Herbivore search behaviour drives associational plant refuge

•Plants can gain refuge from herbivores by association with plant neighbours.•How do refuge-providing neighbours alter foraging behaviour at the patch-level?•Swamp wallabies reduced search time in patches with manipulated plant neighbours.•Palatable seedlings with manipulated neighbours escaped herbivory for longer.•Palatable plants may persist because neighbours influence patch foraging behaviours.

Herbivore foraging decisions leading to consumption of a plant are complex and multi-faceted, shaped both by the plant itself and by its neighbours. Associational plant refuge arises when neighbours reduce focal plant susceptibility to herbivory. The specific foraging behaviours generating refuge patterns have rarely been examined in free-ranging systems, yet these are key to understanding why such refuge works or fails. We aimed to integrate herbivore foraging and associational plant refuge theories by linking foraging decisions directly to browsing outcomes on focal plants and their neighbours. We tested whether obstructive, unpalatable neighbours reduce the number of patch visits and/or interrupt searching, leading to associational refuge of focal plants. We compared visits by and behaviours of free-ranging mammalian browsers, swamp wallabies (Wallabia bicolor), in control and manipulated plant patches using cameras. Patches (7 m2) comprised a central focal plant (palatable native tree seedling, Eucalyptus pilularis) with neighbours of either existing or manipulated vegetation (unpalatable native daisy, Coronidium elatum). Wallabies made fewer visits to control than manipulated patches, but always browsed the focal plant during the first visit to a control patch. In contrast, wallabies often visited manipulated patches multiple times before browsing the focal plant. These ‘futile’ visits were both shorter and involved less searching time than visits when the focal plant was browsed. Focal plants escaped browsing for longer in manipulated than in control patches, and although none had escaped browsing after one year, survival was significantly greater in manipulated patches. We demonstrate that reduced investment in searching during visits to manipulated patches drove the associational plant refuge, but this refuge was eventually surmountable. Understanding the behaviours underpinning refuges allows better prediction of outcomes, and explains why refuge can collapse. By shaping foraging behaviour in patches, neighbouring vegetation can increase the probability that palatable plants persist despite high herbivore pressure.