Predators can induce behavioral responses in prey that lead to significant community level impacts. Although prey state and density may mediate such interactions, few studies examine these factors simultaneously. This omission is surprising given that prey state and density often co-vary across habitats (e.g. urchin barrens are dominated by high densities of starving urchins). Also, most studies manipulate prey state rather than compare anti-predator responses of prey collected from habitats that vary naturally in prey state. We used mesocosms to compare the responses of purple sea urchins (Strongylocentrotus purpuratus) from either kelp forests or barrens to waterborne cues from lobster predators (Panulirus interruptus). When tested at ecologically relevant densities (low or high density, for kelp forests and barrens, respectively), urchins from forests, but not barrens, reduced kelp grazing by 72.7% when exposed to lobster cues. Isolated barren urchins starved for up to 64 days strongly reduced grazing in the presence of predation risk. Interestingly, barren urchins failed to respond to lobster cues only in the presence of waterborne cues from high densities of conspecifics. Such behaviors should reduce the non-consumptive effects of lobsters on urchins in barrens, which could impair kelp recovery in deforested habitats, and may generally contribute to the persistence of alternative stable states.