P-values to determine whether natural selection operates in nature: seizing their potential

After 34 years of the development of Lande and Arnold’s (1983) model to quantify natural selection on multivariate quantitative phenotypic traits, researchers still heavily rely on p-values to determine whether natural selection operates or not on biological populations. I discuss some issues that may arise from using individual p-values obtained from selection gradients, instead of computing a single p -value to address an overall pattern of selection. Selection gradients, standard errors and an overall measure of these effect sizes should be reported when quantifying the regime and strength of natural selection.