Character displacement drives trait divergence in a continental fauna

Abstract

Coexisting (sympatric) pairs of closely related species are often characterized by exaggerated trait differences. This widespread pattern is consistent with adaptation for reduced similarity due to costly interactions (i.e., "character displacement")-a classic hypothesis in evolutionary theory. But it is equally consistent with a community assembly bias in which lineages with greater trait differences are more likely to establish overlapping ranges in the first place (i.e., "species sorting"), as well as with null expectations of trait divergence through time. Few comparative analyses have explicitly modeled these alternatives, and it remains unclear whether trait divergence is a general prerequisite for sympatry or a consequence of interactions between sympatric species. Here, we develop statistical models that allow us to distinguish the signature of these processes based on patterns of trait divergence in closely related lineage pairs. We compare support for each model using a dataset of bill shape differences in 207 pairs of New World terrestrial birds representing 30 avian families. We find that character displacement models are overwhelmingly supported over species sorting and null expectations, indicating that exaggerated bill shape differences in sympatric pairs result from enhanced divergent selection in sympatry. We additionally detect a latitudinal gradient in character displacement, which appears strongest in the tropics. Our analysis implicates costly species interactions as powerful drivers of trait divergence in a major vertebrate fauna. These results help substantiate a long-standing but equivocally supported linchpin of evolutionary theory.

Keywords: character displacement; coexistence; trait evolution.

Fig. 1.

Alternative models of trait divergence. A through C replicate evolutionary walks (n = 15) for an allopatric pair (similar bill shape) and sympatric pair (diverged bill shape) under CD, SS, and AD models of trait divergence. SS and AD models are shown in a Brownian motion context for clear illustration, but we also tested versions of these models in which paired lineages evolved under selection toward either a shared optimum or separate optima. Y-axes denote change in mean beak shape with respect to a pair’s most recent ancestor. ${\theta }_x$ represents the optimal trait value to which lineage x is pulled by selection, and $\psi$ represents the distance between adaptive optima for a pair ($\psi =\left|{\theta }_2-\hspace{0.17em}{\theta }_1\right|$). Under CD (A), allopatric pairs diverge under a given evolutionary process until sympatry begins, when the newly sympatric lineages are forced to adapt to the more distantly spaced optima. Pairs that do not become sympatric continue to evolve under the allopatric regime. Under SS (B), pairs that evolved greater trait differences in allopatry have a higher probability of becoming sympatric upon secondary contact. In the AD model (C), allopatric and sympatric pairs evolve under the same stochastic process, but allopatric pairs are younger, on average, and thus have had less time to accumulate trait differences.

Fig. 2.

Signatures of three alternative processes. Rows display expected trait differences under a given model. Graphs in the first column show changes in trait differences through time, where thick lines are expected values and thin lines are 10th to 90th quantiles. “Allopatric 1” refers to allopatric pairs that are too young to be sympatric. “Allopatric 2” refers to pairs that are old enough to be sympatric but have yet to establish sympatry (in the SS model, they have not established sympatry because they are too phenotypically similar). Graphs in the second and third columns show the expected distributions of trait differences at 4 Myr (i.e., at the first onset of sympatry) and 8 Myr. In rows A through C, evolution occurs under an identical regime of weak divergent selection prior to sympatry. (A) Under CD, a shift toward greater divergent selection in sympatry causes pairs to evolve greater trait differences, while allopatric pairs continue to evolve under the original regime. (B) Under SS, pairs that diverged more prior to contact become sympatric at a higher rate. Differences in the expected disparity between allopatric and sympatric pairs thus take time to evolve in the CD model and are immediate in the SS model. Sympatry and expected trait divergence are unrelated in the AD model (C and D). (D) Evolution occurs under a pure random walk and all pairs have greater expected trait differences (vertical black line) at later ages. If sympatric pairs are older than allopatric pairs in nature, they will have greater average trait differences under this model.

Fig. 3.

Geographic variation in sympatric versus allopatric bill shape divergence. Tropical sympatric pairs are more differentiated in bill shape than their allopatric counterparts at each age class. * denotes P values < 0.05 from phylogenetic two-sample t test; ** denotes P values < 0.01.

Fig. 4.

The latitudinal gradient in CD. ${\psi }_{allo}$ (orange line, main panel) is roughly constant with latitude, while ${\psi }_{sym}$ (blue line, main panel) declines with distance from the equator. The difference $\mathrm{\Delta }\psi ={\psi }_{sym}-\hspace{0.17em}{\psi }_{allo}$ represents the magnitude of CD. Light-colored polygons denote estimates within two log-likelihood units of the maximum likelihood estimate (present but extremely narrow in the case of ${\psi }_{allo}$). Boxes A through C illustrate the expected values (thick lines) and 10th to 90th quantiles (thin lines) of bill shape differentiation through time for allopatric and sympatric pairs in tropical, middle, and temperate latitudes, respectively, where T_sym denotes the time at which sympatric lineages begin to overlap geographically and is arbitrarily assumed to be 4 Myr in this graph for illustrative purposes.