Your face gives clues to your name, suggesting your name has affected your appearance

Screenshot 2017-07-25 09.49.22
Example of test material from Zwebner et al. See footer for correct name.

By Alex Fradera

What can we tell about someone from their face? Their favoured facial expressions can hint at their temperament, the weathering of their skin at their life history, their facial hair and makeup at their aesthetic taste. But now, new research in the journal Attitudes and Social Cognition suggests that we can also intuit their names, because a person’s given name influences their facial appearance in adult life. On the face of it (sorry) this is hard to believe, but the case, made across eight studies, is based on plenty of careful evidence, and also proposes a plausible explanation.

In the first of several experiments, Yonat Zwebner at the Hebrew University of Jerusalem and his team showed student participants 20 colour headshots of a person with a neutral expression and asked them to choose which of five given names they would assign to each face based on sheer intuition. The given names included the true name of the person in the photo and four alternatives. When participants picked the name that genuinely belonged to the photo’s subject, they were scored as correct. Chance performance (i.e. random guessing) would mean getting 20 per cent correct, on average, but the participants performed better, at 28 per cent. For 17 of the 20 faces they achieved greater accuracy than you’d expect based on random guessing.

Another study was similar but this time Zwebner and his colleagues used photos only belonging to people from an Ashkenazi Israeli ethnic background and who were in their mid-twenties, and the mix of actual and filler names were all drawn from the same ethnic population (the intention was that this would avoid potential confounds due to participants choosing the correct name because of extrapolation from the name or ethnicity). This study also threw in a control condition – participants had the same menu of names to pick from, but the photos were covered up – the idea was to check whether participants’ were being drawn to the correct names because they were for some reason simply more pickable. In fact, participants in the control condition showed no tendency towards picking the true names, whereas participants who could see the photos managed an average face-name matching accuracy of 25 per cent (again better than you’d expect if they’d just guessed at random).

In further studies, the researchers made sure they took account of the possible influence of the relative popularity of different names. They also conducted a replication in France with photos of French people and featuring French names. Whatever changes they made, the researchers found the same thing: based purely on the appearance of someone’s face, participants had an inkling of their name.

The same proved true when the researchers programmed a computer to match names to faces. Through machine learning processes using a training set of nearly 100,000 faces drawn from an online networking site, a computer programme learned to distinguish between true and false names of faces at better than chance levels. The data was rich enough to evaluate specific comparisons, so we know that in this dataset the algorithm found it impossible to tell from her photo whether a woman was called Emilie as opposed to Jeannie, but found it almost no trouble to tell whether another woman was called Emma rather than Veronique (in this case achieving 95 per cent accuracy).

So what’s going on? One possibility is that we give babies names that somehow reflect their physical properties, and observers are simply drawing on these same associations to guide their judgments. This would be an example of the wonderfully titled bouba-kiki effect, wherein across cultures certain sounds seem soft or round, others hard or sharp. However this seems a bit of a reach, given that at the time most babies are named, they are still pretty similar-looking (sorry parents).

Moreover, when the researchers looked for cross-cultural effects by testing Israeli students with French stimuli and vice-versa, they found none: the participants lost the ability to match names at better than chance levels. If the bouba-kiki effects was behind it, the feel of the names – independent of their cultural meaning – would be sufficient to make the associations. So despite some earlier coverage, this isn’t a compelling explanation for the new data.

Zwebner and his team think it is likelier that their findings reflect some sort of self-fulfilling prophecy. Within a given culture, names carry associations – historical and celebrity figures, “types” (a Kevin versus a Brad), and cues related to class or background (contrast Tony with Tristan). Once a person is named, these associations form a filter through which all their future interactions take place – starting with parental guidance and expectations, then first encounters with schoolyard peers, career guidance counselling – and even if the filter is a subtle one, over a lifetime these effects can accumulate to impact how we act in the world, and ultimately this ends up written in our faces –  by our preference for smiles over frowns, big mugging expressions over wry understated ones, or an open relaxed visage rather than one tight-lipped and controlled.

If this explanation is correct, we’d expect to see a stronger influence of names upon those parts of the face and appearance that are more under voluntary control. Consistent with this, in one study the researchers found that removing hair style (arguably one of the aspects of our appearance over which we have most control) from the photos significant reduced participants’ ability to identify the correct names for the photos, though they still performed better than chance.

To investigate this more deeply, Zwebner’s team turned to their algorithm, investigating which pixels of each photo were most useful in making accurate name-matches, and using this data to create heat-maps of the faces. These clearly showed the margins of the faces (aside from hair itself) were next to useless and the nose and ears were also of limited value. The most important facial areas for correct name identification were the brows, eyes and mouth – areas that are habitually engaged in voluntary facial expressions.

