Through their nuanced ability to reinforce, reassure, and judge, smiles accomplish many tasks in daily interactions. Do male- and female-identifying individuals use these smiles equally? Although American women smile more often than men overall, it is possible that gender differences in smile behavior are more nuanced. For instance, since it is more acceptable for men to convey status, men may produce more smiles with features of dominance smiles than women. Conversely, since women are socialized to perform tasks associated with care and nurturance, women may produce more smiles with features of affiliation than men. We filmed same- and different-gender participant pairs interacting while watching humorous videos with content related to reward, affiliation, and dominance. We extracted all visible smiles produced by participants and quantified their physical features using automated face coding software. Female participants smiled more often when watching affiliation videos and less when watching dominance content, compared to male participants. Women also displayed smiles with more affiliation features than men overall. Furthermore, participants’ smiles when discussing dominance content contained more features characteristic of dominance when they were interacting with an opposite-gender (as opposed to same-gender) partner. This study—the first to examine naturally elicited smiles while discussing reward, affiliation, and dominance content—suggests the relationship between gender and smiling norms is nuanced and depends on the smiler’s communicative intent.
Norms guide how, when, and why we smile (LaFrance, 2011). There exist norms for how much and when to smile as a function of specific professions (e.g., salespeople should smile at customers) and social context (e.g., one should generally suppress smiles at funerals). There are also norms that dictate who should smile more often. For instance, North American women are expected to smile more than men (Hess et al., 2005), and they do tend to smile more (Lafrance & Hecht, 2000). An understanding of the smiling behavior of individuals who identify as female versus male is limited, however, because past work has not examined gender differences in the display of different types of smiles. While smiles are defined by an upward turning of the lips, produced by the activation of the zygomaticus major or smile muscle, which serves to lift the corners of the mouth into the shape of a U, smiles take many forms and can involve the contraction of a variety of other facial muscles (Krumhuber et al., 2021; Rychlowska et al., 2019).
A recent social functional theory of smiles holds that there are at least three basic tasks of social living—rewarding others, signaling non-threat, and conveying dominance—that can be accomplished with smiles that take specific forms beyond the contraction of the smile muscle (J. Martin et al., 2017). Because women and men are also socialized to pursue tasks of reward, affiliation, and dominance with different frequencies and different reward structures, it is likely that they also vary in the degree to which they signal reward, affiliation, and dominance with their smiles. This was the present research question.
Reward, affiliation, and dominance smiles
A social functional account of facial expression classifies such signals according to their effects on the social environment (J. Martin et al., 2017; Rychlowska et al., 2017). Born from the behavioral ecology view (Fridlund, 2014) and related social functional theories of emotion (Keltner & Haidt, 1999), this account conceptualizes facial displays, including smiles, as tools individuals can use to change their social environment favorably. Research using techniques of reverse correlation has identified the physical forms smiles take to accomplish the social tasks of reward, affiliation, and dominance (J. Martin et al., 2017; Niedenthal et al., 2010; Rychlowska et al., 2017). Figure 1 provides some prototypical examples of these three smile types, taken from videos of participants in the present study.
The first proposed social functional smile is the reward smile, sometimes referred to as a “true” smile because it is most strongly associated with positive affect (J. Martin et al., 2017). The reward smile is thought to reinforce desired behavior (Furl et al., 2012; Heerey, 2014) because it is pleasurable to both produce (Kraft & Pressman, 2012) and perceive (Otterbring, 2017). Reward smiles expressed by babies activate dopamine-associated regions of their mother’s brains, which in turn reinforces their caretaking behaviors (Strathearn et al., 2008). The reward smile resembles the “play face” of chimpanzees, a relaxed and wide smile that helps sustain positive, playful interaction (Parr & Waller, 2006). This smile prototypically involves the lip corners pulled up in a symmetrical way, exposing the teeth and often creating “crow’s feet” in the outer corners of each eye.
Affiliation smiles, a second type of smile (J. Martin et al., 2017), invite social interaction by signaling harmless intentions (i.e., safety), acknowledging the recipient, or conveying appeasement (Keltner, 1995). Such smiles are functionally similar to the silent bared teeth (SBT) display performed by low-status primates, which reduces antagonistic behavior in the higher-status recipient (Parr & Waller, 2006; Waal & Luttrell, 1988). Interestingly, the function of SBT display in primates varies based on the social structure of the species. Rhesus macaques and similarly hierarchical species use the display as a gesture of appeasement, whereas stumptail macaques and more egalitarian species frequently use it as a non-threatening greeting (Flack & Waal, 2007). Although some argue that these smiles are manipulative or disingenuous (Heerey, 2014), the social functional approach recognizes that affiliation smiles are not misleading and instead provide an honest signal of non-threat. The smile can be identified by tightly closed lips that are pressed in and upwards (J. Martin et al., 2017)—think of a smile you flash at a passing stranger to acknowledge them and signal your nonthreatening intentions.
