The generalism–specialism debate: the role of generalists in the life and death of species

Authors

  • ROGER L. H. DENNIS,

    Corresponding author
    1. School of Life Sciences, Oxford Brookes University, Headington, Oxford OX3 0BP, UK
    2. Institute for Environment, Sustainability and Regeneration, Staffordshire University, Mellor Building, College Road, Stoke-on-Trent, ST4 2DE, UK
      E-mail: rlhdennis@aol.com
    Search for more papers by this author
  • LEONARDO DAPPORTO,

    1. Istituto Comprensivo Materna Elementere Media Convenevole da Prato via 1° Maggio 40, 59100, Prato, Italy
    Search for more papers by this author
  • SIMONE FATTORINI,

    1. Water Ecology Team, Department of Biotechnology and Biosciences, University of Milano Bicocca, Piazza della Scienza 2, 20126 Milan, Italy
    2. Azorean Biodiversity Group, Universidade dos Açores, Departamento de Ciências Agrárias CITA-A, Pico da Urze, 9700 Angra do Heroísmo, Portugal
    Search for more papers by this author
  • LAURENCE M. COOK

    1. Faculty of Life Sciences, The University of Manchester, Manchester M13 9PT, UK
    Search for more papers by this author

E-mail: rlhdennis@aol.com

Abstract

Specialisms on resources and for niches, leading to specialization, have been construed to be tantamount to speciation and vice versa, while the occurrence of true generalism in nature has also been questioned. We argue that generalism in resource use, biotope occupancy, and niche breadth not only exists, but also forms a crucial part in the evolution of specialists, representing a vital force in speciation and a more effective insurance against extinction. We model the part played by generalism and specialism in speciation and illustrate how a balance may be maintained between the number of specialists and generalists within taxa. The balance occurs as an ongoing cycle arising from turnover in the production of specialists and generalists, speciation, and species extinction. The nature of the balance depends on the type of resources exploited, biotopes, and niche space occupied. These vary between different regions and create taxonomic biases towards generalists or specialists. We envisage that the process may be sympatric/parapatric, although it is more likely initiated by allopatry driven by abiotic forces. © 2011 The Linnean Society of London, Biological Journal of the Linnean Society, 2011, 104, 725–737.

INTRODUCTION

In their recent study on ‘The evolutionary improbability of “generalism” in nature’, Loxdale, Lushai & Harvey (2011) pose the question: ‘does generalism truly occur in nature and, if so, where?’. This question emerges after an extensive review, largely focused on insects, in which they conclude that ‘ecological specialization ↔ speciation’ (where ‘↔’ indicates a causal path with direct feedback). In their paper, both the terms specialism and specialization are used and distinguished. Even so, ‘specialism/generalism’ and ‘specialization/generalization’, having the same ‘root’, can be easily confused. Here, we argue that generalism is a common feature of nature, that specialization is not a necessary progenitor of, or sole condition for, speciation, and that generalism is important in an evolutionary (i.e. gene selection, species making, extinction avoiding) context.

We define specialism/generalism in the traditional sense as a gradient in resource use, specialism describing use of limited (typically one to few related) resource components and generalism numerous (many/varied, more distantly related) components. For phytophagous insects, the key resource is the larval host plant(s), which acts as the key consumer resource in the terminology of the resource-based habitat concept (Dennis, Shreeve & Van Dyck, 2003; Dennis, 2010). Because larval consumer specialism/generalism involves relationships among host plants, as well as diet breadth, numerous operational definitions can be applied (Symons & Beccaloni, 1999; Janz, Nylin & Nyblom, 2001; Nyman, 2009). These do not affect the nature of our arguments because the relationships are hierarchical and can be abstracted at different levels. Relationships among plant foods obviously play a key part in phytophage speciation and extinction (Ehrlich & Raven, 1964; Janz & Nylin, 1998). We concentrate here on insect–plant trophic relationships because they play a key part in phytophage speciation and extinction, although our arguments obviously can be applied to other resources, such as utility (nonconsumer) resources and environmental conditions, as well as consumer resources that make up habitats (Dennis, 2010). After Loxdale et al. (2011), and in accordance with a modern interpretation of the traditional definition (Lincoln, Boxshall & Clark, 1982), we accept the definition of specialization as the ‘evolutionary adaptation to a particular (narrow) niche and resource base’. With specialization used in this way, there is no counterpart term (generalization) in general usage that can usefully form a polar alternative, although an evolutionary process leading to the enlargement of the resource base of a species can occur (Thomas et al., 2001).

