TitleVisualizing primate evolution: reification of a statistical model (In Proceedings)
inClassification Society of North America 2005 meeting
Author(s) F. James Rohlf, Nina Amenta, Eric Delson, David F. Wiley, Will Harcourt-Smith, Steven Frost, Alfred L. Rosenberger, Dan A. Alcantara, Lissa Tallman
Year 2005
LocationWashington University in St. Louis
DateJune 8-12
Abstract Evolutionary biologists are often concerned with identifying intermediate shapes in an evolutionary sequence or inferring, in a rigorous way, what a hypothetical ancestor might have looked like. Consider the past 150 years during which time scientists have pondered how the first members of the human lineage might have looked - after departing phylogenetically from an African ape ancestor. We are developing a set of tools to address this matter: how to mathematically constrain and visualize shape transformation based on knowledge from phylogenetics. As a case study we examine the evolution of Old World monkeys. The locations of biological landmarks were identified on laser surface scans (LSS) of the skulls of five extant species of Old World monkey (Family Cercopithecidae). In addition, semi-landmark points were located to capture additional shape information along surfaces between the landmarks. The locations of the semilandmark points were slightly adjusted to minimize difierences in shape resulting from arbitrary difierences in the spacing of the semi-landmarks. Generalized Procrustes analysis was then used to align the configurations of landmarks and to project each specimen into a multivariate space that is tangent to Kendall's shape space. This allows variation among shapes to be analyzed using conventional multivariate methods. A phylogeny was imbedded in this space so that the squared length of the implied trajectory was minimized. Points along this trajectory were then projected back into the physical space of the organism and visualized as 3-dimensional representations of the surfaces of hypothetical skulls which are estimates of ancestral forms along the evolutionary trajectories. The 3-dimensional representations are produced by warping all of the original surface scans so that their landmark points coincide with the landmark configuration corresponding to the point of the trajectory, and then merging the surface scans into a single surface. These methods can also be extended to allow the visualization of shape variation implied by other types of multivariate analysis such as multivariate multiple regression and related methods.