Threat generalization to novel instances is central to adaptive behavior. Most previous work has investigated threat generalization based on the perceptual similarity between past and novel stimuli. Few studies have explored generalization based on abstract, non-perceptual relations despite their importance for cognitive flexibility. In order to measure such rule-based generalization of threat without perceptual similarity, we developed a novel paradigm that prevents perceptual features from gaining predictive value. Our results demonstrate that participants responded according to the correct abstract rule and used it to successfully generalize their anticipatory behavioral threat responses (expectancy ratings, sudomotor nerve activity, and heart rate responses). Our results further show that participants flexibly adapted their responses to an unsignaled mid-session contingency reversal. We interpret our results in the context of other rule-based generalization tasks and argue that variations of our paradigm make possible a wide range of investigations into the conceptual aspects of threat generalization.

对新颖实例的威胁概括对于适应性行为至关重要。以前的大多数工作都基于过去和新颖刺激之间的知觉相似性研究了威胁归纳。很少有研究探索基于抽象的非知觉关系的概括，尽管它们对认知灵活性很重要。为了在没有感知相似性的情况下测量这种基于规则的威胁泛化，我们开发了一种新颖的范式，可防止感知特征获得预测价值。我们的结果表明，参与者根据正确的抽象规则做出了反应，并用它成功地概括了他们的预期行为威胁反应（预期等级，运动神经活动和心率反应）。我们的结果进一步表明，参与者可以灵活地调整自己的反应，以应对信号中期的意外事件冲销。我们在其他基于规则的泛化任务的背景下解释了我们的结果，并认为我们范式的变化使对威胁泛化的概念方面的广泛研究成为可能。