The influence of attorney anger on juror decision making

Abstract

We investigated the interactive effect of attorney anger expression and attorney gender on juror decision-making. Jury eligible Amazon MTurk participants (N = 455) were recruited. They listened to an audio recording of a male or female prosecutor delivering a closing statement with varying levels of authenticity (authentic, inauthentic or no anger). Then, they rendered a verdict. After the verdict, participants filled out various measures: guilty verdict confidence, current feelings of anger, and perceptions of prosecutor trust and competence. We found that the prosecutor’s authentic displays of anger provoked anger in the participants, which, subsequently, increased the odds of a guilty verdict and guilty verdict confidence. Moreover, authentic displays of anger improved perceptions of the prosecutor’s competence, which also increased the odds of a guilty verdict. However, perceptions of the prosecutor’s trustworthiness did not vary, and attorney gender did not play a moderating role. The implications of these findings are discussed.

In contrast to the pressure that judges have to manage their emotions in the courtroom (Bergman Blix & Wettergren, 2016), books on trial strategies advise attorneys on how to deliver their case to juries in a convincing and often emotional manner. Much of this advice mentions the use of emotion to sway the jury to vote in the attorney’s favor. Mathis’s book (Mathis, 2011), Acting Skills for Lawyers, recommends attorneys to be outraged when cross examining witnesses and to be interested when directly examining their client. In addition, it includes a list of emotions, such as anger, to use during closing statements. Further, Lubet’s (2004), Modern Trial Advocacy, claims that in a closing statement the ‘absence of emotion may be taken as a lack of belief in the righteousness of your cause’ (p. 435). These manuals assume that attorneys can captivate jurors through a passionate emotional appeal. Most of this advice, however, is largely intuitive and unsubstantiated, and may be limited to legal training in the United States. A study of legal training in Sweden noted that instead of teaching budding attorneys to use emotional appeals, students are taught to distance themselves from emotions and to dispassionately argue the law (Flower, 2014).

Whether attorneys are encouraged or discouraged from using emotions in the courtroom, little is known about how jurors will interpret and react to emotions such as an attorney’s anger. Without empirical evidence, attorneys may use emotions inappropriately (Van Doorn et al., 2015) or waste their time rehearsing a certain emotion to little effect. Further, if an attorney feigns an emotion, and jurors do not believe the emotion is sincere, the strategy might backfire and make the attorney less effective. Therefore, the purpose of the current study was to determine how jurors are psychologically influenced by information about an attorney’s anger. We examined whether a mock juror’s knowledge about the authenticity of the prosecutor’s anger would affect their verdict. As a caveat, we were not interested in deception detection per se. Rather, we were interested in understanding the outcomes following a juror’s awareness of an attorney’s inauthentic or authentic anger expression. Finally, we investigated the mediating roles of the mock juror’s emotional state and their beliefs about the prosecutor.

Why might authenticity be important in the courtroom?

To our knowledge, there is little to no literature on the influence of attorney authenticity on juror decision-making. Nevertheless, studies on the reliability of evidence presented at court may provide some context as to why authenticity may be important in jury decision-making. Ascertaining the truth is important for various decisions made in the legal system. In jury decision-making, it is the juror’s legal responsibility, as a trier of fact, to find the truth and render a verdict based on what they believe to be the truth (Cornell Law School, n.d.). Jurors may find the truth by evaluating the reliability of evidence presented at court, in which certain pieces of evidence may be found to be more important or reliable than others (Schweitzer & Nuñez, 2018). Research on the reliability of evidence shows that jurors react negatively when they are told that a piece of evidence may be unreliable. That is, jurors – who are led to believe that a piece of evidence is questionable – are less likely to vote favorably on the side where the questionable evidence comes from. McCabe et al. (2011) examined the impact of functional magnetic resonance imaging (fMRI) presented as evidence of defendant deception in mock jury trials. They manipulated whether an fMRI expert admitted to the jury that the fMRI technique was not always accurate, yielded incorrect conclusions in various occasions and was a controversial technique. Mock jurors who were told that the fMRI was not completely dependable were less likely to vote guilty than mock jurors who were not told about the dependability of the evidence.

On the other hand, mock jurors tend to respond auspiciously when they have reason to believe certain facts are telling of the truth. For example, jurors rate DNA evidence as highly reliable and accurate. As a result, they are more likely to be persuaded by such evidence when presented in court and, therefore, vote guilty (Lieberman et al., 2008). Jurors may also respond favorably to a witness’s testimony when they have reason to believe the witness is being sincere. For instance, research shows that a child’s sexual abuse testimony is the most believable and elicits more guilty verdicts when the child is rated as being emotional (Cooper et al., 2014; Golding et al., 2003; Regan & Baker, 1998). Jurors may take the child’s expression of emotional distress (e.g. crying) as an indication of honesty (Regan & Baker, 1998) – albeit honesty is not necessarily a reflection of correctness.

While knowledge about the reliability of evidence is not exactly conceptually equivalent to the authenticity of emotion, it highlights the importance of convincing jurors of the truth. Jurors are tasked with using their beliefs about the truth to make crucial decisions about a defendant's fate. The importance of convincing jurors of the truth is evident in their favoring of the side (via verdict) that is perceived to be truthful with more reliable evidence (Cooper et al., 2014; Golding et al., 2003; Lieberman et al., 2008; Regan & Baker, 1998). In a similar way, we argue that a juror’s decision-making will be swayed by the truthfulness or the authenticity of an attorney’s emotional expression. As the emotional expression of child witnesses is used as an indicator of honesty (Cooper et al., 2014; Golding et al., 2003; Regan & Baker, 1998), it is possible that jurors use an attorney’s emotion to gauge the truth. Although there is no direct evidence for how the authenticity of an attorney’s emotions influences juror decision-making, the psychological literature indicates that emotions (Hunsaker, 2017; Lelieveld et al., 2013; Sinaceur et al., 2015) and emotional authenticity (Campagna et al., 2016; Côté et al., 2013; Potworowski & Kopelman, 2008; Tng & Au, 2014) play key roles in decisions made in negotiation scenarios.

Emotional authenticity: anger

Angry people are compelling. In the appropriate context, those who use their anger instrumentally (Andrade & Ho, 2009; Tamir et al., 2008; Van Kleef et al., 2011) can get what they want from others. This is prominently shown in negotiation research. Counterparts who express anger at an offer are able to elicit more concessions (Van Kleef et al., 2004) and reduce demands from negotiators (Sinaceur & Tiedens, 2006). However, when emotion expression is perceived by a receiver as intentional and instrumental, anger can backfire. That is, negotiators – who are led to believe that a counterpart is purposely regulating their emotion to take advantage of them – tend to show resistant behaviors (e.g. conceding less) in retaliation (Côté et al., 2013; Tng & Au, 2014).

