Compiling forty years of guppy research to investigate the factors contributing to (non)parallel evolution.

Examples of parallel evolution have been crucial for our understanding of adaptation via natural selection. However, strong parallelism is not always observed even in seemingly similar environments where natural selection is expected to favour similar phenotypes. Leveraging this variation in parallelism within well-researched study systems can provide insight into the factors that contribute to variation in adaptive responses. Here we analyse the results of 36 studies reporting 446 average trait values in Trinidadian guppies, Poecilia reticulata, from different predation regimes. We examine how the extent of predator-driven phenotypic parallelism is influenced by six factors: sex, trait type, rearing environment, ecological complexity, evolutionary history, and time since colonization. Analyses show that parallel evolution in guppies is highly variable and weak on average, with only 24.7% of the variation among populations being explained by predation regime. Levels of parallelism appeared to be especially weak for colour traits, and parallelism decreased with increasing complexity of evolutionary history (i.e., when estimates of parallelism from populations within a single drainage were compared to estimates of parallelism from populations pooled between two major drainages). Suggestive - but not significant - trends that warrant further research include interactions between the sexes and different trait categories. Quantifying and accounting for these and other sources of variation among evolutionary 'replicates' can be leveraged to better understand the extent to which seemingly similar environments drive parallel and nonparallel aspects of phenotypic divergence.