The Universality of Facial Expressions

In brief

Paul Ekman (born 1934) argued, in studies beginning in the late 1960s, that a small set of basic emotions produce the same facial expressions in every human culture, and that people everywhere can read those expressions in strangers. The claim rests most heavily on fieldwork among the South Fore of Papua New Guinea, a population with almost no exposure to outside media at the time, who matched photographs of facial expressions to short stories at rates far above chance. The theory reshaped psychology, animation, security screening, and popular writing for fifty years, and it has never stopped being contested. Lisa Feldman Barrett and other psychologists argue the underlying data are weaker than the popular six-emotion picture suggests, and a 2019 scientific review found the evidence insufficient to support many of the strongest claims built on Ekman's work.

The full treatment

The problem it answers

Before Ekman, an influential strand of American social science, associated with anthropologists such as Margaret Mead and the kinesicist Ray Birdwhistell, treated facial expression as learned convention, arbitrary and culture-bound like a spoken word. If that view is correct, a smile carries no fixed meaning, and reading a stranger's face across a cultural divide is guesswork. Ekman tested the rival hypothesis, sketched a century earlier by Charles Darwin in The Expression of the Emotions in Man and Animals (1872): that certain expressions are inherited products of evolution and should appear the same way regardless of upbringing.

How the research worked

Working with Wallace Friesen, Ekman first ran recognition studies in which literate adults in the United States, Japan, Brazil, Argentina, and Chile matched photographs of posed expressions to an emotion word or a short story. Agreement ran well above chance in all five, but shared exposure to Western film and photography remained a plausible rival explanation.

What Ekman claimed

Ekman argued for a small set of basic emotions, originally six: anger, disgust, fear, happiness, sadness, and surprise, each carrying its own universal facial signature installed by evolution rather than taught by culture. He later proposed adding contempt, a seventh, on thinner evidence. With Friesen he built the Facial Action Coding System (1978), an anatomical catalogue of every visible facial muscle movement, or "action unit," letting researchers score any expression without reference to emotion words. FACS shaped the animator's model of a believable face at Pixar, whose staff consulted Ekman while developing Inside Out, and it underlies the automated software that later claimed to read emotion from a face in real time.

The key study: the Fore of New Guinea

The decisive test needed a population that could not have absorbed Western images: the South Fore of the Papua New Guinea highlands, studied with the anthropologist and filmmaker E. Richard Sorenson, most of whom had never seen a foreigner or a photograph. Told simple stories such as "his child has died and he feels very sad," they picked the matching face from a set of photographs well above chance, and posed their own faces to fit the stories; those photographs, shown later to American students, were identified correctly too. This fieldwork is the load-bearing evidence for the whole theory: it removed the obvious rival explanation. Critics have since pressed on how isolated the Fore really were, and, more forcefully, on the design of the matching task itself.

Related distinction: expression versus display rule

Ekman's own data forced an early complication onto the theory: the display rule, a culturally specific norm governing when and how an inner expression is shown, muted, masked, or amplified. In research with Friesen, formalized in Friesen's 1972 doctoral dissertation, American and Japanese subjects watched a stressful film alone and showed nearly identical facial reactions on hidden camera. When an authority figure entered partway through, the Japanese subjects masked their distress with a smile far more often than the Americans did. Ekman's answer to the observation that expressions visibly differ across cultures was never that they don't; the underlying program stays constant, and culture governs only the permitted output.

Lineage

The theory descends directly from Charles Darwin's Expression of the Emotions (1872), which treated expressive movements, the sneer, the wide eye of fear, as evolutionary remnants shared with other animals and inherited rather than invented, the same logic natural selection applies to any trait. Ekman's more immediate influence was Silvan Tomkins (1911 to 1991), whose affect theory proposed a small number of innate affects, each with its own facial pattern, and who pushed both Ekman and, independently, Carroll Izard toward cross-cultural testing.

The strongest case for it

The Fore studies solved a genuine methodological problem: how do you rule out shared media as the explanation for shared recognition? A remote highland population in the 1960s is close to a natural experiment, and the core finding, that recognition ran above chance, held up under decades of replication. FACS also proved durable, giving the field a common, emotion-neutral vocabulary for describing any face, still used in developmental psychology, primatology, and computer vision regardless of what a researcher believes about basic emotions. The largest quantitative synthesis of the debate, Hillary Anger Elfenbein and Nalini Ambady's 2002 meta-analysis, found accuracy for the Ekman emotions reliably above chance across cultures, even if people read expressions from their own group somewhat better, an in-group advantage. Above chance and cross-culturally consistent, even if imperfect, is a real signal, not noise.

The strongest case against it

James Russell, in a detailed 1994 review, argued that the classic studies rested on a forced-choice design: hand someone a short menu of emotion words, and even a weak signal will look inflated, because subjects pick the least-bad label rather than freely reporting what they see. When researchers let people describe or sort faces in their own terms instead, cross-cultural agreement drops.

Lisa Feldman Barrett has pressed the critique furthest. In How Emotions Are Made (2017) and earlier papers, she argues emotions are not discrete, hard-wired programs with one dedicated facial output each, but categories the brain constructs moment to moment from physiological arousal and learned conceptual knowledge, closer to how a mind assembles a category like "weed" than to how a reflex fires. A scowl, on this view, does not mean anger the way a fever means infection; it is one variable expression among several an anger episode can produce. Maria Gendron, Debi Roberson, and Barrett tested this among the Himba, a remote pastoralist population in Namibia, in a 2014 study: when participants freely sorted or labeled photographs instead of choosing from a short list of Western emotion words, they did not reliably converge on Ekman's six categories.

The dispute reached a formal reckoning in 2019, when the Association for Psychological Science (often loosely called an APA review, though APA and APS are distinct bodies) commissioned Barrett, Ralph Adolphs, Stacy Marsella, Aleix Martinez, and Seth Pollak to assess the evidence in Psychological Science in the Public Interest. The panel asked three questions: do facial configurations reliably recur with a given emotion, are they specific to that emotion, and does the pattern generalize across cultures? Their answer, on all three, was that the evidence was weak: people do not scowl every time they are angry, a scowl is not produced only by anger, and the mapping between face and feeling shifts with context and culture more than the popular picture allows. The review carried practical stakes, since automated emotion-recognition software built on Ekman's model was already used in hiring, security screening, and marketing.

Ekman and defenders of the classical account have not conceded. They argue the review demanded an unreasonably strict one-to-one mapping between a single pose and an emotion, when the original theory always allowed for blending and suppression through display rules, and that Elfenbein and Ambady's meta-analytic evidence for above-chance recognition still stands.

Where it stands now

Almost no researcher on either side denies that facial movement carries emotional information or that some regularities recur across cultures. The dispute is over how tight, how universal, and how usable for real-world inference that signal is. The applied stakes are sharp: the United States Government Accountability Office reported in 2013 that the Transportation Security Administration's behavior-detection program, built partly on Ekman-derived cues, lacked scientifically validated evidence that officers could flag threats above chance, and firms selling automated facial emotion-recognition software have faced the same objection from Barrett's camp since. The center of scientific gravity has shifted from Ekman's confident six toward a more qualified picture: expression as a real, partly universal, but noisier and more context-dependent signal than the version that reached the public through television, airport security, and animated film.

Test yourself

Next time you catch someone's face and feel certain you know what they are feeling, ask what you are actually reading: the shape of their mouth and brow on its own, or the situation, the tone of the room, what you already expect them to feel. Ekman's critics would ask you to notice how much of the "obvious" emotion you are supplying from context rather than receiving from the muscles themselves.