In Part 1 of a study, Tng and Au (2014) aimed to see how a negotiator’s perceptions of a counterpart’s strategic (inauthentic) anger would influence the course of negotiations. The experimenters had participants play the role of a seller. The participant’s goal was to sell the cellphone at the best price, warranty and service contract to the buyer (a confederate) within a set number of rounds. Importantly, the experimenters manipulated the type of information participants received about the ‘personality’ of the buyer who expressed angry statements in each round. In the authentic condition, the buyer was described to have a reliable personality and to be emotionally sincere. ‘Your counterpart likes to treat others sincerely . . . frank about how he/she feels’ (Tng & Au, 2014, p. 310). Conversely, in the inauthentic condition, the buyer was described as being unreliable and emotionally manipulative and insincere. ‘Your counterpart . . . has a tendency to be insincere and even manipulate others . . . also tends to hide his/her emotions . . . ’ (Tng & Au, 2014, p. 310). The outcomes of the negotiation revealed that participants who were warned that the angry negotiator could be inauthentic made significantly fewer concessions than participants who were told that the angry negotiator was authentic. Therefore, if a counterpart discovers that a negotiator’s anger is strategic or inauthentic, then they may retaliate against this negotiator, making fewer concessions (Tng & Au, 2014) and more demands (Côté et al., 2013).

A potential reason why counterparts may express these retributory behaviors could be due to psychological reactance to inauthenticity. Psychological reactance refers to arousal triggered by threat to individuals’ behavioral freedoms (Brehm, 1989). When individuals believe that external forces are trying to control their autonomy, they are aroused and motivated to resist this control (Brehm, 1989; Shoenberger et al., 2021). For example, psychological reactance can be triggered for people who find out that companies are lying about their consumer choices, such as instances where brands advertise their products under false pretenses. This type of reactance is known as inauthenticity aversion in consumer research (Silver et al., 2021). Inauthenticity aversion can lead customers to boycott brands that have used deceptive tactics to sell their products (Chiu, 2016). While this line of consumer research may not be completely relevant to the topic of jury decision-making, inauthenticity aversion portrays the psychological underpinnings behind human beings’ inherent distaste for and unfavorable responses to inauthenticity.

Taken together, the literature implies that human beings do not process emotions at face value but are thoughtful about the merits of another’s emotional expression. If an individual is given an inkling that another person’s anger may be unreliable, untrustworthy or inauthentic, the individual could experience psychological reactance and push back against the insincere person. People may exhibit retributory behaviors, such as conceding less (Tng & Au, 2014) and making tougher demands (Côté et al., 2013), against a person who feigns emotions. Therefore, the strategic use of emotions could backfire if observers perceive those emotions to be inauthentic.

In a similar way, we argue that the efficacy of a prosecutor’s emotion in the courtroom depends on a juror’s perceptions of that emotion. When perceived as authentic, a prosecutor’s angry expressions may confer an advantage by increasing a juror’s likelihood to vote guilty. In contrast, when perceived as inauthentic, a prosecutor’s anger may become a liability by decreasing a juror’s likelihood of voting guilty. To support these claims, we review the mechanisms underlying interpersonal expressions of emotion and how they may impact juror decision-making below.

Mechanisms: emotional contagion

The emotion as social information (EASI) model proposes that one person’s emotions can influence another person’s behavior via two mechanisms: emotional contagion and social information (Schwarz & Clore, 1983; Van Kleef, 2009; Van Kleef et al., 2004, 2010). Emotional contagion refers to how one person could ‘catch’ another’s emotions through mimicry and feedback (Hatfield et al., 1993). With some notable exceptions (Lanzetta & Englis, 1989; Van Der Schalk et al., 2011), human beings, for the most part, have the natural proclivity to mimic other’s emotions through various means, including vocal expressions (Hatfield et al., 1993, 2009; Rueff-Lopes et al., 2015). Mimicked behavior, such as vocal expressions, become information for the brain about how people feel at the current moment (Strack et al., 1988). Hence, according to emotional contagion, an attorney’s anger has the potential to spread to other people in the courtroom, namely jurors.

Once jurors become angry, they can become ‘intuitive prosecutors’, manifesting punitive biases in their legal decisions (Goldberg et al., 1999; Tetlock et al., 2007). Psychology and law research has shown that angry jurors are more likely to convict in criminal trials (Bright & Goodman-Delahunty, 2006; Salerno & Peter-Hagene, 2013) and are more likely to choose the death over life sentence in simulated capital jury trials (Georges et al., 2013; Nuñez et al., 2017; Nuñez et al., 2015; Semmler & Hurst, 2017). Furthermore, the Appraisal Tendency Framework posits that anger increases certainty about life events (Lerner & Keltner, 2000; Lerner et al., 2015), which would suggest that angry jurors would also be more confident in their guilty verdict (Salerno & Peter-Hagene, 2013). Thus, if prosecutors manage to appeal to a juror’s emotion through anger expression, they could reap the benefits of having a punitive, confident and angry jury.

Nonetheless, there is an important caveat. Whether a juror actually ‘catches’ anger through contagion may rely on their beliefs about the attorney’s authenticity. Research has shown that emotional authenticity is linked to the contagious nature of emotions (Keller & Becker, 2020). Specifically, inauthentic emotions can be unsuccessful in eliciting emotion contagion if it is interpreted to be insincere (Hennig-Thurau et al., 2006). Thus, jurors should only ‘catch’ anger if they believe the prosecutor is truly angry.

Mechanisms: emotion as social information

While anger can influence behavior via contagion, it can also influence behavior by conveying social information. Social information refers to how an individual makes inferences about another person’s thoughts, feelings and intentions based on emotions (Keltner & Haidt, 1999). In the context of a negotiation, anger conveys a sense of toughness and dominance (Côté et al., 2013). Negotiators perceive tough customers as less willing to make concessions. As a consequence, negotiators tend to reduce their demands (Sinaceur & Tiedens, 2006) and make large concessions (Van Kleef et al., 2004); they will make more compromises to accommodate their tough and angry counterpart to prevent negotiations from deteriorating. However, a reversal can occur when individuals believe their counterpart is feigning anger. Instead of seeing their counterpart as a tough customer, negotiators may view their counterparts as untrustworthy (Côté et al., 2013). Negotiators, faced with an untrustworthy counterpart, become tougher – increasing demands from their counterparts and making fewer concessions. In other words, negotiators resist an inauthentic counterpart’s attempt to ‘dupe’ them.

There is little explicit literature on the type of social information an attorney’s anger communicates to jurors. However, consistent with negotiation research, we theorized that the authenticity of anger expression will tell jurors something about an attorney’s trustworthiness. That is, jurors will lose trust in an attorney who they believe expresses insincere anger. On the other hand, jurors will attribute positive characteristics to an attorney who they believe is genuinely angry. In a study examining attorney-hiring decisions, lay participants were asked to rate their perceptions of an attorney’s competence based on how a closing statement for a murder case was delivered. The experimenters manipulated whether the attorney expressed anger or calmness during the closing statement delivery. Participants rated the angry male attorney as more competent (i.e. confident, commanding and knowledgeable) than the calm male attorney. They also preferred hiring the angry male attorney over the calm one (Salerno et al., 2018).

Source credibility, a concept closely related to competence, refers to the characteristics of a professional – like sociability, friendliness, trustworthiness, extroversion, calmness – which has no bearing on the actual validity or merits of a case. Yet, it has the potential to sway the opinions of lay people (Cramer et al., 2009; McGinnies & Ward, 1980). Expert witness credibility has been shown to persuade jurors to vote in favor of the witness, above and beyond the actual probative value of a case (Cramer et al., 2009; Cutler et al., 1989). In line with Salerno et al. (2018) and research on expert credibility, we expect an attorney’s anger to be an indication to jurors of the prosecutor’s competence. In turn, the jurors’ elevated perception of the prosecutor’s competence would increase the likelihood that they would vote guilty, in favor of the prosecutor. However, when the prosecutor feigns anger, mock jurors will, instead, distrust the prosecutor. This lack of trust – and, perhaps, negative reactance – should backfire against the prosecutor and decrease the odds of a guilty verdict.