Consequently, we can agree with Loxdale and colleagues that an organism such as a butterfly species can use a number of larval host plant species, genera or families but still has evolved through ‘specialization’. It could, for example, feed on certain plants or parts of plants in distinct ecological associations (dry or wet conditions, light or shade, etc.). Moreover, as Loxdale et al. (2011) point out, the habitat of a species is composed of a multidimensional array of resources (larval and adult host, mating and egg laying sites, micro-habitat, etc.) and, at least for some of these, the species could show strong specialisms. Here, we see how confusion may arise; the insect in question, although still unquestionably exhibiting specialization, is both a generalist and a specialist. However, we do not agree that it is sufficient to be specialized in one trait to be considered as a specialist, whereas it is necessary to be generalist on all/many traits to be considered as a generalist, because the habitat comprises a number of resources, all of which may impinge on persistence (Dennis, 2010). Moreover, there can be hidden relationships, such as so-called faux specialists and faux generalists (Brooks & McLennan, 2002): generalists may appear to be specialists owing to ecological or geographical restriction, whereas specialists may appear in the guise of generalists owing to their occurrence on a diverse but phylogenetically limited plant taxon or because they exist as yet unrecognized cryptic taxa (sibling species) (Loxdale et al., 2011). We agree that the term generalism as traditionally cast is typically beset with implications of scale dependency, at least at the individual level and increasingly larger as one progresses through siblings, populations, and regions to the full species' range. Also, it is very possible that it ‘cannot be maintained over the longer evolutionary term’ (Loxdale et al., 2011), although that perhaps is a questionable issue from the vantage of the ecological fitting hypothesis (Agosta, Janz & Brooks, 2010), which describes the variance in tolerance of species to conditions and resources, considered to be greater for generalists than specialists.

Before considering ecological situations, it is necessary to examine another aspect of the specialism–generalism issue not explicitly discussed by Loxdale et al. (2011). The two evolutionary processes of adaptation and speciation are not necessarily coupled. Opinion on how coupled they are has varied over the years (Gittenberger, 1991). If species are defined as reproductively isolated entities, then adaptation and speciation would be inseparable when conditions permitted sympatric or parapatric speciation. In the words of Loxdale et al. (2011), ‘speciation involves ecological specialization and vice versa’. They go on to say that the huge diversity of taxa is testament to this fact. But what proportion of this huge diversity is created by parapatric or sympatric species formation? Opinions vary, and are perhaps influenced by the type of material and level of observation employed; for example, compare Bridle, Pedro & Butlin (2004) and Evans, Cannatella & Melnick (2004). From the perspective of biogeography, it might be considered to be a modest one; species richness is associated with a complex of geological/climatic/habitat instability, which may usefully be referred to under the term ‘geodetic’ (Cook, 2008). High diversity may result from a fortuitous coincidence of geodetic periodicity on the one hand, and migration rate on the other, whatever further adaptation takes place in the new species. The neutral model of biodiversity (Hubbell, 2001; Rosindell, Hubbell & Etienne, 2011) can then explain why some geographical areas and taxa are species-rich, whereas others are poor. Applied to faunas of the tropical rain forest, this type of argument goes back to discussion of consequences of Pleistocene fractionation into refuge islands (Haffer, 1969; Cook, 1978; Legg, 1978) and continues to be useful (Fattorini, 2007; Basset et al., 2011). If geodetic processes are a common generator of new allopatric species, it is reasonable to assume that the parental species that would do best in the hurly-burly of colonization, establishment and subsequent population increase, local adaptation, and reproductive isolation, etc., would be generalists rather than specialists. In that case, over a period of time, there would be a tension between increasing specialization, which would improve fitness and competitive ability during phases of relative stability, and the favouring of generalists at times of geographical fractionation and climatic instability. The result could be a sort of equilibrium condition where generalism is protected and so is not an evolutionary improbability. Whether it arises from allopatry or parapatry, we shall follow Loxdale et al. (2011) in considering the adaptive features of this tension, especially among butterflies.

