Vigilance is a key to the early detection of predators, but may be costly if it impairs foraging efficiency. Hence, we would expect vigilance to be suppressed and/or counter-selected in predator-free environments, although this might depend on the environmental drivers influencing perceived predation risk. We studied vigilance in two populations of Sitka black-tailed deer (Odocoileus hemionus sitkensis) on Haida Gwaii (Canada) which have not been exposed to predators since they colonized the study islands approx. 60yr ago. In this context, anti-predator behavior should not have any obvious current benefit. Moreover, its maintenance should be particularly costly in our study populations because these deer have depleted their food resources and, thus, anti-predator behaviors should interfere with time spent searching for scarce resources. We used bait stations equipped with camera traps to assess vigilance under standardized feeding conditions. We expected to observe lower vigilance levels than those observed elsewhere in locations with predators. We investigated how vigilance varied in relation to the amount of bait, the level of visibility, and between day and night. During the day, deer spent, on average, 14% of their time in overt vigilance during foraging bouts, a level similar to, although in the lower range of, values reported at sites where predators are present. Levels of vigilance were lower at night, and decreased with increasing visibility, but not during the day. Deer were less vigilant when bait availability was high, but only when visibility was also high. We discuss why the maintenance of vigilance is here best explained by the ghosts of predators past, and how, at the temporal scale of a few generations, the ecological factors driving vigilance levels might override the absence of significant risk from large predators.