In the absence of predation large herbivores build up to high abundance often leading to biodiversity loss and biotic homogenization of their environment. Although these populations are expected to decline due to density-dependent processes caused by resource depletion, they are often maintained at densities which preclude habitat recovery. Estimating the contribution, and nature, of behavioural adjustments, allowed by predation relaxation, to the maintenance of high herbivore abundance is the focus of this project. The main project partner (CEFE) has previously used the unique situation created by the introduction of Black-tailed deer Odocoileus hemionus in Haida Gwaii (aka Queen Charlotte Islands, BC, Canada), originally devoid of large forest herbivores and predators, to show that after 50 years of uncontrolled browsing the vast majority of plants were virtually eliminated from the understory. We build on this knowledge to propose the first study that links the “fear biology” of a key ecosystem driver (deer) to its response to resource and biodiversity loss. We will address this question experimentally and empirically by taking advantage of the remarkable set up we have in this “natural laboratory”. The variety of islands with different browsing histories and the presence of islands where we have reduced deer populations and restored vegetation for the last 12 years provide a unique and quasi experimental context for such a project. We will study vigilance, diet selection, social interactions, movement, habitat use and demography of deer in contrasted situation of risk and resources, using individual-based data collected via direct observation of marked animals and GPS/activity loggers. The project brings together international experts of the methods used in these research fields (CEFE: diet selection, movement and habitat use, demography; CEFS: diet selection, movement and habitat use; EDB: vigilance and foraging; Canadian partners: behaviour, black-tailed deer ecology and manipulation). Because risk and resources are by nature negatively correlated due to deer overbrowsing in risk-free environments, we go beyond the simple comparison of behaviour across risk and resource gradients and will manipulate deer environment using translocation and chronic low-intensity hunting. We will thus be able to assess the behavioural adjustments induced by the relaxation of predation and its potential effect on herbivore performance. Results will be largely evaluated within the context of the ability of predation risk to limit herbivore abundance and thus affect their effects on ecosystems. Thus, to date, the project represents one of the most comprehensive attempts to study the non-consumptive effects of predation in the context of ecosystem functioning and biodiversity conservation.