, the species richness of parts of the South African flora is comparable with the most diverse tropical areas and much greater than expected from its area and latitude. This discrepancy can, however, largely be explained by extremely high levels of environmental diversity (Linder 2003). That natural selection may lead to the evolution of divergent traits, and ultimately species, adapted to different environments is termed ecological speciation and is supported by a number of other studies on a range of taxa and environments (e.g. Peck, Wigfull & Nishida 1999; Kruse, Strasser & Thiermann 2004; Williams & Reid 2004) and by mathematical models (Doebeli & Dieckmann 2003). While ecological speciation certainly occurs, its prevalence is unknown, and the extent to which spatial variation in habitat heterogeneity will weaken relationships between mutation rates and speciation is thus unclear (Schemske 2000; Schluter 2001; Via 2002; Servedio & Noor 2003). Studies of relationships between environmental heterogeneity and speciation rates should avoid spurious correlations that may arise from using a measure of heterogeneity that is itself influenced by speciation rates. The use of habitat diversity as a measure of heterogeneity may generate such circularity as more habitats may be defined in species-rich areas (Rosenzweig 1995), which may have higher speciation rates.

Speciation appears to be promoted by environmental instability such as tectonic movements, sea-level change and habitat fragmentation arising from glaciation (Barraclough & Vogler 2002; Mercer & Roth 2003; Ricklefs 2003; Weir & Schluter 2004; Hewitt 2004; Ribera & Vogler 2004; Willis & Niklas 2004). There is, however, no clear relationship between energy availability and environmental instability; if anything, high-latitude, low-energy regions are more unstable owing to frequent glaciation events, although the latter may still influence tropical assemblages (Jackson & Sheldon 1994). Moreover, the role of instability in promoting diversification is equivocal; some studies report negative relationships between environmental instability and speciation rates (Fjeldså 1994; Fjeldså, Lambin & Mertens 1999). In summary, it seems likely that speciation is influenced by numerous factors (including isolation barriers, environmental heterogeneity and environmental stability) that will add considerable noise to the relationship between speciation and mutation rates.

Strong correlations between genetic variation and the phenotypic variation which it causes, and upon which natural selection acts, will increase the extent to which mutation rates limit speciation. We are aware of only one study that conclusively demonstrates positive correlations between molecular and phenotypic evolutionary rates (flowering plants, Barraclough & Savolainen 2001). Omland (1997) reported a similar conclusion in vertebrate taxa, but re-analysis of the data reveals equivocal results (Bromham et al. 2002). Conversely, there is much evidence that rates of molecular and phenotypic evolution are often poorly correlated. A review of marine taxa indicates that morphologically similar species are often quite distinct genetically (Knowlton 2000), and the rate at which such cryptic species are described is rapidly accelerating (Sáez & Lozano 2005). Poor correlations between molecular and phenotypic evolution are apparent in Foraminifera (Kucera & Darling 2002), reptiles and mammals (Bromham et al. 2002) and numerous taxa that have undergone adaptive radiations (cichlids, Meyer et al. 1990; columbine monkeys, Hodges & Arnold 1994; and most oceanic island radiations, Bromham 2003). It also appears likely that adaptive trait divergence can be accomplished with limited divergence of allele frequencies (McKay & Latta 2002; Latta 2004; but see Cmokrak & Merilä 2002; Hendry 2002). Moreover, speciation may sometimes be driven by changes in very few genes or traits (Wu & Palopoli 1994; Doi et al. 2001; Orr 2001; McKinnon et al. 2004; Wu & Ting 2004). Such processes may explain why species-rich clades can contain less genetic variation than species-poor clades (mosses, Shaw et al. 2003) and suggest that while fast rates of molecular evolution and resultant high levels of genetic variation may promote speciation, they may not be essential. Moreover, and while much more research is needed, it has been proposed that environmentally induced novel traits have greater evolutionary potential than ones arising through mutation (West-Eberhard 2005).

