To set the stage for this symposium, here I provide a brief overview of the philosophical roots of approaches to science ranging from the standard, hypothetico-deductive (H-D) method to information theoretic and Bayesian epistemologies, I identify several key epistemological problems related to the process of how theory is tested in ecology, and I offer my own commentary and potential solutions, including examples from studies of food webs. I begin with review of the so-called “problem of induction,” the fact that inductive reasoning cannot provide certainty of knowledge; necessary truths can be demonstrated only via valid deductive arguments, but inductive reasoning when cast in such a framework is revealed as a fallacious (i.e., the fallacy of “affirming the consequent”). In essence, various scientific epistemologies contrast in the way this problem is treated; from the severe falsificationist H-D prescription which attempts to exclude induction altogether and maintain certainty of knowledge as the goal, to Bayesian approaches that embrace inductive reasoning as part of a process whose goal is continuous reevaluation of “degrees of belief.” Conceptually, the way that any theory is evaluated differs dramatically on these different accounts, but several well known problems confront a strict adherence to H-D method alone.
Results/Conclusions
Evidence from any H-D test is insufficient to determine what belief should be held with respect to an ecological theory. For instance, any attempt to test theory directly via the H-D method depends upon a translation and logical equivalency from theory to hypothesis, from hypothesis to a null, a prediction, and statistical test (commonly via frequentist techniques); however these are all steps involving abstraction; translation is rarely straightforward and, hence, equivalency is far from certain. Moreover, ecological theories are typically not testable via single hypotheses, but involve consideration of bundles of related hypotheses. Indeed, in ecology, theories are not ultimately tested by any one study; rather theory evaluation is something that occurs at the scale of the disciplinary community. Yet, if a “weight of evidence” approach is to be applied at this community scale, the H-D method does not provide the formal basis for strong inference regarding such judgment. Using examples from my own studies of food webs, I outline the application of a contextual and pluralistic approach, whereby different epistemological modes (and associated statistical toolsets) may be applied as the scale of individual studies, whereas a Bayesian technique is used to periodically update assessment of knowledge regarding particular theories.