The key questions we investigate are: (1) what part does generalism play in furnishing specialists and thus specialization and speciation; (2) what are the comparative advantages of specialists and generalists in forestalling extinction; and (3) what is the overall contribution of generalism and specialism to speciation, clade furcation, and persistence? These three issues are dealt with in turn below.

IS THERE A ROLE FOR GENERALISM IN SPECIATION?

We argue that generalism may well provide the ‘fast route’ to specialism and thus to specialization and speciation. This argument implies that there are many greater opportunities through generalism (than there are from specialism) for clade furcation and thus for species with specialisms (Janz, Nylin & Wahlberg, 2006; Janz & Nylin, 2008). Specialisms (and increased specialization) can arise in generalists as much as generalism can arise in specialists (e.g. when novel plant use could occur spontaneously in both specialists and generalists) (Agosta et al., 2010). The second issue focuses on the dark side of species' existences: extinction. Is specialism likely to be associated with a higher extinction load than generalism when, with changing abiotic conditions, there are fewer outlets to fall back on? Specialization, with its vulnerable (more limited) web of niche/resource links, tends to be associated with extinction on the geological time scale (McKinney, 1997). According to Loxdale et al. (2011), species showing generalism in some traits should show specialism in others. The important questions, as examined below, are: does this deny generalism and generalists, and how do combinations of generalism and specialism traits emerge in evolution?

If there is a role for generalism, this should become apparent in several traits of phytophagous insects such as butterflies. Generalism in butterflies is usually related to larval host plant use and, if generalism is commonplace, there should be many butterfly species that exploit more than one host plant species. This is typically the case among European butterflies, both for Britain and the continent (Fig. 1) (Dennis, 2010). In British butterflies, at least, larval host plant numbers correlate with the other main consumer resource, comprising nectar-providing plants exploited by adults, and the number of utility resources used by different developmental stages (Dennis et al., 2005; Dennis, 2010). For host plant–nectar sources Pearson r = 0.48, for host plant–utility resources r = 0.58 and for the relationship of nectar sources with utility resources, r = 0.58 (P < 0.001 in each case) (Dennis, 2010). As the correlations indicate, there is room for species that are referred to as generalists on the basis of one resource to be specialists in another.

Figure 1.

Frequency distribution of the number of larval host plants used by 387 European butterfly species.

If generalism allows species to escape extinction and to spread over large areas and, in doing so, to differentiate over larger areas, the number of used host plants should relate to some factor(s) that ensure: (1) the persistence of species and (2) the evolution of novel species. Again, this is the case for butterflies. Among British butterflies (N = 60 species), increased use of host plants correlates significantly with number of occupied biotopes (r = 0.54), niche breadth exploited (r = 0.78), and geographical range size (r = 0.57); with the latter relationship also being supported by a wider array of species over Europe (r = 0.73, P < 0.001 in each case; Fig. 2) (Dennis, 2010). Generalism may enhance evolution through an increased potential for novel resource use, and thus novel habitats and niches, as well as ensuring potential for increasing persistence of a species. It does so, we argue, by creating a variety of opportunities in space–time that counteract the tendency to specialization. In Polygonia (Nymphalidae), taxa using other or additional host plants than the urticalean rosids were found to be more species-rich than their sister clade restricted to the ancestral host plants (Weingartner, Wahlberg & Nylin, 2006). This finding is consistent with the theory that expansions in host plant range underpin diversification in butterflies and other phytophagous insects, and correlate with speciation either by driving it (West-Eberhard, 2003) or simply making it more likely.

Figure 2.

Relationship between distribution of European butterflies and their host use breadth. Distributions, proportion of mapped squares (Kudrna, 2002) occupied, arcsine transformed; number of larval host plants, log transformed. Symbols: cross, Hesperiidae; triangle, Papilionidae; circle, Pieridae; diamond, Lycaenidae and Riodinidae; small square, Nymphalidae (Satyrinae); large square, Nymphalidae (remainder). F1385 = 446.87, R2 = 53.7%, P < 0.0001, N = 387. Differences (residuals from regression line) between butterfly families; analysis of variance: F5381 = 1.58, P = 0.16.

