Of Mice and Monkeys: Expressions Reveal Cognitive States

Long-standing research in human psychology has claimed that certain facial expressions of emotion are universal across cultures. For example, classic cross-cultural studies by Paul Ekman and colleagues argue that certain "basic emotions" (happiness, sadness, anger, fear, surprise, and disgust) are not only shared across human cultures, but they are also signaled by very similar facial expressions. Moreover, facial expressions can also signal internal cognitive states (such as attention or inattention), which can be used to infer subsequent behavior.

Facial expressions in mice and monkeys

In a new research study published in Nature Communications, Alejandro Tlaie and colleagues suggest that the relationship between facial expressions and internal cognitive states extends beyond the human species. In their study, the researchers trained two monkeys (Macaca mulatta) and six mice (Mus musculus) to navigate a naturalistic virtual foraging environment using trackballs. The animals were trained to search for certain targets while avoiding others, receiving drops of diluted juice (for the monkeys) and drops of vanilla soy milk (for the mice) when they reached a desired target. A camera tracked the animals' facial expressions as they navigated the virtual environment.

Using a Markov switching linear regression (MSLR) model for each species, the researchers inferred potential cognitive states as a function of the animals' facial expressions and then used those inferred cognitive states to predict their subsequent behavior (such as faster or slower reaction times, or their direction of motion in the virtual environment). The models revealed the "optimal" number of hidden states for the two species to predict their behavior. For the monkeys, the data could be best predicted by four unique states. For mice, three unique states were sufficient. The researchers validated their model by successfully predicting behavior in new animals that did not contribute to the model's training data.

Holistic facial changes, not isolated features

Overall, the study showed that pretrial facial expressions could indeed predict subsequent performance in the foraging task. In mice, pupil size and whisker movements predicted reaction times during target trials, but nose movements were better at predicting reaction times in nontarget trials. Similarly, monkeys' pupil sizes and eye movements predicted reaction times in nontarget trials and target trials, respectively. Importantly, the information extracted from the animals' facial expressions could not be reduced to a single feature; rather, behavior was best predicted by a holistic combination of facial features.

The findings highlight the importance of considering the complex changes in animals’ faces rather than focusing on isolated features such as pupil size. In addition, the link between facial expression and behavior is not fixed but rather depends on the internal state of the animal. Therefore, much like in humans, facial expressions in mice and monkeys reflect internal factors that directly relate to subsequent behavior.