Issue 2 – Science Feature 1

Contributed by Dr. Katie Hoemann


Learning and reversing: How the brain navigates threat in an ever-changing environment

Science spotlight on 2022 Best Dissertation in Affective Science Award winner Dr. Hannah Savage


Once, I spent the better part of a year as a professional pet sitter and dog walker. It started out as the perfect job. Until one day when administering medicine to a sick cat resulted in a pretty deep scratch to the arm. I was fine but shaken. I decided to avoid cats for the next while and focus on bonding with dogs. As luck would have it, I was leaving a house one day when a dog chased after and bit me. It left a bruise through my jeans but didn’t break the skin. Still, after a lifetime of happy-go-lucky petting and nuzzling of dogs: I was nervous. I cautiously went back to spending more time with cats.

In the scientific literature, this narrative arc could be described as an episode of threat learning and threat reversal. I formed an association between one stimulus (the cat) and an aversive experience (getting scratched) — an example of threat learning — which I then had to reassess when I continued to have good experiences with cats — an example of threat reversal. This happened at the same time that my love of dogs was challenged by a frightening experience. Thankfully for me, my response suggests I have an ability to respond flexibly to changing sources of threat and safety. Being able to do this well is associated with adaptive emotional functioning and well-being; in fact, inflexibility in this process might be related to the development and maintenance of anxiety disorders. Yet scientists are still working out exactly how the brain accomplishes this feat and what role subjective and autonomic responses play.

2022 Best Dissertation in Affective Science Award winner Dr. Hannah Savage tackled these questions in a series of fMRI studies using a novel threat-safety reversal task. During an initial baseline phase, participants were presented with a blue and a yellow sphere. During the conditioning (‘learning’) phase, one of these spheres was paired with a burst of white noise. Then, during the reversal phase, the pairing of the sphere color and the white noise was switched. Ratings of valence and anxious arousal were collected at the end of each phase, and skin conductance responses were collected throughout, allowing Dr. Savage to track not only the neural, but also the subjective and autonomic components of learning.

In her first study (Savage et al., 2020a), Dr. Savage found participants’ subjective ratings indicated successful threat and safety reversal learning. In terms of neural responses, threat reversal was associated with activation in regions of the salience network (anterior insular cortex [AIC], rostral dorsal anterior cingulate cortex [dACC]) and safety reversal associated with activation in regions that overlap with the default mode network (DMN; anterior ventromedial prefrontal cortex [vmPFC], posterior midline). In her second study (Savage et al., 2020b), Dr. Savage found that, contrary to expectations, this learning process (and corresponding patterns of neural activation) was not disrupted in people with social anxiety disorder.

In her third study (Savage et al., 2021), Dr. Savage dug deeper, to unpack the brain’s involvement in the subjective and autonomic responses to threat. She found that the brain systems generally thought to represent threat learning (including AIC, dACC, and vmPFC) mostly reflected the subjective experience of being anxiously aroused during this learning process, while threat reversal relied on systems associated with valence processing. In contrast, a different subset of regions was responsible for mediating autonomic (skin conductance) responses.

In other words: how people reported feeling was more strongly and broadly predicted by the neural response to threat than their bodily response. This finding is in line with growing evidence showing that the subjective and physiological components of emotion may not correlate as strongly as has traditionally been assumed (e.g., Siegel et al., 2018). It further suggests that subjective (conscious) experiences may be a better, or more comprehensive, predictor of emotional functioning and well-being than their physiological (unconscious) counterparts – a suggestion with profound implications for understanding and treating mental health problems (Taschereau-Dumouchel et al., 2022).

Ultimately, Dr. Savage’s work shows the strides affective science can make by examining emotional phenomena through multiple lenses. There are a lot more threat- and safety-related contingencies out there than stories conveying the (stretched) truth about cats and dogs. These contingencies have consequences for navigating our everyday, ever-changing environments. But by considering the complex interrelations between brain, body, and mind, we can come to better understand human emotions and their relation to mental health.



Savage, H. S., Davey, C. G., Fullana, M. A., & Harrison, B. J. (2020a). Clarifying the neural substrates of threat and safety reversal learning in humans. NeuroImage, 207, 116427.

Savage, H. S., Davey, C. G., Fullana, M. A., & Harrison, B. J. (2020b). Threat and safety reversal learning in social anxiety disorder – an fMRI study. Journal of Anxiety Disorders, 76, 102321.

Savage, H. S., Davey, C. G., Wager, T. D., Garfinkel, S. N., Moffat, B. A., Glarin, R. K., & Harrison, B. J. (2021). Neural mediators of subjective and autonomic responding during threat learning and regulation. NeuroImage, 245, 118643.

Siegel, E. H., Sands, M. K., Van den Noortgate, W., Condon, P., Chang, Y., Dy, J., Quigley, K. S., & Barrett, L. F. (2018). Emotion fingerprints or emotion populations? A meta-analytic investigation of autonomic features of emotion categories. Psychological Bulletin, 144(4), 343–393.

Taschereau-Dumouchel, V., Michel, M., Lau, H., Hofmann, S. G., & LeDoux, J. E. (2022). Putting the “mental” back in “mental disorders”: A perspective from research on fear and anxiety. Molecular Psychiatry, 27(3), 1322–1330.