Do emojis represent the whole gamut of human emotion?

Emojis have become part of our everyday communication online, allowing us to succinctly communicate how we're feeling in a way that written language cannot. Psychologists are even beginning to use emojis in research, to allow children or other participants to respond without the need for traditional questionnaires.

But is the library of emojis that is available to us truly representative of the range of emotions that we feel? A new study in Scientific Reports suggests that, broadly, it is — but that there are some important gaps too.

Many psychologists have moved away from the idea that we have a handful of discrete emotions. Instead, they see our emotional experiences as falling along continuous scales of both valence — how positive or negative an emotion is — and arousal. So, for instance, "sadness" has a negative valence but is fairly low in arousal; "anger" is also negatively-valenced but high in arousal; and "excitement" is positively-valenced but still high in arousal.

In their new study, Gaku Kutsuzawa and colleagues at the National Institute of Advanced Industrial Science and Technology in Kashiwa, Japan, looked at whether emojis can be classified according to valence and arousal in a similar way to our own emotions. The team asked more than 1,000 Japanese participants aged 20 to 39 to rate facial emojis on 9-point scales for both valence (ranging from "displeasure" to "pleasure") and arousal (ranging from "weak" to "strong"). Emoji libraries currently contain 94 face emojis, though the researchers excluded 20 that would be hard to classify (for instance, emojis related to sickness like 🤒). Each participant rated a subset of 30 emojis from the remaining 74.

When the the team plotted these ratings on a chart representing valence and arousal, they found a u-shaped pattern (see below). Some emojis like 😡 were negative in valence and high in arousal. As an emoji's valence rating became less negative, arousal ratings dropped too, so that neutrally-valenced emojis like 😶 tended to be rated low in arousal. Then, as an emoji's valence rating became more and more positive, arousal ratings tended to increase again, with extremely positive emojis like 😍 also being rated high in arousal.

Using a statistical method that looks for distinct groups within data, the team found that the emojis could be separated into six clusters that fell at different points along this u-shaped plot. Overall, the emojis within each cluster corresponded pretty well with the human emotions we tend to experience at that level of arousal and valence.

These results are not that surprising: we've developed emojis precisely because we want to be able to convey emotion, so it makes sense that they reflect our emotional experiences. Perhaps more interestingly, the team found that the emoji library didn't quite represent all emotions. In particular, there were no emojis that were neutral in valence but high in arousal — think emotions like being "tense" or "alert".

The team also identified features within emojis that were related to their ratings. An emoji's valence seemed to be influenced by eye and mouth shape: an upward-turned mouth was related to more positive valence ratings, for example. High arousal ratings, on the other hand, seemed to be associated with the presence of "accessories": stars in the place of eyes in 🤩, for instance, or icicles and a blue face for 🥶. We often use facial movements to convey arousal, but emojis are static — so these accessories might be fulfilling that function instead.

The researchers conclude that emojis display a broad range of human emotional states. This is useful knowledge for emotion researchers, who may want to allow participants to respond with emojis rather than text in order to make questionnaires less taxing, or to survey participants whose language they do not speak. However, it seems there is also room for the development of new emojis to represent states like alertness.

That said, the way that emojis are interpreted could be influenced by factors like language, age and culture, so the meaning assigned to emojis by these relatively young Japanese participants may not be universal. Take 😂, for instance: there are clearly generational divides in how the emoji is used, so this could potentially influence people's ratings. Before emoji-based questionnaires become a common part of emotion research, it seems like more work is needed to investigate the potential influence of these kinds of demographic factors.

Further reading

– Classification of 74 facial emoji's emotional states on the valence-arousal axes

About the author

Matthew Warren (@MattBWarren) is Editor of BPS Research Digest