Female sand tilefish (Malacanthus plumieri) inhabiting a deep channel in the fringing reef at Glover's Atoll, Belize (channel females) spawned planktonic eggs more frequently than those occupying the shallow sand-rubble slopes adjacent to patch coral reefs inside the atoll lagoon (reef-slope females). We tested five non-exclusive hypotheses to explain habitat differences in female spawning frequency. We found no evidence that spawning frequency variation was a consequence of differences in food availability, variation in male fertility, or the intensity of predation on spawned eggs. On reef slopes, barracuda stalked tilefish near their benthic burrows, whereas these piscivores attacked channel tilefish by diving suddenly from higher in the water column. Differences in the hunting behavior of barracuda suggested that the behavior of tilefish females might be influenced by temporal variation in predation risk (risk allocation hypothesis). Consistent with this hypothesis, reef-slope females had a much higher frequency of retreats to burrows in response to barracuda, spent more time burrowing and ventured less far from these refugia in response to a simulated predatory threat. As predicted by the risk allocation hypothesis, reef-slope females also had lower and more variable frequencies of foraging bites, and shorter and more variable travel distances to forage than channel females. Estimated mortality from predation was over nine times higher in channel tilefish. Consistent with the hypothesis that investment in current vs. future reproduction is influenced by rates of adult mortality (mortality risk hypothesis), channel females invested more supplemental energy in egg production whereas reef-slope females invested more in growth. Our results indicate that behavioral and life-history traits of female tilefish show phenotypic plasticity depending upon the nature and intensity of localized predation risk.