Poor correlations between rates of genetic evolution and speciation may arise because the latter may be promoted by changes in sexual selection preferences. Speciation through sexual selection is strongly indicated in many taxa including insects (Mendelson & Shaw 2005), birds (Ellsworth et al. 1994) and, assuming that selection of traits that attract pollinators is a form of sexual selection, plants (Sargent 2004). The extent to which sexual selection processes contribute to speciation events is, however, uncertain (Panhuis et al. 2001; Arnegard & Kondrashov 2004; Kirkpatrick & Nuismer 2004). The apparent lack of correlation between genetic and phenotypic diversity is perhaps more likely to arise because only a small section of the genotype, such as regulatory genes, contribute to phenotypic expression (Barrier, Robichaux & Purugganan 2001). Regardless of its cause, the lack of correlation significantly reduces the probability that faster mutation rates generate greater genetic diversity that increases the amount of phenotypic variation upon which natural selection can act. The case that mutation rates limit speciation rates is far from being proved.

Assumption 3: historic and current energy levels are strongly positively correlated

The evolutionary-rates mechanism argues that the rate of past speciation was controlled by former energy levels, and can explain current species–energy relationships. It therefore assumes that there is a strong correlation between current and former energy levels. This assumption has not yet been tested directly, but evidence is available with which to assess its validity.

Tectonic drift has changed the location of many landmasses and thus caused a marked shift in their exposure to solar radiation. For example, Northern Australia is currently a tropical region, but occupied more temperate latitudes before the early Miocene. Similarly, the Indian subcontinent has been located in the tropical and subtropical zones for the last 50 million years, but was formerly at much higher latitudes (Smith, Smith & Funnell 1994). Are changes on such time-scales relevant to current species richness patterns? Speciation events occur approximately once every million years in avian (Avise & Walker 1998; Ricklefs 2003) and flowering plant lineages (Magallón & Sanderson 2001), and once every two million years in old world deer lineages (Pitra et al. 2004). Some species of these taxa, on these continents, thus probably currently experience different energy levels from those present when they originated. However, the proportion to which this applies is unlikely to be sufficient to weaken markedly the strength of correlations between current energy levels and those experienced by their ancestral populations during speciation; most species only persist for a few million years (May 2000).

The nature of correlations between current and relevant historical energy levels may be more sensitive to temporal changes in the spatial distribution of energy than tectonic drift. The world has experienced many marked warming and cooling episodes, with the magnitude of change generally increasing with increasing latitude (Lockwood 2001). Such changes are likely to influence the precise form of correlations between current and historic climates, but are less likely to alter broad patterns of spatial variation in temperature; the tropics have remained warmer than more temperate regions. Spatial patterns in exposure to UV radiation will change with other factors, such as the atmospheric concentration of ozone. While these factors are known to have varied historically, data are currently insufficient to investigate their impact on correlations between current and historical exposure to UV (Rozema et al. 2001; Visscher et al. 2004). Some insight, however, may perhaps be gained from recent anthropogenic effects on ozone levels. Despite significant reductions in ozone at high southern latitudes that have increased exposure to UV radiation, this has had relatively little influence on broad latitudinal gradients in UV radiation (Solomon 2004).

Despite tectonic drift and historical changes in energy availability, latitudinal species richness gradients appear to have persisted over numerous geological time periods (Crane & Lidgard 1989; Crame 2001, 2002; Buzas et al. 2002; but see Anderson et al. 1999). Such persistence may have arisen, however, because species adjusted their ranges in a manner that maintained species-energy relationships. For example, North American mammal species richness currently peaks in the west but, during the Pleistocene, was probably higher in the east due to greater plant productivity in this region (Cannon 2004). The rapid adjustment of species richness patterns to spatial variation in energy availability demonstrates that the latter can influence species richness through pathways other than that described by the evolutionary-rates hypothesis. This is further demonstrated by the presence of species–energy relationships within taxa introduced by humans to areas outside their native range (Chown, Gremmen & Gaston 1998; Evans, Warren & Gaston 2005b).

Assumption 4: a region's current species richness is positively correlated with the rate at which species evolved in that region

The evolutionary-rates hypothesis assumes that an area's current species richness is positively correlated with the rate at which species evolved there in the past. This is likely to be true if, following origination, species ranges are relatively constant. In contrast, correlations between an area's current species richness and speciation rates will be weakened if newly evolved taxa disperse to areas with lower speciation rates, and thus presumably energy levels, than those in which they evolved. Such correlations will also be reduced if species generally become extinct in their area of origin while persisting elsewhere.

