Sexism in teams: Exposure to sexist comments increases emotional synchrony but eliminates its benefits for team performance

Significance

Sexism, defined as acts of prejudice, stereotyping, discrimination, or harassment, primarily targeting women, is a pervasive phenomenon that has been shown to have detrimental effects on individuals. Extending this research to the team level, we argue that sexism would undermine the effects of emotional synchrony—a key mechanism underpinning team functioning—on team performance. The results revealed that exposure to sexist comments heightened synchrony in team members’ facial expressions (a form of emotional synchrony), yet this synchrony failed to facilitate team performance. Overall, this research demonstrates that by sabotaging the effects of synchrony on team outcomes, sexism disrupts the role of emotional synchrony as the social glue of collective action.

Abstract

In a world where teams serve as the backbone of collaboration and innovation, women must feel safe when contributing to teamwork. Unfortunately, an increasing number of women report experiencing sexual harassment in workplaces and other collective settings. This research included an examination of how exposure to sexist insults affects collaborative efforts. We argue that in addition to its well-documented effects on individuals, sexism within teams undermines team performance. Hence, emotional synchrony—temporally coordinated emotional facial expressions between individuals—loses its ability to enhance collaboration when team members are exposed to sexism. Under the threat of sexist comments, emotional synchrony signals social bonding rather than focusing team members on performance goals. To test this theory, 177 woman dyads interacting on a video-conferencing platform received instructions for a cooperative task with/without sexist comments from an actor-experimenter in sexism/control conditions. Emotional synchrony was assessed through temporal alignment in facial expressions between dyad members and was correlated with team performance. Our findings revealed a significant increase in facial expressive synchrony among teams in the sexism condition. Whereas facial expressive synchrony predicted better performance in the control condition, its classic positive effects on team performance vanished under sexism. These results suggest that exposure to sexism, while enhancing social cohesion, eliminates the benefits of emotional synchrony, which is considered the social glue of collective action. These findings suggest that zero-tolerance policies for sexual harassment are not only more ethical but can also promote effective teamwork.

Does sexism sabotage team performance, and if so, how? To address this question, we examined the impact of sexist comments made by an authority figure on the performance of women in a dyadic cognitive task and proposed that sexism eliminates the typical positive effects of emotional synchrony (i.e., synchronized facial expressions) on team performance. Specifically, whereas emotional synchrony is generally key to effective collective functioning (1), we argue that its effects subside when team members are exposed to sexism. Whereas previous research has focused on the detrimental effects of sexism on individual women (2), the present study extends this research to examine the effects of sexism on interpersonal dynamics. Thus, one important contribution of the present study is the explanation it provides for the adverse effects of sexism on collective functioning.

Sexism, often in the form of derogatory gender-based stereotypical comments, sexist humor, objectification, and other gender-based mistreatment (2–4), is highly prevalent in many settings across society (5, 6). Sexism concerns one’s social identity and is a particularly malignant form of mistreatment, as it builds on and reinforces traditional societal power structures, such as male dominance (7). Because it undermines their credibility, women are often forced to overemphasize their competence and dissociate it from common discriminatory assumptions (8).

The profound undesirable effects of sexism on individual outcomes are well established and include reduced well-being and impaired individual work performance (9)^*. One could similarly expect that sexism would also have detrimental effects on team processes and outcomes, although very little research has addressed the effects of sexism on teams. An exception is a correlational study (10) that reported negative relationships between perceptions of sexual hostility and indices of team financial outcomes. Given that exposure to sexism often occurs in social settings, such as team discussions, presentations, and meetings, it is particularly important to better understand the effects of sexism on team outcomes. This is our focus in the present study.

Specifically, we examine how sexism derails team performance and propose that it undermines the effects on performance of a key factor in joint action — interpersonal synchrony—known to enable cooperation and support collective performance (11, 12). Our study focuses on emotional synchrony, defined as the “concurrent and time-lagged covariation of the same emotion” (13). Emotional synchrony is often assessed in the form of facial expressive synchrony, which reflects the coordinated temporal alignment of facial expressions between at least two individuals (i.e., a dyad). Facial expressive synchrony, almost exclusively considered and assessed within the framework of dyads, is an important predictor of desirable collective processes and outcomes (1). For example, different forms of facial expressive synchrony predict team productivity within newly formed teams (14), audience engagement in concerts (15), and even social connections between viewing partners (16).

In addition to its main effects on collective outcomes, we theoretically argue that the effects of synchrony may vary based on the context in which it occurs (17–19). Specifically, in the context of sexism as the background to the emergence of synchrony, the focus of team members may shift from common goals to social bonding. This shift will reflect individuals’ inclination to bond in the face of the threat they experienced, consistent with classic theories of human bonding (20, 21). Furthermore, evolutionary approaches suggest that collectives respond to threat through synchrony, aimed at coordinating their actions to promote survival (22). Although limited, some support for these arguments comes from experimental evidence showing an increase in neural synchrony as a response to exposure to external sources of threat (23, 24). Similarly, facial expressive synchrony signals efficient information transfer, coregulates emotions, and builds social connections among individuals (1, 25). Therefore, we expect sexism threat to increase emotional synchrony, but emphasize bonding among team members rather than signaling common goals. Consequently, under exposure to sexist comments, synchrony will no longer contribute to team performance.

