I Can See it in Your Face: Facial Expressions in Mice
How do you know what an animal is feeling? Unfortunately, you can’t ask a dog, mouse, or fly how they are feeling or what they are thinking. This is one of the chief problems in animal research – we just don’t have ready access to their internal thoughts and feelings. This inability to access the thoughts and emotions of animals led radical behavioral psychologists like B.F. Skinner and John Watson to assert that there was no internal experience in animals, and that emotions were purely physical responses. Nowadays scientists generally accept that animals do indeed experience emotions or “emotional states”, but it’s still very tricky to measure them or to determine if they are at all similar to what humans experience.
Of course, for some animals it’s a bit easier to tell how they are feeling. We can see a dog wag its tail, or hear a cat purr contentedly in our lap and have at least some idea of what they might be feeling. But how do we know the emotions going on inside a mouse? Until recently, researchers have used all sorts of complicated methods to try to figure out what a mouse is feeling or thinking. For instance, since mice are prey animals, they generally prefer to spend their time in enclosed, dark places. So, we can put mice in an enclosure with both light, open spaces and dark, enclosed spaces and then measure how much time they spend in either the light or the dark sides. If a given mouse spends more time in the safe, dark area, perhaps that mouse is experiencing more anxiety or fear than a mouse who explores the light area readily. We can even manipulate this tendency by giving mice drugs that we know reduce or exacerbate anxiety in humans. Still, this is, at best, an oblique and imprecise measurement.
I can see it in your face
You can look at someone’s face and generally have a pretty good idea how that person is feeling. Emotions like pain, surprise, disgust, and joy are readily apparent in facial expressions and consistent across cultures. But look into the face of a mouse and you’ll likely be pretty puzzled. However, just because we can’t readily identify a mouse’s emotion from looking at their face doesn’t mean that mice don’t have emotions, or that they don’t have facial expressions – we might just be bad at reading a mouse’s poker face.
A recent study in Science set out to determine if mice do indeed have facial expressions that reveal their emotional state. Researchers used cameras to record a mouse’s facial expressions both during neutral conditions as well as when they experienced various stimuli. These stimuli were chosen since they should evoke emotional states; they included sugar water (pleasure), bitter water (disgust), mild electric shock (pain), and injection of a chemical that induces mild sickness (malaise). The researchers then took these recordings and used machine learning techniques to compare the facial expressions of mice in the neutral condition vs. during the various emotional stimuli. Essentially, these machine learning techniques are able to compare individual pixels of the video frames and see if there are consistent changes (both within an individual mouse and across mice) in facial expressions when mice are exposed to the stimuli. Interestingly, these machine learning algorithms revealed that mice do indeed display prototypical facial expressions in response to the various stimuli (Fig. 1). Further analysis revealed that these facial expressions were mostly unique to the stimuli – that is, that pleasure doesn’t look like pain, nor does pain look like disgust.
Of course, from this data it’s still not clear if mice are actually experiencing emotions. Old school behavioral psychologists like John Watson and B.F. Skinner would tell you that it’s possible that these facial expressions are in some way reflexive – like blinking your eye shut in response to a puff of air. One of the qualities of emotions is that they can vary in intensity; if you find a quarter on the ground it might make you a little bit happy, but probably not quite as happy as if you won the lottery. Researchers varied the concentration of the sugar or the bitter agent (quinine) in the water and found that this could change the intensity of the facial expressions, with, for instance, higher concentrations of sugar making facial expressions appear more prototypically “pleasurable”.
So it seems like these facial expressions can change based on the intensity of the stimulus, but that still leaves open the possibility that they are just direct responses to the stimuli. Another characteristic of emotions is that they are state-dependent; if you are already having a bad day, someone cutting you off in traffic might really make you angry while on a good day you might not even notice. The researchers recorded facial expressions of mice drinking water when they were either thirsty or quenched and found that mice who were thirsty displayed much more prototypical pleasure responses.