Species are certainly capable of long-distance dispersal events, even over inhospitable habitat types, as demonstrated by the continental origins of many species on volcanic oceanic islands (Vargas, Baldwin & Constance 1998; Sato et al. 2001). Many closely related terrestrial species have geographical distributions separated by oceans. Genetic data from numerous taxa indicate that such species generally diverged long after the land masses were connected, strongly suggesting that long-distance dispersal across oceans is not a rare event (de Queiroz 2005). For example, over half of the rodent and artiodactyl species in the Siwalik fauna of the Indian subcontinent probably originated in other biogeographical provinces (Flynn et al. 1995). Rapid dispersal events across land masses can clearly influence spatial variation in species richness; glaciations typically reduce species richness to almost zero, yet after just 10–13 000 years the influence of such events on current species richness is negligible or undetectable (Hawkins & Porter 2003a,b). Long-distance dispersal also appears to be common in the marine environment; analysis of the phylogeography of 94 marine calyptraeid gastropod species suggests that there have been 27 major dispersal events (Collin 2003). More generally, most species probably have small ranges immediately after speciation and then disperse outwards to attain much larger ranges (Gaston 2003). Such patterns have been confirmed in the fossil record of African mammals (Vrba & deGusta 2004), and by an analysis of geographical range size across avian species of different ages which suggests that range expansion is relatively rapid (Webb & Gaston 2000). Moreover, a quarter of the tree species that occur in non-tropical regions of North America evolved in the tropics (Fine 2001). It is currently unclear, however, if species dispersal in general is biased towards high- or low-energy areas.

There is conflicting theory concerning the nature of the relationship between mutation rate and extinction risk, and thus whether many species are likely to become extinct in their region of origin while persisting elsewhere. High rates of mutation may enable species to adapt to changing environments, reducing the extinction risk. Alternatively, as mutation rates increase, the number of deleterious mutations in a population may increase and thus elevate extinction risk, particularly in small populations (Huerta-Quintanilla & Rodriguez-Achach 2004; Visscher et al. 2004). A unimodal relationship between mutation rate and extinction risk may thus be more likely (Huerta-Quintanilla & Rodriguez-Achach 2004).

If temporal changes in the spatial distribution of energy availability and species ranges limit the applicability of the evolutionary-rates hypothesis then relationships between current species richness and speciation rates should be weak. Molecular phylogenies for Papionini primates suggest that species richness and speciation rates are positively correlated at very broad geographical scales (Böhm & Mayhew 2005). We are aware of two other studies that have looked at such correlations, and although speciation rates were measured using circumstantial evidence, their results are noteworthy. Species richness and speciation rates appear to be positively correlated in Mexican rodents and small mammals in the Philippines, but uncorrelated in Mexican bats (Heaney 2001; Sanchez-Cordero 2001). One potential reason for these differences is that bats have greater dispersal ability and thus their ranges may change more rapidly following speciation (certainly bats have ranges that are on average larger than those of terrestrial mammals). This suggests that speciation rates are more likely to be correlated with species richness when studies are conducted at larger spatial scales as species are less likely to disperse away from the region in which they evolved. Temporal scale may also be important, with younger taxa that have had less time to disperse being more likely to exhibit positive correlations between speciation rates and species richness. More numerous and detailed studies on the nature of the spatial association between speciation rates, current energy levels and current species richness are clearly needed across a range of taxa that differ in their dispersal abilities and age.

Future research

The explanatory power of the evolutionary-rates hypothesis rests on the validity of each of its four assumptions. Relatively large efforts have already been invested in some of them (e.g. the assumption that energy availability influences mutation rates). However, while the assumption that mutation rates control speciation rates is particularly central to the hypothesis rather little evidence is available with which it can be assessed.

Discerning the nature of the relationship between mutation and speciation rates is undoubtedly difficult, not least because of the number of confounding secondary factors. Experimental techniques provide the benefit of controlling for such factors, and although field experiments that manipulate energy levels and measure resultant mutation and speciation rates are clearly impractical, similar studies could be conducted under laboratory conditions. Indeed, much data on the speciation process has been provided by experiments on species such as Drosophila (Ödeen & Florin 2000; Rundle 2003; Martin & Hosken 2004). Similar experiments could also be conducted using digital organisms, i.e. self-replicating computer programs that mutate and evolve. Such techniques have, for example, been used to show that mutation can promote speciation in assemblages that would otherwise be evolutionarily static (McFadden & Knowles 1997; also see Chow et al. 2004). Additional experiments in more realistic environments would be useful, for example ones in which gradients in energy availability influence both mutation rates and population size.

