Résumé
Organisms were understood by Georges Cuvier as functionally integrated units. This statement made possible his functional and paleobiological predictions. After Cuvier, the theory of evolution has shown both that natural selection is a major process to understand the correlations between the features of organisms, and that organisms are not perfect machineries but mirror also phylogenetic constraints. The correlations between the features of organisms are hierarchical (at individual, population and interspecific levels), have multiple causes, and constitute an assessment of morphological integration. The need to understand the evolution of the patterns of integration at a macroevolutionary scale is not new. However newly developed methods allow the evolution of both isolated characters and integrative patterns to be estimated in a quantitative way, taking into account phylogeny. Indeed, current phylogenetic comparative methods have allowed to depict and infer relationships between characters shared by species, which cannot be seen as independent observations to infer (paleo)biological predictions. The success of phylogenetic comparative methods relies on the partitioning of the character variance as resulting of phylogenetic and other effects (e.g. ecology, non-additive genetic effects). These methods should allow in turn to infer the processes underlying the patterns. Phylogenetic comparative methods have been mostly used to study the evolution of one character independently of the other characteristics of organisms, leaving out the problem of correlation among characters. Several perspectives are given showing that these methods may allow however possible applications in the study of the evolution of morphological integration, and its corollary: modularity. In this paper I emphasise perspectives on the evolution of developmental integration using morphological and paleontological data.