Vol. 48, No. 4, 384-392
1992 Award Addresses
Facial Expression and Emotion
Paul Ekman
Cross-cultural research on facial expression and the developments of methods to measure facial expression are briefly summarized. What has been learned about emotion from this work on the face is then elucidated. Four questions about facial expression and emotion are discussed. What information does an expression typically convey? Can there be emotion without facial expression? Can there be a facial expression of emotion without emotion? How do individuals differ in their facial expressions of emotion?
In 1965 when 1 began to study facial expression,1 few thought there was much to be learned. Goldstein (1981) pointed out that a number of famous psychologists—F. and G. Allport, Brunswik, Hull, Lindzey, Maslow, Osgood, Titchner—did only one facial study, which was not what earned them their reputations. Harold Schlosberg was an exception, but he was more interested in how to represent the information derived by those who observed the face than in expression itself.2 The face was considered a meager source of mostly inaccurate, culture-specific, stereotypical information (Bruner & Tagiuri, 1954). That this contradicted what every layman knew made it all the more attractive. Psychology had exposed the falseness of a folk belief, a counterintuitive finding.
The late Silvan Tomkins (1963) was virtually the only contrary voice. He convinced me to extend my studies of nonverbal behavior from body movement to the face, helping me design my initial cross-cultural studies. Tomkins also advised Carroll Izard in the design of similar studies at the same time. He did not tell either of us about the other, which helped the science because it provided independent replications but was an unwelcome surprise when we learned that we had not been alone in our discoveries.
We each found high agreement across members of diverse Western and Eastern literate cultures in selecting emotion terms that fit facial expressions. Izard (1971) added evidence that cross-cultural agreement was preserved for most emotions when subjects were allowed to choose their Own words to describe the feelings shown in the expressions. We (Ekman & Friesen, 1971) extended the findings to a preliterate culture in New Guinea, whose members could not have learned the meaning of expressions from exposure to media depictions of emotion. We also found agreement about which expressions fit with different social situations, such as the death of a child, a fight, and seeing friends.
Friesen and I (Ekman, 1972; Friesen, 1972) also extended the findings of how people interpret expressions to the study of how and when people show expressions. We found evidence of universality in spontaneous expressions and in expressions that were deliberately posed. We postulated display rules—culture-specific prescriptions about who can show which emotions, to whom, and when—to explain how cultural differences may conceal universals in expression, and in an experiment we showed how that could occur.
In the last five years, there have been a few challenges to the evidence of universals, particularly from anthropologists (see review by Lutz & White, 1986). There is, however, no quantitative data to support the claim that expressions are culture specific. The accounts are more anecdotal, without control for the possibility of observer bias and without evidence of interobserver reliability. There have been recent challenges also from psychologists (J. A. Russell, personal communication, June 1992) who study how words are used to judge photographs of facial expression. However, no one to date has obtained strong evidence of cross-cultural disagreement about the interpretation of fear, anger, disgust, sadness, or enjoyment expressions. There is no instance in which 70% or more of the people in one cultural group judged a picture as showing one of these emotions and a similar percentage of the people in another cultural group judged the same expression as showing a different one of these emotions. (See Ekman, 1989, for a review of the evidence on universality; see also Brown, 1991, for an analysis of the wider issues and arguments about universals of any kind.)
This evidence of universality both required and justified nearly a decade of work to develop methods for measuring the movements of the face. We (Ekman & Friesen, 1976, 1978) developed the Facial Action Coding System, which was the first, and still is the only, comprehensive technique for scoring all visually distinctive, observable facial movements. A few years later, Izard (1979) published his own technique for selectively measuring those facial movements that he thought were relevant to emotion. A number of investigators have also measured electrical activity in the facial muscles (see Ekman, 1982, for a review on facial measurement).
The findings on universality, the development of methods to objectively measure the face, and the many studies of facial expression that were done subsequently have taught us not just about facial expression but also about emotion. I will explain nine different contributions that the research on facial expression—in particular, the universals finding—has made to our understanding of emotion. Then I will raise four major questions about facial expression in emotion, some of the possible answers, and directions for research.
1In some of my earliest writing, I avoided the term expression and instead used the more awkward phrase facial behavior to avoid the implication that an inner state is being manifested externally. I have reverted to facial expression because it is more felicitous, although it should he clear that in my view (Ekman, 1977) expression is a central feature of emotion, not simply an outer manifestation of an internal phenomena.