Thirdly, dominance smiles are proposed to assert superiority over their recipient in order to increase or maintain social standing (J. Martin et al., 2017). This smile type was first identified by Darwin, who labeled it as the “derisive or sardonic smile” (Darwin, 1916), but also relates to descriptions of the “scheming” smile and the expression of pride (Öhman et al., 2001; Tracy & Robins, 2004). And in everyday language, people use the label “smug smile” or “smirk.” We may see this form of smile in more hierarchically structured environments wherein expressors can assert dominance without provoking further aggression. However, they are also found in everyday interactions—even in egalitarian relationships—and can convey disapproval of the recipient or personal pride (Wood et al., 2017). Despite being a relatively unthreatening expression (compared to, say, a scowl), dominance smiles can cause an increase in heart rate and salivary cortisol in the recipient, relative to reward and affiliation smiles, reflecting the dominance smile’s ability to give indirect negative feedback (J. Martin et al., 2018). Recent work suggests dominance smiles are morphologically related to displays of disgust and contempt and tend to be asymmetric (J. Martin et al., 2017). Asymmetric smiles, in which only one side of the face is activated, can add ambiguity and negativity to the otherwise friendly signal of a smile.
Gender norms influence how and why people smile
The social functional account of smiles posits that although the tasks of reward, affiliation, and dominance are universal, the smiles vary in frequency across cultures and groups, reflecting the social tasks individuals often encounter. An important social shaper of expressive behavior is gender (Wood et al., 2017).
Women smile more often than men (Hall, 1984; Hall & Halberstadt, 1986; Henley, 1977; LaFrance & Hecht, 1999), and this difference arises in children as young as eleven years old (Wondergem & Friedlmeier, 2012). Gender differences in smiling partly reflect general emotional display rules, which encourage women to be more emotionally expressive than men (Brody & Hall, 2008). Smiling also reflects gender differences in which states and intentions are appropriate to express: in particular, women are expected to support and affiliate with others, for instance by expressing warmth, happiness, and embarrassment (Keltner, 1995; Smith et al., 2015). In addition to potentially explaining why women smile more often than men, it also gives us some indication of the forms of smiles women may be more likely to express. More specifically, due to their pronounced affiliation role, women may be more likely to produce smiles that involve affiliative features, particularly when the situation calls for it.
Men, on the other hand, are encouraged to display powerful states like anger and pride and to dampen all other expressions of emotion (Fischer, 2001), which may influence their motivation behind smiling, as well as the form that it typically takes. Notably, men have been found to smile to assert dominance more than women. Ansfield (2007) had groups of individuals watch amusing and distressing films and found that men smiled more often than women only when watching distressing films, despite reporting more distress. Ansfield reasoned that, just as women typically smile more than men to conform to gender norms, men smiled more in order to mask expressions of sadness and to project masculinity by “laughing in the face of danger”. These findings support the notion that although women generally smile more than men, there may be situations in which men may be more likely to smile – such as when they are moved to signal dominance. Given Ansfield’s interpretation, it is also possible that men in their study not only smiled more often than women, but expressed smiles that involved a stronger activation of features that are associated with signals of dominance.
Men and women in interaction
Facial expression is not only guided by gender norms for the producer but can also be moderated by the gender of the recipient. Prior research reports conflicting findings relevant to whether individuals conform more to gender norms when they are in the presence of individuals of the same gender or different gender. Some findings suggest that individuals perform more normatively for their gender when interacting with members of their own gender (Buhrke & Fuqua, 1987). For example, groups of men tend to establish stable social hierarchies whereas groups of women tend to be more concerned with community and affiliation (Aries, 1976). In a mixed gender group, then, we might expect both men and women to conform to each other and become less gender-stereotypical in their nonverbal behavior. At least in certain contexts, women smile less often, express less verbal affirmation, and act more assertively when interacting with men versus women (Athenstaedt et al., 2004; Lafrance et al., 2003; Maccoby, 1990). And men smile more, express more warmth, and discuss their feelings more often when with women versus men (Aries, 1976; Athenstaedt et al., 2004; Davidson et al., 2016).
But some accounts of inherent power dynamics in gendered interaction suggest that people conform to their gender roles more readily in a mixed gender setting. According to status characteristics theory, indicators of status such as gender, race, education, age, and physical attractiveness inform the behavior of group members by offering those with a culturally privileged status more social power (Berger, 1977). Additionally, expectation states theory argues that gender differences in communication occur mainly where status is relevant (Berger et al., 1980). Therefore, as gender status is more relevant in mixed-gender groups, we might anticipate that individuals conform more to their gender roles when interacting with a person of a different gender. This prediction is bolstered by findings that both men and women act more deferentially towards men (Ellyson et al., 1992; Sayers & Sherblom, 1987) and act more assertively towards women (Ellyson et al., 1992; Killen & Naigles, 1995). Furthermore, women tend to assume a more authoritative communication style when in a leadership role or when participating in more stereotypically feminine tasks (Johnson, 1994; Yamada et al., 1983), which further supports the status characteristics theory, since the shift in power dynamics seems to give women the freedom to act less deferentially and more assertively.
The present work
Given the complexity and limitations of prior findings and theories, it is unclear how gender norms extend to a more refined and nuanced theory of the smile. In the reported study, participants watched and discussed humorous videos with a partner of the same or different gender. The videos were selected because their content was relevant to tasks of reward, affiliation, and dominance, with the aim of eliciting associated smiles. We analyzed both the number of spontaneous smiles and the features present within them using automated facial recognition software, comparing the computer-measured action units to the patterns associated with reward, affiliative, and dominance smiles (Rychlowska et al., 2017).
