Abstract
Species may evolve on a reticulate network due to hybridization or other gene flow rather than on a strictly bifurcating tree, but comparative methods to deal with trait evolution on a network are lacking. We create such a method, which uses a Brownian motion model. Our method seeks to separately or jointly detect a bias in trait value coming from hybridization (β) and a burst of variation at the time of hybridization (vH) associated with the hybridization event, as well as traditional Brownian motion parameters of ancestral state (μ) and rate of evolution (σ2) of Brownian motion, as well as measurement error of the tips (SE). We test the method with extensive simulations. We also apply the model to two empirical examples, cichlid body size and Nicotiana drought tolerance, and find substantial measurement error and a hint that hybrids have greater drought tolerance in the latter case. The new methods are available in CRAN R package BMhyd.
