Abstract

Estimates of the direction, strength and shape of selection in natural populations are increasingly frequent, but we generally have limited knowledge about the ecological processes that underlie them. Here, we present a multi-level analysis of selection on offspring fledging mass over 51 years from a population of great tits Parus major, with the aim of identifying environmental modulators of the mass-fitness relationship within this multi-level framework. We show that the trait-fitness covariance depends on the level at which it is estimated, with the between-year covariance being 3 times as large as the between-brood covariance, which itself is 2.6 times larger than the within-brood covariance. We identify food availability during, and following, reproduction as an environmental factor underlying part of the non-causal covariance between fledging mass and recruitment probability, and explain variation in within-brood selection with the density of conspecifics and the presence/absence of a principle predator of tits (the Eurasian sparrowhawk Accipiter nisus). We show that, at the individual level, sparrowhawks indeed selectively prey on great tit fledglings with low mass, providing compelling evidence for a predator to act as a causal agent of selection. Estimates of directional selection on fledging mass differ by 67% when calculated on the different levels, while estimates of stabilising selection differ by 30%. These analyses therefore provide an example of how correlations between traits, fitness and the environment can influence estimates of selection, pinpoint aspects of the environment that explain covariance between traits and fitness at different levels, and show how partitioning covariance between levels of selection and environmental factors is a promising approach to identify causal agents of selection.