American Society of Naturalists

A membership society whose goal is to advance and to diffuse knowledge of organic evolution and other broad biological principles so as to enhance the conceptual unification of the biological sciences.

“A comparative test for divergent adaptation: inferring speciation drivers from functional trait divergence”

Posted on

Sean A. S. Anderson and Jason T. Weir (Oct 2020)

A new framework to test for signatures of divergent selection between sister species and other paired lineages

Read the Article

Testing for general signatures of adaptive trait divergence between lineages

Biologists have long been intrigued by the creative power of adaptation. More than 160 years ago, Darwin and Wallace suggested that populations adapting in different environments could evolve into distinct ‘varieties’, which may later become species of their own. Today we say that ‘ecological speciation’ occurs when populations adapt to exploit different resources such as habitat or diet, and the resulting changes cause a substantial reduction in their ability to successfully interbreed.

The underlying process of divergent adaptation has received considerable study in recent decades – but much of this work has focused on just a handful of natural model systems, including benthic-vs-limnetic stickleback and red-vs-blue Pundamilia cichlids. These studies have yielded important insights into the early stages of lineage divergence, but crucial questions remain unanswered. In particular, we still don’t know if divergent adaptation is a generally important driver of speciation in nature or how it is influenced by changes in the ecological theatre.

Anderson and Weir present a new tool to address these questions by modelling the evolution of differences in ecomorphology – specifically, continuous traits with known ecological function – between sister species and other lineage pairs. The authors show that divergent adaptation imparts a unique signature on the distribution of trait differences in datasets comprised of many paired lineages. Empiricists can test for this signature by providing a measure of trait differentiation and an estimate of the divergence time for each pair. Users can also test for changes in the strength of divergent selection across continuous variables like latitude and elevation or categorical variables like ‘sympatric’ versus ‘allopatric’. With this new tool, empiricists with a variety of research questions can begin to more generally characterize an adaptive process of long-hypothesized importance in evolutionary ecology. The tool is encoded as the R package diverge, now available on CRAN.


Ecological differentiation between lineages is widely considered to be an important driver of speciation, but support for this hypothesis is mainly derived from the detailed study of a select set of model species pairs. Mounting evidence from non-model taxa, meanwhile, suggests that speciation often occurs with minimal differentiation in ecology or ecomorphology, calling into question the true contribution of divergent adaptation to species richness in nature. To better understand divergent ecological adaptation and its role in speciation generally, researchers require a comparative approach that can distinguish its signature from alternative processes such as drift and parallel selection in datasets containing many species pairs. Here we introduce the first statistical models of divergent adaptation in the continuous traits of paired lineages. In these models, ecomorphological characters diverge as two lineages adapt toward alternative phenotypic optima following their departure from a common ancestor. The absolute distance between optima measures the extent of divergent selection and provides a basis for interpretation. We encode the models in the new R package diverge and extend them to allow the distance between optima to vary across continuous and categorical variables. We test model performance using simulation and demonstrate model application using published datasets of trait divergence in birds and mammals. Our framework provides the first explicit test for signatures of divergent selection in trait divergence datasets, and it will enable empiricists from a wide range of fields to better understand the dynamics of divergent adaptation and its prevalence in nature beyond just our best-studied model systems.