American Scientist

The Geographic Mosaic of Coevolution. John N. Thompson. xii + 443 pp. University of Chicago Press, 2005. $75 cloth; $28 paper.

The ability of any species to survive and reproduce is determined to a large extent by other species—prey, predators, pathogens, competitors and mutualists. This fact suggests that if we are to have any hope of understanding why a given species has the characteristics we observe today, we need to understand coevolution, the coupled evolutionary changes that occur in sets of interacting species.

Explaining traits based on the mutual evolution of interacting species dates back to Darwin, who noted that faster deer would select for increased speed in wolves, and vice versa. Evolutionary biologist John Thompson has arguably contributed more to our understanding of coevolution than has any other living scientist. In The Geographic Mosaic of Coevolution, his third book on the topic, he elaborates an idea introduced in his second book, The Coevolutionary Process (1994): the notion that coevolution must be understood in the context of a physical environment that is heterogeneous and often spatially discontinuous. These characteristics mean that the evolutionary forces acting on a given species will usually differ from place to place and that evolutionary outcomes will depend on these local differences in selection combined with movements of individuals and/or genes.

Of course, selection on any characteristic in any organism will vary in space and time. Tolerance for high temperatures is selected for more strongly in warmer locations and in hot years. No evolutionary biologist can completely ignore spatial variation in conditions. Evolution of characteristics involved in interactions with other species—coevolution—is certainly no exception to this general rule.

To give a purely hypothetical example, some deer may find themselves in a suburban neighborhood without a wolf within hundreds of kilometers, whereas others may live in a Canadian national park with a large and vigorous population of predators. Some may find themselves coexisting with wolves but also with some other prey species that the wolves prefer. Others may live in places where there are no wolves but great speed is required to avoid heavy traffic. If longer legs confer greater speed at a cost of increased chance of broken legs, the length of a deer's legs will depend on local selection for greater maximum speed and on the movement of individuals to or from populations where speed is less of an advantage. Wolves experience their own spatially varying selection for traits that make them faster. Because any selective pressure that affects speed in wolves is likely to also affect selection on traits that determine speed in deer that are exposed to those wolves, understanding the evolution of leg length in either species requires an understanding of all factors affecting speed in both species. There is no doubt that leg length or other traits related to speed should vary spatially across a heterogeneous environment, provided the genetic exchange between different areas is not so high as to overwhelm adaptation to local conditions.

Thompson reviews a number of case studies in which the outcome of coevolution differs from one area to another. For example, some populations of lodgepole pine trees have cones that resist predation by crossbills, birds specialized to extract conifer seeds from cones. However, the cones' adaptations for resisting these birds are much less developed where red squirrels are the main source of mortality for pine seeds. Different adaptations are required to defend cones against red squirrels. Thompson chooses to classify different locations dichotomously as being coevolutionary "hotspots" or "coldspots." Areas with many red squirrels are a hotspot for pine-squirrel coevolution, but a coldspot for pine-crossbill coevolution.

I have to admit that when Thompson first introduced his "geographic mosaic" idea, I was rather puzzled by it. Ecologists and evolutionary biologists have long known that any species with a significant geographic range will experience different conditions in different places—different physical variables, different sets of interacting species and different characteristics of almost any particular interacting species. Yet there is no demand that every ecologist or evolutionary biologist must study geographical variation to understand every single local population. The key question is, Why should coevolution require a special "geographic mosaic theory," when the purely ecological study of between-species interactions does not have this requirement, nor does the evolutionary study of traits that are not involved in interactions? What is it about coevolution that makes it different from these other fields? I am afraid I found no convincing answer to this question in Thompson's new book.

In any evolutionary study, it is clearly important to be aware of potential local differences in selection pressures and movements of individuals for different populations. However, there have been successful studies of coevolution of predators and prey and of competitors in environments (such as small lakes) where gene flow and spatial differences in selection were not key attributes of the process. This is the case with Dolph Schluter and J. Donald McPhail's well-known work on evolution in the three-spined stickleback, work that is summarized quite nicely in Thompson's book.

On the other hand, Thompson is no doubt correct in arguing that many biologists studying coevolution have devoted too little attention to spatial variation. Until recently, evolutionary biologists studying species interactions had not studied spatial variation to the extent that population ecologists have done. The body of work summarized in Thompson's new book provides a strong case for correcting this imbalance by considering the impacts of spatial variation when thinking about coevolution.

Thompson's own research on the interaction and evolution of a variety of butterflies and moths with their host plants represents a model of how fieldwork can help one to understand the nature and magnitude of current-day interactions. His studies of the coevolution of yuccas and yucca moths and of swallowtail butterflies and their host plants are featured in many evolution textbooks. His work has revealed several cases where interactions that had at one time been classified as mutualistic have turned out to vary across the full spectrum from mutualism to neutrality to parasitism. His recent collaborations with Richard Gomulkiewicz and Scott Nuismer have laid a firm basis for mathematical modeling of evolving traits in two or more species distributed across a heterogeneous landscape.

All of this work is presented in a clear and simple manner in the book under consideration here. However, The Geographic Mosaic of Coevolution is far more than a review of Thompson's own work or a discussion of how spatial variation alters the evolution of interacting species. It is a review of the majority of empirical and theoretical work on almost all facets of coevolution that has appeared over the past decade.

Although such a review is needed, and Thompson does a thorough and insightful job, trying to fit the diverse body of research on coevolution into a single theory with codified assumptions, hypotheses and predictions may not be the best way to advance research in the field. It is not necessary to study everything as a geographic mosaic. There are also reasons to question the generality of the five assumptions and three hypotheses used to define the Geographic Mosaic Theory, and the classification of adaptive outcomes into seven major trajectories seems restrictive and somewhat arbitrary. These trajectories are actually reflective of the rather limited number of species assemblages in which coevolution has been studied.

Take the example of "coevolutionary displacement," one of Thompson's seven trajectories. Thompson defines this to be divergence of competing species, although several researchers in the field (I am one) have argued for definitions of "displacement" that include any direction of change, rather than just divergence (species becoming more dissimilar). Given that there are easily a dozen mechanisms that in theory are expected to produce convergence or parallel change in response to competition, it seems premature to define the only coevolutionary response to competition to be divergence. In Thompson's scheme, convergence is restricted to being a consequence of mutualistic interactions. Both convergence and divergence have also been predicted as potential outcomes of predator-prey interactions. Coevolutionary trajectories are likely to be far more diverse than the limited number of previous studies has suggested.

Other aspects of Thompson's framework can be criticized as well. The prediction that there will often be hotspots and coldspots of coevolution appears to be almost tautological; every ecological or evolutionary process is stronger in some places than in others. And I would like to know why Thompson chose to retain the historical restriction of the term "coevolution" to cases with reciprocal evolutionary responses within both species of a pair of interactors. The same models and the same message about the potential importance of space apply to the evolution of any trait that influences interspecific interactions of a single species within a biological community, even if there is no coevolutionary response.

Although the entire complex edifice of geographic mosaic theory may not stand the test of time, there is no doubt that Thompson's emphasis on the implications of spatially varying selection will help us to better understand the evolution of interspecific interactions. His fieldwork and his theoretical collaborations will be cited for many years to come. There is no more authoritative source for the latest research in one of the most important areas of evolutionary biology.