The present study involved 177 dyads of women. We chose to focus on dyads, a common form of a team that is defined as two or more individuals working interdependently toward shared goals in either face-to-face or virtual settings (26). Participants were randomly assigned to one of two experimental conditions: a sexism condition, where the experimenter (an actor) made several sexist derogatory comments regarding women’s ability to succeed in this task, and a control condition, in which the instructions given by the same actor were neutral. Participants then joined a Zoom meeting in which they engaged in a cooperative cognitive task (a version of the guess-who game). The faces of the participants were continuously video recorded throughout the experiment for later analysis of their emotional facial expressions. To facilitate an environment that highlights the role of facial expressive synchrony, this research was conducted in a virtual setting, where participants were able to focus continuously on each other’s facial expressions. According to media naturalness theory (27), video conferencing amplifies participants’ attention to specific cues, including facial expressions. The reason is that the framing of the screen minimizes distractions by prompting participants to focus more closely on each other’s faces. Whereas in-person meetings often involve exposure to multiple stimuli (e.g., body language and environmental factors), video conferencing fosters concentrated engagement with facial expressions, enhancing nonverbal communication. In addition, because of the practicality of video-conferencing platforms, individuals have become familiar with these platforms (28).

The experimental procedure and the stages of sexism manipulation are depicted visually in Fig. 1. After completing the experiments, each participant’s emotional facial expressions were analyzed via AI-based software (29). This study focused specifically on negative emotions (anger, contempt, disgust, fear, sadness, and surprise), consistent with the sexist context that can elicit such responses. The level of facial expressive synchrony between participants in each dyad was then calculated, indicating that in the sexism condition, the experimenter’s sexist comments indeed led to increased facial expressive synchrony between team members compared with the control condition. Additionally, as expected, teams in the control condition performed significantly better than those in the sexism condition did on the task. Finally, whereas in the control condition, facial expressive synchrony positively predicted team performance, in the sexism condition, it did not. These findings emphasize the intricate effect of emotional (facial expressive) synchrony on team performance when moderated by verbal sexism.

Fig. 1.

Experimental stages and sexism manipulation. Note. This figure illustrates the experimental procedure across five sequential stages. Stage 1 commences immediately after the task instructions, where the sexism condition includes a sexist remark, while the control condition includes neutral commentary. Stage 2 follows with a second set of sexist/neutral comments, further differentiating the conditions. In Stage 3, participants perform the first round of the “guess-who” task, with no additional manipulation. Stage 4 serves as a “threat booster” in the sexism condition and includes a third sexist remark, while the control condition maintains neutrality. Finally, Stage 5 involves the second round of the task, mirroring the third stage. Once Stage 5 is over, participants are asked to complete a cohesion questionnaire, which includes items assessing social cohesion and task cohesion. During each stage, participants’ facial expressions are recorded to calculate facial expressive synchrony, providing insight into the effects of sexist remarks on team performance. The quotes from the experimenter displayed in the figure articulate the verbal manipulation implemented at each stage.

Results

Sexism Increases Emotional Synchrony.

The facial expressions of 177 pairs of participants across the five stages of the experiment were measured and analyzed. To calculate facial expressive synchrony, Cross-correlation functions (CCFs) were used. This technique involves the assessment of the co-occurring temporal alignment of participants’ emotional expressions over time. Lags of 3 s around each point in time were examined to assess how emotions aligned when there was a slight delay between participants’ expressions. The resulting synchrony score for each emotion was then averaged to obtain an overall facial expressive synchrony score. These scores ranged from −1 to +1, with higher scores indicating that participants’ emotional responses were more in sync (Materials and Methods).

To test the hypothesis that sexism enhances facial expressive synchrony, a repeated-measures ANOVA was employed to examine variations in emotional synchrony throughout the stages of the experiment (Analytical Strategy). The ANOVA results revealed significantly greater levels of facial expressive synchrony in the sexism condition than in the control condition (mean difference = 0.02, F _{(1, 140)} = 8.11, P = 0.005, η² = 0.02). The interaction effect between the stage of the experiment and the experimental condition was also significant (F _{(4, 560)} = 5.40, P = 0.005, η² = 0.03). Post hoc comparisons demonstrated that the difference in facial expressive synchrony levels was significant only when sexist comments were repeated by the experimenter (mean difference in Stage 2 = 0.05, Tukey’s t ₍₁₄₀₎ = 4.69, P < 0.001; see Fig. 2). During this experimental stage, facial expressive synchrony peaked in the sexism condition but not in the control condition. The levels of each emotion across the experimental stages are detailed in SI Appendix, as is a detailed analysis of the synchrony of different emotions throughout the five stages of the experiment. The results of these supplementary analyses suggest that the metrics of contempt, disgust, fear, sadness, and surprise were more synchronized in the sexism condition than in the control condition.

Fig. 2.