Finally, another characteristic of emotion is that it can be learned. If you drink alcohol, there might be a particular type of drink that you have had a really bad experience with – for me it’s tequila. Now, whenever I smell, or sometimes even think of tequila I feel visceral disgust. This is what’s known as a conditioned taste aversion. The same thing can be done in mice by pairing a previously pleasurable experience with a drug that induces feelings of sickness. When the researchers paired sugar water with this aversive state, they saw that animals shifted from displaying pleasure-typical facial expressions to disgust-typical facial expressions in response to sugar water (Fig. 2). This does an excellent job of showing that mouse facial expressions are not purely responses to stimuli, as in this case the stimuli – sugar water – remains the same but the facial expressions themselves completely change.
Emotion in the brain
Collectively, all this data suggests that mice do indeed experience emotions and that they have stereotyped facial expressions in response to these emotional experiences. But a big question remains: how is emotion controlled by and represented in the brain? To begin addressing this question, the researchers used optogenetics (prior Neuwrite post) to activate specific regions in the brain. In particular, they focused on the insula, a region heavily implicated in emotions in both humans and animals. When they stimulated different subregions of the insula they were able to induce either “pleasure” (the anterior insula) or “disgust” (the posterior insula) -typical facial expressions. Finally, the authors also used calcium imaging (see previous Neuwrite post) to image the activity of individual neurons in the insula. Here they found neurons whose activity correlated with specific emotional facial expressions. Collectively this suggests that the insula is involved either in the emotional states themselves or in the facial expression of said states.
Let me emphasize that, if this study replicates, this is a huge deal. This study suggests that 1) mice do indeed experience distinct, consistent emotional states, 2) that these states can be inferred from their facial expressions, and 3) that we can use these facial expressions to understand how the brain might be involved in controlling and representing emotion. This study employs a powerful new method to investigate emotion in animals. Whereas previous methods were indirect and often reduced down to a single variable (e.g., time spent in a well-lit location as a measure of anxiety), facial expressions are not only much more direct, they are much more nuanced. By recording a mouse’s facial expression, we would have instant access to their current emotional state (rather than having to infer it after the fact), and have access to all sorts of interesting data points like how various aspects of a facial expression changes (e.g., ears vs. whiskers), when the expressions change, and how long they persist. All this rich data can then be examined in concert with recording or manipulating activity in the brain.
A further question that this study raises is why do mice have facial expressions that correlate with their emotions? While the evolutionary benefit of emotions themselves is relatively clear (for example, pain motivates an animal to avoid or escape a bad situation), why would they be revealed by facial expressions in mice? With the initial caveat that evolution often operates on the “if it ain’t broke, don’t fix it” principle, it’s possible that these facial expressions merely reflect some sort of preparatory motor or sensory plan for the mice to orient to (e.g., pleasure) or escape from (e.g. pain) some stimuli. Another possibility is that these facial expressions might be a means of social communication. Clearly we humans use our facial expressions to communicate to others, but might mice do so as well? Like humans, mice are also social animals, though unlike humans, mice rely much more on senses other than vision. Indeed, previous work has found that mice use pheromones to socially communicate pain (see this previous Neuwrite article). Using the techniques in this study, it would be possible to see if mice use facial expressions to communicate with one another.
Regardless of why mice have facial expressions, that they do indeed have them makes it much more feasible to study emotions in mice and opens up all sorts of fascinating questions about emotions in animals.
 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. https://doi.org/10.1126/science.aaz9468
Featured image of smiling hamster from: https://www.pinterest.com/pin/197384396142580942/
Figures from  (adapted from Figures 1 and 2)
do you think FACS (Animal Facial Action Coding System, an anatomical approach to classify animal facial muscle movements) can help to improve this approach using camera (computer vision and neural network) ?
Pingback: Can Botox make us happier? | NeuWrite San Diego