A General Framework for Inferring the Developmental Causes of Modularity of Morphological Variation with Applications to the Craniomandibular Complex in Morphological Variation with Applications to the Craniomandibular Complex in Rodents.

Abstract

Modularity is a principle of construction whereby individual units are internally cohesive and relatively autonomous from other such units. Modularity thus confers a degree of evolutionary autonomy to the sets of traits integrating a module, a feature hypothesized to enhance evolvability by allowing selection to optimize individual parts without interfering with others. Detecting modularity in morphological traits requires analyzing the structure of covariation because traits integrated by development into modules are expected to show stronger mutual covariation. However, unambiguous patterns of modularity are rare. That is because the developmental processes underlying most phenotypic traits share regulatory elements and/or have spatially overlapping effects. Pervasive interactions can produce the appearance of statistical integration among biologically modular traits. Herein, a statistical framework is provided that confronts these limitations on methods for inferring modularity from morphological data. The theoretical basis of this new method states that modules are subsets of dimensions embedded in phenotypic space, an approach that differs from previous ones by not defining modules as anatomical parts but rather as aspects of the variation of these parts. This abstraction allows traits to be integrated into more than one module and also suggests a natural approach for testing a priori hypotheses of modularity by fitting competing hypotheses to observed covariance matrices, searching for the best-supported causal explanations. A comprehensive method is developed and tested using simulated data, then used to address a major outstanding issue in evolutionary biology: whether the developmental processes that structure variation within populations bias the direction of long-term divergence. This hypothesis is tested by fitting multiple developmental models to both intraspecific and interspecific craniomandibular data obtained from a clade of ecologically diverse rodents. Results reveal a remarkable congruence among patterns within and between species, and they also suggest that there are different mechanisms by which modular variation arises within different parts of the skull, i.e., cranium and mandible. That these structures have different dynamics both within and among species suggests that whether intraspecific variation constrains the direction of divergence may depend on mechanisms structuring modularity within populations.