Contagion versus social information

According to the EASI model, contagion and social information are not necessarily mutually exclusive (Van Kleef, 2009; Van Kleef et al., 2010). Both mechanisms can work simultaneously to influence human behavior. However, depending on the motivation to deeply process emotion, one mechanism can have a stronger influence on behavior than the other. Specifically, when the motivation to process the social meaning behind emotions is high, social information takes precedence over contagion. To elaborate, Van Kleef et al. (2004) posits that observers’ motivation to process emotions is high when they have: a low need for cognitive closure (Webster & Kruglanski, 1994), low power (Van Kleef, De Dreu, Pietroni, et al., 2006) and plentiful time to think about the meaning behind emotions (De Dreu Carsten, 2003). Conversely, contagion takes precedence when there is a higher need for cognitive closure, higher power and little time to think about the meaning behind emotions. Cognitive closure is characterized by an impatient need to arrive at a quick conclusion without considering all relevant facts (Webster & Kruglanski, 1994). While there is a dispositional component to cognitive closure, the levels of this need can vary based on situational constraints. For instance, the urgency of completing a task may increase the need for cognitive closure (Webster & Kruglanski, 1997).

Applying these principles to a criminal trial scenario, we expect that mock jurors will have a high motivation to process a prosecutor’s anger. Mock jury members should be low on cognitive closure (i.e. do not have a need to arrive to a quick conclusion) because of the high stakes of a criminal trial. If jurors come to an incorrect verdict, they can either place an innocent individual in jail or set a dangerous criminal free. Thus, criminal trials demand a high degree of thought and deliberation. Furthermore, during trial proceedings, jurors have ample time to sit in the jury box and think about why an attorney is angry. Together, low time pressure and low cognitive closure should highly motivate the jury to process the social message communicated by a prosecutor’s anger. Nevertheless, jurors, who are in the position to decide the culpability of a defendant are still high in power. In this respect, jurors should not have an all-consuming motivation to process the social meaning behind an attorney’s anger. The contagious component of anger should still have a hand in how the jury decides the fate of a defendant.

Although we would expect both contagion and perceptions of the prosecutor (i.e. social information) to be reliable predictors of juror verdict, we were interested to see whether one mechanism would have a stronger influence on decision-making behavior than the other. The EASI model says that either one could have a stronger effect depending on the situation (Van Kleef et al., 2010). We wondered whether a mock juror trial would be one of those situations in which social information would be greater than contagion. We did not make any particular a priori predictions regarding this question but conducted a multiple mediation model to test this theory in our exploratory analyses.

The potential moderating role of attorney gender

Our primary contribution was to determine whether perceptions of an attorney’s anger expression would influence juror decision-making via the two routes proposed by the EASI model (Van Kleef, 2009; Van Kleef et al., 2010). However, we were also curious to see whether an attorney’s gender could play a moderating role. In general, jurors are more likely to render a verdict in favor of male than female attorneys (Hahn & Clayton, 1996), and in certain situations, U.S. courts are more likely to rule in favor of male than female attorneys (Szmer et al., 2013; Szmer et al., 2010). Further, the psychological literature suggests that this gender discrimination could be exacerbated by certain expressions of emotion, particularly anger. This is because gender expectations are different in the United States. Men are expected to express anger and pride, whereas women are expected to show sadness and fear (Plant et al., 2000). In situations where women are angry, they are penalized for violating social norms. For instance, female leaders who display anger are viewed as less competent than female leaders who display either no emotions or sadness, whereas male leaders are viewed equally as competent displaying both anger and no emotion (Lewis, 2000). Similarly, female leaders are seen as significantly less competent if their anger is considered dispositional, but for males, regardless of disposition, perceptions of their competence remain the same when they display anger (Schaubroeck & Shao, 2012). During jury deliberations, male mock jurors command a high degree of social influence when expressing anger, whereas a female mock juror loses social influence when expressing anger (Salerno & Peter-Hagene, 2015).

The most recent and relevant study by Salerno et al. (2018) explored how the interaction between attorney and attorney gender would impact employment. The experimenters had participants watch a video of an attorney’s closing statement, in which gender and anger expression were manipulated. The results revealed that participants drew positive inferences about the angry male attorney (i.e. competence, persuasive, knowledgeable), increasing the chances he would be hired, whereas participants drew negative inferences (e.g. shrill, hysterical, awkward) about the angry female attorney, decreasing the chances she would be hired. These results suggest that female attorneys may be seen as less competent when they express anger, whereas male attorneys may be seen as more competent when they express anger. Taken together, women are penalized for expressing anger in a variety of contexts. As a result, we accounted for attorney gender in our methodology.

The current study

Our goal was to investigate how a juror’s knowledge or awareness of an attorney’s authenticity of anger would impact their decision-making. Consistent with Tng and Au (2014) methodology, we provided mock jurors with background information about the authenticity of the attorney’s anger. Then, we used a closing statement as a means for the prosecutor to communicate said anger. As mentioned previously, the closing statement is the most suitable place for an attorney to express emotion. To boost the strength of the manipulation, actors were also instructed to deliver the closing statement in an authentic, inauthentic or non-angry way. After listening to closing statements, jury-eligible participants rendered individual verdicts. After the verdict, participants completed measures of confidence, their own feelings of anger, attorney competence and trustworthiness. We opted to exclude any back-and-forth evidence between the defense attorney and prosecutor for our trial scenario. This is because the current study aimed to establish the basic effect of attorney anger – unfettered by trial evidence. A real-world trial would indeed include evidence from both sides. However, given that attorney emotion was our focus and a novel area of study, we decided the literature would benefit from first establishing the base effect of a prosecutor’s anger before adding other moderators or covariates (e.g. testimony).

Hypotheses

Total direct effects

The effect of prosecutor authenticity by gender on verdict

We hypothesized that mock jurors would be more likely to vote guilty when they were informed that the prosecutor expressed authentic anger than when they were informed that the prosecutor expressed inauthentic or no anger. And mock jurors would be less likely to render a guilty verdict when they were informed that the prosecutor expressed inauthentic anger than the control condition in which the prosecutor expressed no anger. Additionally, we expected this relationship between authenticity of the prosecutor’s anger and verdict to be moderated by the prosecutor’s gender. The rationale for this expectation is described later in the hypothesis regarding the effect of prosecutor authenticity by gender on competence.

The effect of prosecutor authenticity on guilty verdict confidence.

We also hypothesized that mock jurors would be more confident in their guilty verdicts when they were told that the prosecutor expressed authentic anger than when they were told that the prosecutor expressed inauthentic or no anger.

Mediation: the indirect effect of prosecutor authenticity on verdict via mock juror anger

Next, the direct effect of prosecutor anger on verdict would be explained by the mediating role of mock juror anger. Genuine anger should be more contagious to mock jurors than fake or no anger. Therefore, mock jurors would become angrier listening to the authentic than to the inauthentic or non-angry prosecutor. And the angrier the mock jurors got, the more likely they would be to vote guilty.

