Japanese smileys vs. Ekman faces

Some medias and the blogosphere (see here, here and here) are celebrating a new study published in Current Biology, allegedly showing that recognition of facial expressions is not universal. Psychological universalists and relativists never seem to get tired of chewing that old bone of contention.

There are two aspects to the study. The first is a very nice exploration (by means of eye-tracking) of the way Asians process facial expressions, replicating the earlier work of Masaki Yuki and colleagues three years ago (read what Karim wrote of it at the time). Japanese subjects tend to focus on the eyes instead of the mouth to decode emotions – as one could have guessed from looking at Japanese Smiley faces :        (^_^)     for 'happy',       (T_T)        for 'sad', and other such          (*_*)       …

Yet the authors don't stop at that fascinating result, and go on to try and prove another point : that because of this difference in face-processing style, East Asian subjects and 'Caucasian' subjects are not equally good at recognizing some of Paul Ekman's supposedly universal facial displays of emotions, like disgust and fear. And indeed East Asian subjects are significantly likelier than Caucasians to misinterpret happy or fearful faces.

As Neuroskeptic points out in his excellent coverage of the experiment, the difference, though, is really tiny…


and East Asians are still quite high above chance in recognizing fearful and happy faces. For two other emotions, anger and sadness, East Asians are no different than Caucasians. If anything, they are better.

According to the abstract, East Asians "use a culture-specific decoding strategy that is inadequate to reliably distinguish universal facial expressions of fear and disgust." Well, maybe their method is not the best possible one, but it does give reliable results, if "reliable" means that East Asians are always much more likely to make the correct decision, out of seven possible options, when recognizing an emotion. Their worse performance is 58 % of good guesses for Caucasian fearful faces (they are better at recognizing emotions on Asian faces, as Karim predicted three years ago).

These are beautiful, but tiny results, and I don't think the magnitude of the effects warrants the author's claims that "Our results question the universality of human facial expressions of emotion, highlighting their true complexity, with critical consequences for cross-cultural communication and globalization." The Neuroskeptic gets it right as usual :

"The differences notwithstanding, the whole task relies upon the fact that the subjects know the meaning of "happy", "fear", and so forth, and associate them with certain face expressions. The fact that the experiment worked at all shows – as Ekman would predict – that both Westerners and East Asians share an emotional understanding. There appear to be some cultural quirks, but the essential universality of facial emotion still stands." (Neuroskeptic)




  • comment-avatar
    Simon Barthelme 31 August 2009 (16:43)

    I wholeheartedly agree with your post and the one by Neuroskeptic. The eye movement data is interesting, but there’s something that bothers me about the behavioural performance differences: raw performance might be affected by reporting bias. It’s a problem that was identified in the Signal Detection Theory literature 50 years ago. Imagine that observer A is able to discriminate disgust from happiness just as well as observer B: if observer A is unwilling to report disgust her performance will be lower than that of B. In the limit if you always say “Happiness” because you don’t like saying “Disgust”, your performance will be at chance. I did not see a discussion of this, but maybe I missed it.

  • comment-avatar
    Neuro Skeptic 2 September 2009 (23:30)

    Thanks for the link! Simon Barthelme makes a good point, although in this task I think an unwillingness to report certain emotions is unlikely since the subjects were instructed to classify faces as one of the seven basic emotions in a forced-choice manner. If they had been able to write down any word they felt appropriate I suspect such factors would come into play. That would be an interesting study actually…

  • comment-avatar
    Simon Barthelme 3 September 2009 (10:49)

    Neuro skeptic, as far as I understand from what I read in the paper the procedure is that observers are shown a single face, and asked to classify it in one of seven emotional categories. The authors call it 7AFC but it’s not really a forced-choice in the classical sense of the word: forced-choice is when you show observers two faces, and ask them to tell you which face looks disgusted. That would probably eliminate the bias problem, but the task used here doesn’t. Also, they mention in the Methods that observers had to say the emotional category out loud so the experimenter could double-check their response. This might have made matters worse, I’d guess.