The Strategy of Model Building in Ecology, Revisited

This paper analyzes strategies of model building in ecology twenty years after Levins advocated the use of simple models to generate supposedly qualitative and general biological insights. Modeling has five aspects: elevation of biological processes, construction and analysis of the model, and of observations, analysis of correspondence between the model and observations, and model-based action. I distinguish four different roles we can assign to models: schemata, exploratory tools, redescriptions, and representations of generative biological relations. In the analysis of correspondence fidelity of fit and accuracy of predictions are insufficient evidence that a model represents the biological processes that generated the observations. In order to confirm the model a variety of “accessory conditions” need to be established; these are often overlooked and difficult to establish, especially in the case of simple models of naturally variable situations. I reinterpret the strategy of using simple models in community ecology as exploratory. The simplicity may ensure mathematical generality but not ecological generality; the models are only suggestive of — not support for — ecological hypotheses. Generality in ecological theory will require much more particularity.