The use of fences in conservation can be controversial, as artificial barriers constrain natural behavior and ecological dynamics. However, in the case of large predators inhabiting protected areas within a hostile human‐dominated landscape, predators may remain at low densities if they face high mortality upon leaving the reserve. In turn, this may compromise the potential for density‐dependent effects such as top‐down regulation of prey species abundance. We simulate the hypothetical reintroduction of gray wolves (Canis lupus) to reserves in their former range (Scottish Highlands), with the objectives of identifying parameters that allow a viable wolf population and the potential for direct top‐down forcing of red deer (Cervus elaphus) densities. We examine the extent to which the number of dispersing wolves leaving the protected area influences whether these objectives are achieved. Our simulations confirm that source‐sink population dynamics can result in a self‐perpetuating wolf population, but one that never achieves densities needed for strong top‐down forcing. When wolf density is weakly controlled by intraspecific competition, strong top‐down forcing occurs when 20% of dispersing wolves or less leave the population. When 20–35% of dispersing wolves leave, the strength of top‐down forcing is highly variable. The wolf population remained viable when 35–60% of dispersing wolves left, but then did not exert strong top‐down forcing. Wolves were vulnerable to extinction at greater than 60% disperser loss. Despite their negative connotations, fences (including semi‐permeable ones) could increase the potential for interspecific density‐dependent processes in some cases, thereby facilitating trophic rewilding.