The potential for species to adopt novel host plants and other resources is central to evolution of host use and the frequency distribution of hosts used by species. Does the use of new host plants increase as a phytophagous insect's geographical range size expands (e.g. response to an abiotic trigger) or is multiple host use a feature, although a lesser one, of the speciation process? Furthermore, is novel host use counteracted by host abundance? Regarding the former, there is recent well documented evidence of increasing use of novel host plants with geographical range expansion in the warmer conditions that actually trigger range expansions in thermophilous insects [e.g. Aricia agestis Denis & Schiffermüller (Lycaenidae), Thomas et al., 2001; Kemp et al., 2008] and sometimes resulting from human land use practices (Singer, Thomas & Parmesan, 1993). Larger populations and increased dispersal provide opportunities for oviposition mistakes and evolutionary experiments (Dennis, 1993; Tammaru, Kaitaniemi & Ruohomaki, 1995; Nylin, Bergström & Janz, 2000). There are examples where the reverse holds: a loss of host plants with geographical range retractions (Hardy et al., 2007). This has been the case for a number of British Lepidoptera [e.g. Papilio machaon L (Papilionidae) Lycaena dispar Haworth, and Lysandra species (Lycaenidae)] during mid-Holocene forest development and Sub-Atlantic climatic cooling (Dennis, 1977) and is clearly a feature of contractions in distributions with current fragmentation and extinctions [e.g. British Carterocephalus palaemon Pallas (Hesperiidae); Thomas, 2005]. It is also evident that novel host use in the wider sense may accompany emergence of new species regardless of whether they develop from a single unit (population or metapopulation), as sympatric/parapatric entities, or allopatric isolates, or from regional units (multiple populations) created by vicariance. In the latter case, multiple host use may already be a feature of the divided species' range with geographical division. The former case is particularly interesting because it could involve de novo resource use and radiation onto novel host taxa [e.g. Euphydryas editha Boisduval (Nymphalidae); McBride & Singer, 2010]. The factors resisting this process are fully accounted by Loxdale et al., involving plant chemical and other defences and, for this reason, attempted novel shifts are not always successful [e.g. Pieris napi L. (Pieridae) on Thlaspi arvense L.; Chew, 1977]. For a new species to spontaneously access new hosts, it must either: (1) choose (related) plants with very similar defences to those of the ancestral host plant; (2) exploit plant families using broadcast general defences easily overcome or (3) possess sufficient genetic variability (thus large enough populations) and/or have sufficient phenotypic plasticity to experiment on more ‘distant’ plant hosts at the risk of much failure (Futuyma, 1983; Nylin & Janz, 2009; Agosta et al., 2010). Key factors will be proximity and physiological and chemical (phylogenetic) similarity because they increase the probability of novel host use. A classic case of transfer related to physical contact is the likely move in Pieridae from Brassicales host trees to mistletoes (Santalales) aerially hemiparasitic on them and, hence, to other tree taxa on which mistletoes abound; such moves are neatly modelled by Braby & Trueman (2006).

So what part does generalism (resource breadth) play in the speciation process? We would argue with Janz & Nylin (2008; Nylin & Janz, 2009) that generalism provides the means for generating new species through creating specialists in isolates that are dependent on singular resource use. Adoption of novel hosts has the capacity to transform life-history parameters, leading to further specialization [e.g. Operophtera brumata (L.) (Geometridae); Vanbergen et al., 2003]. This can occur during environmental upturns and range expansion (with novel resource adoption) or during environmental downturns causing range/distribution contractions (with residual resource dependence). The result can be significant positive correlations between resource use, population size, range size, biotope occupancy, and niche breadth (Dennis et al., 2004). New genes (or changes in gene regulatory networks) emerging in a population that increase host resource use (generalism) ensure greater opportunities and a greater number of ‘experiments’ (niches) in a wider variety of biotopes over a larger region subject to greater variation in environmental conditions (Fig. 3A). During range expansion, generalists have greater opportunities to develop novel associations; specialists are likely to be more dependent on limited host chemistry for defence against predators and this relationship would tend to deter host shifts [e.g. Junonia coenia Hübner (Nymphalidae); Camara, 1997]. In the terminology of Agosta et al. (2010), specialist individuals have a lower capacity for ecological fitting than generalists. Emergent generalist genes may spread in the population, allowing a specialist species to become a generalist one, a process facilitated by increasing dispersal capacity (Fig. 3B). For species characterized by moderate dispersal ability and in hard climatic conditions such as during glacial maxima, this process could also lead to local speciation in refuges and then to successive spreading of the new generalist species in the area occupied by its ancestral taxon. This appears to be the case for Leptidea butterflies (Pieridae). The original Leptidea sinapis (L.) has been recognized to comprise a pair of species (Leptidea sinapis and reali), with the former showing generalist traits and the latter being more specialized (Friberg & Wiklund, 2009). Recently, this clade has been shown to involve a triplet of species (two specialists: L. reali Reissinger and Leptidea juvernica Williams; one generalist: L. sinapis) whose ancestor was probably closely related to one of the two specialists. Leptidea sinapis originated in Western Europe at the extreme margin of the ancestral distribution, and then expanded north and east into the territory of its ancestor (Dincǎet al., 2011) by a mechanism very similar to that shown in Figure 3B.