Mediation: the indirect effect of prosecutor authenticity on guilty verdict confidence via mock juror anger

We also hypothesized that the direct effect of prosecutor anger on guilty verdict confidence would be explained by the mediating role of mock juror anger. As aforementioned, mock jurors would be angrier listening to the authentic than to the inauthentic or non-angry prosecutor. And the angrier the mock jurors got, the more confident (i.e. certain) they would be in a guilty verdict.

Moderated mediation: the indirect effect of prosecutor authenticity by gender on verdict via prosecutor competence

As mentioned previously, we predicted that the authenticity of the prosecutor’s anger would predict mock jury verdict. However, this time, we hypothesized that this link could be mediated by two social–informational variables: perceived prosecutor competence and trustworthiness. Mock jurors would perceive the authentically angry prosecutor as more competent than the non-angry prosecutor. Subsequently, the more competent the prosecutor was perceived, the more likely mock jurors would be to vote guilty. Based on a lack of evidence, we did not have a specific hypothesis for how mock jurors would perceive the competence of an authentically angry prosecutor in contrast to an inauthentically angry prosecutor.

We were also interested in determining whether the prosecutor’s gender would play a moderating role in perceptions of the prosecutor’s competence. Salerno et al.’s (2018) study demonstrated that angry male attorneys drew positive inferences in regard to competence in contrast to angry female attorneys who elicited negative inferences in relation to competence. Then, competence subsequently impacted whether either attorney would be hired. Based on this study, we were curious to see whether the prosecutor’s anger authenticity would interact with the prosecutor’s gender to influence perceptions of the prosecutor’s competence, and whether perceptions of competence would, in turn, predict verdict. However, because hiring is a less serious decision than a criminal verdict, and because of the lack of research in this area, we did not make any strong or specific predictions of how prosecutor gender would moderate perceptions of competence.

Mediation: the indirect effect of prosecutor authenticity on verdict via prosecutor trustworthiness

In terms of trustworthiness, we predicted that an inauthentic prosecutor would be trusted less than a non-angry prosecutor. And the less mock jurors trusted the prosecutor, the less likely they would be to render a guilty verdict.

Method

Participants

A sample of 455 jury-eligible (i.e. over the age of 18, U.S. citizens, with no prior convictions) participants (141 males and 314 females; age ranging from 18 to 83 years), were recruited and compensated $0.35 through Amazon MTurk, an online crowdsourcing service (Litman et al., 2017). The racial background of the sample consisted of 75.2% Caucasian/White, 9.5% Black/African American, 7.0% Asian/Pacific Islander, 4.6% Hispanic/Latinx, 0.4% Native American/Alaska Native and 3.3% Other/Multiracial people. We used Fritz and MacKinnon’s (2007) article to find the required sample size for a small a and b path mediation. The article indicated that about 462 participants would be needed to detect an effect with .8 power. We were only 7 participants short of this cutoff. Also, this study received approval from the institutional review board affiliated with the first author’s university.

To confirm that the mock juror participants successfully received the manipulation in the main study, we asked them, amongst other distractor questions, the gender of the prosecutor and about the authenticity of the prosecutor’s anger. The survey was programmed to immediately remove those who failed the manipulation checks. More specifically, of 691 participants, five participants were removed for incorrectly identifying the prosecutor’s gender, and 98 participants were removed for incorrectly identifying the authenticity of the prosecutor (based on condition). Plus, three outliers were removed. See Results for more details. Further, attention checks were embedded throughout the survey to prevent participants from arbitrarily responding to items. Hence, the remaining 130 disqualified participants were those who failed attention checks, leaving us with N = 455. Finally, ReCAPTCHA was put in place to prevent bots from taking the survey. Written response questions were also utilized to ensure that any bots that breached ReCAPTCHA would be identified and removed from the study. Bots often provide nonsense responses to written response questions (Moss & Litman, 2018).

Materials and personnel

Actors

Our male and female voice actors were both young, White, undergraduate Drama and Theatre majors paid $100 to perform the audio recordings. On average, the participants guessed that the male actor was 36.87 years old and guessed that the female actor was 31.82 years old.

Audio transcript

A prosecutor’s closing statement was adapted from Salerno et al.’s (2018) experiment. The closing statement involved a real case about a mother being brutally murdered in the presence of her one-year-old son. There were points demarcated in the transcript where the actor acted angrily, like banging on the table or raising his/her voice (see Appendix). Salerno et al. tested that this transcript was anger appropriate, addressing the potential problem that the attorneys might be seen as incompetent due to expressing anger at an inappropriate moment.

We justified using an audio recording for a few reasons. Emotion simulation research has shown that people’s mirror neurons fire (Warren et al., 2006) and skin conductance increases (Ramachandra et al., 2009) in response to angry vocalizations – even in the absence of physical facial expressions. In addition, studies have found that people are very effective at recognizing emotions from their nonverbal acoustic signals (Lima et al., 2013). Finally, a number of studies have used audio recordings as stimuli to assess participant perceptions of the authenticity of emotions (McGettigan et al., 2015; Neves et al., 2018). Therefore, an audio recording was deemed as a sufficient method to stimulate anger contagion.

Authenticity manipulation

To manipulate the authenticity of emotion in their study, Tng and Au (2014) provided participants with descriptions about their counterpart’s emotional authenticity: ‘Your counterpart likes to make a point to . . . be frank about how he/she feels’ (p. 310). In line with their methodology, we gave participants three different descriptions explaining the prosecutor’s emotions – authentic, inauthentic or no anger – prior to them listening to the closing statement recording. Additionally, we told participants about the prosecutor’s gender: male or female. The overall design consisted of a 3 (prosecutor emotion: authentic, inauthentic or no anger) × 2 (prosecutor gender: male or female) factorial design, which yielded six different descriptions of the prosecutor’s characteristics.

The authentic description read as:

You are about to listen to an audio recording of Stephen (Stephanie) Warren, the prosecutor, delivering his (her) closing statement regarding the murder of Tiffany Lindsey, a mother of a 1-year-old child. The closing statement is a space where an attorney can passionately weave together a story about how and why a crime occurred. Though Stephen (Stephanie) Warren did not plan to become angry, he (she) became enraged while recounting his (her) version of the events leading up to the murder of Tiffany Lindsey. The thought of a one-year-old child growing up without a mother made him (her) furious.

The inauthentic description read as:

You are about to listen to an audio recording of Stephen (Stephanie) Warren, the prosecutor, delivering his (her) closing statement regarding the murder of Tiffany Lindsey, a mother of a 1-year-old child. The closing statement is a space where an attorney can passionately weave together a story about how and why a crime occurred. In order to move the jury, Stephen (Stephanie) Warren planned to fake anger while recounting his (her) version of the events leading up to the murder of Tiffany Lindsey. He (she) believed that the strategic use of anger could help him (her) gain a persuasive edge in trial.

Lastly, the no anger description read as:

You are about to listen to an audio recording of Stephen (Stephanie) Warren, the prosecutor, delivering his (her) closing statement regarding the murder of Tiffany Lindsey, a mother of a 1-year-old child. The closing statement is a space where an attorney can passionately weave together a story about how and why the crime occurred. Stephen (Stephanie) Warren, nonetheless, maintained a neutral tone while recounting his (her) version of the events leading up to the murder of Tiffany Lindsey.

