We thank Shackman et al. for their correspondence on our Review article (Tseng, Y. T., Schaefke, B., Wei, P. & Wang, L. Defensive responses: behaviour, the brain and the body. Nat. Rev. Neurosci. 24, 655–671; 2023)1, which raises interesting issues about the functional relationship between the human central nucleus of the amygdala (CeA) and the bed nucleus of the stria terminalis (BNST) (Shackman, A. J., Grogans, S. E. & Fox, A. S. Fear, anxiety and the functional architecture of the human central extended amygdala. Nat. Rev. Neurosci. https://doi.org/10.1038/s41583-024-00832-y (2024))2. Here, we briefly address claim 2 (“the CeA alone triggers responses to more certain-and-immediate dangers”2) drawn by Shackman and colleagues from our Review and discuss the proposed role of the BNST in responses to higher-imminence threat, which we did not present in our piece.

In our Review, we focused on the established neurocircuits activated within the predatory imminence continuum, as confirmed in rodent studies, and their interplay with peripheral systems. Causal studies in mice have clearly shown that the CeA is required for responses to high-imminence threat in the circa-strike phase under ethologically relevant conditions1,3. However, corresponding research on the BNST is limited, with no study clearly establishing a causal role of the BNST in regulating the response to high-imminence predator (or equivalent) threat. Shackman et al. cite a study which shows that optogenetic stimulation of serotonin-responsive neurons in the BNST during fear acquisition strengthens responses to Pavlovian threat cues during fear recall4. This is an exciting finding; however, it does not show BNST activation under ethologically relevant conditions in direct response to a high-imminence threat. Our Review consequently focused predominantly on causal studies, which have well characterized the BNST’s role in regulating rodent responses to lower-imminence threat.

Shackman and colleagues have shown that the BNST and CeA are engaged by anticipation of both temporally certain and temporally uncertain threat in a human fMRI study5. It is, however, debatable whether the anticipatory response to ‘certain’ and ‘uncertain’ threat in this experimental context would correspond to ‘high-imminence’ and ‘low-imminence’ predator threat, respectively. Both measures concern anticipation of unpleasant, but arguably not ecologically realistic and certainly non-lethal, stimuli, and in our view could therefore be mapped to the pre-encounter (low-imminence) phase within the predator imminence continuum.

In addition, although we value the notion that the CeA and BNST are tightly interconnected6,7, their functional heterogeneity cannot be easily dissected with the temporal resolution offered by human fMRI7. For example, the statistically indistinguishable BNST and CeA responses in Shackman and colleagues’ human fMRI study5 and co-varied metabolisms of these brain areas during sustained exposure to intruder threat in a monkey study8 from their group were based on an oversimplified ‘average response’ over space and time. As a result, temporal dynamics of the BNST and CeA responses as well as their multivoxel patterns of activation were missing, obscuring potential differences in spatial activation patterns or time courses, as shown by Buff et al.9 A very useful approach to understand their functions will be an unbiased estimate of their causal influences, for example, by stimulation experiments with high spatiotemporal precision.

Human fMRI studies have limitations in temporal and spatial resolution, and mouse studies present challenges in generalizing findings to humans10. As Shackman et al correctly point out, there is a need for an increased investment in coordinated cross-species studies in neuroscientific research, and interdisciplinary collaboration, including computer-assisted automated and unbiased behavioural analysis, to better understand brain function and disease mechanisms.