Replication of treatment effects and differences among populations during experimental evolution of sex allocation in an annual plant

Variation among replicates of experimental evolution studies that begin by sampling from a single source population can be attributed to measurement error or to random differences in the environment experienced by measured individuals. On the basis of measurements in a large common garden, Cossard et al. (2021: Current Biology 7, 1-7) reported rapid evolution of enhanced ‘leaky’ sex allocation by females (which produced male flowers) in experimental populations of the dioecious herb Mercurialis annua from which males had been removed. Their study thus demonstrated the rapid dissolution of dioecy via changes in sex allocation in response to mate limitation and strong competition for siring success. But their study also found substantial variation among replicate populations. Here, we replicated their common garden, growing plants from the same populations but under different conditions and assaying the measured variables differently. The effects (significance, magnitude and direction) of selection treatment and generation on the sex expression of females with inconstant sex expression met five definitions for successful replication. Importantly, comparisons between the two studies revealed relatively consistent differences among the replicate populations, suggesting that the among-population variation reported by Cossard et al. cannot be attributed to random noise or local differences in growing conditions in the common garden. Our study represents a rare example of the replication of a common garden study and raises interesting questions about the nature of interpopulation divergence within treatment groups of studies using experimental evolution.