Watching Aggression Triggers Same Brain Changes as Experiencing It

Researchers at Princeton University have demonstrated that merely observing aggression in mice induces the same shared behavioural and neural changes as directly experiencing it. Published in Nature, the study reveals that both direct experience and passive observation recruit a widespread network of excitatory and inhibitory signals across the brain. This network-level activation predicts a shared adaptive behaviour-control policy when the animals are later placed in a novel aggression context. The findings challenge previous models that suggested aggression observation only activates specific hypothalamic neurons, instead pointing to a network-wide phenomenon involving multiple cell types across an extended aggression-control circuit.

The study, led by Annegret L. Falkner and colleagues, utilized a strategy for large-scale, cell-type-specific recordings across subcortical networks. The team recorded neural activity longitudinally while mice trained through direct aggression experience or observation. They then probed shared differences in behaviour and neural activity in a new aggression context. The research aimed to determine whether observation and experience engage common circuit mechanisms that promote a new shared behavioural policy. By developing a multi-fibre photometry approach, the team could simultaneously monitor up to twelve distinct brain regions in freely moving animals.

The experimental design involved three groups of male mice: an experience group (EXP), an observation group (OBS), and a non-social control group (NON). Mice in the EXP group underwent eight consecutive days of resident-intruder experiments, receiving unconstrained access to three different male intruders. Mice in the OBS group were placed in a cage extender with a clear plexiglass barrier, allowing them to visually and olfactorily observe a paired mouse undergoing the same resident-intruder sessions. The NON group was exposed to an inanimate toy resembling a mouse for the same duration. On day nine, all animals were subjected to a hard fight against a unique, undefeated CD1 male aggressor screened for aggression.

Behavioural analysis revealed that both the EXP and OBS groups exhibited signatures of resilience in the novel context. These included increased initiated attacks and decreased latency to attack, a classic signature of aggression instigation. Both groups also developed a new shared strategy characterized by two interpretable behavioural signatures. First, individuals in both groups spent more time in aggression preparatory postures and similarly transitioned between aggression-preparatory clusters, indicating readiness to fight. Second, both groups spent less time in distant inattentive states where they might be vulnerable to attack. The researchers interpret these changes as a common defensive strategy of being ready to respond.

Neural recordings showed that during aggression observation, the observed behaviour recruited widespread activation across subcortical networks for behavioural generation. This activity mimicked the activity evoked by experienced behaviour. Multiregion modelling revealed shared changes to a core aggression network across both experience and observation, including a widespread reduction in inhibition. The team found that this reduction in inhibition is necessary for adaptive behavioural change. Specifically, they observed a shared motif across the core hypothalamic aggression circuit where inhibition becomes less predictive of neural signals in the ventromedial hypothalamus ventrolateral division (VMHvl).

The study also examined the role of dopaminergic signals in the nucleus accumbens (NAc). The NAc-DA signal, known for encoding action value, vicariously responded to observed actions during naturalistic social behaviour. It encoded differences in the behaviour of the aggressor and the subordinate animal. The response during a simulated attack was substantially reduced when the subordinate was absent. This suggests that the observer does not automatically take the point of view of the aggressor but learns vicariously from both the animal being attacked and the aggressor. This mix of perspectives is consistent with previous work on witness defeat stress, suggesting that vicarious learning emerges from attending to the actions of both participants.

To test the causal role of inhibitory populations, the team used chemogenetic perturbation in a separate cohort. Mice that underwent observation training and received clozapine-N-oxide before the hard fight showed altered behavioural outcomes compared to controls. This provided direct circuit-level evidence for the decreased influence of inhibition in aggression-control circuits that emerges after repeated experience or observation. The results demonstrate that these changes can generalize to novel and uncertain contexts, potentially facilitating adaptive self-defensive strategies when aggression levels in nearby conspecifics are high.

The researchers used advanced machine learning techniques to analyse the behavioural and neural data. They employed SLEAP software for pose tracking and UMAP for unsupervised behaviour embedding and clustering. A supervised multilayer perceptron was trained to classify attack behaviours during the training period, achieving a balanced accuracy of 0.92 on a held-out test set. For neural analysis, a shift-only time-warping approach was used to identify a population-wide single timepoint for observed attack alignment. This method addressed temporal uncertainty arising from variability in the timing of neural responses relative to classifier-defined attack onsets. The team also developed a deep neural network to classify observer behavioural states directly from raw video frames, achieving approximately 86% validation accuracy.

