Exploring the Structure of Human Defensive Responses from Judgments of Threat Scenarios

How humans react to threats is a topic of broad theoretical importance, and also relevant for understanding anxiety disorders. Many animal threat reactions exhibit a common structure, a finding supported by human evaluations of written threat scenarios that parallel patterns of rodent defensive behavior to actual threats. Yet the factors that underlie these shared behavioral patterns remain unclear. Dimensional accounts rooted in Darwin's conception of antithesis explain many defensive behaviors. Across species, it is also clear that defensive reactions depend on specific situational factors, a feature long emphasized by psychological appraisal theories. Our study sought to extend prior investigations of human judgments of threat to a broader set of threats, including natural disasters, threats from animals, and psychological (as opposed to physical) threats. Our goal was to test whether dimensional and specific patterns of threat evaluation replicate across different threat classes. 85 healthy adult subjects selected descriptions of defensive behaviors that indicated how they would react to 29 threatening scenarios. Scenarios differed with respect to ten factors, e.g., perceived dangerousness or escapability. Across scenarios, we correlated these factor ratings with the pattern of defensive behaviors subjects endorsed. A decision tree hierarchically organized these correlation patterns to successfully predict each scenario's most common reaction, both for the original sample of subjects and a separate replication group (n = 22). At the top of the decision tree, degree of dangerousness interacted with threat type (physical or psychological) to predict dimensional approach/avoidance behavior. Subordinate nodes represented specific defensive responses evoked by particular contexts. Our ecological approach emphasizes the interplay of situational factors in evoking a broad range of threat reactions. Future studies could test predictions made by our results to help understand pathological threat processing, such as seen in anxiety disorders, and could begin to test underlying neural mechanisms.