Abstract
Modularity is an appealing concept in evolutionary and developmental biology to explain complex adaptations and body plans. In a quantitative context, modularity usually is assessed in terms of phenotypic correlations among measured traits (morphological integration). Sets of traits with high mutual correlations are referred to as variational modules and are supposed to reflect a modular genetic and developmental system. I demonstrate that this common assumption does not hold in the presence of overlapping pleiotropic factors and allometric growth processes, and further ignores the spatial configuration of measurements. It is thus not possible to reliably infer genetic or developmental modularity from standing phenotypic variation alone.
In spite of these difficulties it is possible to estimate multivariate patters of covariation in a morphometric data set and to model underlying developmental interactions. I will demonstrate approaches based on partial least squares analysis and Sewall Wright's common factor analysis. By applications to hominoid cranial morphology I will show their significance for evolutionary research.
Philipp Mitteroecker
Dept. of Theoretical Biology, University of Vienna, Austria