Figure 3.

Schematic diagram to illustrate the envisaged advantages created for a novel gene/species, from uptake of a distinct resource, over its monophagous sibling. A, process-response model. Polyphagy, linked to increased migration capacity, leads to more rapid colonization of sites and increased experimentation/uptake of additional host resources. The system involves a positive feedback in which polyphagy, once established, has further opportunities of expanding. B, schematic illustration. (i) The system is represented by four areas characterized by the presence of two resources: black and white. Initially, the gene/species occurs in the central area and it is specialized in a single (black) resource. Then it spreads to the surrounding areas [grey arrows (i)]. In the area comprising an alternative resource (white), a new gene/species (white) emerges, allowing use of both resources (white and black) (ii). The new gene/species can then spread to the other areas where the individuals can use both resources [black arrows (iii)]. The white gene/species thus has a higher probability to grow and to disperse among patches in the metapopulation [large black arrows (iii), (iv)]. As such, the proportion of white gene/species can increase rapidly, compared to the grey gene/species in the metapopulation until the ancestral, specialized gene/species is completely replaced by the generalist (iv).

There is evidence of a correlation between generalism (number of resources used) and migration capacity in British butterflies (r = 0.43, P < 0.001, N = 60 species), with the latter measure tested in the field (Cook, Dennis & Hardy, 2001; Stevens, Turlure & Baguette, 2010). In butterflies, there should also be a substantial association between colonization ability and migration capacity, something that has been previously demonstrated (Dennis et al., 2000, 2005). During range contraction (or distribution fragmentation), however, generalists have greater opportunities for survival by occupying a wider variety of biotopes, and thus more sites, than specialists. Range contraction probably leads to specialization, with growing isolation and adaptation to changes in conditions and resource availability. Janz & Nylin (2008) describe a model, the oscillation hypothesis, which accounts for how combinations of environmental upturns and downturns can lead to specializations and speciation. Generalism with range expansion is an essential feature of the model; below, we draw together the factors that support this notion and incorporate links to extinction (Scriber, 2010).

WHAT IS THE ROLE OF GENERALISM IN SPECIES' SURVIVAL?

In Europe, it is likely that generalism has been critical for surviving glacial-interglacial cycles in refuges (Dennis, Williams & Shreeve, 1991; Dennis, Shreeve & Williams, 1995). The larger geographical range alone ensures a greater number of populations, metapopulations, and regional populations. The link between generalism and increased mobility also ensures that population units over a wider area remain intact, linked by gene flow; thus, metapopulations have more local population units and will persist longer. That is one reason why generalists may continue to be generalists; wide resource use selects for mobility and greater mobility ensures that isolated resources can be accessed and that a newly-evolved generalist gene can spread inside the original population. Migrants incapable of locating limited, suitable resources are rapidly selected against (Futuyma, 1998).

