Flip it and reverse it: Reasonable changes in designated controls can flip synergisms to antagonisms.

Ecological systems are subjected to multiple stressors that can interact in complex ways resulting in "ecological surprises". We examine the pivotal role of 'control' assignment in the categorization of stressors into five classes: additive, +synergistic, -synergistic, +antagonistic, and -antagonistic. We demonstrate if an alternate treatment can reasonably be considered the experimental control, nonlinear interaction classifications change, both in sign (+/-) and in direction (synergistic/antagonistic). Further, switching of interaction classifications is not predictable as changing control can result in multiple possible alternate nonlinear classifications. To explore the magnitude of this problem, we evaluate publications gathered for a recent meta-analysis to 1) explore rationales for choice of controls and 2) quantify how frequently it is reasonable to reassign the control. We found controls were designated with a variety of implicit and explicit justifications, with two overall rationales: 1) controls based on 'natural' conditions (historic, current, or future); 2) controls based on direction of impact, such that stressors always have negative impacts. We reasoned that control re-assignment was justified if an alternate treatment met one of these rationales. Of the 844 interactions classified in the meta-analysis, we determined >95% could be reassigned. Based on these findings, we recommend a new approach to meta-analyses, where the 'control' is strictly and consistently defined by the authors of the meta-analysis. These controls should be based on their broader question, rather than following the common practice of defaulting to controls assigned by the authors of each study, as we found these rationales vary broadly based on the specific questions of each study. Consistent control designation within the ecological or toxicological framework of each meta-analysis may provide deeper and more consistent insight into the nature of interactive effects between multiple stressors. Gaining this insight is crucial because stressor interactions are certain to increase in the Anthropocene.