To be consistent with the manipulation, we also instructed the actors to express real, fake or no anger. We followed Côté et al.’s (2013) protocol for manipulating authenticity in emotional acting. Prior to the inauthentic recording, we directed the actors to externally express anger (e.g. clench their teeth, bang on a table and speak with an angry tone) but maintain a neutral internal state (i.e. to engage in surface acting). Prior to the authentic recording, we had the actors internally experience anger – by having them think about a time someone made them angry – and externally express anger (i.e. deep acting). Bergman Blix (2015) conducted a study of stage actors and demonstrated how highly trained actors engage in surface acting in order to express emotions at rehearsal that are different than the deep acting used during a performance. For example, one of her participants describe the difference in this way: ‘During rehearsal period I live the character’ but during performance they have ‘embodied the character’ (Bergman Blix, 2015, p. 1–15).

For the non-angry recording, the actors simply read the transcript without changing their tone of voice (i.e. they did not raise their voice and emphasize any words marked in the text) and without making extra sound effects (i.e. they did not make any motions with their arms). To prevent feelings of anger from spilling over to subsequent recordings, we instructed the actors to perform the recordings in the following order: no anger, inauthentic and authentic. We also made sure that the actors did not recently experience any angering events prior to the recording.

Verdict

Mock jurors were asked to fill in a guilty or not guilty statement: ‘If I were a juror, I would find Steven Dixon (the defendant) [blank] of murder.’ In addition, an abridged burden of proof statement was presented before the verdict.

Verdict confidence scale

A single-item percentage scale, ranging from 0% to 100%, was used to assess how confident mock jurors were in their verdict.

Positive and Negative Affect Schedule Expanded Form

The Positive and Negative Affect Schedule Expanded Form (PANAS–X) was used to assess current levels of juror anger (Watson & Clark, 1994). The PANAS–X contained six hostility items – angry, hostile, irritable, scornful, disgusted and loathing – which were used to assess participant anger on a scale ranging from 1 (very slightly or not at all) to 5 (extremely; α = .91). Recent studies have used the hostility items as proxy to appraise anger (Nuñez et al., 2017; Nuñez et al., 2015). Other emotions in the PANAS–X, such as sadness, were included to divert attention away from our target emotion, anger.

Prosecutor competence

A modified 10-item (i.e. commanding, strong, confident, knowledgeable, powerful, convicted, competent, influential, charismatic and passionate) positive inferences measure from the Salerno et al. (2018) study (p. 392) was utilized to assess perceived competence (α = .94). Two items, convincing and persuasive, were removed from the original 12-item scale because they appeared redundant with verdict, our main dependent variable. The scale ranged from 1 (strongly disagree) to 7 (strongly agree).

Prosecutor trustworthiness

Van Kleef, De Dreu, & Manstead (2006) 3-item scale was adapted to assess overall trustworthiness (α = .95). On a scale ranging from 1 (not very trustworthy) to 7 (very trustworthy), the first item asked, ‘To what extent do you believe the attorney is trustworthy?’. On a scale ranging from 1 (unreliable) to 7 (very reliable), the second question asked, ‘To what extent do you believe the attorney is reliable?’. On a scale ranging from 1 (very dishonest) to 7 (very honest), the third item asked, ‘To what extent do you believe the attorney is honest?’.

Procedure

The study was conducted through Qualtrics, an online survey tool. The first page of the online survey contained the terms of consent. After agreeing to consent, participants were randomly assigned to listen to the closing statement of one of six prosecutors: authentic-male, authentic-female, inauthentic-male, inauthentic-female, non-angry-male or non-angry-female. Prior to hearing the recording, participants read a description of the prosecutor, consistent with their assigned condition. Thereafter, mock jurors listened to the prosecutor’s closing statement. They then answered two manipulation check questions. The first question asked about the gender of the prosecutor, and the second question asked about the authenticity of the prosecutor’s anger. After the mock jurors passed the manipulation check, they rendered a verdict. Following the verdict, they rated: their confidence in their verdict; their level of anger on the PANAS–X; and their perceptions of the prosecutor, respectively. Lastly, mock jurors were asked to provide demographic information and were given a secret code to claim their monetary compensation.

Results

Data screening and corrections for multiple contrasts

There were two issues with the data. First, there were violations of homogeneity of variance for a few group-based analyses. Wherever there were violations of homogeneity for planned contrast t tests, we referred to Levene’s test for p values and degrees of freedom. Second, we made an a priori decision to exclude any outliers exceeding the absolute value of 3.29 standard deviations (|z| > 3.29; Tabachnick & Fidell, 2013). Three outlying cases – two cases reporting extreme trustworthiness and one case reporting extreme anger – were completely removed from all analyses.

For analyzing planned contrasts following a significant analysis of variance (ANOVA), we used a method endorsed by Howell (2013): for a small number of specific planned contrasts, they recommend employing independent samples t tests and utilizing the Dunn Šidák correction to account for Type I error. For any two planned contrasts following significant ANOVAs, we used the Dunn Šidák correction. The Dunn Šidák correction indicated that a criterion of p < .03 should be used to claim statistical significance with two planned contrasts involving independent samples t tests.

Focal analyses

Total direct effects

The effect of prosecutor authenticity by gender on verdict

While accounting for prosecutor gender and the interaction between prosecutor authenticity and gender, we hypothesized that mock jurors would be more likely to render a guilty verdict if they were told that the prosecutor expressed authentic anger than if they were told that the prosecutor expressed inauthentic or no anger, and that mock jurors would be less likely to render a guilty verdict if they were told that the prosecutor expressed inauthentic anger than if they were told that the prosecutor expressed no anger. This hypothesis was partly supported. A binary logistic regression revealed that authenticity reliably predicted verdict when the authentic was contrasted with the inauthentic (reference group) condition (b = 1.25, SE = 0.46), χ²(1) = 7.41, p = .006; Exp(β) = 3.49, 95% confidence interval, CI [1.42, 8.58].¹ Mock jurors’ odds of a guilty verdict were 249% greater when they knew the prosecutor conveyed authentic anger than when they knew the prosecutor conveyed inauthentic anger. But authenticity did not predict verdict when the authentic was contrasted with the non-angry (reference group) condition (b = 0.87, SE = 0.46), χ²(1) = 3.60, p = .06, and when the inauthentic was contrasted with the non-angry (reference group) condition (b = −0.38, SE = 0.38), χ²(1) = 1.00, p = .32. See Table 1. Also, the main effect of prosecutor gender (with male as the reference group) was not significant (b = 0.41, SE = 0.45), χ²(1) = 0.83, p = .36. The interaction between authenticity and prosecutor gender was not significant for the authentic versus inauthentic (reference group) condition contrast (b = −1.19, SE = 0.64), χ²(1) = 3.47, p = .06, for the authentic versus non-angry condition (reference group) contrast (b = −0.70, SE = 0.58), χ²(1) = 1.45, p = .23, and for the inauthentic versus non-angry condition (reference group) contrast (b = 0.49, SE = 0.57), χ²(1) = 0.72, p = .40. Therefore, prosecutor gender and the interaction did not have a significant effect on verdict.

Table 1.

Cross-tabulation table of observed counts and percentages across the authenticity and verdict.