Experimental approaches will, however, reveal only the potential for mutation rates to limit speciation; they will not demonstrate whether this potential is met in the wild. Such data are essential, and both the fossil record and molecular phylogenies may supply them. Studying speciation in the fossil record is difficult as information is lost in its gaps, and species with short life spans or restricted ranges are unlikely to be recorded. Despite this the fossil record has revealed much useful information on speciation rates, and is likely to continue to do so (Benton & Pearson 2001; Lieberman 2001; Munoz-Duran 2002; Alroy 2003; Jablonski & Roy 2003).

The accuracy of speciation rate estimates derived from molecular phylogenies is influenced by sampling in three ways. First, the accuracy of phylogenies is proportional to the number of individuals and genes sampled per species, but many phylogenies are based on a handful of individuals/genes (Barraclough & Nee 2001; Nichols 2001). Second, molecular phylogenies rarely include all extant species which creates a bias against recent speciation events and underestimates speciation rates (Nee et al. 1994). Third, the exclusion of extinct species may bias estimates of speciation rates (Chown & Gaston 2000). While much sampling remains to be done, continued reductions in the resources required for DNA sequencing will reduce the extent of inadequate sampling (Kocher 2004). Moreover, numerous advances are being made in the use of molecular phylogenies including the development of techniques that enable (i) the use of those resolved at the family level (Paradis 2003); (ii) take into account biases created by missing nodes that arise from incomplete sampling and extinctions; (iii) distinguish chance variation in speciation rates from that which requires deterministic explanation; and (iv) make explicit assumptions regarding the form of molecular evolution (Pybus & Harvey 2000; Barraclough & Nee 2001; Pybus et al. 2002; Chan & Moore 2002; McConway & Sims 2004). Molecular phylogenies will thus undoubtedly reveal much useful information on spatial variation in speciation rates. In addition, the same molecular data can also be used to measure mutation rates, allowing a direct assessment of the assumption that mutation rates control speciation rates.

The availability of more accurate data on speciation rates will facilitate testing whether they correlate closely with species richness (see Heaney 2001; Sanchez-Cordero 2001; Böhm & Mayhew 2005). Such tests should compare the ability of speciation rates to predict current species richness with that of other potential drivers of spatial variation in species richness, such as the amount of productive energy available measured by variables such as net primary productivity (see Davies et al. 2004; Kaspari et al. 2004).

Turning to the other assumptions, most research addressing the relationship between energy availability and mutation rates has been conducted under controlled laboratory conditions, and focuses on UV radiation rather than temperature (but see Lutzoni & Pagel 1997; Tuteja et al. 2001; Wright et al. 2003). It would thus be profitable to investigate spatial variation in mutation rates across free-living populations along an energy availability gradient. Such studies should attempt to distinguish the effects of temperature and UV radiation on mutation rates. This may be achievable by careful selection of the focal taxa, such as the comparison of mutation rates in diurnal and nocturnal species that experience similar temperatures, but markedly different levels of exposure to UV radiation; a similar comparison could be made between species that live in the forest canopy, with high exposure to UV, and those on the forest floor where exposure is reduced. Such studies could also take advantage of the fact that mutations induced by UV radiation may be identified on the basis of the exact nature of the associated base substitutions (Lutzoni & Pagel 1997).

The assumption that current and past energy levels are positively correlated has not yet been tested systematically, but could be achieved with the development of paleoclimatic databases (see Beerling & Woodward 2001). Testing the assumption that temporal changes in species distributions do not alter relationships between current species richness and speciation rates requires data on the number of species that (i) evolved in the focal region and still occur there; (ii) evolved in the focal region but are now extinct there; and (iii) that colonized following speciation elsewhere. Obtaining such detailed information is difficult and will require continued exploration of the fossil record, and the use of species-level molecular phylogenies. If such data are absent alternative approaches may be of merit. The centres of diversification are known, at a coarse resolution, for many genera and combining such data with information on current distributions may reveal the extent to which species disperse from their points of origin (Fine 2001).

In summary, a number of approaches are available that will enhance future assessment of the validity of the evolutionary-rates hypothesis. The one prevailing theme that connects most of these approaches is the use of molecular data and species level phylogenies, although continued exploration of the fossil record will also be important.


This work was funded by the Leverhulme Trust. R.G. Davies provided the temperature maps and, along with J. Slate and two anonymous referees, comments on an earlier version. We thank T. Ephraim (NASA Earth Observatory) for permission to use the maps of ultraviolet radiation.