Facial expressive synchrony levels throughout the five stages of the experiment in the sexism vs. control conditions. Note. This graph illustrates variations in facial expressive synchrony across different experimental stages, computed via CCFs between participants and averaged over 3 s segments (Computing Emotional Synchrony). The red line indicates emotional synchrony in the sexism condition, whereas the blue line shows emotional synchrony in the control condition. Error bars represent SE of the mean at each stage. Dashed horizontal lines indicate the average of facial expressive synchrony for each condition. An asterisk (*) denotes whether the results are statistically significant. A repeated-measures ANOVA revealed a significant within-subjects effect for the five stages (F _{(4, 560)} = 4.78, P < 0.001, η² = 0.02) and a between-subjects effect for the experimental condition (F _{(1, 140)} = 8.11, P < 0.001, η² = 0.02). Higher levels of facial expressive synchrony were observed in the sexism condition (mean = 0.02, SD = 0.03) than in the control condition (mean = 0.01, SD = 0.01). The interaction effect between the stage of the experiment and the experimental condition was also significant (F _{(4, 560)} = 5.40, P = 0.005, η² = 0.03). Post hoc comparisons revealed a significant increase in facial expressive synchrony under sexism following the second stage of the experiment (Tukey’s t ₍₁₄₀₎ = 4.69, P < 0.001).

Sexism Attenuates the Relationship Between Emotional Synchrony and Team Performance.

To answer the core research question, which is whether exposure to sexism attenuates the favorable effects of facial expressive synchrony on team performance, an adapted version of the “guess-who” game, a task previously used in social dynamics research, was employed. Participants worked in pairs to correctly identify as many words as possible within a limited time frame (Materials and Methods). The team performance score represents the average number of words identified by each dyad. The data were analyzed via general linear model analysis (Analytical Strategy). As expected, teams that were exposed to verbal sexism significantly underperformed compared with those in the control condition (mean difference = 0.62, F _{(1, 172)} = 4.93, P = 0.028, η² = 0.03). In line with our theoretical argument about the adverse effects of sexism on the role of emotional synchrony in team performance (Fig. 3), a significant interaction emerged between facial expressive synchrony and the experimental condition (F _{(1, 172)} = 4.84, P = 0.029, η² = 0.03). Facial expressive synchrony had typical positive effects on performance only in the control condition (β = 0.28, t ₍₁₆₉₎ = 2.28, P = 0.024, η² = 0.03). In contrast, the favorable effect of facial expressive synchrony on team performance disappeared in the sexism condition (β = −0.07, t ₍₁₆₉₎ = −0.69, P = 0.489, η² = 0.01).

Fig. 3.

Facial expressive synchrony predicting team performance under the sexism vs. control conditions. Note. The blue line represents the control condition, whereas the red line corresponds to the sexism condition. Shaded bands around each regression line represent the SE of the estimated regression lines. The beta values of each condition are presented above the line, where an asterisk denotes statistically significant simple effects. Team performance represents the average number of words identified by each dyad. The y-axis begins at 8 to better visualize the relatively small variations in team performance, as all observed scores fall within a narrow range. The general linear model analysis revealed a significant model (F _{(3, 172)} = 3.15, P = 0.027, η² = 0.05). There was a main significant effect of the experimental condition on team performance (F _{(1, 172)} = 4.93, P = 0.028, η² = 0.03). The team performance scores were greater in the control condition (mean = 10.32, SD = 2.16) than in the sexism condition (mean = 9.70, SD = 2.01). The effect of facial expressive synchrony on team performance was not significant (F _{(1, 172)} = 1.73, P = 0.190); however, a significant interaction emerged between facial expressive synchrony and the experimental condition (F _{(1, 172)} = 4.84, P = 0.029, η² = 0.03). Facial expressive synchrony had a significant positive effect on team performance in the control condition (β = 0.28, t ₍₁₆₉₎ = 2.28, P = 0.024, η² = 0.03) but had no significant effect in the sexism condition (β = −0.07, t ₍₁₆₉₎ = −0.69, P = 0.489, η² = 0.01).

Additional Analysis: Sexism Increases Social Orientation but Not Goal Orientation.

Post hoc analyses were conducted to explore whether participants exposed to the sexism manipulation and those in the control condition perceived their connection with the other member of the dyad as goal- or socially oriented. To examine this question, a survey measure of cohesion in groups (30), adapted for this study, was used. Cohesion typically encompasses two dimensions: social cohesion, which stems from members’ attraction to the group, reflecting positive interpersonal relationships, and task/goal cohesion, reflecting shared commitment to achieving group objectives (31–33). Participants’ self-reported levels of social and task-related cohesion in the sexism and control conditions were analyzed. Social cohesion was assessed by items reflecting interpersonal bonding and mutual support, whereas task-related cohesion was measured by items capturing commitment to shared group objectives (Materials and Methods). Because the cohesion data were obtained from individuals, we accounted for individuals’ dyad membership by utilizing mixed-effects models with condition (sexism vs. control) as a fixed effect and team as a random effect. We found that social cohesion was significantly higher in the sexism condition than in the control condition (β = 0.18, SE = 0.08, t ₍₁₇₂₎ = 2.45, P = 0.015). In contrast, task/goal cohesion did not significantly differ between conditions (β = 0.09, SE = 0.08, t ₍₁₇₂₎ = 1.07, P = 0.285). These findings suggest that exposure to sexist remarks enhanced social cohesion but did not affect goal cohesion, indicating that exposure to sexism did not impact the participants’ shared commitment to their group goals.

