AbstractThe theory of model simplification is presented as a means of increasing model understanding. Simplification is based on a selection of the behavior modes defined by the linearized representation of the model and results in a smaller and more easily understood model. To allow understanding, the variables in the simplified model must be easy to interpret relative to those in the original model. This interpretation is complete in an exact simplification, a concept used to derive measures of the importance of different variables in generating selected behavior modes. These measures are used to select which variables to retain and which to omit in forming the simplified model. Issues in the application to nonlinear models are considered, and software that facilitates model simplification is discussed.