We explore the use of Fisher Information as a basis for an index of sustainability. Sustainability of an ecosystem refers to the robustness of a preferred dynamic regime to human and natural disturbances. Ecosystems under perturbations of varying regularity and intensity can either remain within the current regime or transition into the neighbourhood of a regime with different (viz. less desirable) characteristics. The Fisher Information index we develop is based on the probability of finding the system in a particular state. We apply the index to a 10-compartment food web model with five functional groups: detritus, primary producers, herbivores, carnivores, and an omnivore. Fisher Information is shown to be sensitive to transients in model generated data. Such transients can be indicative of a transition to a new dynamic regime. Early detection of transitions to undesirable regimes may permit management intervention. (c) 2004 Elsevier Ltd. All rights reserved.