ALGORITHMIC APPROACHES FOR SPECIES' DELIMITATION IN MULTIDIMENSIONAL MORPHOSPACE (Invited Presentation)

There are many ways of differentiating species from variation in data rather than an investigator's subjective discrimination, but fewer which start from the traits routinely used by stratigraphers and taxonomists. Here, we highlight four key steps (rotation to orthogonal axes, dimension reduction, clustering and diagnostic checks) that utilize robust variance estimation and statistical thresholds to assess the evidence for one or more clusters without a priori classification of individuals to species. We illustrate the approach to resolve taxonomic speculation on the development of the planktonic foraminiferan genus Turborotalia. We use data on 12 morphometric traits on each of 200 individuals in each of 51 samples over ~11 million years. For most of the record, contemporaneous specimens form a single morphological cluster which we interpret as an evolving species showing quasi-continuous, but non-directional, gradual evolutionary change (anagenesis). In the upper Eocene from ~36 to ~34 m. yr ago there are two clusters that persistently occupy two distinct areas of morphospace, from which we infer that speciation (cladogenesis) must have occurred.