The 82nd Annual Meeting of the American Association of Physical Anthropologists (2013)

Exploratory multivariate analysis of shape in commingled fossil assemblages


1Department of Basic Medical Science, University of Missouri - Kansas City, 2Department of Anatomy, Kansas City University of Medicine and Biosciences, 3Department of Anthropology, Texas State University

Friday Afternoon, 200DE Add to calendar

Fossil assemblages are often small, fragmentary, and commingled, posing considerable challenges to the multivariate study of size and shape. In this context, we must find or develop new methods appropriate to the nature of the data. We focus here on exploratory methods, searching for interesting patterns in shape variation. We use a novel method where complete observations (“composites”) are simulated by randomly combining isolated fragments. We scale the data by defining logshapes, which are ln-transformed ratios of a composite’s measurements to their geometric mean. Principal components analysis summarizes logshape variation in relatively fewer dimensions. We use interactive graphics methods, including scatterplot matrices and parallel coordinates plots, to search the PCA scores for patterns. These are linked together and patterns are sought by interactively highlighting specific composites with distinctive colors and symbols.

To illustrate, we study dental shape variation in Homo erectus sensu lato from China, Indonesia, Africa, and Europe. Mesiodistal and buccolingual diameters of permanent teeth were analyzed in 1000 random composite dentitions. The first ten principal components of logshape were explored using the ggobi software package. Most components describe contrasts between the relative sizes of incisors and molars and between different molars. In comparison, there appears to be relatively less variation in the shapes of individual teeth. When the distribution is explored graphically, the distinctiveness of European specimens from Dmanisi is effectively highlighted. We conclude that these methods show promise for detecting patterns in fossil assemblages that could reflect possible variations in sex, time, geography, or taxonomy.

comments powered by Disqus