1. Competition, niche differences, and chance all contribute to community assembly, yet the role of reproductive interactions between species is often less appreciated. Closely related plant species that share floral form, phenology, and habitat often interact through pollination. They potentially facilitate pollinator attraction, compete for pollination services, and/or exchange pollen. If reproductive processes are important to co-occurrence, we predicted that fitness costs of heterospecific pollen transfer or pollen limitation should result in lower rates of co-occurrence among outcrossing congeners. In contrast, selfers, which may be less exposed to heterospecific pollen, and/or less negatively affected by it, should co-occur more frequently. 2. Flower size is an excellent proxy for mating system in clovers. Using herbarium records and three independent field datasets, we documented co-occurrence patterns of Trifolium at 1m2 - 1km2 scales in California. Using a randomization procedure to reshuffle matrices of community membership, we generated null hypotheses for the expected composition of large- and small-flowered species in Trifolium communities of different sizes. 3. Across all spatial scales, large-flowered outcrossers were over-represented at sites lacking congeners, but under-represented in communities with multiple congeners. Conversely, small-flowered selfers often occupied sites with multiple other Trifolium species. Patterns for plant height and leaf size, which are weakly or strongly correlated with flower size, did not explain co-occurrence patterns as robustly. Regression and model selection corroborated the null model analyses, indicating that the likelihood of co-occurrence decreased as flower size, and thus reliance on outcrossing, increased. 4. Synthesis. This study suggests that reproductive traits and processes may be significant contributors to community assembly and co-occurrence in flowering plants. 

We gathered Trifolium co-occurrence data from four independent sources: 1) at 11,447 randomly-sampled points from across California, documenting occurrences of all Trifolium species as inferred from georeferenced herbarium records, and 1-kilometer buffers around verified records; 2) in 107 500m² plots sampled from across California, as described in Harrison et al. 2006; 3) in 423 4m² quadrats sampled in 2015 and 2016 at the Bodega Bay Natural Reserve, as described in Siefert et al. 2018; and 4) from 372 1m² quadrats sampled annually from 2000-2017 at the McLaughlin Natural Reserve, as described in Harrison et al. 2015.

All scripts are fully functional with the accompanying .Rdata objects (which were generated as described in the manuscript through 10,000 randomizations/re-shufflings of the observed co-occurrence data).