Habitat selection is a multi-level, hierarchical process that should be a key component in the balance between food acquisition and predation risk avoidance (food – predation trade-off ). However, to date, studies have not fully elucidated how fine- and broad-scale habitat decisions by individual prey can help balance food versus risk. We studied broad-scale habitat selection by Newfoundland caribou Rangifer tarandus , focusing on trade-off s between predation risk versus access to forage during the calving and post-calving period. We improved traditional measures of habitat availability by incorporating fine-scale movement patterns of caribou into the availability kernel, thus enabling separation of broad and fine scales of selection. Remote sensing and fi eld surveys served to create a spatio-temporal model of forage availability, whereas GPS telemetry locations from 66 black bears Ursus americanus and 59 coyotes Canis latrans provided models of predation risk. We then used GPS telemetry locations from 114 female caribou to assess food – predation trade-offs through the prism of our refi ned model of caribou habitat availability. We noted that migratory movements of caribou were oriented mainly towards habitats with abundant forage and lower risk of bear and (to a lesser extent) coyote encounter. These findings were generally consistent across caribou herds and would not have been evident had we used traditional methods instead of our refined model when estimating habitat availability. We interpret these fi ndings in the context of stereotypical migratory behaviour observed in Newfoundland caribou, which occurs despite the extirpation of wolves Canis lupus nearly a century ago. We submit that caribou are able to balance food acquisition against predation risk using a complex set of factors involving both finer and broader scale selection. Accordingly, our study provides a strong argument for using refined habitat availability estimates when assessing food – predation trade-offs.