We predicted that gender norms in social signaling map onto the social functional framework for smiles. Specifically, we expected that women would smile more often than men when viewing affiliation (rather than dominance) content and that men would smile more often than women when viewing dominance (versus affiliation) content. We also expected that women would show more pronounced features of the affiliation smile when smiling and that men would show more pronounced features of the dominance smile when smiling.
We did not have a clear prediction about how smiling behavior changes as a function of same- versus mixed-gender dyads. If male-identifying participants’ smiles contain more pronounced dominance features and female-identifying participants’ smiles contain more pronounced affiliation features in same- versus mixed-gender pairs, this would support the notion that gender dynamics are particularly salient within groups composed of the same gender (Aries, 1976). However, if male-identifying participants’ smiles instead contain fewer dominance features and female-identifying participants contain more dominance within mixed- vs. same-gender pairs, this would provide support for the idea that power dynamics guide the use of gendered expression (Berger et al., 1980).
Method
We employed a video-viewing paradigm designed and utilized in a related study on laughter (Wood, 2020). Pairs of participants sat opposite one other, each in front of their own computer screen, and watched three sets of humorous videos. The sets were composed of 20 YouTube videos previously selected to elicit feelings associated with reward, affiliation, or dominance. The participants first watched the videos alone and selected ones which, to them, best fit the description for each category. They were then filmed interacting with their partner while watching the selected videos. Figure 2 provides a bird’s eye view of the experimental set-up. We used Qualtrics software (Qualtrics, Provo, UT) for all video presentation and survey questions.
Participants
Participants were 122 undergraduate students at a large midwestern university enrolled in Introductory Psychology, with a relatively equal number of women (N = 62) and men (N = 60). Sixty-five percent of the participants identified as White, 24% as Asian, 5% as Black or African American, and 8% as Hispanic or Latino/Latina. The mean age was 18.7 years. Students participated in exchange for extra credit. Pairs of previously unacquainted participants completed the study together. Participants were paired based on their availability, resulting in quasi-random same- and different-gender pairs. Thus, the study had three between-dyad conditions: all women, all men, and mixed gender. We advertised the study as three different studies: one calling for women, another for men, and the third with the gender unspecified. Although we aimed to get mixed-gender groups from the third posting, if both participants identified as men or women, we categorized them as such. Participants without study partners completed a separate task not reported here and were not included in the final sample.
YouTube Clip Stimuli
We used 120 clips that had previously been compiled and tested in prior work for use with the same paradigm (Wood, 2020). The videos had an average length of 9.9 seconds. Fifty of the clips came from an emotion-eliciting stimulus database (Cowen & Keltner, 2017), and the remainder came from YouTube. Clips had to be understandable and humorous without audio. The clips were selected to be relevant to the expression of reward, affiliation, or dominance (40 of each). Reward videos were humorous in a straightforward sense, affiliation videos communicated care and empathy, and dominance videos involved ridicule or a feeling that the person or people got “what they deserved”. Henceforth we refer to reward videos, affiliation videos, and dominance videos to refer to the experience of co-watching the relevant content as described below.
Procedure
Participants arrived at the laboratory in pairs and were escorted to a laboratory room with two computers positioned back-to-back, which allowed participants to face each other and view each other’s expressions throughout the study. After the participants took a seat opposite one another, the experimenter began by telling them that, “we are interested in how people understand and talk about humor”. They then left the participants to perform an icebreaker activity for seven minutes. The activity was designed to be relaxing and lighthearted, affording the participants an opportunity to get familiar with one another to maximize future social interaction and expressiveness. Specifically, after sharing their names, hometowns, and majors, participants took turns answering questions provided to them on a sheet of paper. Questions included, “What is your favorite line from any movie?” and, “What would be the worst ‘buy one get one free’ sale of all time?”
When seven minutes had elapsed, the experimenter re-entered the room and informed the participants that they would be “watching 3 playlists that each contain 20 very short videos…In each playlist, you’ll be asked to pick at least 5 videos (out of 20) that fit a description”. Experimenters informed participants that the videos selected would later be shared with their dyad partner. After watching each video, participants answered ‘true’ or ‘false’ to one of the following prompts:
I feel happiness, amusement, or joy towards the person/object in this video. (Reward)
I feel warmth, friendliness, or care towards the person/object in this video. (Affiliation)
I feel derision, disdain, or mocking towards the person/object in this video. (Dominance)
After viewing the videos and making their responses, participants then (re-)watched their 30 selected videos together (five videos per social function per participant). If participants selected more than five videos for a specific social function, the software randomly picked five of them. Note that this procedure further ensured that the videos discussed by the dyads indeed had reward, affiliation, and dominance-related content as perceived by the participants. This procedure validated the 120 videos, which were originally chosen by the research team, and simultaneously accounted for individual differences in what people find funny, warm, or worthy of derision.