Supplementary Results.

Please refer to SI Appendix for an analysis of individual-level emotions and a comparison of the mean levels of each emotion across conditions. Detrended fluctuation analysis was further used to examine and control for potential effects of autocorrelations in the time series data with respect to the synchrony calculation. Finally, emotion-specific synchrony analyses and tests for mood as a control variable in our analysis of synchrony and performance across conditions are also provided.

Discussion

The dire effects of sexism on individuals are well documented (2). For example, in the United States, 22% of working women reported experiencing sexual harassment (5), whereas in Canada, 44% identified authority figures as a main source of gender discrimination, with 62% viewing gender discrimination as a major barrier for promotion to leadership roles (34). Our findings provide insight into these consequences by demonstrating how exposure to sexism, from an authority figure, disrupts the classic positive effects of emotional synchrony on team functioning. Using an experimental design, we found that whereas exposure to sexist comments increased facial expressive synchrony, it eliminated the positive effects of emotional synchrony on task performance. Our findings validate and augment previous findings concerning the negative implications of sexism for individuals (9, 10, 35) by indicating that individuals, separately and within teams, subjected to sexism, tend to underperform.

Our finding that exposure to sexism increased facial expressive synchrony aligns with the evolutionary role of synchrony in fostering connections and facilitating information transfer among individuals (19). Indeed, sexism is considered a threatening context that increases the need for bonding (36), which may manifest through enhanced emotional synchrony. The experience of synchrony potentially enables team members to efficiently share information and connect when facing a threat; thus, it is a crucial bonding mechanism that promotes solidarity and mutual support among individuals facing shared adversity (1). Indeed, experimental evidence has demonstrated that interpersonal synchrony enhances expressive and reciprocal communication between individuals, thereby improving emotional information transmission within groups (37–39). Bonding can also facilitate affiliation and cooperation under stressful conditions, described as a “tend and befriend” strategy (40). Our findings provide support for this theoretical argument; that is, the participants exposed to sexist remarks not only demonstrated increased synchrony but also reported greater social cohesion.

Our results demonstrate the role synchrony has in team interactions, highlighting the complex interplay between emotional synchrony (measured here as facial expressive synchrony) and team dynamics under threat. Increased synchrony seems to reflect an adaptive response aimed at maintaining group cohesion and the collective processing of a threat. Whereas synchrony is typically triggered by contexts that have a simultaneous social- and task-related focus, exposure to sexism appears to confine it to marking primarily social support and threat management functions at the expense of facilitating task performance. The attenuated relationship between synchrony and team performance under sexism can also be understood from a cognitive stress perspective. Exposure to sexist behavior increases cognitive and emotional demands (41), forcing team members to divide their attention between processing the discriminatory experience and maintaining task performance. This cognitive burden likely explains why increased emotional synchrony in the sexism condition failed to enhance performance as it typically does in nonthreatening situations.

Our findings challenge the prevailing literature regarding the positive effects of synchrony (42). This extensive body of knowledge posits that synchrony among team members is a hallmark of successful collaboration (43), often leading to enhanced communication (38, 39) and prosocial outcomes (44). This theoretical stance is well supported by a wealth of literature that documents the myriad ways in which synchrony—be it in movements (45), facial expressions of emotions (13), or physiological responses (46)—facilitates a shared understanding and rapport among individuals (47). Although synchrony has been linked to a range of positive team-based outcomes (48), our findings align with those of other studies that uncover the limitations and nuances of synchrony, suggesting that it does not always serve all interests of the team and may come at the expense of performance in certain contexts (18, 45). Future research should aim to uncover additional contexts in which “social glue” may consciously or subconsciously redirect team efforts away from the tasks of the team (17). Such a dynamic may thus point to a deeper interplay between responses to threats, where an automatic, perhaps instinctual drive for emotional alignment (13) may conflict with a more deliberate strategic approach to task execution.

Unexpectedly, we found that emotional synchrony when exposed to sexism increased only after the second instance of verbal sexism, rather than immediately following the first exposure to sexist comments. Several explanations may account for this observation. First, the participants may have needed more exposure to sexism to sense it collectively and, consequently, experience increased emotional synchrony. In other words, the impact of sexist comments could be cumulative, requiring multiple instances to trigger a significant increase in the synchrony response. Second, the first sexist comment may have served as a primer, heightening participants’ sensitivity to subsequent sexist remarks. This increased vigilance to sexist comments could have led to a more pronounced synchrony response to the second threat. Finally, the delay in response may reflect the suspension typically associated with the onset of biological coping mechanisms (49). These explanations are not mutually exclusive and may work in concert. Future research may further explore how emotional synchrony evolves over time in response to sexist comments (or other threatening comments) by varying the timing and frequency of such comments. Furthermore, incorporating physiological measures, such as the hormone cortisol, can more precisely capture and track the temporal dynamics of stress responses. Overall, understanding these dynamics is crucial for comprehending how and when teams become emotionally aligned and how members respond to sexism in collective settings.

