Resolving temporal threat uncertainty by observational learning involves the amygdala, hippocampus and anterior insula

While the neurobiological distinction between temporally predictable (cued) and unpredictable (contextual) threats has been well-characterized in direct learning using the NPU paradigm, it is poorly understood how these processes unfold during observational learning, a key mechanism of human threat-learning. In this study, we developed a novel observational paradigm based on the NPU paradigm (Experiment 1, n=20, male and female) and combined it with fMRI (Experiment 2, n=23, male and female) to investigate how the brain encodes predictable and unpredictable threat cues observed in others. Participants learned predictable (P), unpredictable (U), and no-threat (N) cues by observation and encountered the same threat cues during an expression phase. Threat expectations indicated that participants successfully learned threat contingencies, showing heightened threat expectations for predictable cues and unpredictable contexts. This converged with neural responses in the anterior insula during the expression phase. Reflecting the dynamic process of learning, the amygdala responded to predictable threat cues with a linear decrease across trials. Interestingly, we found that learning to predict threats from others’ pain was accompanied by enhanced neural response within the amygdala, insula and hippocampus, as compared to unpredictable conditions. Our findings suggest that humans learn to resolve temporal uncertainty relying solely on observation. The ability to learn shapes the neural response to an observed threat.