Specialization is tantamount to extinction on geological time scales (e.g. Cope's Law; Kaiser & Boucot, 1996) and sometimes on shorter historical ones (e.g. primates: Harcourt, Coppeto & Parks, 2002; Coleoptera, Carabidae: Kotze & O'Hara, 2003; corals: Munday, 2004; plants: Cousins & Vanhoenacker, 2011); the chance of extinction is enhanced by the low resilience of a narrow niche, vulnerability of a specific biotope, and the precariousness of dependence on a limited site or narrow geographical range. However, not all generalists have large ranges and not all specialists have small ones. Then there is the question of why some specialists with narrow distributions and single biotopes do not contract further and become extinct. Sometimes, there may be a problem distinguishing true generalists (i.e. mutliple resource use on same sites) from polyspecialists (i.e. species comprising regional populations, each population a resource specialist) (Nylin & Janz, 2009), although an essential component lies in the multiplicity of factors making up colonization ability and migration capacity. Clearly, specialists exist that have extensive ranges based on single larval host plants; at best, these plants are typically ubiquitous, abundant where they occur, and available throughout the ‘growing’ season [e.g. Vanessa atalanta L and Aglais urticae L. (Nymphalidae) on Urtica dioica L.]. A striking example occurs among longhorn beetles. Phoracantha semipunctata (Fabricius) (Cerambycidae) feeds only on Eucalyptus and was originally Australian, although it is now a cosmopolitan species because of artificial plantation of these trees (Paine, Steinbauer & Lawson, 2011).There are also specialists that are polyphagous (Dennis et al., 2004); classic examples being those on temporally limited plant parts, such as flowers and young growth [e.g. Plebejus argus L. (Lycaenidae) on Cistaceae, Ericaceae, and Fabaceae; Dennis, 2010]. This observation points to greater complexity in specialism and generalism than basic resource variety. In a detailed study of British butterflies, specialists were deemed to be stress tolerators in the triangular plant competitive, stress-tolerant, ruderal strategy system (Grime, Hodgson & Hunt, 1988) and dependent on highly synchronized or narrowly available food of slow-growing plants in environmentally more extreme conditions (Dennis et al., 2004).

A key to the survival of specialists is the link with population size. Specialists are best able to focus their defences and digestibility on the plant (or plant parts) used so as to build large and dense populations (Loxdale et al., 2011); this relationship extends to other forms of specialism (e.g. terrestrial vertebrates in the rainforests of the Australian Wet Tropics; Williams et al., 2009; but see also Hobbs, Jones & Munday, 2010). On the other hand, as Loxdale et al. (2011) explain, a generalist can only adapt broadly to the defences encountered; losses will be higher on the suite of plants used and populations less synchronized, less dense, and smaller. This is compensated for by greater mobility to seek out new resource sites and to track suitable habitats with climate change (Dennis, 2010), and also by wider ecological fitting (Agosta et al., 2010). The inverse relationship in butterflies between population size and migration is well established (Bink, 1992). A final, ironic, point is that during environmental downturns, generalists ultimately have the ‘option’ of surviving by becoming specialists in isolates (refuges), sometimes to become novel species. This is the likely reason for the predominance of specialists. Such an option is less open to (already) specialist species during environmental downturns; generally, they must specialize further or become extinct. During environmental upturns, their persistence depends on exactly how the environmental changes impact on their limited resources. If their resource spreads, they too have the opportunity of doing so. For species with generalist tendencies, the opportunities during environmental upturns are much greater and extinction is less likely. Moreover, if they have spawned specialists during the downturn, then the more such isolates there are, the greater the chance of some clade continuing.

GENERALISM, SPECIALISM, SPECIATION, AND EXTINCTION: CONCATENATIONS

The links between specialism, generalism, speciation, and extinction are not evident without recourse to other variables affecting a species' demography and population dynamics. Here, we consider how the constellation of factors surrounding the specialism/generalism issue integrate to influence speciation and species' extinction (Fig. 4). Specialization leads to speciation, although it is subject to an increased likelihood of extinction. Generalists have reduced probability of extinction as a result of enhanced resource opportunities and greater colonization ability. Generalism may also provide some insurance against viral infections and other enemies. Even though polyphagy of a vector may well present viruses with opportunities for transmission and extending their host range (Keldish et al., 1998), a generalist phytophagous arthropod is likely to encounter a wider range of pathogens than a specialist and, accordingly, may develop immune responses that are capable of dealing with a wide range of pathogens, whereas a specialist is likely to encounter a narrower pathogen range and have more specialized immune responses (T. G. Shreeve, pers. comm.). Compared to specialization and specialism, this should increase species' persistence. Exceptions only become possible where extinction with specialization is countered by potential for increase in population size. However, more typically, the single host resource of a specialist is rare, sparse, and ephemeral.