This suggests that the effect due to given names is an extension of the Dorian Grey effect, the phenomenon where personality influences facial appearance. In terms of links between facial appearance and personality, we already know that short-tempered people tense certain facial muscles more frequently than do relaxed people, which can make their resting face appear angry, for example. Or that adolescent women who are more sociable are perceived as more attractive later in life, presumably because smile lines are construed as more attractive than a tight-set face.  The bundle of associations that make Celina’s name reflected in her face are arguably doing the same thing, but through a more circuitous route (for example, the name label leading to differential treatment that then contributes to personality, and finally onto the face), which would explain why the effects were statistically significant but not large in size.

So, if you’re struggling to pick the ideal appellation for your newborn, these new findings suggest that on top of everything else you have to worry about, your decision will likely influence what they look like as an adult – but only to a small extent. If you want to predict their future appearance, you won’t find the answers by gazing at their potential namesakes. But you might get a clue.

We look like our names: The manifestation of name stereotypes in facial appearance

Image is from Zwebner et al, 2017. The correct name for this example is Dan.

Alex Fradera (@alexfradera) is Staff Writer at BPS Research Digest

12 thoughts on “Your face gives clues to your name, suggesting your name has affected your appearance”

  1. Surely it’s possible that parents that look a certain way (e.g. hairstyles/facial hair/friendliness), etc., tend to pick certain names common to their particular lifestyle, and then the children are influenced by their parents in terms of appearance, which is then seen on their children’s faces – rather than appearance being directly caused by the name. Did the ethnicity study control for SES/lifestyle etc.?

    1. Exactly, thank you Kara for adding what seems to me the obvious reason. I can’t understand why it sounds like the researchers effectively assumed that an ethnic group and age range within once country constitutes a homogenous culture and therefore homogenous offspring name conventions. Subcultures by SES are well understood by the population within a country; we can all guess SES for many given names; there’s reams of evidence that successive generations of one family are unlikely to move far in SES. Desirable appearance is also linked to SES, hence names and appearance must be somewhat correlated *through* SES.

      1. Hi Naomi and Kara

        The researchers do indeed put their SES focus on only two factors, repeatedly saying things in the text like
        “Additionally, we tightly controlled for socioeconomic factors (such as age and ethnicity).” Their focus on these factors seems to be based on previous research – summarised here – that suggests that age and ethnicity were reliable socioeconomic cues used to match names to faces. I agree that there are yet more comprehensive angles that could be covered.

        That said there is another study that I didn’t have space to report, but it does cut against the socioeconomic claim. It goes as follows:

        When individuals go by a unique nickname day to day, rather than their given name, it’s harder to predict their given name from their face. Participants still detect their given name better than chance, but the effect is attenuated relative to people only known by their given name. This suggests that it’s the process of living under a name that changes the face, rather than the name being a marker of predictable SES facial features.

  2. It would be great if a BPS reader who is literate in statistics and research methods could look at the original study, because this seems pretty implausible.

    1. Hi DZ
      Do you find the concept of Dorian Grey effects – that social effects can produce changes in the face (chiefly through use of musculature) over time – implausible in its entirety? Or this version specifically?

      1. For some research showing the relationship between personality characteristics and facial structure, see Squier, R.W. & Mew, J.R.C. (1981) The relationship between facial structure and personality characteristics. British Journal of Social Psychology, 20, 151-160.

  3. Why is the example showing 4 names when the article talks about 5?

    Different cultures have different norms for how much people smiles. So statistically, a woman with lots of smile lines should more likely have an American sounding and not a Slavic name.

    This writeup only talks about Israelis being able to evaluate fellow countrymen but not being able to distinguish among people from other cultures – although being able to match a French sounding name compared to an English sounding name to a picture. It would be interesting to see this replicated with other populations.

    1. Hi Charlotte
      Across the various studies sometimes four names were used and sometimes five. And yes, it would be interesting to do the replications across other nationalities.

  4. Given that the Israeli subjects were better at “detecting” the names of Israeli subjects, but dropped back to random chance for French subjects, could this not simply be a matter of knowing the relative odds of the names? The computer test example given supports this, in my view: Emilie and Jeannie are both relatively common names, and hence hard to guess which might be right, while I would bet that Emma is much more common than Veronique, allowing the computer to go with the odds. Improving from 20% to 25% does not seem significant enough to me to jump to the conclusions asserted in the article. I think it more reasonable to look at the distribution and familiarity of names regardless of faces. I suggest the researchers use exactly the same test set, but randomly assign the pictures to the name lists, so that there is no correlation between them, but subjects will still be assessing more familiar or less familiar names. Without this check on the data, I question whether there is enough information to make the much more abstract and less justified conclusion that “personality affects personal appearance.”

    1. Hi Nora, this exact criticism was addressed in the second study (the one referred to in the third paragraph of my piece). To expand, in the author’s words:

      “Study 1B ruled out the possibility that the frequency of the names (or other feature related to the targets’ names) drives the observed effect, because participants correctly matched the names to their corresponding faces even when the same names appeared with different faces”

Comments are closed.