Numerous hypotheses have been proposed to explain the dynamics in abundance of individual species, how species interact, how communities assemble, and how interactions between biotic and abiotic processes shape ecosystem stability. Many if not most of these hypotheses find some degree of support, but often only within relatively narrow spatial and temporal ranges. This is because conditions vary over time and from place to place, and so the strength and extent of processes that were the focus of a given a hypothesis become altered by other forces. Ecologists have confronted variability from two perspectives; conceptual and statistical. Conceptually, spatial and temporal variability are now recognized as being scale dependent and hierarchical. Statistically, there are many models that ecologists readily use that account for the hierarchical and scale-dependence of variability present in many datasets. But linking the two perspectives into a meaningful understanding of what variability means in real systems has been much less successful. For example, it is common to see studies where the fixed effects of a generalized linear mixed model are reported, but very often random effects are completely ignored or, at best, given scant attention. The likelihood of this being a significant problem increases greatly in what are rapidly becoming more common studies that utilize datasets spanning long temporal and/or large spatial scales, or when extreme and often unpredictable events (gray and black swans) occur.