Ecosystems are archetypal complex systems: perturbations to one component—through pollution, fertilization, targeted culling, or other interventions—can propagate through the network and generate distant unforeseen consequences. Despite this, ecological analyses, like most classical approaches to causation, typically represent such systems as static networks of independent pairwise interactions, whether trophic (food webs) or non-trophic (mutualistic networks). This representation reflects equilibrium reasoning and its local linearization, as in the community matrix) under which causal interactions are treated as independent and pairwise. The talk examines that assumption in light of nonlinear dynamical theory and ecological time-series data.