2 Schlosberg told me that to avoid bias he had never looked at the faces he asked his subjects to judge.
1. Study emotion. The most important effect of the evidence on universals in facial expression was to contribute to reawakening interest in emotion. Dormant for many years, research on emotion now is one of the most rapidly growing areas, with activity in clinical, developmental, personality, physiological, and social psychology. Much of the current work rightfully does not focus on the face, although some investigators who focus on other matters use the face as a marker of when an emotion occurs.
2. Consider both nature and nurture. The findings on facial expression also encouraged some of those who became interested to view emotion as a psychobiological phenomenon, influenced by our evolutionary heritage as well as by our current circumstances. Even within the more narrow confines of expression, an explanation of what was found required consideration of the influences of both nature and nurture (Ekman, 1972, 1977, 1992a). Twenty-seven years ago, when the work on universals began, psychology was still focused nearly exclusively on what is learned, considering only nurture and largely ignoring the influence of nature. Although the findings on universals in expression were inconsistent with that frame, they did appeal to another prejudice then fashionable—to credit only that which is palpably observable.
3. Search for emotion-specific physiology. A focus on universals in expression was inconsistent with the then-reigning view that all that differentiates one emotion from another is our expectations about what we should be feeling. Despite failures to replicate Schacter and Singer’s (1962) experiment, flaws in the design of that study, and contrary evidence, it was very influential. All that distinguished one emotion from another, they proclaimed, was cognition about the social setting; physiological activity varied only in the extent not in the nature of the emotional arousal. But once expressions were found to be emotion specific, it made sense to reexamine the issue of whether there might also be emotion-specific physiological changes.
Although the evidence on universals in expressions could not prove that these expressions have evolved, those findings, together with the observation of similarities in some expressions between humans and some other primates, certainly increased the viability of an evolutionary perspective on emotion. Such a perspective would expect that emotion-specific changes in autonomic physiology would have evolved to serve the quite-different adaptations that are likely in emotions such as fear and anger. A new generation of investigators are examining again the possibility of emotion-specific autonomic and central nervous system activity. I have been a collaborator in some of this work, in which we use facial measures to identify when emotions occur. (For a review of current work on the biology of emotion, see Davidson & Cacioppo, 1992.)
4. Specify the events that precede emotions. Most controversial in our study of emotion-specific physiological activity was our discovery (Ekman, Levenson, & Friesen, 1983) that voluntarily making one of the universal facial expressions can generate the physiology and some of the subjective experience of emotion. Of course, making a face is not how emotions usually are brought forth. Emotions typically occur in response to an event, usually a social event, real, remembered, anticipated, or imagined. The findings of both universals and cultural differences in the situations in which facial expressions occur focused attention on the events that call forth emotion. There is now cross-cultural data on what people report are the antecedent events for specific emotions (Boucher, 1983; Scherer, Summerfield, & Wallbott, 1983). Observational data on the antecedents of emotion are much more limited but are growing in studies of early development and in studies of marital interaction. Measures of facial behavior are a central part of those endeavors.
Any close observer of emotional expression must develop an account that allows for both commonalities in the events that call forth an emotion and the enormous individual differences in which events call forth different emotions. Not every event calls forth an emotion, nor does an event call forth the same emotion across individuals, and yet there are some common features. Tomkins’s (1963) proposal that emotional events produce changes in the density of neural firing that parallel features of the event has been regarded skeptically by neuroscientists. Other quite different accounts of how events are appraised is one of the most active current areas of theory and research (see Lazarus, 1991, for a review).
5. Examine ontogeny. It was consistent with an evolutionary account of universals in facial expression to expect that emotions might appear much earlier in infancy than had been previously thought. The tools for measuring the face provided the means for identifying when emotions, or at least expressions, might be occurring. This is another very active area of research, although there is still argument about just when each emotion is first evident (Camras, Malatesta, & Izard, 1991; Izard, Huebner, Risser, McGinnes, & Dougherty, 1980; Oster, Hegley, & Nagel, 1992).
