• Schamp, Brandon S.
  • Aarssen, Lonnie W.
  • Piggott, Gillian S.J.
  • Dante, Sneha K.



Studies of community structure and co-occurrence patterns rely on the premise that community data reflect where species successfully grow and which species they grow with. However, plant censuses generally do not distinguish between species with reproductive individuals and those only represented by non-reproductive individuals. We tested whether inclusion of non-reproductive species, which may not reflect success in that location, significantly impacts evaluations of community structure and co-occurrence.


Queen’s University Biological Station, Ontario, Canada, old-field plant communities.


We quantified the impact of non-reproductive species in two plant communities by comparing community structure and co-occurrence patterns when non-reproductive species were included or excluded.


Including non-reproductive species significantly increased plot-level species richness in both communities (54% and 13% increases), altered species evenness in both communities, significantly impacted beta-diversity among plots in one site, and disproportionately impacted assessments of diversity in species-rich plots. Excluding non-reproductive species resulted in reduced negative co-occurrence patterns in both communities, with a substantially larger impact in one community. In that community, the impact of non-reproductive species was even more pronounced when abundance data were used in analysis, and when pair-wise co-occurrence patterns were assessed. Additionally, including non-reproductive species drastically decreased the number of species pairs with perfect negative co-occurrence across sites, indicating that these species can add‘noise’ to co-occurrence patterns. We examined possible explanations for the presence of non-reproductive species. In one community, non-reproductive species were 22 times less abundant (per plot) than reproductive species within plots, although they were not rare overall. Differences in the number of non-reproductive species per plot across our focal communities were not clearly driven by differences in clonality, stress from extreme weather or low N. While these patterns are consistent with the interpretation that non-reproductive species are present due to mass effects, this possibility requires further research.


Including non-reproductive plant species in censuses can significantly impact assessments of community structure and species co-occurrence.The divergent impact of their inclusion on our two communities highlights the possibility that excluding non-reproductive species from surveys may remove noise from community data and clarify theories of plant species co-existence.


Study sites

The two focal communities were located at the Queen’s University Biological Station, Chaffey’s Locks, Ontario, Canada (44°34' N, 76°20' W). Both were old-field communities on abandoned agricultural fields that had not been ploughed or tilled for more than 70 yrs. Community plot censuses were performed by setting up grids of 1 m 9 1 m plots within an 80 m 9 80 m area in Site 1 and a 50 m 9 50 m area in Site 2. Each study field was bordered on three sides by dense forest and on one side by a road, and in each case the study grid was centrally located in the field such that there were no edge effects; in both cases, plot grids were over 100 m from the road. The study communities were chosen to be relatively homogeneous topographically. The two communities were approximately 750 m apart, separated by a road as well as a matrix of mixed deciduous forest. Site 1 contains 61 vascular plant species, the majority of which are perennial (e.g. Poa pratensis, Elymus repens, Asclepias syriaca) including some forage species typical of the region (e.g. Phleum pratense, Vicia cracca, Trifolium pratense). Site 2 is also dominated by herbaceous perennials, and contains 37 plant species, the most common of which are P. pratensis, P. pratense, Cerastium arvense, Potentilla recta, V. cracca and Rumex acetosella.

Data collection

Data collection took place between May and August for both sites; Site 1 data were collected in 2006 and Site 2 in 2009. Fifty focal plots in each site were chosen randomly from the available grid of plots; one plot from Site 1 was excluded from analysis because vegetation was repeatedly disturbed (i.e. flattened) by a deer that chose to sleep there. In both communities, two methods of plot-level census were conducted on a bi-weekly basis. The first plot census method included all species found in each plot, regardless of whether they flowered (Full census), and the second method included only species that flowered that season (Flower census). The Full and Flower censuses represent liberal and conservative estimates of species co-existence respectively. The Flower census should reduce the impact of mass effects on plot-level species composition by excluding all individuals for which evidence of successful reproduction in that plot does not exist