Question: Are competitive hierarchies, which are typically based on the results of pair-wise competition experiments, sensitive to the level of species interaction in the underlying competition experiments?

Location: Controlled greenhouse study using vegetation typical of old-fields in East Tennessee, USA.

Methods: We extend traditional competitive effect/response methods to incorporate data from competition experiments featuring any level of species interaction (i. e., 2,3,.,., n species interacting simultaneously) and develop an ordinal technique that makes hierarchies more robust to variation in the numerical values of relative yield. We apply these methodological techniques to empirical data from a greenhouse experiment wherein four old-field plant species were grown in pair-wise and tri-wise combination. We also demonstrate how resampling can be used to determine the variability of data and its consequences for development of competitive hierarchies.

Results: Different hierarchies were produced when we used different evaluation methods, different levels of species interaction, and different levels of replication. More acute resampling distributions and wider ranges of target/neighbor scores revealed that higher levels of species interaction lead to more distinct hierarchies.

Conclusions: Hierarchies developed from interactions among subsets of species may inadequately characterize relationships among the full community because of indirect or higher-order interactions within multi-species assemblages. Different evaluation methods can yield different hierarchies, and resampling is an effective tool to determine the sensitivity of resultant hierarchies to the level of replication. In sum, our new methodology can be used to control uncertainty in poorly-replicated experiments.

Nomenclature: Terrell (1967); Wofford & Kral (1993).