Is it possible that your face gradually comes to reflect your name as you age? New research suggests that the name you’re given at birth might subtly shape your appearance as you grow older. Researchers found that adults often look like their names, meaning people can match a face to a name more accurately than random guessing. But this isn’t true for children, which suggests that our faces grow into our names over time. The findings have been published in the Proceedings of the National Academy of Sciences.
The idea that names might influence facial appearance draws from the broader concept of self-fulfilling prophecies—where expectations about a person can influence how they behave and, in this case, perhaps even how they look. If society has certain expectations of how someone named “John” or “Emma” should appear, it’s possible that over time, individuals might unconsciously shape their physical appearance to align with those expectations.
However, it’s also possible that babies might be born with certain facial features that subconsciously influence parents to give them a name that “fits” their appearance. This study sought to determine which of these two scenarios is more plausible: Are people’s faces influenced by their names as they age, or do people already look like their names from birth?
“We noticed that even though we are good with names, there are few people that we can’t remember their name or even call them by mistake by a different name,” said study author Yonat Zwebner, an assistant professor of marketing at Reichman University. “It is also common to hear that someone’s name ‘really suits’ them while sometimes you hear that someone’s name really doesn’t suit them. So it got us thinking – could it really be that most people look like their name?”
To answer this question, the researchers conducted a series of five studies that combined human perception tests with machine learning techniques to analyze whether faces could be matched to names more accurately than by chance alone.
In the first two studies, the researchers aimed to test whether people could accurately match names to faces more often than would be expected by random chance. They did this by using a straightforward experimental setup where both adult and child participants were shown a series of headshot photographs of unfamiliar faces. Each photograph was accompanied by a list of four possible names, one of which was the correct name for the person in the photograph. The task for the participants was to choose which name they believed matched the face.
Study 1 included two groups of participants: 117 adults (aged 18 to 30 years) and 76 children (aged 8 to 13 years). These participants were asked to match names to both adult and child faces. The researchers wanted to see if participants could more accurately match names to adult faces compared to child faces, suggesting that adults might “grow into” their names over time.
To ensure the robustness of their findings, Study 2 replicated the experiment with a different sample of participants and faces. This study included 195 adult participants (aged 20 to 40 years) and 168 child participants (aged 8 to 12 years). All participants completed the experiment online, and the faces used were sourced from a professional database to ensure consistency in image quality and background characteristics.
In both studies, the researchers found that participants were able to match names to adult faces more accurately than by random chance, but they were not able to do the same for children’s faces. These findings suggest that the face-name congruence—the idea that people might look like their names—seems to develop over time, as it was present in adults but not in children.
In the third study, the researchers used machine learning to examine facial similarities among people with the same name. They employed a Siamese Neural Network, which was trained on a large dataset of facial images from both adults and children. The dataset included 607 adult faces and 557 child faces, with each group featuring the same 20 names (8 male and 12 female names).
The neural network was trained using “triplet loss,” where the model was presented with an anchor image, a positive image (same name as the anchor), and a negative image (different name). The model learned to identify the positive image as more similar to the anchor than the negative image. This training process helped determine whether there was a detectable pattern of facial similarity among individuals with the same name.
The findings from Study 3 were consistent with those of the earlier studies. The Siamese Neural Network found that adults who shared the same name had more similar facial representations compared to those with different names. The model showed a “similarity lift” of 60.05% for adult faces with the same name, significantly higher than the random-chance level of 50%. In contrast, for children’s faces, the similarity lift was only 51.88%, which did not differ significantly from chance.
In the final two studies, the researchers tested whether the name-face matching effect could be observed in faces that were artificially aged to resemble adults. They used Generative Adversarial Networks (GANs) to digitally age photographs of children, creating “artificial adults.”
Study 4A involved 100 adult participants (aged 19 to 39 years) who were asked to match names to a mix of real adult faces and digitally aged faces of children. Participants saw these images and selected the correct name from four options, just as in the earlier studies. The goal was to determine if the artificially aged faces would exhibit the same name-face congruence as real adult faces.
The participants were able to match names to real adult faces with an accuracy of 27.98%, significantly above chance. However, when it came to the artificially aged faces, their accuracy dropped to 24.25%, which was not significantly different from chance. This result suggested that merely aging a face digitally does not produce the same name-face congruence seen in real adults.
