Are emotional expressions universal?

• Understanding both universal and culturally specific expressions is crucial for developing AI that can accurately interpret emotions, ensuring its effectiveness in global interactions.

• Recent studies using machine learning and large global datasets have confirmed that expressions in areas like facial cues, vocal bursts, and music are shared across cultures, but with nuanced variations.

• Hume’s expression measurement models use the same information that humans use to learn how to interpret expressions, based on findings from computational emotion science showing that expressions have largely shared meanings worldwide.

Do people around the world express themselves in the same way? Does a smile mean the same thing worldwide? And how about a chuckle, a sigh, or a grimace? These questions about the cross-cultural universality of expressions are among the more important and long-standing in behavioral sciences like psychology and anthropology—and central to the study of emotion.

In an increasingly globalized world, these questions are no longer just academic. How can we be sure we’re understanding each other when doing business internationally? If we're using tools like sentiment analysis or expression measurement, how can we ensure they deliver accurate results across different cultures? Ensuring accuracy in expressive communication and emotional understanding will be critical for the future of diplomacy, customer service, health care, and beyond.

Thankfully, research shows that a large number of expressions have broadly shared meanings across all cultures studied to date. However, there are important nuances in how intensely or overtly feelings are expressed in different cultures, making it crucial to dig deeper into the science of cross-cultural expressions and the history of how these questions have been investigated.

The Evolution of Cross-Cultural Emotion Studies

The study of facial expressions across cultures dates back to early claims by Charles Darwin that facial expressions were universal, having evolved for specific social functions, laying the foundation for hundreds of studies on emotion recognition. Paul Ekman, a key figure in the field, conducted groundbreaking research in the 1960s by studying the Fore people of Papua New Guinea, an isolated group with minimal exposure to Western culture. His findings, showing that they could identify emotions like happiness and sadness from facial expressions, reinforced the idea of universally recognized emotions and established the basic playbook for cross-cultural research on expressions.

Ekman and David Matsumoto later expanded this research by investigating how cultural norms, or “display rules,” govern the intensity and context of emotional expressions. Their work revealed that while certain emotions may be universally recognized, cultures differ in when and how those emotions are expressed. For example, in a study involving Japanese and American participants, both groups reacted similarly to emotional videos when alone, but the Japanese masked their emotions when observed, while Americans continued to express them openly.

Further meta-analyses by psychologists Nalini Ambady and Hilary Elfenbein added depth to this understanding, finding that people tend to recognize emotions more easily within their own cultural group. This suggests that while basic emotional expressions may be universal, subtle cultural “accents” influence how emotions are expressed and perceived, highlighting the complex interplay between universal human emotions and culturally specific display norms.

Modern Studies on Cross-Cultural Emotional Expressions

This research laid the foundation for our understanding of how different cultures express emotion, but had important practical limitations. These studies typically focused on a small number of emotions, typically the “basic 6”—anger, disgust, fear, happiness, sadness, and surprise—and studied only a few cultures at a time.

More recently, advances in data collection and computational methods have provided a more nuanced understanding of emotional expression across cultures. By gathering large amounts of data from around the world and analyzing it using data-driven statistical techniques, researchers can gain insight into how a wider range of emotions are expressed and perceived globally. These advances also reveal subtle differences in cultural display tendencies at a finer level of granularity than previously possible.

Moving beyond comparisons between two or three cultures, recent research has leveraged advances in large-scale data collection to run studies comparing much larger numbers of diverse cultures than previously possible. One study used machine learning to analyze millions of videos from 144 different countries. The researchers examined the emotional contexts–weddings, funerals, sports wins–in which 16 specific facial expressions occur, showing that there is substantial similarity in the social contexts in which individuals make similar facial expressions, even in very different cultures. Each type of facial expression was linked to specific situations, and these connections were the same about 70% of the time across 12 different regions around the world.

These studies have also moved beyond facial expressions, showing that a large number of expressions have shared meanings across cultures in domains such as speech prosody, vocal bursts, and music. For speech prosody–the tune, rhythm, and timbre of speech–one study used over 2,000 speech samples collected from 5 cultures, and showed that individuals from different cultures reliably recognized 12 distinct emotions conveyed by variations in prosody. A similar study found that there are at least 13 subjective dimensions of feeling evoked by music are shared across cultures–showing that what we share across cultures goes beyond our expressions and into how we process and appraise the world around us.

For vocal bursts – nonlinguistic sounds like laughs, sighs, oohs, and ahhs – researchers at Hume recently ran a study with a large sample of over 15,000 individuals from 5 countries and showed that at least 21 distinct vocal bursts have shared emotional meanings in these different cultures. The emotions people associated with different vocal bursts were 79% preserved across cultures, even considering the fact that people in different cultures spoke different languages.

A Data-Driven Approach to Understanding Universality

The specific number of emotions that studies find overlap between cultures in these studies is less important than the general, consistent finding: expressions have largely shared meanings worldwide, and we are newly equipped with the computational tools to understand these meanings as well as the subtle variations and nuances present between cultures.

Recent research conducted at Hume demonstrates this principle. We collected facial expression data from over 5,000 participants in six different countries and trained a model to recognize the expressions that were consistent across cultures. This approach showed that 28 distinct facial expressions are shared worldwide, even when controlling for factors like physical appearance and context. A figure from this paper demonstrates that, despite having shared expressions that largely overlap in meaning, there are subtle variations in the meanings inferred from facial expressions in different cultures. The important thing is to measure these subtleties and account for them in our expression measurement models.

 

Figure reproduced courtesy of Cell Press

Each of the 28 facial expressions is arranged in a column, with the average examples of the facial expression depicted in each row. The emotion concept that was most strongly associated with each facial expression in each country is overlaid on the corresponding image. Our approach demonstrates that there is not a one-to-one mapping between facial actions and specific emotions, but that there are some cases where the same facial expression could convey different emotions to different individuals or cultures, or some cases where multiple facial expressions could be interpreted as conveying the same emotion.

Conclusion

Every day, we make use of expressions to perceive, understand, and communicate with one another. Even when someone is speaking a different language, we can tell when they might be angry or annoyed by observing the way their face looks and their voice sounds in context. Our AI models use the same information that humans use to learn how to interpret expressions, a capability based on findings from computational emotion science showing that expressions have largely shared meanings worldwide.

While emotional expressions are largely universal, cultural nuances shape the way they are displayed and perceived. The emerging ability to collect and analyze data on a global scale allows us to appreciate both the universal and culturally specific aspects of emotion. This understanding is essential as we continue to develop AI systems that interact with diverse human populations, helping ensure that emotional expressions are accurately interpreted across cultures and contexts. 

Resources

Developer Platform

Our Research

API Documentation

Follow us on Twitter and Linkedin