“ADAPTIVE SPECIATION”—IT IS NOT THAT EASY: REPLY TO DOEBELI ET AL.

The picture of the world of speciation according to Doebeli et al. (2005) is very romantic. There are the innocent, oppressed victims: "evolutionary biology" in general and "'speciation research" in particular. There are the shady villains: "past dogma," "old geographic classification of speciation," and "traditional population genetic approaches" as well as their "adherents" (Mayr, Dobzhansky, and others including the author of this response). There are also the brave, young revolutionaries: theories of "adaptive dynamics" and "adaptive speciation" that finally bring freedom to the masses of biologists after decades of the "shackles of allopatry." Here I suggest that this is not an accurate picture of reality and comment on the claims that adaptive speciation is theoretically plausible and that sympatric speciation is all but inevitable. A few clarifications are in order. (1) I gladly admit that research on "adaptive dynamics" has significantly improved our understanding of how genetic variation can be maintained under disruptive frequency-dependent selection. I also believe that work on "adaptive speciation" has been very stimulating for the field of speciation in general. (2) I do not think I really have a stake in the old arguments on allopatric versus parapatric versus sympatric speciation. In my work, I have studied models of all these processes. In fact, as a theoretician I am probably biased towards sympatric speciation rather than against it. Indeed, my work has already lead to analytical results on the conditions for sympatric speciation in eleven different models, both classical and new (e.g., Gavrilets 2004). I am puzzled how my Perspective article (Gavrilets 2003) can be viewed as a defense of "'allopatric dogma" given that it devotes four pages to demonstrating how sympatric speciation can occur in mathematical models. (3) The clearly stated goal of my article was to give a sample of speciation models for which analytical progress has been possible, which meant that a great deal of numerical papers, including those published under the umbrella of "'adaptive dynamics" or "adaptive speciation," were not discussed. Models of reinforcement were not discussed for exactly the same reason. The potentially promising analytical approaches being developed by Kirkpatrick and Servedio (e.g., Servedio and Kirkpatrick 1997; Kirkpatrick and Servedio 1999; Kirkpatrick 2000; Servedio 2000) assume weak selection and, thus, in my opinion, currently do not allow us to evaluate the real theoretical significance of reinforcement (because weak selection will only result in weak effects). The analytical results on "adaptive dynamics" cited by Doebeli et al. (2005) are concerned with the maintenance of genetic variation in asexual populations rather than with the evolution of reproductive isolation and speciation. For those interested in more details on numerical models, Gavrilets (2004) reviews general theoretical speciation research, and Waxman and Gavrilets (2005) provide a critical review of adaptive dynamics. (4) I agree that extensive numerical simulations can indeed provide a good understanding of a mathematical model. The problem is that theoreticians often do not perform extensive simulations. Moreover in many situations, particularly if there are more than five or six parameters, extensive numerical simulations are simply not feasible. Before discussing the plausibility of "adaptive speciation," it is necessary to clarify the meaning of this term. One straightforward interpretation is that this is a speciation process in which genetic changes underlying divergence and reproductive isolation are driven by selection (as opposed to changes driven by mutation and random genetic drift). This is however not how Doebeli et al. define adaptive speciation. What they mean by adaptive speciation is a much narrower category of "speciation processes in which the splitting is an adaptive response to disruptive selection caused by frequency-dependent biological interactions" (Dieckmann et al. 2004, p.4). "Adaptive speciation" requires ecological contact. This implies that allopatric speciation cannot be "adaptive.'' Another crucial feature of ''adaptive speciation'' is that its driving force is frequency-dependent rather than constant selection. This implies that speciation driven by adaptation to discrete ecological niches is not "adaptive speciation" according to Doebeli et al. In particular, if two races of a phytophagous insect speciate via adaptation to two different host plants (e.g., Bush 1969; Feder 1998; Berlocher and Feder 2002) or two different fish morphs adapt to benthic and limnetic environments (e.g., Schluter 2000) or a number of different lizard ecomorphs adapt to different parts of trees on a Caribbean island (e.g., Losos 1998; Losos et al. 1998) or Hawaiian spiders adapt to different ecological niches (e.g., Gillespie 2004) it is not "adaptive speciation." Unfortunately, Doebeli et al. (2005) fail to provide biological examples of "adaptive speciation." I suggest that the disparity between a narrow focus of adaptive speciation 'a la Dieckmann et al. (2004) and Doebeli et al. (2005) and a much broader interpretation of this term will probably result in a lot of confusion. From a theoretical point of view "adaptive speciation" can potentially occur both in sympatric and parapatric geographic settings. However because most of the new "revolutionary" claims have so far focused on sympatric speciation, I will concentrate on this geographic mode, in general, and on sympatric speciation driven by frequency-dependent selection, in particular. A general conclusion that emerged from several decades