1. Our aim was to explore plant-soil feedback in mixed grassland communities and its significance for plant productivity and community composition relative to abiotic factors of soil type and fertility.

2. We carried out a 4-year, field-based mesocosm experiment to determine the relative effects of soil type, historic management intensity and soil conditioning by a wide range of plant species of mesotrophic grassland on the productivity and evenness of subsequent mixed communities.

3. The study consisted of an initial soil conditioning phase, whereby soil from two locations each with two levels of management intensity was conditioned with monocultures of nine grassland species, and a subsequent feedback phase, where mixed communities of the nine species were grown in conditioned soil to determine relative effects of experimental factors on the productivity and evenness of mixed communities and individual plant species performance.

4. In the conditioning phase of the experiment, individual plant species differentially influenced soil microbial communities and nutrient availability. However, these biotic effects were much less important as drivers of soil microbial properties and nutrient availability than were abiotic factors of soil type and fertility.

5. Significant feedback effects of conditioning were detected during the second phase of the study in terms of individual plant growth in mixed communities. These feedback effects were generally independent of soil type or fertility, and were consistently negative in nature. In most cases, individual plant species performed less well in mixed communities planted in soil that had previously supported their own species.

6. Synthesis. These findings suggest that despite soil abiotic factors acting as major drivers of soil microbial communities and nutrient availability, biotic interactions in the form of negative feedback play a significant role in regulating individual plant performance in mixed grassland communities across a range of soil conditions.