Our understanding of natural and sexual selection is too limited to allow predicting evolutionary responses in the wild. This knowledge gap is due in part to lack of temporal and spatial replication, of non-morphological studies, and of statistical power.This project offers to integrate measures of selection at multiple levels (selection acting on phenotypic, additive genetic and genomic variations) for multiple potentially correlated traits (morphology, life-history, sexual ornament and personality) and across space and time. We will use data from a long-term monitoring project of blue tits Cyanistes caeruleus in multiple Mediterranean study sites. First, a phenotypic approach will measure present selection on combinations of traits, and explore spatio-temporal fluctuations of selection. We will especially test the importance of selection during extreme climatic events. Second, a quantitative genetic approach will estimate how fluctuating selection can impact evolutionary potential, and will predict the future evolutionary response to selection, independently of possible confounding environmental covariances. Third, an ecological genomics approach will analyze the genomic regions found to be associated with phenotypic variation in terms of genetic structure and footprints of selection, thereby providing insights into past selection that has shaped phenotypic variation. We will explore the habitat-specific genetic architecture of two fundamental fitness-related traits: timing of breeding and personality.The originality and strength of this project lay in the unique opportunity to develop refined analyses of selection in the wild in a multi-site setting for a vertebrate living at the species’ range-edge. Its main assets are my expertise on the various analytical approaches and the availability of long term phenotypic and genetic data. Its ambition is to create a textbook example to improve our understanding of the dynamics of selection in space and time.