Aim: Different species assemblages of annual killifish possess replicated body size distributions yet have unique sets of species in each area of endemism. Here, we use models of trait evolution and historical biogeography to discover how size variation originated and has been restructured.

Location: South America.

Taxon: Austrolebias (Cyprinodontiformes).

Methods: We sampled 63 individuals from 26 Austrolebias species. Using phylogenetic trees (BEAST2), data on environmental variables at sampling locations and size data we compare different models for trait evolution (SURFACE, l1OU) of body size and niche traits. We model the historical biogeography of the areas of endemism (BioGeoBEARS) and use both analyses in combination to reconstruct the history of four species assemblages.

Results: We present new phylogenetic trees for Austrolebias and use them to show that large size principally arose within a single area driven by a shifted selection optimum for a subset of the species in that area. We suggest that ecological interactions triggered size divergence and that this large-bodied lineage subsequently spread to two other areas. A second assemblage may have been shaped by adaptation to a new environment without an associated increase in size divergence. A third assemblage, which has the smallest size range and the most recent origin, is phylogenetically clustered and we found no evidence of environmental filtering.

Main conclusions: Assemblage similarity in Austrolebias is the result of contrasting ecological, evolutionary and historical processes. Modelling trait evolution together with historical biogeography can help to disentangle the complex histories of multispecies assemblages. This approach provides context to commonly used tests investigating the role of ecological processes from phylogenetic data and generates new testable hypotheses on the processes that generated trait diversity and assemblage similarity.