Another line of research may investigate how other sources of discomfort and mistreatment affect facial expressive synchrony and collective action, particularly in contexts where synchrony may either support or detract from collective outcomes. Specifically, future studies could examine whether the duration and intensity of synchrony vary between identity-based discrimination and general discomfort and whether these different forms of discomfort lead to distinct team outcomes. Researchers could investigate how having multiple identities (e.g., mixing gender and race) influences synchrony patterns, building on work showing differential effects of sexism and racism (8). Future studies may also examine other types of work-related mistreatment beyond sexism to reveal their unique effects on synchrony and team outcomes (7)

Notably, whereas our study focused on emotional synchrony as a mechanism through which sexism impacts team performance, other team-level processes may also play important roles. Ambient sexual harassment in teams was associated with increased relationship and task conflicts and, ultimately, reduced team financial performance (50). These findings suggest that sexism can impair team functioning through multiple pathways. Future research may examine the role of sexist behaviors in creating interpersonal tensions that distract team members from focusing on their tasks or diminish their willingness to work together to achieve their tasks. Furthermore, additional research is needed to examine how the cognitive and emotional energy devoted to coping with or responding to sexism may leave fewer resources available for task-related efforts. Future research may also explore how various team processes, including emotional synchrony, conflict, or cohesion, interact in the face of sexist behaviors, offering a more comprehensive understanding of the impact of sexism on team dynamics and performance.

In the present study, we utilized the Zoom platform because it offered several methodological advantages, including controlled facial expression viewing, reduced external distractions and a familiar environment. While synchrony can indeed occur in virtual interactions, as found in our study and in previous studies (51, 52), the dynamics of emotional synchrony and team processes may manifest differently in face-to-face settings. In such settings, participants can perceive a broader range of nonverbal cues and engage in more natural interactions. Future research could explore the effects of exposure to sexism in various contexts, including face-to-face interactions, hybrid settings, and different virtual platforms, while incorporating additional indicators of synchrony (e.g., body movements, physiological responses, and vocal patterns).

Our study examines an interpersonal mechanism through which exposure to sexism may hamper team performance and as such it is a steppingstone for research on the dire effects of sexism on collective functioning. Nevertheless, given that the integration of synchrony into research on sexism and team dynamics is still in its early stages, better specification of the key theoretical links and mechanisms that explain the effects of sexism in teams is needed. To alleviate this limitation, future studies can better specify important theoretical links in more detail. For example, it might be possible to carefully design an experiment to test the effects of sexism on synchrony by having participants engage in different tasks that highlight meeting goals (e.g., business task) vs. enhancing social bonding (e.g., survival task) to better delineate the effects of different forms of synchrony on team outcomes. Furthermore, our work did not involve testing mediation models. Future research can utilize mediation analyses to more effectively specify and validate the role of synchrony and its causes as mediators of the effects of sexism on team outcomes. For example, one possible avenue for future research is to investigate whether the impact of sexism on performance is mediated by goal versus social orientation. Such studies could examine whether sexism first influences goal versus social orientation, which in turn affects synchrony, ultimately shaping team performance. Whereas our study was not designed to disentangle all possible pathways, it represents an important step in understanding how sexism impacts team processes and outcomes.

Another limitation of our study concerns the relatively small effect size of the manipulation. There are several reasons for this limitation. First, besides the advantages of having a controlled experimental environment, lab settings are, by definition, less ecological, potentially reducing the extent to which a manipulation is perceived as powerful (53). This may be particularly salient in the case of deviant behavior, like sexist comments, which may be less potent outside its typical social context. Indeed, sexist remarks often come from individuals with whom the victims have a more meaningful relationship, such as from their bosses, who have financial and/or long-term emotional impacts over the victims. Second, because manipulations that involve adverse effects are often rightfully unacceptable for Institutional Research Boards, we toned down original versions of our manipulation. Nevertheless, despite these limitations, we observed statistically significant effects, suggesting that even a relatively weak manipulation was sufficient to trigger emotional synchrony and its consequences. Future research may examine real-world encounters with sexism, where the stakes are higher, potentially leading to obtaining more pronounced effects.

Furthermore, future research could explore the effects of team size and composition. As in previous studies of team synchrony (42, 45, 54, 55), we focused on dyads as the fundamental unit for analysis. In larger teams, multiple dyadic relationships exist simultaneously, potentially creating complex patterns of emotional synchrony and social support. Studies could examine whether the negative effects of synchrony on performance found in our study strengthen or weaken as team size increases. It would also be intriguing to investigate how different team characteristics (e.g., dyads of men and women) and individual differences modulate the impact of different forms of sexism (e.g., sexism targeted at men) on team processes (56). These studies may offer deeper insights into creating supportive environments that mitigate the adverse effects of sexism and promote resilient team dynamics.

In addition to fostering a safe environment at the workplace and meeting legal requirements, our research suggests that when organizations adopt a zero-tolerance policy for sexual harassment, they contribute not only to employees’ well-being but also remain goal focused and, thus, more competitive. In addition to their contribution to team performance, the findings of this study indicate that synchrony in team dynamics marks the team’s emotional responses to stress, which is vital for team well-being (57). Beyond the many other reasons for which organizations should strive to eliminate sexism, our findings also underscore the importance of utilizing emotion-focused training (e.g., emotional intelligence) to harness the potential of emotional synchrony as a catalyst for improved teamwork and productivity.

