Zoo animals were tested to see if they perceived the scary nature of Halloween masks, using a procedure that measured the avoidance response latency to take food from a masked human experimenter. Human perception of the masks was also assessed using a rating scale, with results showing that a Bill Clinton mask was rated not scary, while a Vampire mask was rated very scary. Animal results showed that primate latencies correlated significantly with the human ratings, while non-primate latencies did not. Taken together, these results indicate that human perception of scary faces does not depend upon human-specific cultural factors, e.g., belief in the supernatural. Rather, it has a more biological basis, shared specifically with other primates, by which scary faces are perceived as predators or threatening conspecifics.