In Study 4B, the researchers used the same machine learning approach from Study 3 to analyze the artificially aged faces. The dataset for this study included 310 artificially aged faces (108 males and 202 females). The Siamese Neural Network evaluated the similarity of these faces to real adult faces with the same names, comparing the similarity scores to determine if the name-face matching effect observed in real adults could be replicated in these digitally aged faces. The model found that the similarity lift for the digitally aged faces was only 51.41%, almost identical to the lift observed for children’s faces and not significantly different from chance.
But how could someone shape their own facial features? The researchers suggest that this might happen both directly, through choices like hairstyle, glasses, and makeup, and indirectly, through life experiences that leave their mark—such as the way repeated smiling can create wrinkles over time.
“We know how belonging to a specific gender can have a strong social structuring impact, but now we know that even our name, which is chosen for us by others, and is not biological, can influence the way we look, through our interactions with society,” Zwebner told PsyPost.
“I’ll elaborate more on how we suggest the effect is developed: An example we all know too well is gender stereotypes. If for example, society expects girls to be gentle and polite while boys to be more assertive and aggressive, through self-fulfilling prophecy processes most boys and girls become exactly like that. We think the same process underlies our face-name matching effect.
“We already know from previous research that names have stereotypes,” Zwebner said. “For example, prior published studies show that in the U.S., you will evaluate a person named ‘Katherine’ as more successful than a person named ‘Bonnie.’ You will also evaluate a person named ‘Scott’ as more popular than ‘Herman.’ Moreover, we know from prior research that people imagine a ‘Bob’ to have a rounder face compared to a ‘Tim.’ All these are name stereotypes that also entail how we think someone with a specific name should look like.”
“Therefore, like other stereotypes, one may indeed become more and more like his/her name expectations, including appearance. This is strongly supported by the fact that our participants chose names according to hairstyle alone. It suggests that people embrace a certain hairstyle, and probably more facial features that fit the expectations of how they should look according to their names. Assuming that within a society all share a similar stereotype for Katherine, then we interact with her in a way that matches our shared stereotype. We treat her with specific expectations. As a result, Katherine becomes more and more like a Katherine is expected to be, resulting with a specific matching look. It could also be a more direct association, if the name stereotype is related to a specific look (e.g., wears a ponytail)—then the person could embrace that look.”
But it is also important to note that while the research suggests that, on average, adults tend to develop a face-name congruence over time, this effect is not uniform across all individuals.
“In our studies, there is of course a diversity – there are people who have a very high face-name match while others have a low match and others are in the middle,” Zwebner explained. “So clearly, some people look very much like their names, and some don’t. We think we demonstrated that this face-name match is something that is part of our social world so these different levels of face-name match should have life implications. For example, would you trust a salesperson who completely does not look like his/her name? Would you hire someone that looks very different from what you imagined according to his/her name? And, of course, are there other factors that are correlated with looking like your name or not?”
While these studies provide evidence for the idea that social expectations tied to names can influence facial appearance, they also have some limitations. For example, the study primarily focused on participants and facial images from specific cultural backgrounds, which might limit the generalizability of the findings. It’s possible that the effects observed in this study could vary in different cultural contexts where names and social expectations differ.
Additionally, the researchers suggested that it would be interesting to investigate the point at which people start to “grow into” their names. At what age do people begin to exhibit the face-name congruence observed in adults? Understanding this could provide further insights into the developmental processes at play. Future research could also investigate the impact of a person’s name on their life outcomes.
“If a name can influence appearance it can affect many other things, and this research opens an important direction that may suggest how parents should consider better the names they give their children,” Zwebner said.
“It is interesting to see a pattern in people’s reaction to our findings: their first reaction is typically ‘no way!’ but then, they comment that it ‘actually seems totally reasonable,’ and proceed to tell us a story about their or a friend’s name, and how it matches their face,” she added. “The fact that the findings relate to any person, anywhere. Anyone can relate to it, whether they are surprised at first or they feel that it is intuitive.”
The study, “Can names shape facial appearance?“, was authored by Yonat Zwebner, Moses Miller, Noa Grobgeld, Jacob Goldenberg, and Ruth Mayo.
