Summary: A new RIKEN study reveals how rats can acquire fear through inference rather than only through direct experience. By first pairing a neutral sound with an image and then pairing that image with an aversive event, researchers showed that rats later exhibited fear to the sound alone. The work pinpoints the medial prefrontal cortex (mPFC) as the neural substrate that links neutral sensory cues to negative outcomes, supporting higher-order emotional inference and offering insight into mechanisms relevant to human anxiety and trauma-related disorders.
Researchers Xiaowei Gu and Joshua Johansen at the RIKEN Center for Brain Science used behavioral experiments combined with calcium imaging and optogenetics to map how the frontal cortex coordinates with the amygdala to produce this inferred emotional response. Their findings, published in Nature on May 14, provide the first direct evidence of how the brain encodes internal models that underlie human-like emotional inference.
Key Facts:
- Higher-order emotion: Rats can infer fear for a cue that was never directly paired with an unpleasant event, demonstrating indirect emotional learning.
- mPFC role: The medial prefrontal cortex tags sensory representations during initial cue pairing and later activates those representations during aversive learning, enabling inference.
- Neural pathway: mPFC projections to the amygdala are required to recall inferred fear; blocking this output disrupts expression of inferred, but not directly learned, fear.
- Clinical relevance: The circuit mechanisms uncovered may help explain the neural basis of anxiety and trauma-related conditions in humans and point to new targets for research.
Experimental design and behavioral model
To model inferred emotion, rats first learned a neutral sensory association: a specific noise was repeatedly presented along with a visual image. Later, researchers paired the same image with an aversive stimulus (aversive conditioning). When tested the following day, animals that had experienced both stages froze in response to the noise alone, indicating they had inferred that the noise predicted an unpleasant outcome even though the noise had never been directly paired with the aversive event.
Neural mechanisms: mPFC encoding and tagging
Using calcium imaging to monitor neural activity in the medial prefrontal cortex, the team observed that before aversive conditioning mPFC neurons responded similarly to the image and noise whether or not they had been paired. After the image was paired with the aversive event, the number of neurons responsive to the noise—and neurons co-responsive to both noise and image—increased substantially, but only when the noise and image had been previously paired. This indicates that initial sensory pairing “tags” a population of co-responsive neurons in the mPFC, priming them to participate in later aversive learning.
Optogenetic experiments established causality: suppressing mPFC activity during the aversive pairing stage prevented the formation of the inferred association, while suppressing mPFC output to the amygdala at test blocked the expression of inferred fear without impairing direct memory recall for the image. In other words, the amygdala retains simple, directly formed emotional memories, but mPFC integrity and its projection to the amygdala are essential for encoding and retrieving inferred emotional associations.
Implications for emotion and psychiatric research
These results clarify how internal models of emotion are built and stored in the prefrontal cortex to guide behavior when animals must infer threat from indirect cues. Johansen and Gu emphasize that while the amygdala is critical for storing simple associative fear memories, the mPFC supports higher-order, model-based emotional processes that resemble human inference. Understanding this circuit—how the mPFC encodes inferred associations and how it communicates with the amygdala—could inform studies of pathological anxiety, post-traumatic stress, and other disorders where internally generated fear and anticipation play a major role.
About this fear and neuroscience research news
Author: Adam Phillips
Source: RIKEN
Contact: Adam Phillips – RIKEN
Image: The image is credited to Neuroscience News
Original Research: Closed access. “Prefrontal encoding of an internal model for emotional inference” by Joshua Johansen et al., Nature. DOI: 10.1038/s41586-025-09001-2
Abstract summary
Anticipating unpleasant events is a core function of emotional brain systems. While simple associative learning links sensory cues directly to aversive outcomes, higher-order emotional learning requires internal models that allow inference about stimuli not present during the original experience. This study demonstrates that neurons in the rodent dorsomedial prefrontal cortex (dmPFC) encode a flexible internal model by linking sensory features to aversive events whether the association is direct or inferred. Encoding emerges through recruitment and stabilization of dmPFC cells that support inference. Although dmPFC populations represent salient associations broadly, dmPFC neurons projecting to the amygdala specifically encode and are required to express inferred associations. Together, these findings explain how internal models of emotion in the dmPFC regulate subcortical systems to enable recall of inferred emotional memories, with clear relevance for understanding complex emotional behavior and psychiatric conditions.