A KVM switch was used to display the same content on both monitors, allowing participants to view clips simultaneously while still positioned opposite one another so that they could see each other’s faces. Participants’ faces were video- and audio-recorded during this section of the experiment with cameras mounted on the screens. After activating the cameras, the experimenter left the room and participants watched the videos in three blocks, one per social function. At the start of each video, participants audibly reported that video’s identification number, which was specified on the screen. While watching the videos, participants freely conversed, describing why they picked each video, how they thought it fit the associated social task, and/or other reactions to the video. Once finished, participants reported demographics, their English-speaking proficiency, whether they watched the videos in their entirety, and whether they were acquainted with their partner. They were then debriefed and thanked for their time. The entire study procedure took approximately an hour to complete.
Smile Measurement
A research assistant watched the recordings of the participants during the joint video-viewing phase and identified every smile that the participants expressed. Three other research assistants, who were blind to experimental hypotheses, verified the presence of a smile. Expressions that the three research assistants unanimously identified as smiles were then extracted as video clips, creating a total of 3,316 videos. We estimated the facial actions present in the smiles using the Computer Expression Recognition Toolbox (CERT; Littlewort et al., 2011). CERT is a software tool trained to process the intensity of 19 different facial actions from the Facial Action Unit Coding System or FACS (Ekman & Friesen, 1978), as well as nine facial expressions, with 90% accuracy.
For each smile instance, we calculated the average activation of facial action units (AUs) or features associated with reward, affiliation, or dominance (Rychlowska et al., 2017), which are described in Table 1: AU 1 (inner brow raiser), AU 2 (outer brow raiser), AU 9 (nose wrinkler), AU 10 (upper lip raiser), AU 12 (lip corner puller), AU 12 asymmetry (the absolute difference in smile intensity for each side of the face), AU 20 (lip stretcher), and AU 24 (lip pressor). Both affiliation and reward smiles involve the symmetrical activation of AU 12, however, affiliation smiles are also comprised of AU 1, AU 2, and AU 24 (Shelde & Hertz, 1994). Dominance smiles, however, involve the asymmetrical activation of AU 12, as well as activation of AU 9 and AU 10.
| Action Unit | Facial Muscle Movement | Smile Type |
|---|---|---|
| AU1 | Inner brow raise | Affiliation |
| AU2 | Outer brow raise | Affiliation |
| AU9 | Nose wrinkle | Dominance |
| AU10 | Upper lip raise | Dominance |
| AU12 | Lip corner pull | Reward/ Affiliation |
| AU12 asymmetry | Asymmetry in lip corner pull | Dominance |
| AU20 | Lip stretch | Reward/ Affiliation |
| AU24 | Lip press | Affiliation |
| Action Unit | Facial Muscle Movement | Smile Type |
|---|---|---|
| AU1 | Inner brow raise | Affiliation |
| AU2 | Outer brow raise | Affiliation |
| AU9 | Nose wrinkle | Dominance |
| AU10 | Upper lip raise | Dominance |
| AU12 | Lip corner pull | Reward/ Affiliation |
| AU12 asymmetry | Asymmetry in lip corner pull | Dominance |
| AU20 | Lip stretch | Reward/ Affiliation |
| AU24 | Lip press | Affiliation |
Results
All analyses were conducted in the R environment using the lme4 package. Complete analysis code and output are available in the Online Resource. We conducted a post hoc power analysis using the pwr library in R (Champely et al., 2017) and determined we had the power to detect a minimal effect size of Cohen’s f = 0.15, which represents a “small-to-medium” effect.
Frequency of smiles
First, we regressed smile counts on dummy-coded participant gender, dummy-coded partner gender, dummy-coded video context variable, the three-way interaction, and all two-way interactions. We then releveled the variables to examine the simple effects. Because our outcome is a within-subject count variable, we used a generalized linear mixed-effects model with a Poisson distribution and nested participants within dyads with a by-participant random intercept. Table 2 includes means and standard deviations of smile frequency by gender, partner gender, and condition. Smile frequency can also be examined by looking at Figure 3. Additional analyses and all model outputs are reported in supplementary materials.