6. Examine more than verbal behavior. The need to measure the face (and voice) is obvious in infancy when speech is not available. However, in the older child and adult it is equally important not to rely only on the more easily obtained questionnaires, on accounts of emotion given when an emotion is not felt, or even on what people say during an emotional episode. This is not to diminish the importance of these sources of information but even what people say when they are in the midst of an emotion may not always reveal what they are actually feeling or thinking, not even what they are aware of feeling or thinking.
My research on deception has shown how convincingly people can misrepresent in their speech the emotions they are feeling. Even though many facial expressions are recruited in a lie, sometimes there is what we termed leakage in facial and vocal expressions of concealed feelings (Ekman, 1985; Ekman & Friesen, 1969; Ekman, Friesen, O’Sullivan, 1988; Ekman, Friesen, & Scherer, 1976). When attention is focused on these often-brief, fragmentary signs of emotional expression, they can betray a lie by contradicting the emotion the person verbally claims to be feeling.
More generally, there is an increasing trend to use multiple measures of emotional response, not only to obtain better reliability and validity but also to understand discrepancies among the different emotional responses and to examine individual differences in the extent of coherence among different emotional responses. Even when focused on expression alone, an investigator is confronted, if not overwhelmed, with the importance of individual differences; this is the last of the four major questions about the emotion process and the face that I will consider.
7. Consider emotions as families. Precise measurement of facial expression suggested a metaphor that may be useful in thinking about emotion. We (Ekman & Friesen, 1978) found not one expression for each emotion, but a variety of related but visually different expressions. The 60 anger expressions, for example, that we have identified share certain core configurational properties, which distinguish them from the family of fear expressions, disgust expressions, and so forth. Variations within a family of facia1 expressions likely reflect the intensity of the emotion, whether the emotion is controlled, whether it is simulated or spontaneous, and the specifics of the event that provoked the emotion.3
Just as it is useful to think of expressions as Constituting families, I have proposed (Ekman, l992a) that we consider each emotion as constituting a family of related affective states, which share commonalities in their expression, physiological activity, and in the types of appraisal that call them forth. These shared characteristics within an emotion family should distinguish one emotion family from another. The anger family, for example, would include variations in intensity stretching from annoyance to rage. It should also include different forms of anger, such as resentment, which is the kind of anger in which there is a sense of grievance; indignation and outrage, which are anger about the mistreatment of someone; vengeance, the anger that retaliates against a misdeed by another; berserk, anger that appears to others to be an uncontrolled response inappropriate to any provocation; and so on.
The characteristics shared by all members of an emotion family constitute the theme for that emotion and are most likely to reflect the contribution of nature. The different members of the family are variations around that theme, reflecting more the influence of nurture and the particulars of the occasion when the emotion occurs. Our common language of emotion words may include many or few descriptions relevant to any of the emotion families. In English, we have many terms for anger, some specifying how the person is behaving (e.g., argumentative, testy, huffy, sulky, spiteful), some that are metaphors (fed up, pissed off), and some referring to changes in physiology (hot, bristling). (See Tomkins, 1981, for a description of how language may incorporate different aspects of an emotion.)
Those studying the lexicon of emotion (Shaver, Schwartz, Kirson, & O’Connor, 1987) have proposed a similar framework, although not using the term family I believe the definitive evidence on what constitutes a family, and in particular the delineation of the theme for each family, will come not from the study of emotion words but from closer examination of appraisal processes, motor responses, and ultimately what is revealed by studies of emotion-specific activity in the central nervous system.
8. Consider emotions to be discrete states. The research on facial expressions has also shown the utility of conceiving of emotions as separate discrete states, such as fear, anger, and disgust, rather than simply as positive versus negative states or even more simply as differing only in respect to arousal. Although some current emotion researchers continue the early (Woodworth & Schlosberg, 1954) conceptualization of emotions in terms of a few dimensions, that approach has not proven as useful in studies that measure facial behavior in early development or social interaction or in many of the studies of physiological changes in emotion.
9. Consider expression in determining how many emotions there are. If our definition of emotion were to require a distinctive expression so that conspecifics can know instantly from a glance how a person is feeling, then we need look only to the evidence on how many emotions have distinctive expressions to determine the number of emotions. Distinctive universal expressions have been identified for anger, fear, disgust, sadness, and enjoyment. Even adding contempt, surprise, and interest, about which the evidence is far less certain, the list of emotions that have a universal facial expression is far shorter than the number of emotions most theorists have proposed, far smaller indeed than the various words for emotion. How are we to deal with this discrepancy?