Figure 4.

Factors in the generalism–specialism cycle linking host plant generalism (specialism) to speciation and extinction in phytophagous insects (butterflies). Shaded boxes and bolder links highlight just two of the integrated effects of generalism in the model, resistance to extinction, and generation of inter-population variability, the path to speciation. The full model envisages a burgeoning of generalism followed by specialization, triggered by oscillations in environmental conditions (typically abiotic fluctuations/cycles, particularly climate; i.e. warmer/colder, wetter/drier) over millennia and geological time, with each cycle uniquely influencing range changes and abundances in different taxa. Upturns describe conditions that lead to higher population density, distribution infilling, and range expansion of any butterfly species; downturns, the reverse. Marginality is the direction in which the niche is the furthest from the average environment (Calenge & Basille, 2008). HTNS, host taxonomic neighbourhood size. The model has its basis in island biogeography in that generalists have greater capacity for colonizing plants, persisting, and producing specialists that become new species than do specialists, with the number of specialists being replaced via clade furcation from generalists.

Species' generation is more complex because it involves the temporal dynamics of abiotic environmental changes (e.g. climate), which may increase or decrease range, distribution, population abundance or number (i.e. the oscillation hypothesis of Wilson et al., 2004; Janz & Nylin, 2008). Whether increasing or decreasing, generalists may have opportunities to specialize during the process of speciation. This is envisaged as occurring directly via ecological fitting (Fig. 4) (Janz & Nylin, 2008; Agosta et al., 2010), although it may more commonly come about through allopatry. Because generalism is positively related to gene flow, the emergence of specialism from generalism will depend primarily on independent agents that increase choice, distinctions in resource use, and overall isolation among population units. Such generalism–specialism turnover depends on abiotic pulses (Fig. 4); the model illustrates the concatenations of these processes and interactions among variables, including those discussed by Janz & Nylin (2008), Scriber (2010), and Loxdale et al. (2011). Each taxon will be affected differently by changing conditions but, over many cycles of environmental changes, generalists will fare differently from specialists. During environmental upturns, increases in range size, occupancy of biotopes, biomes and environments, and genetic variability facilitate ‘experimentation’ on new hosts (Craig et al., 2011). Whether the uptake of new resource opportunities leads to specialism depends on factors allowing independent development, such as changes in voltinism or mating time. Shifts and changes are more likely to increase towards the margins of expanding ranges, causing them to differ increasingly from the geographical core (Shreeve, Dennis & Pullin, 1996; Bridle & Vines, 2007; Dincǎet al., 2011). Increased distance of novel habitats from natal habitats involves adaptations to contrasting conditions. Temporal isolation resulting from asynchrony between daughter populations on the parental resource can trigger sympatric/parapatric evolution (Drès & Mallet, 2002; Fitzpatrick, Fordyce & Gavrilets, 2008). With enhanced colonization opportunities and migration capacity, generalists can more easily occupy isolated host patches. These may be on oceanic islands or terrestrial islands well beyond the range margin. The processes illustrated in the model (Fig. 4) have close affinity with island biogeography concepts, with an example being the highly nested structure of butterflies on British islands. In a nested pattern, the species composition of small assemblages is a nested subset of the species composition of large assemblages. When this occurs, the inference is that different species do not have the same probability of dispersing to, and persisting on, different areas (islands) contrasting for characteristics such as physical geography (Ulrich, Almeida-Neto & Gotelli, 2009). It is evident that generalists, using more resources, also occur on more islands (Fig. 5).

Figure 5.

Relationship of generalism in larval host use and adult nectar source exploitation illustrated against nestedness of species for British islands. Rows: species (N = 60 species, with each depicted by five letters: one for genus and four for species from Dennis, 2010); columns: islands (N = 103 islands unnamed); in the table: shaded cells, species present; unshaded cells, species absent (for individual islands). Columns to left of species' labels: left column, correlation of host plants versus nestedness order: r = −0.45, P = 0.001; right column correlation of nectar source versus nestedness order: r = −0.73, P < 0.001; darker tones indicate lower classes of nectar sources and host plants. Nestedness analysis on British offshore islands was undertaken using the BINMATNEST based on the algorithm proposed (third null model) by Rodriguéz-Gironés & Santamaría (2006), which implements the procedure of packing the presence/absence matrix. British butterfly species have a highly nested structure on islands (temperature metric 12.20°, P < 0.001).