| Participant Gender | |||||||
|---|---|---|---|---|---|---|---|
| Female | Male | Total | |||||
| Condition | Partner Gender | Mean | SD | Mean | SD | Mean | SD |
| Affiliation | Female | 11.23 | 5.27 | 8.05 | 7.91 | 10.27 | 6.29 |
| Male | 9.63 | 7.16 | 7.78 | 3.57 | 8.42 | 5.11 | |
| Dominance | Female | 8.93 | 4.67 | 10.11 | 9.98 | 9.29 | 6.66 |
| Male | 8.05 | 5.13 | 7.50 | 4.49 | 7.69 | 4.68 | |
| Reward | Female | 10.52 | 4.54 | 11.63 | 8.93 | 10.86 | 6.14 |
| Male | 10.84 | 6.90 | 8.80 | 3.34 | 9.52 | 4.93 | |
| Total | 10.01 | 5.40 | 8.68 | 6.16 | 9.39 | 5.80 | |
| Participant Gender | |||||||
|---|---|---|---|---|---|---|---|
| Female | Male | Total | |||||
| Condition | Partner Gender | Mean | SD | Mean | SD | Mean | SD |
| Affiliation | Female | 11.23 | 5.27 | 8.05 | 7.91 | 10.27 | 6.29 |
| Male | 9.63 | 7.16 | 7.78 | 3.57 | 8.42 | 5.11 | |
| Dominance | Female | 8.93 | 4.67 | 10.11 | 9.98 | 9.29 | 6.66 |
| Male | 8.05 | 5.13 | 7.50 | 4.49 | 7.69 | 4.68 | |
| Reward | Female | 10.52 | 4.54 | 11.63 | 8.93 | 10.86 | 6.14 |
| Male | 10.84 | 6.90 | 8.80 | 3.34 | 9.52 | 4.93 | |
| Total | 10.01 | 5.40 | 8.68 | 6.16 | 9.39 | 5.80 | |
We predicted, first, that female participants would smile more often when discussing affiliation compared to dominance videos. Female participants with a female partner did smile more often when discussing affiliation (M = 11.23, SD = 5.27) vs. dominance videos (M = 8.93, SD = 4.67), b = -.23, SE = .07, z = -3.40, p \< .001, 95% CI [-.36, -.10]. However, female participants with a male partner smiled an equal amount between affiliation (M = 9.63, SD = 7.16) and dominance videos (M = 8.05, SD = 5.13), b = .18, SE = .11, z = -1.64, p = .100, 95% CI [-.40, .04] In accordance with our hypotheses, women did smile more often when viewing affiliation rather than dominance videos, but only when their partner was also a woman.
We also hypothesized that male participants would smile more often when discussing dominance rather than affiliation videos. Male participants with female partners smiled more often to dominance videos (M = 10.11, SD = 9.98) than to affiliation videos (M = 8.05, SD = 7.91), b = .23, SE = 0.11, z = 2.11, p = .035, 95% CI [.02, .44]. However, male participants with a male partner were equally likely to smile while watching dominance (M = 7.50, SD = 4.49) and affiliation videos (M = 7.78, SD = 3.57), b = .04, SE = .09, z = .43, p = .668, 95% CI [-.20, .13]. As predicted, men did smile more often when viewing dominance videos than when viewing affiliation videos, but only when they watched these videos with a woman.
Facial action units involved in women’s and men’s smiles
Next, we regressed the mean activation of each relevant facial action units (AUs) for each smile on the three-way interaction between participant gender, partner gender, and video type, along with all lower-order effects. Again, we nested participants within dyad and included a by-participant random intercept. Since AUs are continuous variables, we used general linear mixed-effects modeling. For each of the eight AU outcome variables, we releveled the three dummy-coded predictor variables to get estimates for all simple effects. Because we examined so many AU outcome variables, we controlled the false discovery rate within each family of participant gender, partner gender, and video context comparisons using the Benjamini-Hochberg procedure (Benjamini et al., 2001) and report adjusted p values. For the sake of space, we report only effects that were statistically significant after we applied the Benjamini-Hochberg procedure. See supplementary materials for complete model estimates.
We predicted that female participant smiles would exhibit more pronounced features of the affiliation smile. These facial action units include outer brow raise (AU2), lip stretch (AU20), lip corner pull (AU12), and lip pressor (AU24).
Female participant smiles had more outer brow raise (AU 2) than male participants. Several simple effects indicate that female participants expressed more outer brow raise, which is a feature of affiliation smiles, than male participants while discussing all three video types. As hypothesized, female participants with a female partner (M = -0.21, SD = .34) showed more outer brow raise than male participants with a female partner (M = -0.52, SD = 0.25) while watching reward videos, t(86.30) = 3.42, adj. p = .008, 95% CI [.16, .59]. Likewise, female participants with a male partner (M = -.10, SD = 0.50) showed more outer brow raise than male participants with a male partner (M = - 0.45, SD = 0.30) while watching reward videos, t(79.54) = 3.22, adj. p = .014, 95% CI [.15, .58].
Female participants with a female partner (M = -0.20, SD = 0.39) showed more outer brow raise to affiliation videos than male participants with female partners (M = -0.49, SD = 0.24), t(95.36)= 3.79, adj. p = .003, 95% CI [.21, .65]. Male participants with a male partner had less outer brow raise to affiliation videos (M = -0.47, SD = .38) than did female participants with a male partner (M = -0.06, SD = .38), t(80.99) = -2.98, adj. p = .004, 95% CI [-.56, -.12]. And male participants with a male partner showed less outer brow raise to dominance video (M = -0.44, SD = 0.37) than did female participants with a male partner (M = -0.10, SD = 0.38), t(83.98) = -2.82, adj. p = .041, 95% CI [-.55, -.10]. Across content types, female participants’ smiles tended to involve more outer brow raise, which is associated with the communication of affiliation. This suggests women conveyed more appeasement than men, regardless of the meaning of the social interaction.
Female participants expressed more lip stretch (AU 20) than male participants. Several simple effects provide evidence that compared to male participants, female participants expressed more lip stretch, a feature of both reward and affiliation smiles, when discussing reward and dominance videos. Male participants with a male partner had less lip stretch to reward videos (M = 1.57, SD = 0.29) than female participants with a male partner (M = 2.00, SD = 0.47), t(81.44) = -3.12, adj. p = .014, 95% CI [-.55, -.13]. Male participants with a female partner had less of a lip stretch to reward videos (M = 1.61, SD = 0.35) than female participants with a female partner (M = 1.85, SD = 0.41), t(93.56) = -3.09, adj. p = .003, 95% CI [-.55, -.13].
