Testing for trait lability at macroevolutionary scales: is life history evolution labile in grasses?

Abstract Trait lability has been defined in a myriad of ways in the macroevolutionary literature, yet there is no clear consensus what exactly is meant by a labile trait or how best to test for it. Here, we compare four frequently used approaches, using life history evolution in grasses as an example: i) phylogenetic signal, ii) ancestral state reconstruction with the hidden rates model, iii) ancestral state reconstruction with the threshold model and iv) simulations. We asked whether inferences of trait lability are consistent among approaches, and whether life history continues a labile trait in grasses. We found that inferences varied across the four approaches, with no two methods consistently yielding the same result. We advocate the use of simulations for comparing observed transition rates to expectations under stable and labile conditions. We infer numerous origins and losses of annuality in grasses, with annuals being unevenly distributed among clades (0-30%), concentrated in (sub)tropical clades comprising C 3 and C 4 species. Surprisingly, life history is not labile throughout grasses, but stable in the C 3 , temperate Pooideae. Our findings highlight the importance of objective tests of trait lability, based on predefined criteria, with implications for the field of macroevolution and beyond (e.g. crop breeding).