Meeting Abstract

P1.138  Jan. 4  New methods for measuring convergence with quantitative data and a comparison of convergence in real and simulated data sets STAYTON, C.T.; Bucknell University tstayton@bucknell.edu

Convergence is a central concept in evolutionary biology, and is often taken as evidence for adaptation or constraint, yet convergent evolution itself remains a poorly defined phenomenon. Existing metrics for measuring the amount of homoplasy in qualitative and quantitative data sets often fail to discriminate between convergence, parallelisms, and reversals, even under simple, pattern-based definitions of these phenomena. This is unfortunate, as convergence, parallelism, or reversal each have different implications for the evolutionary history of species and clades. Moreover, tools for measuring these phenomena in quantitative data sets are not commonly available. Here, I study the phenomenon of geometric convergence, in which two relatively distantly related taxa are closer to one another in multidimensional trait space than each is to their sister taxon or other close relatives. I develop three new metrics to evaluate convergence in a given phylogenetic tree, and investigate the expected amounts of convergence in randomly simulated data sets. Finally, I compare expected levels of convergence to those actually observed in a few empirical data sets. Large amounts of convergence (sometimes approaching the theoretical maximum) are expected even in randomly simulated data sets, especially in trees with high tree balance and numbers of taxa, and low numbers of traits. However, real data sets can show significantly higher amounts of convergence than expected, especially if the data sets include species with non-overlapping geographic ranges. Caution should be used when interpreting instances of convergence that are discovered in an investigation, especially in studies that use few measured variables, as some amount of convergence is expected in all data sets.