Authenticity	Verdict
	Not guilty		Guilty		Total
	Observed count	Percentage of total	Observed count	Percentage of total	Observed count	Percentage of total
No anger	40	23.0	134	77.0	174	100
Authentic	27	16.2	140	83.8	167	100
Inauthentic	30	26.3	84	73.3	114	100
Total	97	21.30	358	78.70	455	100.0

A chi-square test of independence analyzing authenticity only in relation to verdict revealed that verdict depended on authenticity, χ²(1) = 4.32, p = .04; mock jurors were 67% more likely to render a guilty verdict when they knew the prosecutor was being authentic than when they knew the prosecutor was being inauthentic.

The effect of prosecutor authenticity on guilty verdict confidence

We also hypothesized that mock jurors who voted guilty would be more confident in their guilty verdict if they were informed that the prosecutor expressed authentic anger than if they were informed that the prosecutor expressed inauthentic or no anger. This hypothesis was not supported. A two-way ANOVA indicated a non-significant mean difference in guilty verdict confidence across authenticity, F(2, 352) = 2.00, p = .14, η² = .01. Also, the main effect of attorney was not significant, F(1, 352) = 1.78, p = .18, η² = .01. However, the interaction between authenticity and prosecutor gender was significant, F(2, 352) = 3.34, p = .04, η² = .02. This result was unexpected because we did not predict that prosecutor gender would moderate the influence of prosecutor authenticity on guilty verdict confidence. See Table 2 for means and standard deviations.

Table 2.

Effect of mean prosecutor anger authenticity by gender on guilty verdict confidence.

Prosecutor anger authenticity	Prosecutor gender
	Male	Female
	M (SD)	M (SD)
Inauthentic	78.52 (18.75)	69.83 (27.60)
Authentic	81.94 (17.35)	76.66 (21.77)
No anger	77.03 (20.93)	81.93 (19.75)

Note: Anger authenticity: inauthentic versus authentic versus no anger.

To probe this unexpected interaction, we used Tukey’s honestly significant difference (HSD) to identify the pairwise differences. Tukey’s (HSD) was used instead of independent t tests because these sets of contrasts were unplanned and exploratory. Tukey’s HSD revealed two marginally significant contrasts. For the first exploratory contrast, mock jurors reported marginally higher levels of guilty verdict confidence when they encountered the authentically angry male prosecutor (M = 81.94, SD = 17.35) than when they encountered the inauthentically angry female prosecutor (M = 69.83, SD = 27.60), M_diff = 12.11, 95% CI [–0.18, 24.40], p = .06. However, the means were crossed (i.e. both authenticity and gender were different), meaning it would be difficult to determine whether it was gender or authenticity driving these marginal differences. For the second exploratory contrast, mock jurors reported marginally higher levels of guilty verdict confidence when they encountered the non-angry female prosecutor (M = 81.93, SD = 19.75) than when they encountered the inauthentically angry female prosecutor (M = 69.83, SD = 27.60), M_diff = 12.10, 95% CI [–0.07, 24.26], p = .052. Given the marginal significance of these contrasts and the fact that the unexpected interaction was not based on theory, we did not make any further inferences or interpretations about these effects.

Mediation: the indirect effect of prosecutor authenticity on verdict via mock juror anger

Next, we hypothesized that mock jurors would become angrier listening to the authentic than listening to the inauthentic or non-angry prosecutor. This hypothesis was supported. A two-way ANOVA demonstrated a significant mean difference in mock juror anger across the authenticity condition, F(2, 449) = 15.98, p < .001, η² = .07. In addition, there was no main effect of attorney gender, F(1, 449) = 0.56, p = .46, η² = .001, and no interaction between prosecutor authenticity and gender, F(2, 449) = 1.46, p = .24, η² = .01. See Table 3 for means and standard deviations. Following the significant main effect of authenticity, two planned contrasts t tests were conducted, which revealed that mock jurors were angrier in the authentic (M = 2.67, SD = 1.11) than in the inauthentic (M = 2.02, SD = 0.91) or non-angry (M = 2.19, SD = 1.06) conditions, t(269.94) = 5.30, p < .001, 95% CI [0.40, 0.88], g = 0.62;² t(339) = 4.05, p < .001, 95% CI [0.24, 0.71], g = 0.44, respectively. Therefore, mock jurors were angrier when they recognized that the prosecutor expressed real rather than fake or no anger.

Table 3.

Effect of mean prosecutor anger authenticity by gender on mock juror anger.

Prosecutor anger authenticity	Prosecutor gender
	Male	Female
	M (SD)	M (SD)
Inauthentic	2.07 (0.90)	1.95 (0.93)
Authentic	2.80 (1.07)	2.56 (1.14)
No anger	2.12 (1.03)	2.25 (1.08)

Note: Anger authenticity: inauthentic versus authentic versus no anger.

After analysis of the prosecutor’s authenticity predicting mock juror anger (predictor to mediator), we hypothesized that elevations in mock juror anger would increase the likelihood of a guilty verdict (mediator to outcome). Supporting this hypothesis, a binary logistic regression showed that a unit increase in anger led to a 78% improvement in the odds of a guilty verdict, while controlling for the prosecutor authenticity (b = 0.58, SE = 0.13), χ²(1) = 19.23, p < .001; Exp(β) = 1.78, 95% CI [1.38, 2.31].

Although the total direct effect of the authentic versus non-angry contrast on verdict was not significant, we still opted to run the overall hypothesized mediation model because finding a total direct effect (predictor to outcome) is not necessarily a pre-requisite for analyzing an indirect effect (Hayes, 2018). Also, given the non-significant interaction between prosecutor authenticity and gender in the predictor-to-mediator path from the analysis mentioned previously, we excluded gender as a moderator from the hypothesized mediation model. We hypothesized that authenticity would indirectly influence verdict via mock juror anger. Given that we were interested in the contrast between authentic and inauthentic anger and the contrast between authentic and no anger, we examined two different indirect effects. Because Hayes’ PROCESS Macro is based on ordinary least squares regression (Hayes, n.d.), and the analysis of verdict required logistic regression, we used the mma (multiple mediation analysis) function from a statistical package in R. The mma function is able to analyze mediation and multiple mediation models with dichotomous outcomes (Yu & Li, 2017). Specifically, its default argument is set to the binomial family (link = ‘logit’) for generalized linear models involving binary outcomes (Yu & Li, 2017). The mma function, with 5000 bootstrapped samples, produced a significant indirect effect, M_{bootstrapped estimates} = 0.25, 95% CI [0.02, 0.50], for the authentic versus no anger (reference group) contrast, and a significant indirect effect, M_{bootstrapped estimates} = 0.39, 95% CI [0.06, 0.70], for the authentic versus inauthentic (reference group) contrast. See Figure 1.³ Consistent with our prediction, knowledge of the authenticity of the prosecutor’s anger had a significant indirect effect on verdict via mock juror anger.

Figure 1.

Mock juror anger mediates the relation between the authenticity of prosecutor anger and verdict. The M values represent the average estimates of bootstrapped samples. *p < .05.

Mediation: the indirect effect of prosecutor authenticity on guilty verdict confidence via mock juror anger

Relatedly, we hypothesized that mock juror anger would predict guilty verdict confidence. In support of this hypothesis, a binary logistic regression demonstrated that a unit increase in mock jurors’ anger subsequently increased guilty verdict confidence by 4 units, while controlling for prosecutor authenticity (b = 4.00, SE = 1.02), t = 3.91, p < .001, r_partial = .20.

