Summary: Recent research challenges the long-held view of the amygdala as merely the brain’s primitive “fear center.” The new study shows the amygdala serves as a sophisticated decision hub that arbitrates between two core learning strategies: action-based learning, which emphasizes motor actions that previously led to success, and stimulus-based learning, which prioritizes features or identities of objects. Under uncertainty, a healthy amygdala evaluates which approach is more reliable and shifts the brain toward the best strategy, revealing a central role in cognitive flexibility and adaptive decision-making.
When outcomes are uncertain, the amygdala helps the brain decide whether to rely on repeated actions or on recognizing and selecting specific stimuli. This mediator function reframes the amygdala as more than an emotion center: it is an adaptive controller that balances exploration and exploitation to improve learning and choice behavior.
Key Facts
- Strategic arbitration: The amygdala mediates between action-based learning (repeating motor responses that worked before) and stimulus-based learning (choosing based on object identity or defining features).
- Role under uncertainty: In ambiguous situations, the amygdala gathers and integrates information to decide which learning strategy is more likely to yield a reward.
- Effects of damage: Amygdala damage impairs this arbitration, causing the brain to default toward rigid, action-based strategies and reducing behavioral flexibility.
- Implications for phobias: Phobias often involve a strong bias toward stimulus-based learning. Encouraging action-based exploration can help override rigid fear responses.
- Complex connectivity: Far from being “primitive,” the amygdala is extensively connected to executive networks and plays a nuanced role in guiding learning and choices.
Source: Dartmouth College
A Dartmouth research team published findings in Nature Communications showing that the amygdala’s role extends beyond simple fear processing. Instead, it functions as a dynamic arbiter that weighs competing learning models and steers behavior toward the model most likely to succeed.
“Although the amygdala has been studied primarily for fear learning, its broad connections to many brain regions suggested it must support additional functions,” says Jae Hyung Woo, the study’s first author and a PhD candidate in psychological and brain sciences. The research explores how those additional functions emerge especially when outcomes are uncertain.
The authors illustrate the difference between learning modes with the example of using an unfamiliar coffee machine. An action-based learner repeats a motor sequence that worked on a similar machine, while a stimulus-based learner focuses on a defining cue—such as a blinking light—and selects accordingly. Both strategies can run in parallel, and the brain needs a mechanism to choose between them.
“The key distinction is whether learning links outcome to a motor action or to the identity of a stimulus,” explains Alireza Soltani, senior author and associate professor of psychological and brain sciences at Dartmouth. Action-based learning ties rewards to specific movements, whereas stimulus-based learning evaluates stimuli directly and can be more flexible in some contexts.
To investigate arbitration between these systems, the team used computational reinforcement-learning models to track how the brain weights each strategy when the correct model is unclear. Behavior in control animals and animals with targeted lesions showed that the amygdala initially promotes exploration between systems, then favors the system whose predictions prove more reliable.
When the amygdala was damaged, arbitration became less systematic. The brain struggled to update its assessment of which learning system was most useful and tended to favor action-based responses from the start. That default produced more repetitive, less flexible behavior and reduced the capacity to shift strategies as new evidence accumulated.
“A healthy amygdala encourages exploration among alternative models, which can lead you to try choices you otherwise wouldn’t and learn from them,” Soltani says. This framework helps reconcile past findings that sometimes linked amygdala damage to impaired stimulus learning and other times to improved performance: the amygdala’s role depends on how it balances competing models over time.
Evolutionarily, the amygdala remains important for threat detection and rapid responses. But as other brain systems developed, the amygdala’s connectivity likely enabled a broader role in coordinating learning and decision processes across regions.
The study also offers practical implications for treating anxiety and phobias. People who respond to feared stimuli by rigid stimulus-based avoidance may benefit from interventions that shift attention toward stepwise actions. For example, rather than attempting to reinterpret a spider’s threat value, practicing a sequence of small, goal-directed actions—like gently covering the spider with a cup and moving away—can engage action-based learning and reduce the stimulus-locked fear response. By favoring action-based exploration, the amygdala can help break strong negative associations and restore flexibility.
The researchers are now examining neural recordings from the prefrontal cortex to clarify how neurons interact during arbitration and are collaborating on animal studies to map pathways between the amygdala and prefrontal regions.
The study was led by Jae Hyung Woo and Alireza Soltani in collaboration with colleagues at the National Institute of Mental Health and others.
Key Questions Answered:
A: The amygdala still processes fear, but that is only one of its functions. Think of it as a central manager that handles fear along with strategic mediation between learning systems. It helps determine how the brain should learn from environmental cues and actions.
A: Phobias often reflect a stimulus-based bias—associating a particular object with danger. The findings suggest that shifting toward action-based, exploratory steps can reduce that rigid link and promote flexible behavior, making it easier to overcome fear.
A: Without intact amygdala-mediated arbitration, the brain tends to default to repetitive actions instead of choosing the most appropriate learning model. This leads to less adaptive, more rigid behavior.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The referenced journal paper was reviewed in full.
- Additional context was added by editorial staff.
About this neuroscience research news
Author: Morgan Kelly
Source: Dartmouth College
Contact: Morgan Kelly – Dartmouth College
Image: The image is credited to Neuroscience News
Original Research: Open access. “Contribution of amygdala to dynamic model arbitration under uncertainty” by Jae Hyung Woo, Vincent D. Costa, Craig A. Taswell, Kathryn M. Rothenhoefer, Bruno B. Averbeck & Alireza Soltani. Nature Communications. DOI: 10.1038/s41467-025-66745-1
Abstract
Contribution of amygdala to dynamic model arbitration under uncertainty
Uncertainty about rewards forces the brain to run multiple predictive models in parallel. Outcomes may depend on specific stimuli or on particular actions, corresponding to stimulus- and action-based learning. How the brain chooses between these models remains unclear.
The authors combined computational approaches to measure concurrent learning while male monkeys performed tasks with varying uncertainty about environmental models. Comparing behavior in control monkeys versus monkeys with bilateral lesions to the amygdala or ventral striatum revealed a dynamic competition between stimulus-based and action-based learning and identified a distinctive role for the amygdala in arbitration.
The amygdala adjusts the initial balance between learning systems and is essential for updating which model should govern behavior. This arbitration shapes the interaction between learning and choice over time. By contrast, ventral striatum lesions reduced overall stimulus-value signals. These results reconcile prior contradictory findings and offer predictions for future studies of circuit mechanisms supporting flexible learning under uncertainty.