Revisiting the Palimpsest: Exploring the Interaction of the Variance and Structure of Developmental Determinants of Phenotypic Covariance

The analysis of patterns of trait covariance is of tremendous importance in the study of evolution and development as well as in quantitative genetics. Multivariate quantitative genetics and morphological integration were combined in the 1980s under the latent variable model for integration developed by Wagner and Cheverud. The palimpsest framework builds on this model. Based on a definition of integration as disposition, it distinguishes between the developmental determinants of covariation (integration) and observed phenotypic covariance patterns. Further, the palimpsest distinguishes between the role of the distribution of variance across latent processes that generate covariance versus the developmental architectural features that connect those processes to phenotypic traits. In this paper, we construct a simple simulation to illustrate how these two different aspects of integration relate to observed phenotypic covariance patterns. We show that: 1) changes in the distribution of variance across latent processes can significantly alter both integratedness and covariance structure, 2) increases in the variance of leading processes tend to increase integratedness and the circumstances under which integratedness decreases are more restricted, 3) changes in the relative variances of latent processes can result in discontinuous change covariance structure, 4) changes in which developmental processes are linked to which traits (the process-trait map) can also alter covariance structure and interact with how variance is allocated among processes in complex ways. These findings can serve to inform and ground hypotheses about covariance structures and their interpretation in real biological settings.