A simple effect shows that male participants with female partners had less lip stretch, a feature of both affiliation and reward smiles, than female participants with female partners. Male participants with female partners had less lip stretch to dominance videos (M = 1.69, SD = 0.37) than female participants with female partners (M = 1.88, SD = 0.41), t(99.14) = -2.84, adj. p = .029, 95% CI [-.53, -.10]. This finding shows that while viewing dominance videos, men displayed reduced evidence of reward and affiliation smiles. Regardless of partner gender, female participants expressed more lip stretch, a feature of both reward and affiliation smiles, than male participants while watching reward and dominance videos. This finding shows that women expressed more smiles—specifically those that support, appease, and reward—than men.
Male and female participants demonstrated an equal activation of lip pressor (AU 24). Male and female participants with a female partner were equally likely to exhibit lip pressing (AU 24) regardless of whether they were watching affiliation, reward or dominance videos, t(100.88) = .85, adj. p = .624, 95% CI [-.03, .07]; t(89.11) = .95, adj. p = .567, 95% CI [-.03, .07]; t(92.60) = -.63, adj. p = .739, 95% CI [-.07, .03], respectively. Male and female participants with a male partner were equally likely to exhibit lip pressing regardless of whether they were watching affiliation, reward, or dominance videos, t(82.29) = -.97, adj. p = .578, 95% CI [-.08, .03]; t(80.41) = -1.10, adj. p = .510, 95% CI [-.08, .02]; t(86.15) = -.85, adj. p = .639, 95% CI [-.07, .03], respectively.
Male and female participants expressed equal lip corner pull (AU 12). Male and female participants with a female partner expressed equal levels of lip corner pull, regardless of whether they were viewing affiliation, reward, or dominance content, t(118.52) = -1.59, adj. p = .298, 95% CI [-.67, .07]; t(99.83) = -2.35, adj. p = .083, 95% CI [-.77, -.07]; t(105.88) = -0.22, adj. p = .830, 95% CI [-.40, .32], respectively. Male and female participants with a male partner expressed equal levels of lip corner pull, regardless of whether they were viewing affiliation, reward, or dominance content, t(89.69) = -1.21, adj. p = .46, 95% CI [-.58, .13]; t(86.75) = -.01, adj. p = .99; t(95.87) = -.03, adj. p = .98, 95% CI [-.36, .35].
Male participants exhibited more asymmetrical lip corner pull (AU 12 lateralization) than female participants when watching dominance videos with a female partner. We also predicted that male participants would exhibit more pronounced features of the dominance smile than female participants. These facial action units include asymmetric lip corner pull (AU 12 lateralization), nose wrinkle (AU 9), and upper lip raise (AU 10).
Male participants with a female partner exhibited more asymmetrical lip corner pull (AU 12 lateralization) than female participants with a female partner (M = .21, SD = .18) when watching dominance videos, t(178.92) = 2.61, adj. p = .041, 95% CI [.04, .26]. However, male participants with a female partner did not exhibit more asymmetrical lip corner pull than female participants with a female partner when watching affiliation, t(205.67) = 1.74, adj. p = .277, 95% CI [-.01, .22], or reward videos, t(161.14) = -.52, adj. p = .796, 95% CI [-.14, .08]. Also, male participants with a male partner did not express more asymmetrical lip corner pull than female participants with a male partner when watching affiliation, reward, or dominance videos, t(136.83) = -1.35, adj. p = .419, 95% CI [-.18, .03]; t(180.83) = 1.12, adj. p = .506, 95% CI [-.04, .17]; t(152.97) = -2.44, adj. p = .070, 95% CI [-.25, -.03].
Male and female participants exhibited an equal activation of nose wrinkle (AU 9). Male and female participants with a female partner were equally likely to exhibit nose wrinkling (AU9) when smiling, regardless of whether they were watching affiliation, reward or dominance videos, t(90.31) = -1.70, adj. p = .286, 95% CI [-.08, .01]; t(84.52) = -1.15, adj. p = .487, 95% CI [-.08, .02]. t(86.09) = -1.27, adj. p = .400, 95% CI [-.07, .01] respectively. Male and female participants with a male partner were equally likely to exhibit nose wrinkling, regardless of whether they were watching affiliation, reward, or dominance videos, t(80.68) = -.07, adj. p = .954, 95% CI [-.05, .04]; t(79.75) = -.04, adj. p = .989, 95% CI [-.04, .04]; t(82.60) = -.46, adj. p = .829, 95% CI [-.05, .03].
Male and female participants exhibited an equal activation of upper lip raise (AU 10). Male and female participants with a female partner were equally likely to exhibit upper lip raises (AU 10) when smiling, regardless of whether they were watching affiliation, reward, or dominance videos, t(102.33) = .24, adj. p = .914, 95% CI [-.06, .08]; t(88.50) = 1.32, adj. p = .418, 95% CI [-.02, .11]; t(92.73) = .49, adj. p = .788, 95% CI [-.05, .08], respectively. Male and female participants with a male partner were equally likely to exhibit upper lip raises (AU 12) when smiling, regardless of whether they were watching affiliation, reward, or dominance videos, t(80.97) = -1.63, adj. p = .333, 95% CI [-.12, .01]; t(78.76) = -1.18, adj. p = .490, 95% CI [-.10, .03]; t(85.51) = -.35, adj. p = .880, 95% CI [-.08, .05], respectively.
