Survival in the world requires the ability to respond appropriately to the things that happen in our environment. We have to be able to go after the things that are rewarding such as food, shelter, or a mate, while avoiding dangerous things like predators or poisons.
All animals have some mechanism that allows them to seek the good and avoid the bad, all with the aim of increasing their ability to survive and pass their genes on to their offspring. Over time, we learn the things that are rewarding and we keep doing them, but importantly, we also learn which things are dangerous and we avoid those things. However, it is evident that some things we’d rather not learn through direct experience because they are too costly and directly threaten our survival. It’s much safer to learn about threats in the environment from someone else’s experience.
In the same way, other animals learn about things like which predators to avoid or what kinds of food might be poisonous through observing the experiences of other animals. This ability to use the experiences of other animals to learn about the world is known as observational learning. For humans, observational learning represents a critical means by which we learn about the world, but we now have lots of data to support that many other animals can utilize this same form of learning to learn about action-outcome relationships, places to approach and avoid, and which objects or agents in the environment are rewarding or aversive.
In this study, our team focused on observational fear learning, the process by which an animal learns about which things in the environment are predictive of aversive or dangerous consequences through observation of the experience of another animal. It is dependent on the detection, processing, and integration of a social signal in order to adaptively change behavior. Despite the experimental evidence detailing the ability of rodents, monkeys, humans, and other animals to learn through observation, surprisingly little is known about its neural basis.
We knew from previous research that two regions, the anterior cingulate cortex (ACC) and the basolateral amygdala are important and necessary for observational learning in both humans as well as rodents. We also knew that these two regions were connected with each other. However, there were many unanswered questions. For instance, we didn’t know what information was actually encoded in these two regions and whether or not the ACC neurons that sent information to the amygdala had a specific function in observational learning. We also wondered if information flow from the ACC to the amygdala was necessary for animals to learn through observation. If so, could these neurons encode social information and be used in other kinds of social behaviors?
To begin answering these questions we designed a paradigm in which one mouse which had previously experienced one foot shock, observed a familiar demonstrator mouse undergo classical cued fear conditioning through a transparent, perforated divider allowing for the observation of auditory, visual, and olfactory information. Each time the light and sound came on, the demonstrator mouse received a footshock while the observer mouse could observe from a safe plastic floor. The next day we placed the observer mice on a shock floor and they had learned to freeze to the cue even though they never had experienced the shock happening with a cue.
The next thing we did was to utilize the power of optogenetics and electrophysiology to record from the brains of observer mice during observational conditioning. We found that in observer mice, ACC neurons that project to the amygdala responded to the cues that predict shock to another mouse and that a higher percentage of the ACC neurons that projected to the amygdala encoded for the cues learned through observation when compared to other ACC neurons. This made us think that perhaps these neurons might be communicating critical socially derived information about the cue to the amygdala. If this was true, then inhibiting the input from this region would affect the amygdala’s ability to encode cue information learned through social information.
We recorded in the amygdala and used optogenetics to inhibit the ACC input to the amygdala during some of the cues. We found that some amygdala neurons, that under normal conditions would respond to the cue, stopped responding to the cue when we took away the ACC input. This demonstrated that at least some amygdala neurons really needed ACC information to respond to the cue. This led us to wonder whether this ACC information was actually necessary for animals to be able to learn through observation. We again utilized our optogenetic toolkit to inhibit ACC input to the amygdala, but this time we did it during all of the cues. We found that observer mice no longer learned through observation. This gave us evidence that the amygdala needed this information from ACC during the cue in order for animals to learn.
Interestingly, when we did the same thing during classical fear conditioning (when animals experienced the shock directly), animals were still able to learn. This suggested that ACC input was only needed when animals were trying to learn from social information.  Could it be that this is because the ACC is important for social information and routing it to the amygdala in order to help shape behavior? If so, then inhibiting the ACC input to the amygdala during social interaction would lead to some impairment. Indeed, we found that animals showed deficits in social behavior when the ACC couldn’t send information to the amygdala.
Our experiments suggest that the ACC is a critical hub for taking social information, making sense of it, and routing it to the amygdala to help shape behaviors that require using social information. At the core of observational learning is the ability to take the experiences of another animal and combine it with external cues to learn about the environment. The ACC encodes this social information and information about the cue and then routes this information to the amygdala so that learning can take place.
In this study, we help to provide the basis for an increasingly advanced understanding of how neural circuits contribute to the generation of social behaviors by elucidating a neural mechanism for a fundamental social behavior: observational fear learning.
Through combining optogenetics, electrophysiology, and rodent behavioral paradigms, we take a fascinating psychological phenomenon and provide the basis for a mechanistic circuit-level understanding of this behavior that is so critical for animal survival. We also show that the brain uses the same circuits for fundamental social behaviors like observational learning that it uses for other behaviors in the social repertoire.
These findings are described in the article entitled Corticoamygdala Transfer of Socially Derived Information Gates Observational Learning, recently published in the journal Cell.
This work was co-led by Stephen A. Allsop, Romy Wichmann, Fergil Mills, and Anthony Burgos-Robles, conducted along with Chia-Jung Chang, Ada C. Felix-Ortiz, Alienor Vienne, Anna Beyeler, Ehsan M. Izadmehr, Gordon Glober, Meghan I. Cum, Johanna Stergiadou, Kavitha K. Anandalingam, Kathryn Farris, Praneeth Namburi, Christopher A. Leppla, Javier C. Weddington, Edward H. Nieh, and Demba Ba, from the Massachusetts Institute of Technology, Anne C. Smith from the University of Arizona, and Emery N. Brown from the Massachusetts Institute of Technology and Harvard Medical School. The senior author of this study was Kay M. Tye, from the Massachusetts Institute of Technology and the Salk Institute.