The 81st Annual Meeting of the American Association of Physical Anthropologists (2012)

Are all humans created equal? A multivariate approach to skeletal asymmetry


1Department of Biological Sciences, University of Notre Dame, 2Department of Applied Forensic Sciences, Mercyhurst College

Saturday 4:30-4:45, Galleria North Add to calendar

Traditionally, skeletal asymmetry has been analyzed through univariate analyses, specifically a ratio that converts measures of asymmetry into percentages. This method of data adjustment has been applied to fluctuating and directional asymmetries so as to remove the influence of body size and facilitate comparison of asymmetries in dimensions of a different size. Exploration of this model reveals that it violates the fundamental and often unrecognized assumption of ratios, namely that the variable of interest must be isometric with body size. Moreover, it is limited to assessing one parameter at a time.

This study measured the humerus, radius, femur, and tibia of 119 adult humans and applied the univariate ratio and novel multivariate methods. An examination of the ratio model revealed that it does not fulfill its intended goal of correcting for body size, as the numerator and denominator are independent. A multivariate methodology using Principal Components Analysis and Euclidean distances was developed with the benefit of quantifying size and shape asymmetries, weighing the contributions of each, and looking at overall asymmetry in one analysis rather than variable by variable. This also enabled the direct analysis of both fluctuating and directional asymmetries. The independence of the principal components and separation of asymmetries dependent and independent of condition makes it appropriate for testing for fluctuating asymmetry, size-related directional asymmetry, and directional asymmetry due to robusticity. This approach reflects meaningful differences in overall asymmetry and the individual types of asymmetry while simplifying the analysis and interpretation and reducing the total number of analyses.

comments powered by Disqus