Dropping gender again, we then ran another mediational model to see whether prosecutor authenticity would indirectly influence guilty verdict confidence through mock juror anger. As previously mentioned, the total direct effect of the authenticity on guilty verdict confidence was not significant. Nonetheless, based on Hayes’ (2018) recommendation, we proceeded to look at the indirect effect. Hayes’ (2018) PROCESS Macro, Model 4, with 5000 bootstrapped samples, generated a significant indirect effect (b = 1.74, SE = 0.68, 95% CI [0.55, 3.24]) for the authentic versus non-angry (reference group) contrast and generated a significant indirect effect (b = 2.30, SE = 0.79, 95% CI [0.95, 4.05]) for the authentic versus inauthentic (reference group) contrast. See Figure 2. Consistent with our hypothesis, awareness of the authenticity of the prosecutor’s anger indirectly influenced confidence in a guilty verdict through mock juror anger.

Figure 2.

Mock juror anger mediates the association between the authenticity of prosecutor anger and verdict confidence. *p < .05.

Mediation: the indirect effect of prosecutor authenticity on verdict via prosecutor competence

While accounting for prosecutor gender and the interaction between prosecutor authenticity and gender, we hypothesized that mock jurors would perceive the authentic prosecutor as more competent than the non-angry prosecutor. This hypothesis was partially supported. According to a two-way ANOVA, here was a main effect of prosecutor authenticity, F(2, 449) = 70.06, p < .001, η² = .24. That is, mock jurors perceived the authentically angry prosecutor (M = 4.29, SD = 0.56) to be more competent than the non-angry prosecutor (M = 3.32, SD = 0.93), t(284.61) = 11.73, p < .001, 95% CI [0.81, 1.13], g = 1.25. However, contrary to expectations, there was no main effect of prosecutor gender, F(1, 449) = 0.22, p = .64, η² < .001, and no interaction between prosecutor authenticity and gender, F(2, 449) = 0.56, p = .57, η² = .002.⁴ See Table 4 for means and standard deviations.

Table 4.

Effect of mean prosecutor anger authenticity by gender on perceptions of prosecutor competence.

Prosecutor anger authenticity	Prosecutor gender
	Male	Female
	M (SD)	M (SD)
Inauthentic	3.43 (0.99)	3.42 (0.81)
Authentic	4.31 (0.55)	4.28 (0.56)
No anger	3.25 (0.88)	3.39 (0.97)

Note: Anger authenticity: inauthentic versus authentic versus no anger.

Next, we hypothesized that a mock juror’s increased perceptions of the prosecutor’s competence would improve the odds of a guilty verdict. Consistent with this hypothesis, a binary logistic regression revealed that a unit increase in perceived prosecutor competence corresponded to a 178% increase in the odds of a guilty verdict, while controlling for prosecutor authenticity (b = 1.02, SE = 0.16), χ²(1) = 41.90, p < .001; Exp(β) = 2.78, 95% CI [2.04, 3.78].

As per our hypotheses, we had originally planned to run a moderated mediation with prosecutor gender as the moderator. However, because the prosecutor’s gender and the interaction (between prosecutor’s authenticity and gender) did not significantly predict the competence (the mediator) in the previous analysis, we dropped prosecutor gender from the model and ran a simple mediation – with competence mediating the relationship between authenticity and verdict. The mma function (Yu & Li, 2017), with 5000 bootstrapped samples, generated a statistically significant indirect effect, M_{bootstrapped estimates} = 1.02, 95% CI [0.48, 1.58], for the authentic versus no anger (reference group contrast). See Figure 3. Therefore, our prediction was once again supported. The authentic anger indirectly affected verdict through juror perceptions of the prosecutor’s competence.

Figure 3.

Perceptions of prosecutor competence mediates the relation between the authenticity of prosecutor anger [authentic vs. no anger (reference group)] and verdict. The M values represent the average estimates of bootstrapped samples. *p < .05.

Mediation: the indirect effect of prosecutor authenticity on verdict via prosecutor trustworthiness

Lastly, we hypothesized that the inauthentic prosecutor would be trusted less than the non-angry prosecutor. This hypothesis was not supported. A two-way ANOVA revealed no main effect of trustworthiness, F(2, 449) = 1.15, p = .32, η² = .01. Plus, there was no significant main effect of prosecutor gender, F(1, 449) = 0.06, p = .80, η² < .001, and no interaction between authenticity and prosecutor gender, F(2, 449) = 0.10, p = .90, η² < .001. See Table 5 for means and standard deviations. Moreover, while controlling for the prosecutor authenticity, a binary logistic regression indicated that trustworthiness did not predict verdict as hypothesized (b = 0.06, SE = 0.09), χ²(1) = 0.44, p = .51. Given the predictor-to-mediator and mediator-to-outcome path of the mediation model was not significant, we did not proceed to analyze the hypothesized mediation model.

Table 5.

Effect of mean prosecutor anger authenticity by gender on perceptions of prosecutor trustworthiness.

Prosecutor anger authenticity	Prosecutor gender
	Male	Female
	M (SD)	M (SD)
Inauthentic	5.31 (1.14)	5.36 (1.18)
Authentic	5.43 (1.24)	5.34 (1.31)
No anger	5.21 (1.50)	5.15 (1.38)

Note: Anger authenticity: inauthentic versus authentic versus no anger.

Exploratory analyses: multiple mediation   model

Given that perceived prosecutor competence and mock juror anger were both reliable intervening variables, we decided to simultaneously include both in a multiple mediation model to see whether one mediator would better explain the association between prosecutor authenticity and verdict than the other. When the perception of the prosecutor’s competence was the mediator, the mma function (Yu & Li, 2017) indicated a significant indirect effect when the authentic was contrasted with the no anger (reference group) condition, M_{bootstrapped estimates} = 0.92, 95% CI [0.31, 1.50]. However, when mock juror anger was the mediator, mma produced a non-significant indirect effect when the authentic was contrasted with the no anger (reference group) condition, M_{bootstrapped estimates} = 0.18, 95% CI [–0.05, 0.46]. See Figure 4. Thus, while controlling for one another, perceptions of the prosecutor’s competence maintained a significant contribution to the relationship between prosecutor anger and verdict, whereas mock juror anger did not. Consistent with the EASI model, this result suggests that social information (i.e. attorney authenticity) exerted a greater influence on verdict above and beyond contagious emotions (i.e. mock juror anger).

Figure 4.

Perceptions of prosecutor competence and mock juror anger mediate the association between the authenticity of prosecutor anger [authentic versus no anger (reference group)] and verdict in a simultaneous fashion. The M values represent the average estimates of bootstrapped samples. M1 = perceptions of prosecutor competence. M2 = mock juror anger. *p < .05.