In accordance with our hypotheses, we did find that women exhibited more outer brow raise, a feature of affiliation smiles, when smiling than men. Also supporting our hypotheses, women expressed more lip stretch, a feature of both reward and affiliation smiles, than men. Contrary to our hypotheses, however, men and women exhibited an equal amount of lip pressing, a feature of affiliation smiles, when they smiled
In support of our hypotheses, men exhibited more asymmetrical lip corner pull than women, but only when they were watching dominance videos with another woman. Also contrary to our hypotheses, men did not exhibit more nose wrinkle or upper lip activation – features of the dominance smile – than women when smiling.
Discussion
In this study, we examined gender differences in smiling. Specifically, we evaluated the frequency and form of the smiles of men and women as they discussed videos relevant to themes of reward, affiliation, and dominance in same- and cross-gender dyads. The smile frequency results largely supported our hypotheses concerning the gendered expression of signs of affiliation and dominance when the interaction partner was female: Female participants smiled more often while watching affiliation videos than dominance videos when their partner was female. Male participants smiled more often while watching dominance videos than affiliation videos when their partner was female. In contrast with our hypotheses, however, participants with a male partner smiled to a similar extent across video types (note, of course, that all null results should be interpreted with caution, as the absence of an effect may be due to a lack of power, reflecting a false negative).
This pattern of findings contrasts with previous literature, which proposes that gender norms are particularly salient within same-gender groups (Aries, 1976), or that the use of facial expressions is guided by power dynamics (Berger et al., 1980), which would be indicated by an increased use of dominance when interacting with a male partner. It is possible, however, that men and women are more likely to act according to their gender roles when they are around women. Historically, mothers in the U.S. have been tasked with the role of enforcing gender roles for their children (Ryan, 1983), and some recent work supports the claim that mothers – not fathers – today follow gender role norms for emotion expression (Thomassin & Seddon, 2019). Maternal behavior also more strongly predicts gendered expression in their children than paternal behavior (Hastings et al., 2007). It may be that, due to the pronounced acquisition of gendered knowledge from their mothers, especially about emotion, young adults are more likely to behave consistently with gender norms when interacting with females.
Our facial action unit findings depict a more complex pattern. Overall, the finding that females used more outer brow raise and lip stretch in their smiles, and smiled more often while discussing affiliation videos than males, provides support for our hypothesis concerning gender differences in displays of affiliation. These facial action units were previously found to be a feature of affiliation smiles, suggesting that female participants smile with more affiliation than male participants. The outer brow raise, or eyebrow “flash,” is especially common in expressions directed at babies, and makes the producer seem more approachable (Shelde & Hertz, 1994). These findings support the notion that women employ the use of affiliation smiles and features to a larger extent than men.
Unexpectedly, additional analyses (reported in Supplementary Materials) found that participant smiles were more asymmetrical while discussing dominance videos with a partner of the opposite gender. This finding is inconsistent with status characteristics theory, which predicts that women use dominance expressions to a larger extent with women than with men (Berger, 1977). At the same time, this finding also contradicts the notion that each gender acts more stereotypically while with members of the same sex (Aries, 1976), as men expressed more smiles which theoretically convey dominance while with a female partner than with a male partner.
As with our smile frequency results, it may be that each gender is more likely to perform their gender role when interacting with a female. Similarly, it could be that this is due to the fact that women have been historically tasked with the role of teaching gender norms (Ryan, 1983), which persists to the present (Hastings et al., 2007; Thomassin & Seddon, 2019).
It is also possible that the experimental paradigm itself may have affected our findings. The nature of this experiment was inherently affiliative, as participants were encouraged to get to know each other through an initial set of icebreaker questions, and interacted over a set of humorous videos. It may be that, through social comparison, each gender experiences more social threat and competition when interacting with someone of the same gender. Social competition theory suggests that competition can elicit two differing strategies: prosocial and antisocial (Gilbert & Basran, 2019). Since our experimental paradigm encouraged prosocial behavior by emphasizing the goal of cooperation and bonding in our participants, participants may have reacted to the social threat of interacting with the same gender by amplifying affiliative behavior. Furthermore, social threat and stress tend to promote prosocial behavior towards in-group members and not out-group members, since it is considered beneficial to affiliate with similar others (Faber & Häusser, 2022). This interpretation would also explain why female participants engaged in more lip pressing, a feature of affiliation, while watching dominance videos with a female partner (reported in Supplementary Materials). If the general nature of this experiment caused them to emphasize affiliation goals with a socially comparable partner, then it would make sense that they engaged with more appeasing expressions while watching videos which instead call for dominance behavior.
It is possible that our participants did not perceive their interaction partner as social competition, but as a potential other to bond with. Notably, although our experimental paradigm promotes affiliation, it was also designed to create a naturalistic social experience. There is, therefore, little reason to believe that the results would not generalize to daily interactions, where norms of affiliation are commonplace.