Limitations of the study include the use of only male mice, which may limit the generalizability of the findings to female animals. The study also focused on a specific type of aggression, namely resident-intruder interactions, which may not capture all forms of aggressive behaviour. The team notes that behaviours involving direct social contact show greater change after direct experience than after observation, suggesting limits to what can be changed through observation alone.

The findings have implications for understanding how extreme levels of aggression exposure in humans may lead to long-lasting maladaptive behaviours. The data may shed light on the underlying circuit mechanism by which witnessing aggression can alter an individual’s behavioural policy. The study highlights the importance of considering both direct experience and observation in the development of social behaviour. By demonstrating that experience-like activity during aggression observation can recruit a shared inhibitory plasticity mechanism, the research provides a new framework for understanding the neural basis of social learning and aggression.

The study was approved by the Princeton University Institutional Animal Care and Use Committee and was conducted in accordance with National Institutes of Health standards. All processed data required to reproduce the analyses have been deposited at FigShare and are publicly accessible. Relevant code is available at GitHub. The availability of data and code supports the reproducibility of the findings and allows for further investigation by the scientific community.

The research contributes to the growing body of evidence on embodied social cognition, where observation of aggression activates neural populations associated with aggressive arousal, physiological signalling, and reward monitoring. It moves beyond the idea of mere action reflection to a more complex model of goal emulation. The observer learns a strategy, such as avoiding being attack-vulnerable, that requires a vicarious reflection of outcomes from both participants.

In addition to monitoring dopamine dynamics in the nucleus accumbens, the investigation tracked neural activity in the lateral habenula, a brain region previously linked to aversive learning. Recordings revealed that lateral habenula signaling was not modulated by differences in the specific actions observed during aggressive encounters. Instead, responses in this structure were amplified when the observer had prior familiarity with the observed individual. To isolate this effect, mice were first defeated by an aggressive male in their home cage and subsequently observed four resident-intruder sessions featuring either that familiar aggressor or a novel one. The authors noted that this pattern aligns with bystander observations of familiar conspecifics in distress, indicating that familiarity specifically modulates vicarious punishment signaling.

To simultaneously capture these diverse subcortical dynamics across twelve target sites, the researchers implemented a specialized dual-fluorophore viral approach in male mice aged twelve to eighteen weeks. They delivered an injection cocktail combining Cre-out GCaMP6f and Cre-activated jRCaMP1b across eleven subcortical regions, alongside a dopamine biosensor infused into the nucleus accumbens. Recordings were conducted using a custom multi-fiber implant housing twelve distinct optic fibers lowered stereotaxically to a depth of 5.55 millimeters. During behavioral sessions, light-emitting diodes interleaved three optical excitation wavelengths at forty hertz, delivering 470-nanometer and 560-nanometer light alongside a 415-nanometer calcium-independent channel to correct for motion artifacts.

The team further evaluated how network connections changed over the eight-day training period using an automatic relevance determination regression framework combined with permutation-based variance decomposition. By analyzing eight predictor regions relative to excitatory neurons in the ventrolateral ventromedial hypothalamus, the researchers tracked changes in unique variance using linear mixed-effects models. These models confirmed that unique variance from inhibitory predictors steadily declined across training sessions in both experienced and observing mice, demonstrating that a measurable relaxation of inhibitory predictive influence occurs without requiring direct physical engagement.

Subject of Research: Neural and behavioural mechanisms of aggression observation in mice

Article Title: Aggression experience and observation drive shared behavioural and neural changes

Article References: Iravedra-García, J. M., Guthman, E. M., Tsin, D., Cherepanova, P., Cuturela, L., Ocasio-Arce, E. J., Pillow, J. W., & Falkner, A. L. (2026). Aggression experience and observation drive shared behavioural and neural changes. Nature. https://doi.org/10.1038/s41586-026-11217-9

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11217-9

Keywords: aggression, neural circuits, social learning, mice, hypothalamus, inhibition, behavioural plasticity, fibre photometry, Princeton University, Nature, experience, observation