Is "Sampling Effect" a Problem for Experiments Investigating Biodiversity-Ecosystem Function Relationships?

Increasingly, those studies which are aiming to study the relationship between biodiversity and ecosystem function are utilising experimental designs in which species diversity is varied, with the species composition at each level of diversity being determined randomly from a predetermined species pool. Studies utilising such designs have been criticised on the basis that they are confounded by sampling effect (SE) or selection probability effect, i.e. that the treatments which have the highest diversity have a greater probability of being dominated by the most productive species of the entire species pool; however it has also been claimed that SE is a legitimate mechanism by which diversity effects may express themselves in nature. Firstly I show, using an example of a recently published study claiming to show a diversity effect, how SE can result in the identification of apparent relationships between diversity and ecosystem properties which have little meaning in the real world. I then point out that if we accept SE is a diversity mechanism operating in nature, it is firstly necessary to assume that biological communities are randomly assembled with regard to the ecosystem property being measured; this assumption is inconsistent with conventional concepts about how biological communities are organised. Finally I discuss other experimental approaches which may remove the likelihood of results of biodiversity studies from being confounded by problems associated with SE. None of those studies in which SE may contribute to the observed outcome have successfully shown a result which cannot be ascribed to artifact, and alternative experimental approaches are required in order to better understand how biodiversity loss affects ecosystems in nature.