Perhaps there are emotions that have distinctive vocal expressions but no facial expressions; however, none have been uncovered so far. Grouping emotions into families may provide a better fit between the list of emotions that have an expression and the number of emotions proposed by various theorists. Another part of the answer is suggested by our findings that a number of positive emotions—amusement, relief, pride, sensory pleasure, exhilaration—share but one facial expression, a particular form of smiling (Ekman, l992b). One could argue that these are all members of one emotion family, but I expect that research on appraisal and physiology will show they are distinctive states that share a signal.
The evidence may require that we postulate emotions that do not have a distinctive universal signal—no distinctive facial, vocal, or bodily action that provides information to those who observe it. I will return to this when I discuss the question of whether there can be emotions without expression. There is a prior question, however. My discussion so far has assumed that the information conveyed by an expression is best captured by words such as anger or fear, but is that what most people typically derive when they see a universal facial expression?
3It is the core expressions that have been studied in the cross-cultural judgment studies of facial pictures. We do not know how many of the other expressions for each emotion would be judged in a similar fashion across cultures.
We know virtually nothing about the type of information people typically derive from a facial expression when they see the expression in situ, accompanied as it usually is by speech, gestural, and postural behaviors, and when the person observing the face has the usual array of expectations about what may be most likely to occur in that situation. The studies that determined the information observers obtain from facial expressions when they are seen Out of context—disembodied—answers the question of what the face can signal, not what information it typically does signal.
Consider the messages that might be conveyed by the expression shown in Figure 1, a photograph that I took 25 years ago of a member of a preliterate, visually isolated culture in Papua, New Guinea. The message conveyed may be about an antecedent event that led to the expression, for example, “someone must have insulted her.” Or the inference drawn may be about what the person is feeling or thinking at that moment; for example, “she must feel very tense” or “she must be planning how to get revenge.” The observer may interpret the expression in terms of what the person is likely to do next, such as “she’s going to hit me.” Still another possible message would refer to an emotional state, using a metaphor such as “she is boiling.” Or, the message could be an emotion word, either a specific one, such as “she is mad,” or a more general one, such as “she doesn’t feel good.” (See Ekman, 1977, for a more complete account of the different messages provided by an expression.)
Scene of Villagers’ Response to on Outsider in the Highlands of New Guinea, 1967
Note. From Face of Man: Universal Expression in a New Guinea Village (p. 34, plate 17) by Paul Ekman, 1980, New York: Garland. Copyright 1980 by Paul Ekman. Reprinted by permission.
I expect that we could find better-than-chance agreement within a cultural group about each of these emotion-related messages—antecedents, simultaneous behaviors, metaphors, and consequent events—just as we have found agreement about specific emotion terms. Lakoff (1987) found similar emotion metaphors in English and Hungarian, but they only examined anger. The question remains as to how much cross-cultural agreement there might be about each type of message for each emotion. It is also not known which type of message participants in a social interaction typically derive and whether this varies with the social context in which the expression occurs, the demographic characteristics of the expresser and the observer, or the personality of these individuals.
If a language has no words for an emotion, as has been reported by some anthropologists (Lutz & Abu-Lughod, 1990), it does not mean that the emotion does not occur in that culture, only that it is not represented by single terms in the lexicon. Levy (1984) argued that although the Tahitians have no word for sadness, he saw sad expressions in people who had experienced a loss. Unfortunately, Levy did not determine whether the Tahitians would have selected a sad expression if he had asked them to identify which face was that of a person who had experienced some loss, such as their child dying. Such studies have not been done in any of the language groups that, reportedly, do not have single terms for some emotions.
We do not know how salient facial expressions are when they contradict what a person is saying or what the observers believe to be normative in a particular situation. One could equally well argue that expressions will be ignored, overwhelmed by other sources of information, or just the opposite, that expressions will stand out because of contrast noteworthy in such circumstances. Probably both will be found to occur, depending on the emotion, the situation, and the characteristics of the observer and the expresser.
“Can there be emotion without facial expression” is really two questions. First, considering just those emotions for which universal expressions have been identified, (e.g., fear and anger), do those emotions occur without any semblance of the expression? And second, are there still other emotions that have no distinctive expression, at all, ever? I think the answer to both questions is yes, but the evidence is fragmentary.
...
Bajlif