During downturns involving range contractions and fragmentation, generalists have greater potential for creating specialisms in refuges. These can then progress towards specialization and speciation (Janz & Nylin, 2008; Agosta et al., 2010). The shift from generalism through specialization to speciation probably underpins the numerous clades that have developed in Mediterranean Europe during each glacial cycle (Dennis et al., 1991, 1995; Dapporto et al., 2009, 2011a, b). An upturn followed by a downturn creates additional pockets of populations with resource use over a wider, increasingly heterogeneous region during the upturn, which, during the downturn, are further isolated and driven to specialize and speciate. Each experiment is further tested at hybrid zones during a subsequent environmental upturn (Descimon & Mallet, 2009; Dapporto et al., 2009, 2011a). The apparent dominance of allopatric events over sympatric ecological speciation in sawflies (Nyman et al., 2010) supports this interpretation. The process must have been seeded with host specialists as envisaged by Janz & Nylin (2008), although increasing numbers of host shifts through geological time generate differences between taxa in both opportunities and ecological fitting capacity. The balance of specialist–generalist potential provides opportunities for further speciation and reduces the chance of extinction. When downturns are severe, as in northern Europe during the last glacial stage, mass extinction of populations and species is to be expected. On the other hand, the generation of greater resource breadth during upturns can be maintained if populations remain intact or develop increased migration capacity during seasonal or glacial–interglacial environmental cycles, as perhaps in Vanessa cardui L (Nymphalidae) (Dennis, 1993; Stefanescu, Alarcón & Àvila, 2007; Stefanescu et al., 2011). This may provide one answer to Scriber's (2011) question: why host-associated divergence does not always lead to new species. At any time, different higher taxa will have different mixes of specialists and generalists depending on the previous impact of environmental conditions and their geographical bias to different regions (Dennis et al., 1995). Generalists provision more specialists than specialists do generalists, whereas specialists are more prone to extinction, so that there should be more specialists than generalists. This is the pattern among European butterflies: fewer species have increasing numbers of larval host plants. Despite this two-way process, the assumption that specialism is always a dead end persists in the literature (Thompson, 1994; Kelley & Farrel, 1998) when clearly generalism for host resources must initially have evolved from specialism for a novel resource item (Janz et al., 2001). Environmental change may present upturns for either category, especially at boundaries of closely-related taxa (Scriber, 2011). Conditions undoubtedly occur in which specialists can take advantage of an increasing but limited resource base; Vanessa atalanta, for example, largely dependent on nettle Urtica dioica L., can be abundant, and is very widespread and an effective migrant (Baker, 1978) (Fig. 5). Bearing in mind that each taxon has deep phylogenetic roots that track multiple host resource shifts (Janz & Nylin, 1998; Janz et al., 2006), this is hardly unexpected; they may have evolved from generalists in the first place. If generalism provides the basis for specialisms, it may also lie at the root of increasing complexity in the geological record (Adamowicz, Purvis & Wills, 2008).

CONCLUSIONS

From the above, we may summarize the present review with three questions and comments. First, is there a role for generalism? Within a population, competition clearly generates a tendency towards specialism. However, specialists are vulnerable to extinction and, in the complex changing metapopulation world of local adaptation and migration, generalism can ensure increased persistence. Second, has generalism a role in speciation? Competition is only intimately associated with speciation in the sympatric/parapatric context. If allopatric speciation is common (we still do not know how common it is), the generalist, being more diffused, may be the most likely to occur in newly separated areas, to survive the initial stages of separation, and furnish more surviving clades than the specialist. Third, how do adaptation and speciation connect? Specialization may be ecologically beneficial in one context but increases the chance of extinction in another, ensuring that a variety of strategies are to be found in nature. Key amongst these, in our opinion, is the occurrence of a balance between the tendency to specialization and to enlarging the resource base, thus producing generalists from specialists.

ACKNOWLEDGEMENTS

We thank Hugh D. Loxdale, Jan Christian Habel, and two anonymous referees for their most helpful comments on the text and Tim G. Shreeve for his personal communication.

Ancillary