MICROCOSM EXPERIMENTS HAVE LIMITED RELEVANCE FOR COMMUNITY AND ECOSYSTEM ECOLOGY: SYNTHESIS OF COMMENTS

There is general agreement that experiments are an essential tool in ecology, as in all sciences. However, there is much less agreement on the types and designs of experiments that are most useful for predicting the impacts of (and the solutions to) the complex, largescale environmental problems that increasingly confront society. The continuing discussion of the appropriate role of aquatic microcosms exemplifies this disagreement, but essentially the same arguments are being waged over the study of biodiversity, ecosystem response to climate change, and probably in other subdisciplines of ecology, as well. It is axiomatic that every experiment should be designed to test a specific hypothesis by evaluating one or more of its predictions. Since a hypothesis is simply a conceptual (or mathematical) model of how something works, experimentation and the development of predictive models are inextricably linked. Every experiment can be interpreted in terms of how it contributes to the two primary components of modeling: model development (model structure and parameterization) and model testing (verification of model design and performance, and validation of model predictions for specific conditions against observations made under those conditions).