In conclusion, as an increasing number of organizations rely on teamwork to perform tasks, our research illuminates the intricate ways through which sexism eliminates the advantages of working in teams for task performance. It shows that even if facial expressive synchrony does not facilitate team performance in the context of sexism, it may be crucial for members’ survival.

Materials and Methods

Participants.

A total of 354 women between the ages of 20 and 30 y were recruited via social media platforms. The participants provided informed consent and then participated in a video-mediated experiment (using the Zoom platform) involving two participants and a paid actor serving as the experimenter and in exchange for $12. The participants provided consent to participate and that the experiment would be recorded, as well as reconsent after the experimenter debriefed them about the true purpose of the study. Before the experiment started, participants were randomly assigned to either the sexism or control conditions. A total of 177 teams of two participants were recruited; 89 teams were in the sexism condition, and 88 teams were in the control condition.

Procedure.

Our study received ethical approval from the Bar-Ilan University Department of Psychology ethics committee (#2021/16). All of the procedures reported were in line with the ethical review board’s approval. At the designated time of the experiment, participants were sent an e-mail invitation to join the Zoom meeting in which the experiment took place. In both conditions, the male experimenter welcomed the participants, introduced himself, and asked them to briefly introduce themselves. Participants were asked to turn off their self-view so that they could see only each other on full screen (using the Pin Video function) while still hearing everyone in the meeting. This virtual meeting setup was specifically chosen to create an optimal environment for observing facial expressive synchrony in a setting that is highly familiar to participants. Since the COVID-19 pandemic, video conferencing has become a ubiquitous mode of interaction, making it a familiar and practical medium for studying social dynamics (28). The virtual setting in this study was specifically designed to emphasize facial expressive synchrony by enabling the participants to focus entirely on their partner’s facial expressions. The structured framing of the video call minimized distractions by limiting exposure to external stimuli, such as body language or environmental factors, and it directed attention primarily to facial cues (27).

The experimental procedure then unfolded in five stages. After the introductions, the first stage of the experiment was initiated:

• Stage 1 (Threat 1): In the sexism condition, the experimenter made the first derogatory comment toward women (“I know—and this is based on research—that girls have a harder time with tasks like this. If you want to succeed in the task, you should pay close attention to my instructions.”), whereas in the control condition, the instructions were neutral (“I know—and this is based on research—that people have a harder time with tasks like this when they don’t pay attention to the instructions, so it’s important that you listen carefully”).

Following this stage, the experimenter provided instructions for the cooperative task and introduced the second stage of the experiment.

• Stage 2 (Threat 2): In the sexism condition, the experimenter made the second derogatory comment toward women (“I’ll give you ladies a friendly advice—try not to think like a girl for a moment and do exactly what I asked you... just kidding… don’t be so serious.”), whereas in the control condition, the instructions were neutral (“I’ll give you a friendly advice—keep your eyes on the prize and remember what the goal is here. Don’t forget that and you should be fine”).

After the second stage of the experiment, the experimenter asked the participants to start the first round of the task, thus initiating the third stage of the experiment:

• Stage 3 (Task 1): The participants performed the first round of the task (described below). In this stage of the experiment, there were no differences between the conditions.

After the third stage of the experiment, the experimenter encouraged the participants to prepare for the second round of the task. Subsequently, the fourth stage of the experiment was initiated:

• Stage 4 (Threat Booster): In the sexism condition, the experimenter made the third and last derogatory comment toward women (“I know it’s hard for you girls—but keep focusing on the task!”), whereas in the control condition, the instructions were neutral (“I know it’s hard at this point—but keep focusing on the task!”).

After the fourth stage of the experiment, the experimenter asked the participants to start the second round of the task, thus initiating the fifth and last stage of the experiment:

• Stage 5 (Task 2): Participants performed the second round of the task. In this stage of the experiment, there were no differences between the conditions.

Participants were then asked to report their social cohesion and goal/task cohesion levels. Upon completion, the experimenter explained his behavior and debriefed participants about the goals of the study. After the debrief, the participants were asked to provide reconsent for their participation in the study.

Sexism Manipulation.

The sexism manipulation in our study was implemented by a professional actor, who was selected through a rigorous audition process due to his exceptional acting skills. This was crucial for maintaining the ecological validity of the manipulation while ensuring that the interactions remained controlled and consistent across sessions. Prior to the commencement of the study, the actor underwent training by the research team, which included an in-depth review of the scripted comments and rehearsal sessions to refine his delivery. The aim was to ensure that the actor could perform the manipulation in a way that was believable yet sensitive to the ethical considerations of the research. This method has been used in previous research (41). The content of the sexist comments used in the manipulation was inspired by real-life experiences shared by women within our community who were briefed about the study’s objectives. These individuals generously shared instances of sexism that they had encountered, which were then anonymized and adapted into the script. This approach was taken to ensure that the manipulation reflected genuine societal attitudes and biases, thereby enhancing the relevance and impact of the findings.

The Task.

