STATISTICAL POWER TO DETECT CHANGES IN RATES OF MORPHOLOGICAL EVOLUTION

Striking patterns in plots of disparity through time have long been used as evidence for the influence of biological factors on dynamics of morphological evolution. Brownian motion and random walk models, in which evolutionary change occurs in random directions at a constant rate, are typically considered null models of morphological evolution. Under such models disparity (variance in morphology) is expected to increase approximately linearly with time. The marked departures from linearity exhibited in paleontological records of disparity through time are sometimes taken as evidence for heterogeneous rates of evolution. This is used to argue for the role of certain biological variables in morphological evolution, such as niche availability and developmental constraints. Previous studies have not examined whether such trajectories fall outside the range of variability expected under Brownian motion. In this study I simulate disparity through time under a model in which the rate of morphological evolution decreases through time and compare it to disparity through time under Brownian motion using several new metrics. Trajectories of disparity through time are highly variable under the Brownian motion model, while those produced under a model of decreasing rates of morphological evolution are not always distinctly nonlinear. Thus the ability to detect a decrease in the rate of morphological evolution is very poor using the tests I propose. Phylogenetic methods for testing for changes in rates of morphological evolution are increasingly used in both paleontological and neontological studies. I suggest that methods using phylogenetic information are preferable when such information is available. However, since the current availability of phylogenetic information for many important fossil groups is poor I also suggest improvements to the approaches I developed and ways of applying them to empirical datasets with the understanding that failure to detect a change in rates of morphological evolution will be difficult to interpret because of low test power.