Comparative Analyses


Pollinator shifts and floral diversification in Iochroma

Differences in floral traits among plant species have often been attributed to adaptation to pollinators. We used phylogenetic least squares  to estimate correlations between pollinator groups (hummingbirds and insects) and floral traits (Smith et al. 2008). This approach allowed us to explore multiple models of trait evolution and to examine how the results are affected by phylogenetic uncertainty.  We found that non-phylogenetic models typically fit the data better than phylogenetic models (Brownian motion, Ornstein-Uhlenbeck), and thus results varied little across trees. Our results indicated that species with high nectar reward and large floral displays are significantly more likely to be pollinated by hummingbirds and less likely to be pollinated by all groups of insects. Corolla length and flower color did not show any consistently significant associations with any pollinator groups. For these two traits, we suggested that alternative evolutionary forces, including phylogenetic inertia and community-level factors, may have contributed to the observed diversity.
flowers


Collaborative Projects

Correlated Evolution

I am involved in several projects aimed at describing patterns of character evolution and detecting correlated changes.  Together with Mark Rausher and Dave Des Marais, I am using published phylogenies of several groups (Ipomoea, Antirrhinum, Iochroma) to identify directional patterns of flower color evolution and to determine if losses of floral pigmentation affect the rate of lineage diversification.  Also, I am examining the relationship between breeding system transitions (outcrossing to selfing, sexual to asexual) and floral evolution in Ipomoea (with Andrew MacDonald) and Oenothera (with Marc Johnson).  Initial results in Ipomoea indicate that the many transitions from outcrossing to selfing have been accompanied by dramatic floral changes, e.g., reductions in corolla size and in anther-stigma distance.  These studies are among the first to use formal comparative methods to test theoretical predictions about how changes in breeding system affect floral evolution on a macroevolutionary scale.

Phylogenetic Mixed Models

Biological studies commonly collect count data, such as number of offspring or days to flowering.  Such data are typically averaged for each species in comparative analyses (independent contrasts, phylogenetic least squares), which assume normally-distributed variables.  Together with Cecile Ane and Stephen Stanhope (Dept. Statistics, UW-Madison), I have been working to develop a phylogenetic mixed model for comparative analysis of count data.  This model estimates the effects of normally-distributed covariates on poisson-distributed count data while compensating for dependence due to phylogeny.   We are applying this method to estimate the effects of environmental variables and plant traits (such as nectar reward and display size) on the number of hummingbird visits to species of Iochroma.