Understanding the nature of interactions between multiple stressors and their effects on communities is critically important because complex interactions are a source of uncertainty, limit our ability to predict outcomes, and may be drivers of widespread ‘ecological surprises’. Generally, ecologists use 3 groups for classifying interactions: additive, synergistic, and antagonistic. Additive (linear) interactions occur when the application of all stressors results in a response equal to the sum of each stressor in isolation. Synergistic interactions occur when the response to stressors in combination is greater than the sum of each stressor alone while antagonistic interactions arise when the response to combined stressors is less than their sum. By definition, these interactions are classified by comparing the individual and combined effects of stressors relative to a designated ‘control’ treatment. Here, we determine how re-designation of the control alters interaction type classification. We then use a recent meta-analysis on ecological interactions to evaluate the fraction of studies that explicitly designated and justified choice of control and how often the control could be reasonably re-designated. We examined the 144 stressors to understand 1) how researchers justified controls for each stressor category and 2) how often controls could be reasonably reclassified.
Results/Conclusions
We first demonstrate that when a non-linear interaction is significant, changing the control treatment can flip synergisms to antagonisms, and vice versa. Crucially, we found that the way the ‘control’ is designated by the researcher is highly variable across studies, and this choice determines how non-linear interactions are classified. In the meta-analysis, few authors (~6%) explicitly justified choice of control. Rather, we found 2 overarching motifs used implicitly or explicitly to justify controls: 1) based on what is found in nature and 2) the conditions where the population or community performed best. We found the majority (>95%) of studies had controls that could be reasonably reclassified, given the diversity of justifications provided for control classification across studies. This generates uncertainty across studies that strongly limits our understanding of non-linear interaction types. We strongly recommend researchers focus on linear versus non-linear interaction types because the designation of controls is so arbitrary across studies. We argue recent claims of synergisms increasing are clouded by inconsistent designation of controls. This topic will only continue to increase in importance as stressors are predicted to increase in the Anthropocene, making resolution of this problem crucial.