The task is an adaptation of the “guess-who” task, which has been used in previous research (58, 59), including in classic studies of racism (60), one of which demonstrated the utility of the task as a platform for capturing stressful effects on task performance. This task was chosen because it is highly interdependent, allowing collective performance to be captured in the context of emotional synchrony. Furthermore, the cognitive requirements that must be met to accomplish this task are consistent with those of common work tasks. Among these requirements are verbal ability, visual-spatial processing, executive functioning, working memory, problem-solving, perspective-taking, abstract thinking, and a need to sustain focus and attention owing to time pressure. The task included two rounds, each taking 2 min to complete. In each round, one of the participants was randomly assigned to be the first to receive random nonvalanced words (chair, ball, hat, etc.) via the private message function of the Zoom platform. This participant then had to describe the word to the other participant without using the actual word. Participants were instructed to collectively present and correctly identify as many words as possible in the 2 min period allocated for the task in each round. The roles were alternated in the second round.

Manipulation Check.

To evaluate the effectiveness of the manipulation, participants were asked to rate the degree to which they experienced the situation as threatening to women via three items from the Symbolic Gender Threat Measure (36). Sample item: “It felt as though the experimenter did not understand me because he is a man”. Participants were also asked to rate the degree to which the experimenter’s behavior toward women was uncivil via three items from an incivility measure (3). Sample item: “The experimenter was disrespectful toward women”. All the items were on a five-point Likert scale. The reliability, assessed via Cronbach’s alpha, was 0.91. To assess the effectiveness of the sexism manipulation, we conducted a mixed-effects linear model with team membership as a random intercept and condition as a fixed effect. This analysis revealed that participants in the sexism condition reported significantly higher levels of sexism threat than those in the control condition (β = 0.53, SE = 0.14, t ₍₁₇₂₎ = 3.86, P < 0.001).

Cohesion.

Following the completion of the task, an adapted version of a well-established cohesion scale (30) was administered. This measure distinguishes between two primary dimensions of cohesion: social cohesion and task-related cohesion. Social cohesion was captured through items reflecting interpersonal support and bonding (e.g., “We supported each other” and “We relied on each other”), whereas task-related cohesion was measured through items assessing shared commitment to collective goals (e.g., “I cooperated with my teammate” and “I would be happy to participate in work/study activities with my teammate”). The responses were recorded on a five-point Likert scale (1 = strongly disagree to 5 = strongly agree), allowing the participants to express their alignment with each item. Reliability analyses demonstrated high internal consistency for both dimensions of cohesion (Cronbach’s α = 0.87 for social cohesion and Cronbach’s α = 0.84 for task-related cohesion).

Emotional Facial Expression Data.

Video data were analyzed via “Affectiva” software (29). The Affectiva algorithm functions by first detecting faces in the video input and then mapping key facial landmarks, such as the corners of the eyes, the mouth, and the outline of the jaw. It captures facial expressions at action points, which are specific movements or positions of these facial landmarks. By analyzing the combinations and intensities of these facial expressions, the software identifies a range of emotional activities. Affectiva software employs machine learning models trained on extensive datasets of labeled facial expressions to interpret combinations of facial features as specific emotions. Each emotion is quantified based on the intensity and frequency of the corresponding facial expressions observed in the video data. The output provides a detailed emotional profile for each observed individual, indicating the percentage likelihood of each emotion appearing in every frame of the video, which had a sampling rate of 30 frames per second. The signals fed into the analysis are the emotion-specific time series derived from the emotional facial expression data captured via the software. As illustrated in Fig. 4, these recorded time series show the percentage likelihood of occurrence of each negative emotion in every frame captured in the video throughout the interaction. These signals were then used to calculate the emotional synchrony between the participants. Consistent with theory and the intended effect of the negative sexist context to elicit negative emotions, we focused on the negative emotions identified by the software, namely, anger, contempt, disgust, fear, sadness, and surprise.

Fig. 4.

Emotional facial expression data quantification via Affectiva software. Note. This figure offers a visual summary of the experimental setup alongside the emotional response data captured during a key moment in the experiment. During the experiment, participants could only hear and not see the experimenter. The two graphs represent the fluctuating emotional states of each participant, as analyzed by Affectiva software, positioned correspondingly to their image on the left. These graphs record the intensity and variation of negative emotions—anger, contempt, disgust, fear, sadness, and surprise—across the experimental timeline. The vertical line that runs through both graphs corresponds to the moment captured in the Zoom session, which occurred during the threat booster stage, thereby aligning a specific instance in time across both the visual and data representations.

Data Analysis.

Preprocessing.

The facial expression data of each participant in the 177 runs of the experiment, in each of the five stages (described above), were analyzed. A total of 1,770 datasets were collected (177 runs* 5 stages in each run* 2 participants in each run). Missing values in each dataset were addressed through a robust linear interpolation method that relied on adjacent nonzero data points for imputation (61). Specifically, for each column with missing values, contiguous sequences of such values were identified, and provided that these sequences were shorter than 150 data points (5 s), these gaps were linearly imputed by interpolating between the adjacent nonzero values. Datasets with contiguous missing value sequences exceeding 150 data points (5 s) were excluded. Furthermore, a filtering process was implemented to preserve only datasets with no more than 15% missing data. Among the total of 1,770 datasets, three teams (30 datasets) were excluded from the analysis due to technical issues in the experiment, and 56 additional datasets were excluded as a result of the filtering process.

