Adaptive dynamics as a mathematical tool for studying the ecology of speciation processes

Abstract

After Ernst Mayr published his seminal book in 1963 (Mayr, 1963), the issue of speciation appeared to be settled: according to the established dogma, biological diversification occurred in allopatry due to the accumulation of genetic differences in geographically isolated populations. Despite repeated challenges, this view still prevails today, although perhaps less dominantly than before. The earliest rigorous theoretical challenge was provided by Maynard Smith (1966), who produced the first models of speciation in sympatry. These models were based on very simple ecological and genetic assumptions, with two resource types (or niches) and two loci, one for ecological performance and one for mate choice. Despite its simplicity, this type of model has formed the conceptual basis for most of the theory of sympatric speciation that has been developed since then (Kawecki, 2004). For sympatric speciation to occur in sexual populations, two processes must unfold. First, frequencydependent interactions must generate disruptive selection. Second, a lineage split in sexual populations requires the evolution of assortative mating mechanisms. Skepticism towards the feasibility of both these processes has led to a dismissal of sympatric speciation as a plausible mode of diversification. For example, based on Felsenstein’s (1981) seminal paper, it has long been thought that recombination between traits under disruptive selection and mating traits responsible for assortativeness can be a significant hindrance to the evolution of reproductive isolation between diverging lineages. Similarly, one of the main reasons why the theoretical developments following in the footsteps of Maynard Smith’s model failed to convince speciation researchers was that these models seemed to rely on rather particular ecological circumstances, such as host race formation (Diehl & Bush, 1989), and that the ecological conditions for the emergence of disruptive selection in these models were rather restrictive (Kassen, 2002; Kawecki, 2004). However, there is another line of thinking about the ecology of speciation that already started – how else could it be? – with Darwin, who concluded: Consequently, I cannot doubt that in the course of many thousands of generations, the most distinct varieties of any one species [...] would always have the best chance of succeeding and of increasing in numbers, and thus of supplanting the less distinct varieties; and varieties, when rendered very distinct from each other, take the rank of species. (Darwin, 1859, p. 155) According to this view, and in modern parlance, frequency-dependent competition between similar ecological types can lead to disruptive selection and diversification. This perspective was embodied in the concept of competitive speciation by Rosenzweig (1978) and further studied by Seger (1985), who presented the first mathematical model showing that frequency-dependent competition for occupation of a niche continuum can induce sympatric speciation under certain conditions. More generally, it was argued by Kondrashov (1986) that frequency-dependent selection on a continuous character can induce bimodal splits in the character distribution, with the two modes representing emerging species. In Kondrashov’s models, the disruptive selection regime giving rise to bimodality is simply a consequence of the a priori assumption that the fitness of common types is low, while that of rare types is high. It is difficult to assess the generality of these models, because it is not clear under what conditions ecological interactions would generate such a frequency-dependent selection regime. In fact, it is known that both competitive interactions (Christiansen, 1991) and predator-prey interactions (Abrams et al., 1993) can generate evolutionary scenarios in which the population mean of a continuous trait (such as body size) evolves to a state in which selection becomes disruptive. However, somewhat surprisingly, these results were never put into the common context of speciation, perhaps because these studies used the framework of quantitative genetics and thus assumed Gaussian phenotype distributions with constant variances (and hence implicitly assumed random mating). Overall, it thus remained questionable whether the emergence of disruptive selection due to frequencydependent interactions would be a general and plausible ecological scenario. In fact, it still seems to be the common wisdom that the origin and maintenance of diversity due to frequency-dependent selection regimes requires a delicate balance of different ecological factors (e.g. Kassen, 2002), and that, consequently, most biological diversification occurs in allopatry. We believe that the advent of adaptive dynamics, and in particular the discovery of the phenomenon of evolutionary branching, will change this perspective fundamentally (Dieckmann et al., 2004). Adaptive dynamics is a general framework for studying evolution of quantitative characters due to frequency-dependent interactions. Within this framework, evolutionary branching points Correspondence: Michael Doebeli, Department of Zoology and Mathematics, University of British Columbia, 6270 University Boulevard, Vancouver, BC, Canada V6T 1Z4. Tel.: +1-604-822-3326; e-mail: doebeli@zoology.ubc.ca