1.Plant-soil feedbacks contribute to species invasions, the maintenance of biodiversity, and climate change impacts on terrestrial ecosystems. Despite their far-reaching importance, we lack a general understanding of the ecological and evolutionary determinants of plant-soil feedbacks. 2.We conducted a large-scale plant-soil feedback experiment using 49 co-occurring plant species from southern Ontario, Canada, representing a wide phylogenetic range. We tested whether effects of soil conditioning varies among these species, and whether different focal species respond similarly to the same soil conditioning. Next, we asked whether plant traits and soil feedbacks depend upon phylogenetic similarity, and which plant traits affect plant-soil feedbacks between pairs of plant species. Finally we used our experimental results to test whether soil feedbacks affect co-occurrence of species in the field. 3.We found evidence of both strong positive and negative soil feedbacks between pairs of plant species. Our soil conditioning treatment explained nearly 20% of the variation in focal species performance. 4.Phylogenetic relatedness and phenotypic similarity between plant species were unrelated to the strength of their soil feedback. However, numerous plant traits of the conditioning species influenced the strength of soil feedbacks on focal species, including specific leaf area and total aboveground productivity. Trait differences between species were also predictive of plant-soil feedbacks, though for some pairs of species, increased trait differences were associated with positive plant-soil feedbacks and for others trait differences were associated with negative plant-soil feedbacks. 5.Plant species co-occurrence in the field was related to their experimentally determined soil feedbacks but only for particular plant species. 6.Synthesis. Our results illustrate how evolutionary history and phenotypic variation shape plant-soil feedbacks and highlight the need for trait-based studies. Due to the unique evolutionary history of individual traits and the variability in their importance across all possible interacting species, we show that indices of overall phenotypic and phylogenetic relatedness are poor predictors of plant-soil feedbacks at large phylogenetic scales. We conclude that a detailed trait-based approach can be used to predict plant-soil feedbacks, and laboratory measures of soil feedbacks can explain patterns of co-occurrence in nature.