The “ecology of fear” posits that predation risk shapes the behaviour of large herbivores, their foraging patterns, their habitat selection and their consequent effect on forest ecology. To test some of these predictions we used the extensive empirical and experimental data on vegetation cover and composition, and on deer anti-predator behaviour, collected at study sites with different histories of hunting and natural predation in the Haida Gwaii archipelago and in nearby areas of coastal British Columbia (Canada). Because these deer make use of intertidal areas for foraging, an habitat hypothetically more exposed to risk, we also analysed variation in marine versus terrestrial stable isotopes in deer bone collagen in populations exposed to different levels of predation risk. In absence of risk, deer had a strong negative effect on the amount of understory vegetation and plant composition. In these populations deer showed a remarkable tolerance to human presence (short flight initiation and travel distances when disturbed), a willingness to consume foreign bait or to investigate baited traps, and a propensity to be active at daytime. In contrast, where deer faced long term hunting, and/or natural predators, understories were denser and more diverse and resembled those of forests never exposed to deer. Severe deer culling in sites initially without risk resulted in a dramatic increase in the amount of understory vegetation although different in composition from the one observed in forests with long-term presence of predators and hunting, or in forests that never had deer. Deer born after culling exhibited longer flight initiation distances and travel distances when fleeing, a reluctance to consume foreign bait or to investigate baited traps, and increased night-time foraging. The experimental translocation of unwary deer from a population without risk to the island where culls had partially restored the vegetation, showed that their unwary behaviour was not significantly modified in the presence of abundant and higher quality forage. This contrasted with the wary behaviour observed in the local deer born after the culls and suggested that behavioural contrasts were not direcly linked to forage abundance. Finally, deer in populations exposed to risk from predators and/or hunters were less likely to forage in exposed intertidal habitats, although this trend might be to some extent affected by resources in the understory. We interpreted our results as evidence that experience of risk was key in shaping behavioural contrasts between deer populations and their effect on plant cover and diversity, ecological networks, and ecosystem complexity.