April 28, 2020
Written by: Claudia Lopez-Lloreda
You see somebody pouting their lips with the corners of their mouth pulled downwards and their eyes looking down. Immediately, you identify that the person feels sad and quickly ask what is wrong. Our face muscles twist and turn to put our emotions—feelings based on inner experience—on display through our expressions.
In humans, studying the link between facial expression and emotions has been pretty straightforward. Scientists ask people to portray the facial expression that they associate with different emotions. Usually, facial expressions associated with the different emotions are pretty consistent from person to person — a smile means happy, a frown means sad.
However, studying the neural foundations of emotions is a little more complicated. Parsing out which circuits in the brain go with which processes is normally done in animal models. But studying emotions in animals and assessing which neuronal circuits cause each emotion has been nearly impossible due to the difficulty in determining an animal’s emotional state. We cannot ask mice what they are feeling or tell monkeys to put on a “happy” face. Now, a new study published in Science finally established a reliable and objective examination: using facial expressions and machine learning, scientists were able to identify what emotion mice were feeling1.
The concept of emotions can be tricky to describe, but researchers agree that emotions are internal states of feeling and patterns of behavioral, hormonal, and bodily responses that occur in response to an event or stimulus and are aimed at promoting survival. There are six staple emotions identified by Paul Ekman in the 1970s that neuroscientists and psychologists have focused on: happiness, sadness, anger, surprise, fear, and disgust2. Of course, with time, these have been slightly expanded.
Scientists at the Max Planck Institute of Neurobiology searched for these emotions in the faces of mice in response to different stimuli including tasting sugar or a bitter compound called quinine. The mice generated stereotyped facial movements like moving their ears, cheeks, and nose (Figure 1). The researchers filmed these reactions and studied them thoroughly. Using machine learning, the researchers fed the data to a computer and asked if it was able to predict the emotional event just from the facial expression. Using this system, they were able to determine which emotion the mouse was feeling, like pleasure, disgust, or fear.
Not only was the facial expression providing information about the specific emotion being felt, but also about the intensity and the valance, meaning whether it was a positive or negative experience. In one test, a slightly salty solution (a positive experience) elicited a facial expression similar to the one designated as “pleasure” while a very salty solution (a negative experience) elicited an expression similar to “disgust.”
The initial internal state also played a role in determining the emotion and its associated facial expression. For example, drinking liquid elicited stronger pleasure-like facial expressions in mice that were thirsty when compared to mice that were quenched. This means that the facial expression of mice is not just in response to the stimulus, but also reflects the emotional state of the animal.
Most importantly, the scientists were able to identify distinct sets of neurons that were associated with different emotions by seeing which cells activated in response to the stimulus. The researchers found that single neurons became activated in response to certain stimuli in an area called the insular cortex, which plays a role in sensory experience and emotional valence. The neurons that responded to the different emotions were highly segregated, meaning they did not overlap— neurons that became activated when the mice were disgusted did not get activated when the mouse was feeling fear or pleasure. Interestingly, this activity correlated with specific facial expressions.
Then the scientists controlled this pathway using a technique called optogenetic manipulation that allows researchers to activate specific cells. With this technique, they were able to activate certain areas, which generated predictable facial expressions. In fact, activating the pathway for one emotion while the mouse was stimulated to elicit an opposite emotion counteracted the facial expression generated by the stimulus. For example, activating the “pleasure” pathway attenuated the facial expression of disgust in response to something unpleasant.
This study is a huge step forward for the study of the neural basis of emotion. Now, scientists can use this reliable readout of facial expressions in mice to better identify what the mouse is feeling, allowing for a more accurate study of the neuronal circuits for specific emotions. Identifying emotion-relevant circuits in mice would allow researchers to study these same circuits in humans to see whether emotional coding is similar across species. Therefore, this model could be used to understand the interaction between the brain, emotions, and the environment. For example, recognizing emotions through facial expressions may vary from culture to culture and in different environments3. Could changes in the environment change emotional responses and be reflected in the brain as well? With this newly generated guide of “what are mice feeling?” we are even closer to figuring out exactly how the brain generates emotion and what neural factors influence feelings.
- Dolensek, N., Gehrlach, D. A., Klein, A. S., & Gogolla, N. (2020). Facial expressions of emotion states and their neuronal correlates in mice. Science, 368(6486), 89–94. doi: 10.1126/science.aaz9468
- Ekman, P. (1971). Universals and cultural differences in facial expressions of emotion. University of Nebraska Press.
- Hareli, S., Kafetsios, K., & Hess, U. (2015). A cross-cultural study on emotion expression and the learning of social norms. Frontiers in Psychology, 6. doi: 10.3389/fpsyg.2015.01501
Cover image by Gino Crescoli from Pixabay, https://pixabay.com/photos/smiley-emoticon-anger-angry-2979107/
Figure 1 from MPI of Neurobiology / Kuhl, https://www.eurekalert.org/multimedia/pub/228182.php