The female participants’ pattern of asymmetrical smiles may reflect shifting gender norms in the U.S – partially cause by cumulative efforts from gender scholars and queer studies to frame gender and gender norms as a social system, rather than essentialized, biological identities (Cislaghi & Heise, 2020). Notably, our subjects are college students living in a primarily liberal city and are therefore more likely to be influenced by this shift. Perhaps, in a context where gender is salient, our female participants are reacting to their male partners by enhancing their use of dominance tactics in order to combat, whether subconsciously or consciously, the gender power dynamics inherent in society. It is also possible that our female participants amplified dominance characteristics to build better rapport with their male partner.
Some argue that gender does not have a status value, at least in mixed dyadic interactions (Foschi & Lapointe, 2002), so perhaps the mixed-gender asymmetric smiling results should not be interpreted through a status lens. Instead, it may be that the unique situation of a mixed gender pair has an inherently sexual dynamic for many participants. The asymmetrical smile in this context could have functioned as a coy smile, or a flirtatious expression. The majority of our subjects were likely heterosexual or bisexual, as approximately 87% of college students identify as such in the US (Association of American Universities, 2020). This mixed dyad setting may have, therefore, elicited more coy smiles, which were intended, or interpreted, to be flirtatious rather than straightforwardly portraying dominance.
Unexpectedly, additional analyses (reported in Supplementary Materials) found that female participants displayed less lip pressing while watching reward or affiliation videos than while watching dominance videos with a female partner. Although lip pressing is categorized as a feature of an affiliation smile, it may reflect an attempt to inhibit expression rather than explicitly express affiliation. Lip pressing hides the teeth, dampening the otherwise high-arousal smile expression. Taken with the finding that women smiled less often while watching dominance rather than affiliation or reward videos, the tendency for women to engage in more lip pressing while watching dominance videos might be a method of inhibiting expression in a context that violates gender norms.
Individuals tend to conform more to gender expectations when they are more aware of observation, such as in a lab setting (Hyde, 2005; Lafrance et al., 2003). More specifically, women smile more and men less often when the presence of a camera is more salient or obvious, as well as during face-to-face interaction (Hall et al., 2001; Lafrance & Hecht, 2000). Given the fact that our participants were facing each other, were explicitly made aware of being recorded on several occasions, and were in a laboratory environment, we should be careful when generalizing the gender differences found within this context to smiles truly produced “in the wild”.
Notably, we failed to find gender differences in the expression of nose wrinkles and upper lip activation – facial muscle movements that seemed to be minimally expressed across participants and video content. Future work should aim to test these hypotheses further to determine whether the hypotheses are incorrect or whether it reflects a lack of power or an issue with the study design.
Additionally, caution should be taken when interpreting results due to the number of statistical tests performed and performance accuracy of the CERT software. As the nature of our experiment resulted in conducting a total of 1,297 tests, we had a higher likelihood of getting false positive results. Although we applied the Benjamini-Hochberg procedure to control the false discovery rate, some caution should be taken when interpreting results. Furthermore, the CERT software that we used to measure facial muscle movement has an average error rate of 10%, which reduces some confidence in our findings. However, prior work has shown that this software is more accurate than human coding (Bartlett et al., 2014), giving us confidence in our measures.
Future work should examine whether or not the specific conversational content relates to the participants’ facial expressions. It would also be useful to examine whether there are differences in same vs. mixed gender groups depending on if those groups are dyads or a larger group. It may be that larger groups provide a setting for men to participate more in the establishment of a “pecking order”. It is also possible that mixed gender dyads provide a setting that is more inherently sexual than larger mixed gender dyad. Future research should also seek examine possible differences in the use of various smile types among different racial and ethnic minorities. Black men, in particular, often act in a particularly deferential manner to counter negative stereotypes about Black male aggression (Santino, 1983), and may therefore have an increased need to use affiliation smiles.
This experiment adds to growing evidence that the smile serves a variety of functions (J. Martin et al., 2018; J. D. Martin et al., 2021; Rychlowska et al., 2017) and is the first to study the variation of naturally occurring smiles. It is also the first experiment to examine gender differences within different smile types, despite the plethora of studies that have examined gender differences in smile rates overall. Although we have confirmed previous findings that women generally smile more frequently than men, we also found evidence that men and women not only differ in the kind of smiles that they express, but that there are contexts wherein men may be more likely to smile than women. More specifically, men may be more likely to smile while discussing dominance videos and to use dominance features when smiling, but only when interacting with a woman. Ultimately, the present work provides a more nuanced understanding of smiling patterns within the US and how they differ across gender and context.
Contributions
Contributed to conception and design: AW, ZH, PN
Contributed to acquisition of data: ZH
Contributed to analysis and interpretation of data: ZH, JM, AW
Drafted and/or revised the article: ZH, AW, PN, JM
Approved the submitted version for publication: ZH, AW, PN
Data Accessibility Statement
All video stimuli, procedural details, participant data, behavior coding data, analysis scripts and results can be found at this link https://osf.io/xt9fs/.
Competing Interests
The authors declare that they have no competing interests.
University of California Press