The structure of the morphospace as representation of a biological model

Abstract

Although multivariate approaches were developed since the 1960s, they became largely employed in morphometrics only in the last decades. The attention has been shifted more and more from single traits, simple characters, and the study of variance, to multiple traits, complex characters, and the analysis of the covariance. Multivariate techniques (mostly Principal Compo-nent Analysis) provide low-dimensional spaces as linear approximations representing systems of relationships, which “form” is the result of actual morphogenetic processes. The genetic program, the structural spatial relationships between tissues, and the functional processes associated with the anatomical components, generate specific networks of correlations between morphological elements. The resulting multivariate space is moulded according to such system of forces; this characterises its size (degree of variations) and shape (combinations of the variations). The geometry of morphospace is therefore the representation of the biological model underlying the observed variability. Its structure and properties should be studied and investigated both in terms of actual organisation (morphogenesis) and perturbation (experimental settings). Such perspec-tive could be a useful starting point on which topics like morphological integration, modularity, and evolvability, can be fruitfully developed.