Prey naïveté, or the failure of prey to recognize non-native predators due to a lack of co-evolutionary history, is thought to underpin the large impact of invasive Indo-Pacific lionfish (Pterois sp.) on coral reef fish populations in the western Atlantic. Most previous studies of lionfish recognition have taken place in experimental tanks that did not mimic natural conditions or used bottle or cage field designs that constrained natural behaviour. To alleviate these issues, we compared the homing patterns of experimentally translocated Caribbean bicolor damselfish (Stegastes partitus) in the presence and absence of standardized models of a lionfish, of an ecologically similar native piscivore (black grouper; Mycteroperca bonaci), and of a native non-piscivore (French grunt, Haemulon flavolineatum) in the field. The native grouper model elicited a strong predator avoidance response: translocated damselfish became unlikely to home when released beyond ~ 2 m from their territory and took longer to do so. In contrast, damselfish facing a lionfish model exhibited similar homing behaviours to those of damselfish in the presence of a non-piscivorous grunt and in the absence of any model. Fish length and translocation distance also influenced homing: damselfish stopped homing when released more than 5.6 m away from their territory and larger individuals crossed wider sand gaps. Overall, our findings are consistent with the idea that bicolor damselfish are naïve to the threat of predation presented by lionfish, but also with the notion that damselfish might be assessing, but deeming to be low, the threat of a stalking predator hunting over open sand. Both mechanisms point to inaccurate risk perception in relation to invasive lionfish. More broadly, we highlight a novel experimental translocation approach to evaluate behavioural responses of native prey species to novel predators under realistic field conditions.