Large herbivorous mammals play a crucial role to the function and structure of terrestrial ecosystems, mainly through herbivory and deposition of dung and urine, which influence plant germination and growth, composition and diversity of plant communities, nutrient cycling and translocation, carbon storage and the frequency of disturbances such as fire. So far, however, the role of large herbivorous mammals in tropical forests has been poorly studied. According to the Optimal Foraging Theory, selection of food resources results from the balance between energy intake and costs to search, capture and ingest the food. At the landscape scale, a process that can generate patches with different nutritional quality for herbivores is vegetation succession. Early successional stages should be preferred by large herbivorous mammals because they are dominated by plant with low height and fast growth, with leaves that are tender, present high nitrogen content and few secondary compounds, and are more palatable. However, for herbivore browsers, which include fruits in their diet, the availability of these items that are more nutritious than foliage should also affect foraging.Here, we investigate how foraging intensity and dung deposition by Tapirus terrestris vary among successional stages, aiming at contributing to the understanding of the role of the largest South American terrestrial herbivore to forest regeneration and nutrient translocation.We investigated if T. terrestris forages more intensively in earlier successional stages, if this preference is stronger when there is no availability of a highly nutritious food resource, the fruits of cambuci (Campomanesia phaea), and if it deposits larger amounts of dung where it forages more intensively, not translocating nutrients. In a 20,000 ha landscape of continuous Atlantic Forest, we allocated 12 sampling sites in patches with different proportion of vegetation in early successional stages. At each site, samplings were conducted in four 160 m long perpendicular transects during six sessions between March and August 2011. Tapir were registered by camera traps and tracks and dung located and cleared at the beginning and end of each session. We also quantified the proportion of transects occupied by early successional stages (pioneer, initial, and mid stage), the number of gaps, and the number of points in transects where the density of the understory , of herbaceous vegetation and of trees with DBH up to 5cm was considered high or very high. The fruiting of cambuci was registered throughout the study period, and was recorded in half of the sampling sessions. Using a model selection approach, we compared a set of candidate models for three dependent variables: (a) the parameter abundance (λ, interpreted as foraging intensity along sampling sessions) in abundance models (which assume no temporal variations in abundance), considering only