Computing emotional synchrony.

CCFs, a common approach to studying concurrent behavior and emotions between two individuals’ time series (1, 46, 62), were used to calculate emotional synchrony. This method of analysis was chosen for its ability to capture the dynamic and fluctuating nature of emotional synchrony, which, by definition, does not occur at a single, isolated point in time but, rather, unfolds over the course of the interaction. An R code (63) was used to calculate the correlation value for positive and negative time lags in a small time window of 3 s (from −3 to +3 s) around the 0-lag position. Lags mean that the data were essentially shifted in steps in both positive and negative directions, one video frame at a time within a window of 3 s. Subsequently, the correlation values for each of the shifts were extracted. This relatively small time window was chosen to capture concurrent synchronization dynamics. Synchrony was represented by the average of all correlation values for all the lags within the segment. Facial expressive (emotional) synchrony was considered the average synchrony value from each identified emotion. This score ranged from −1 to +1, where a higher score indicates greater alignment in the participants’ emotional responses. The R script, which details the procedures and scripts used in our analysis, is available in the OSF repository.

Comparing actual and pseudosynchrony.

To investigate the nonrandomness of synchrony scores within our experimental context, an analysis comparing synchrony levels in actual teams versus pseudoteams, which were constructed from members of different teams, was conducted. This comparative analysis between actual and pseudosynchrony is crucial to demonstrate that the observed emotional alignment is not merely a byproduct of shared context or random fluctuation but, rather, is a result of the genuine interactive processes taking place within real teams. Whereas actual synchrony was calculated between paired participants within the same experiment representing real team interactions, pseudosynchrony was generated by calculating the synchrony between participants from different dyads. Through this approach, we were able to simulate interactions in “teams” whose members did not actually interact with each other but were exposed to the same context. Welch’s independent-samples t test was used to compare the synchrony levels between actual and pseudoteams, and a significant difference was found, with actual teams showing higher levels of synchrony than pseudoteams (Fig. 5). This disparity indicates that synchrony was more pronounced in real interactions within actual teams than in simulated interactions and was, at least in part, a result of the social interactions that occurred within teams.

Fig. 5.

Distributions of facial expressive synchrony in actual vs. pseudoteams. Note. This figure illustrates the distribution of facial expressive synchrony levels, with the teal area representing actual teams and the pink area representing pseudoteams, where synchrony was calculated between individuals who did not interact with each other. Synchrony within actual teams was computed between participants who interacted in the same experiment, whereas pseudosynchrony was calculated between all the participants from different experiments in the same experimental condition. There was a significant difference in synchrony levels (Welch’s t _(173.575) = 4.20; P < 0.001, Cohen’s d = 0.35), with actual teams exhibiting a higher mean level of synchrony (mean = 0.014; SD = 0.03) compared to pseudoteams (mean = 0.005; SD = 0.02). This significant disparity highlights that synchrony is more evident in real team interactions than in those simulated among pseudoteams, emphasizing the unique significance of synchrony in authentic team dynamics.

Analytical strategy.

To ascertain the effectiveness of the sexism manipulation, a mixed-effects linear model was conducted, with team membership included as a random intercept and condition as a fixed effect. This approach accounts for the nested structure of the data while assessing differences in manipulation check scores between the sexism and control conditions. We also conducted pseudosynchrony analysis, which entailed computing the synchrony between each participant and all others within the same experimental condition. To test the difference in facial expressive synchrony levels between actual teams and pseudoteams, Welch’s independent-samples t test was employed. Higher real synchrony scores than pseudosynchrony values suggest that synchrony is not random and is not strictly due to the shared context. For the assessment of facial expressive synchrony levels across different stages of the experiment and the potential differences between the two experimental conditions, repeated-measures ANOVA was employed. In this analysis, the experimental condition was the between-groups factor, while facial expressive synchrony within each stage of the experiment served as the repeated measure. This approach facilitated an in-depth exploration of the temporal dynamics of facial expressive synchrony and its modulation by the sexism manipulation. To test the relationship between facial expressive synchrony and team performance and to investigate the potential variations in this relationship between the two experimental conditions, a general linear model analysis was conducted. In this analysis, team performance was the dependent variable, the average facial expressive synchrony value in all stages of the experiment served as the independent variable, and the experimental condition was the moderator. Not only did this analysis reveal the statistical significance of the main effect of the experimental conditions on team performance, highlighting the impact of sexism on team performance, but it also uncovered a significant interaction between facial expressive synchrony and the experimental condition. This interaction indicates that the relationship between facial expressive synchrony and team performance differs depending on the experimental condition. Subsequently, simple effects analyses were conducted to further dissect the interaction, allowing us to examine the specific relationship between emotional synchrony and team performance within each experimental condition separately. All statistical tests were conducted as two-tailed analyses, with the corresponding P-values reported throughout the manuscript.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The data and analysis scripts supporting the findings of this study have been deposited in the Open Science Framework and are publicly available at: https://osf.io/kxay2 (64). The repository includes cleaned time-series datasets and R scripts used for cross-correlation analysis of facial expressive synchrony. These materials correspond to the analyses reported in the manuscript and are described in the accompanying README.