The common belief about smiling in photographs is that it does not really matter, that a smile is a small social performance we all know how to fake, and that whatever a face does for a fraction of a second in a college portrait cannot possibly say anything meaningful about the next three decades of a life. A well-known study out of the University of California, Berkeley suggests this belief is at least partly wrong. Not because the smile itself causes anything, but because a certain kind of smile appears to reveal something about the person behind it that a posed one cannot.
The paper is Harker and Keltner, published in 2001 in the Journal of Personality and Social Psychology. LeeAnne Harker and Dacher Keltner, both then at Berkeley, coded photographs from Mills College yearbooks and matched them against decades of follow-up data. Keltner, now a Distinguished Professor of Psychology at Berkeley, still describes the finding plainly on his department page: differences in positive expressivity, captured from senior yearbook photographs, predict personality traits across time, well-being, and marital satisfaction up to thirty years later.
It is a single study, and its scope is narrower than the way it is usually retold. What it does show, though, is worth reading carefully.
What the researchers actually measured
The photographs came from the Mills Longitudinal Study, a project begun in the late 1950s at Berkeley’s Institute of Personality Assessment and Research. As the study’s own history records, its participants were drawn from the senior classes of 1958 and 1960 at Mills College, a private college for predominantly white, middle-class women in Oakland, California. Those women have been followed ever since, which is what made a yearbook photograph usable as data at all.
Harker and Keltner coded each portrait for the presence or absence of a Duchenne smile, the genuine expression named after a nineteenth-century French physician. As Psychology Today summarises the physiology, a Duchenne smile involves two muscle groups working together: the zygomatic major, which raises the corners of the mouth, and the orbicularis oculi, which raises the cheeks and produces the small creases around the eyes. The zygomatic major can be moved on command. The orbicularis oculi mostly cannot.
That is why a posed smile tends to look flat around the eyes and why a genuine one crinkles them. The two expressions are produced by different parts of the brain. The posed smile is a motor-cortex action, something a person does; the genuine one is limbic, something that happens to them when they feel something.
The women were then reassessed at ages 27, 43 and 52. Those whose yearbook photographs had shown Duchenne smiles reported, on average, more stable marriages and higher wellbeing three decades later than those whose smiles were posed or absent.
Why the finding is easy to misread
The tempting reading is that smiling more, or more convincingly, produces a better life. This is not what the paper supports.
What the study shows is a correlation between an expression captured in one moment at around age 21 and self-reported outcomes decades later, within a specific cohort: women who attended a private liberal arts college in California at the end of the 1950s. That is a narrow sample, in a particular era, from a particular social class, measured on outcomes that were themselves partly defined by the norms of that era. Marriage stability meant something specific in 1960s and 1970s America that it does not mean everywhere or now.
It is also worth knowing that the pattern has not travelled well. In 2007, Jeremy Freese, Sheri Meland and William Irwin ran the same test on the Wisconsin Longitudinal Study, a far larger and more socially varied sample, using high school yearbook photographs. Writing in the Journal of Research in Personality, they reported that despite the apparent advantages of their design, they were largely unable to reproduce the earlier findings. That does not erase the Mills result. It does place a firm ceiling on how far it should be stretched.
The finding also does not mean that women who smiled with their eyes were happier because they smiled. It is far more likely, as the authors themselves argue, that a Duchenne smile at 21 was a small readable trace of an underlying disposition: warmth, ease around other people, a tendency toward positive emotion that would go on to shape thousands of subsequent interactions. The photograph did not cause the marriage. It reflected something about the person that the marriage would also encounter.
This distinction matters because the study is often repackaged as advice, as if practising a better smile in the mirror might change the trajectory of a life. The paper offers no such claim.
What a face gives away in a fraction of a second
What is more interesting is what the finding suggests about the readability of character. A yearbook photograph is a highly constructed image. The subject knows they are being photographed. They have chosen their clothes, their pose, their expression. They are performing a version of themselves for an audience that will include, eventually, everyone they went to college with. And yet, within that heavily managed frame, something involuntary still gets through. The orbicularis oculi either fires or it does not. The eyes either crinkle or they do not. Thirty years of behaviour was, in some small statistical sense, already visible.
This connects to a wider pattern in how people read each other. Attraction, warmth, competence and trust are all judgements formed extremely quickly and revised more slowly than most of us like to admit. Attraction in particular behaves less like a fixed physical fact than a running judgement, one that shifts as a person learns more about the face in front of them. The Mills finding sits at the other end of the same story: even before anyone learns anything, a face is already offering information, and some of that information is more reliable than one might expect.
Why the effect probably lives in relationships
If the correlation is taken seriously without being overread, the mechanism is almost certainly social. People who experience more positive emotion in ordinary interactions tend to elicit warmer responses. Warmer responses reinforce the underlying disposition. Over decades, small differences in how a person is met by others compound into large differences in the quality of friendships, work relationships and marriages.
In the Mills cohort, the women with genuine smiles were not necessarily luckier. They were, on the available evidence, more likely to be met with the kind of response that made partnership easier. That is a subtle claim, and it lives at the level of averages across the sample, not at the level of any individual life.
It also tells us very little about the women whose smiles were posed. A posed smile at 21 does not predict unhappiness. It predicts, in this dataset, slightly less of a particular kind of self-reported wellbeing in later life. Plenty of women in the posed-smile group presumably had rich, stable, contented lives. The averages are averages.
The limits of the photograph
There is one more reading worth naming. A yearbook portrait sits at a very specific moment in a person’s biography: young enough that the face has not yet accumulated much history, old enough that a habitual way of being in the world is already forming. It is a strange kind of evidence. A person’s expression in a photograph reflects who they are, but also what they were told about being photographed, what the photographer said to make them relax, what they thought the picture was for.
There is a familiar pattern in the way some people learn to make themselves smaller in front of a camera, and it is rarely simple modesty. The Mills study points at something similar from the opposite side: the women who could let their face relax into a genuine smile in a formal yearbook context were, on the available evidence, women who had learned that being seen was safe. That is not a personality trait so much as a piece of accumulated history showing up in the eyes.
What the paper does and does not support
To summarise the reading carefully. Harker and Keltner found that in one cohort of women photographed for their college yearbooks at the end of the 1950s, those coded as showing a Duchenne smile reported, on average, better marital and personal outcomes over the following decades than those who did not. The distinction between the two kinds of smile is real and reasonably well established in facial-expression research. The predictive relationship is real within this dataset.
What the paper does not show: that a smile causes a good marriage, that faking a better smile would improve anyone’s life, that women with posed smiles were destined for unhappiness, or that the same pattern would replicate in a different generation, culture or population. On that last point, the Wisconsin attempt suggests it may well not.
What it does suggest, quietly, is that the face is a more honest document than we usually treat it as, and that some of what people spend decades trying to hide was already visible at 21, to anyone who knew where to look.
