From the formal inception of ecology to the present, ecologists have proposed mathematically explicit theories for comparison and evaluation with empirical data. Perhaps most familiar are organismal-level scaling rules including Kleiber’s rule, the -3/2 self-thinning rule, and the metabolic theory of ecology. Parallel theories, however, have been proposed for all ecological hierarchies from populations to communities to landscapes. This symposium will address the status of theory for several focused ecological perspectives. More importantly, however, the symposium will confront general issues in the evaluation, refinement, and confirmation of ecological theories. Many theoretical frameworks in ecology, including null and neutral models, are attempts to represent particular conditions or assumptions; for instance, “default”, or “no effect” patterns. As a result, investigators have generally quantified the validity of said rules and models by setting them as null hypotheses in frequentist significance tests. This is done because an explicit effect (generally 0) can be set for H0, whereas HA can only define only “some effect” distinct from H0. Frequentist significance tests, however, do not allow empirical confirmation of null hypotheses. Instead, under the conventional severe-falsificationist framework we “reject” or provisionally “fail to reject” H0. This reflects the fact that, given a fixed significance level, the probability of rejecting H0 will increase with increasing sample size. A number of the symposium panelists have pointed out other logical/statistical issues in the generalizations of ecological systems and the evaluation of these frameworks. For instance, community ecology null models often rely on Monte-Carlo randomizations whose algorithmic characteristics must be carefully considered if results are to be generalizable. Extremely large, high dimensional and autocorrelated datasets, common in modern ecology, also require particular considerations for valid inference. What, then, is the best approach for evaluating new theories and refining older theories in ecology going forward? The symposium will be comprised of experts to address just this question. Session presenters include developers of broadly utilized ecological theories and models, and leaders in theoretical statistics and statistical ecology. These include influential researchers in organismal-level scaling rules, population-level mathematical models, community-level null models, and ecosystem-level food web models. Panelists, particularly session statisticians and biostatisticians, will address the potential of new statistical methods and perspectives, including frequentist, information-theoretic, and Bayesian approaches.