Process and Pattern in the Evolution of Species Associations

Abstract

We discuss the use of phylogenetic patterns in testing the predictions of coevolu- tionary scenarios or process models. For this purpose we develop a general method for recon- structing the coevolutionary history of an association from the phylogenies of the component groups and an association matrix describing the associations between the species. We focus on the evolution of the traits that determine which species are associated, and therefore set up a null model with no change in these traits. Furthermore, we distinguish among three different processes that change association patterns: (1) colonization, whereby new associations are created; (2) exclusion, whereby old associations disappear; and (3) successive specialization, through which ancestral associations split into two or more descendant associations. In contrast to earlier methods for coevolutionary analysis, the method we propose explicitly includes all these pro- cesses as possible explanations of observed deviations from the null model. The parsimony principle can be used to choose between alternative reconstructions of the history of an asso- ciation if successive specialization, colonization, and exclusion events are given weights relative to the probability of each of these events occurring. However, it is often possible to take advantage of the distribution of traits in the phylogenies of the associated groups to choose between alternative reconstructions; then, the weights need not be specified. Although the method we propose is explicitly developed for coevolutionary analysis, it should also be applicable to biogeographical problems with suitable modifications. The method is compared to parsimony mapping (Brooks parsimony analysis) and found to be superior in several respects. To illustrate the potential of the method, we first discuss different coevolutionary process models for insect- host plant associations with emphasis on the patterns they predict and how reconstructions of the coevolutionary history of an association can be used to test these predictions. We then use the method to analyze the evolution of an association between gall midges (Semudobia spp.) and birches (Betula spp.). (Species associations; coevolution; parsimony mapping; Brooks parsimony analysis; insect-host plant relations.)