Cross-cultural investigation of Smileys
This post, written by Karim N'Diaye, was first published in 2006 on the Alphapsy blog. I republish it here because it is relevant to the discussion I was having with Helen de Cruz, on the cultural specificity of cartoon faces. Below the posts are some comments posted on the Alphapsy blog in 2006 – Olivier
As you might have noticed, japanese emoticons and american ones differ (see illustration below). An empirical study by psychologists from these two countries suggests that people from the two cultures differ in the way they perceive emotions as expressed on faces: while easterners focus on eyes, westerners look at the mouth. Although weak, this difference might prove sufficent to have lead each culture in using different styles of emoticons, Masaki Yuki and his colleagues argue.
Illustration: a japanese smiley on the left and an american/european one on the right
To provide evidence for these culutural biases in facial emotion recognition, the researchers used two sets of stimuli which participants had to judge on a happy/sad scale. One set consisted of schematic faces. The other set was made of cut-and-paste face photographs taken from an oft-used database in emotional research: sad eyes were pasted above an happy mouth, or vice versa. These manipulations did led to significant differences in ratings by the two culture (e.g., the americans rated a hybrid faces made of sad eyes above a happy mouth as more happy than japaneses did while, conversely, japaneses "happiness" ratings were higher for happy eyes above a sad mouth).
However their effects are small and mainly due to the fact that happy mouths trigger much higher happiness ratings by americans relatively to japanese. Considering the fact that Ekman's database of emotional faces is made of caucasian faces I wonder if they didn't loose some power due to the fact that japanese were made rating people from another race. According to the argument of "dialects" in facial expression defended by the authors themselves, this would not be unexpected.
On my second concern the authors did comment, though rather quickly. Here it is: if you look at the pattern of ratings in Exp. #2, faces made of happy eyes and sad mouths were actually perceived **less happy** than the same sad mouths paired with neutral eyes (and this is true in both groups). You may even feel it yourself (at least I did) with example stimuli they display just below the graph. Uh-oh… Wait, don't throw the paper with the baby bath. This one effect simply reveals that eyes and mouth are not independently processed, rather they seem to be appraised in a holistic manner. It doesn't ruin the whole argument by Yuki et al that Japanese and Americans differ in their processing of facial expression however it might render conclusions not so straightforward. At least, the data call for confirmation.
Now, regarding the japanese smileys and their american counterparts, I must say the hypothesis is fun but as always with this type of single-case evolutionary scenarios, it has a slight smell of just-so story: alternative explanations might consider how the smiley face initially entered the late 60's american culture and got then rotated when used as an emoticon, while japanese verticon are mainly inspired from the anime drawing style (see example). But I let this issue be dealt with by internet anthropologists.
Via Le Monde.
1. On Sunday 8 October 2006 by Julian
What is exactly meant by "single-case evolutionary scenarios"?
2. On Sunday 8 October 2006 by olivier
Hasn't it occured to the authors that japanese and european eyes do not have the same shape?
3. On Sunday 8 October 2006 by karim
By this I refer to the kind of explanations which provide (possibly convincing) scenario for an observation, but which can hardly be tested otherwise because phenomena sharing this evolutionary process rare. Most famous example: the origins of human language. This doesn't mean evolutionary approach is flawed, it simply ask for additional data (and in the case of language, see the recent posts by Hugo and Nicolas showing how such model can be constrained by empiraical results from psychological experiments).
Olivier, I also wondered about this but: ¦-)
Interesting study. However, I want to draw your attention to an important difference. If we compare the graphs for the first experiment with those for the second experiment, we observe that, while those on the right are somewhat similar overall, those on the left differ significantly when it comes especially to the 'happy eye'/'neutral mouth' and, even more so, the 'happy eye'/'sad mouth' displays, with respect to the ratings of Japanese subjects and, interestingly, of US subjects for the first display (emoticon, resp. face).
This may be because of the following: in the second experiment, it seems to me that it is easier to evaluate the faces in the right-hand side group (or at least the last two, the first being the most unproblematic) with respect to sadness due to the contextual cue offered by the eyebrows (cf. the unproblematic depiction of 'sad eyes' in the first experiment), while the photographies on the left are quite difficult to rate- the 'sad mouth' and 'happy eyes' depictions are not obvious in the absence of a third (contextual) element (such as eyebrows); the asymmetry observed in the results for 'neutral eye' vs. 'happy eye' coupled with a 'sad mouth', in the sense that faces made of happy eyes and sad mouths were actually perceived as 'less happy' than the same sad mouths paired with neutral eyes, is explainable via this awkwardness of the depiction of 'happy eyes'. Now, on the first experiment, we find the results in the left graph enhanced for the 'happy eye'/'neutral mouth' and, even more so, the 'happy eye'/'sad mouth' display (the ratings for this last one significantly differ only with respect to Japanese subjects). Why is this? It seems Japanese are more prone to recognize the emoticon as happy due to the symbol for happy eyes; it seems to me difficult prima facie to interpret the symbol for happy eyes as a symbol for happy eyes, if presented outside the context of a neutral or happy mouth; secunda facie, due to knowledge of a convention, this might be easier, even outside that context. It seems to me that it is this knowledge of a convention that influences the results here, namely the Japanese representation for happy eyes. The performance of US subjects on the 'happy eye'/'neutral mouth' emoticon might be boosted by the let's call it 'compatibility' between the two, compared with the 'incompatiblity' between the 'happy eye' and 'sad mouth' schematic representations; otherwise put, the 'happy eye' symbol is more amenable to an interpretation of it as such in combination with a neutral or happy mouth, than with a sad mouth.
In conclusion, while this difference does not refute the hypothesis advanced by the authors, a comparative one (Japanese rating happy-eyes faces as happier and sad-eyes faces as sadder than Americans, and Americans rating happy-mouth faces as happier
and sad-mouth faces as sadder than Japanese), it shows how contextual cues and putative knowledge of conventions may significantly influence the ratings of emoticons/faces. The final question here is whether this difference between the two graphs on the left is a difference that makes a difference with respect to the validation or warranting of the conclusion of the authors or not. They acknowledge in the last paragraph the great difference in effect sizes for emoticons and faces, but only to infer that this suggests "emotion interpretation is more complex, and perhaps
meaningful, using real faces rather than illustrated
ones" (p. 8). As for the overall results, the differences in ratings do not seem to be that striking, once we clear up the aforementioned points, at least not to the extent allowing one to draw the conclusion that the cultural differences are significative. They only support a weak comparative thesis. Perhaps other, more persuasive experiments could be devised to confirm (or refute) the starting hypothesis.
One small point: I could not follow why the rating of the emoticons is on a scale of 1 to 9, while the rating of faces is on a scale of 3 to 8.
5. On Tuesday 10 October 2006 by Jason Malloy
[[alternative explanations might consider how the smiley face initially entered the late 60's american culture and got then rotated when used as an emoticon, while japanese verticon are mainly inspired from the anime drawing style]] Sure but this just pushes the exact same iconography under examination a little further back.
The anime style of the icon, with the large eyes and tiny mouth, is to look cute – kawaii* – a pervasive aesthetic in Japan. So I don't think the emoticon is related to the finding in this paper.
The prominence of cute in Japan, I would speculate, is because Japanese people and related populations are more psychologically aroused by neotenous (child-like) stimuli. I think this is because, on average, they physically develop slightly slower, and are therefore adapted to milk out, and be milked out of, slightly more paternal investment.
Jason: I think maybe what you're talking about is amae.