Post hoc analyses

There was a concern that participants would perceive a difference in the level of anger expressed between the male and female prosecutor and between the authentic and inauthentic prosecutor. These differences in perceived anger could serve as alternative explanations to the results of our study. To address this concern, we conducted a follow-up study with 177 MTurk participants. Without having the prompt, participants were asked to indicate to what degree they thought the prosecutor in the voice recording felt anger (along with other distractor emotions) on a 1 (very slightly or not at all) to 5 (extremely) scale. In the no-anger condition, there was no statistically significant difference in perceived anger between the male (M  =  2.72, SD = 1.31) and female prosecutor (M  =  2.79, SD = 1.26), t(55) = 0.18, p = .86, 95% CI [–0.74, 0.62], g = 0.05. In the authentic condition, there was no statistically significant difference in perceived anger between the male (M  =  4.04, SD = 1.14) and female (M  =  4.28, SD = 1.19) prosecutor, t(55) = 0.78, p = .44, 95% CI [–0.86, 0.38], g = 0.20. For the inauthentic condition, there was no statistically significant difference in perceived anger between the male (M  =  3.94, SD = 1.12) and female prosecutor (M  =  4.31, SD = 0.82), t(54.84) = 1.52, p = .14, 95% CI [–0.88, 0.12], g = 0.37. Lastly, we found no statistically significant difference in perceived anger between the authentic (M  =  4.16, SD = 1.16) and inauthentic condition (M  =  4.13, SD = 1.00), t(118) = 0.16, p = .88, 95% CI [–0.36, 0.42], g = 0.03. These results confirm that there were no significant differences in the perceived anger. Therefore, it is unlikely that perceived differences in anger had a reliable impact on our findings.

Discussion

The purpose of the current study was to determine whether mock jurors’ knowledge of the prosecutor’s emotional authenticity would predict their verdict. We used the EASI model to describe two possible paths – contagion and social information – in which mock jurors could be influenced by the prosecutor’s anger (Van Kleef, 2009; Van Kleef et al., 2010). Accordingly, we found evidence for both. When mock jurors learned that the prosecutor was legitimately angry, they became angry. In turn, this anger drove them to render a guilty verdict and increased their confidence in this verdict. In addition, when mock jurors were aware that the prosecutor’s anger was authentic, they perceived the prosecutor as competent. Elevated perceptions of the prosecutor’s competence increased odds of conviction from mock jurors. Furthermore, the exploratory multiple mediation analysis showed that competence was a more significant mediator than mock juror anger. However, contrary to our predictions, mock jurors’ awareness of the prosecutor’s inauthentic anger did not decrease their perceptions of trustworthiness. Also, counter to expectations, attorney gender did not moderate the perceptions of attorney competence. The theoretical and practical implications of these findings are discussed below.

Theoretical implications

Anger, social information and contagion

Past research has found that anger triggers punitive tendencies in potential jurors. However, most of this research has been based on anger aroused from case facts/evidence presented during a trial (e.g. Bright & Goodman-Delahunty, 2006; Nuñez et al., 2017; Salerno & Peter-Hagene, 2013). To our knowledge, there has been only a handful of studies that have examined the interpersonal effect of anger on jury psychology (e.g. Salerno & Peter-Hagene, 2015; Salerno et al., 2018). The current study was the first to demonstrate how the interpersonal aspects of perceived attorney anger (i.e. social information) could drive potential jury members to be more punitive. Specifically, it showed that improving the perceptions of prosecutor competence produced higher conviction rates. Future research should explore the interpersonal impressions of other emotions (e.g. sadness) upon jury psychology.

In addition, this study showed that contagious anger could arise even when participants were informed of attorney anger. It could drive people to convict and be more confident in their guilty verdict. This increased confidence provides support for the Appraisal Tendency Framework, which posits that anger can make people more certain about the world around them (Lerner & Keltner, 2000). Furthermore, as seen in the exploratory multiple mediation model, prosecutor competence (social information) provided a greater contribution to mock juror verdict than intrapersonal, contagious anger. In accordance with the EASI model (Van Kleef, 2009; Van Kleef et al., 2010), the jury verdict scenario may have created the conditions (i.e. low cognitive closure and plentiful time) where social information would have a greater impact than contagion.

We also investigated whether the prosecutor’s gender would moderate the indirect effect of anger expression on juror verdict through perceptions of competence. However, we did not find a statistically significant result. This finding is not totally surprising. Some studies on the characteristics of expert witnesses have found gender effects on jury verdicts (e.g. Maeder et al., 2012), but others have not (e.g. Couch & Sigler, 2002). These mixed findings suggest there may be unaccounted factors that might better predict when we might see gender prejudice in jury decision-making.

For example, in the present experiment, we used Salerno et al.’s (2018) closing statement, which told a story of a mother being murdered in front of her child. Although gender bias was found in their study, such bias was not found in the present study. However, the context of their study was different. Their outcome of interest was hiring decisions, whereas our outcome was legal decision-making. Legal decisions are different from hiring decisions as they have direct and immediate consequences for a criminal defendant (e.g. being convicted of a crime and being incarcerated). Given this difference, legal decisions may be dependent on different moderators. One study on the Supreme Court revealed that legal decisions or rulings regarding women’s issues were often made in favor of female attorneys (Szmer et al., 2010). Though Supreme Court decision-making is not the best exemplar of decisions made in the typical trial court, it illuminates the importance of considering the type of case when studying gender biases against attorneys. Because a mother was killed by a man in the present scenario, it is possible that participants regarded this murder as a ‘women’s issue’ in favor of the female prosecutor, counteracting the effects of emotion-based gender biases on legal decision-making.

Inauthentic versus authentic anger expression

Business management research has shown a link between authenticity and emotional contagion. One finding is that an employee’s genuine happiness tends to evoke happiness in customers, while faking happiness does not (e.g. Hennig-Thurau et al., 2006). The present study expanded on what was previously found by generalizing authenticity and contagion to anger and applying these two concepts to a psycho-legal context. Specifically, the present results showed that mock jurors became angry when they were informed that the prosecutor delivered a closing statement with real rather than fake anger. In addition, real anger increased perceptions of the prosecutor’s competence, consistent with social information and Salerno et al.’s (2018) findings.

However, there was not enough evidence to conclude that the inauthentic expressions of anger would backfire on the attorney. Inconsistent with previous work (Côté et al., 2013; Tng & Au, 2014), mock jurors who knew about the prosecutor’s fake anger did not trust this prosecutor any less than the prosecutor who was genuinely angry or not angry. One possible explanation for failing to reject the null hypothesis is that most research on trust in relation to the authenticity of anger is grounded in negotiations. Negotiation is a circumstance that is not exactly analogous to a jury trial scenario. Jury members may not rely on the perceptions of attorney trust to render a verdict as much as a buyer may rely on perceptions of seller trust to make a purchase. To elaborate, jury members have nothing to lose in their one-way interactions with the prosecutor and sit in the position of decision-making power. Jury members can choose to ignore the prosecutor’s argument if they wish and decide whether a defendant should be punished. The onus is on the prosecutor to convince the jury that his or her version of the story is correct. In contrast, in a negotiation situation, both the buyer and seller are invested in a two-way interaction, which will have direct consequences for both parties. The trust between the buyer and seller is essential for both parties breaking even. If one deceives the other, the buyer or seller may lose out on a trade. For example, if the seller lies about the accident history of a car being purchased, the buyer will have problems with the car down the road. Therefore, these discrepancies may explain why the effects of trustworthiness were not found in the trial setting.

Practical implications

What do the results mean for the legal world? Remember the angry mock jurors were more confident in their guilty verdict. This confidence may have downstream implications for jury deliberations. Research has shown that the least confident mock jurors are most likely to change their opinions after deliberations (Pritchard & Keenan, 2002). Thus, prosecutors may benefit from having angry jury members confidently in favor of a guilty verdict prior to deliberations.

With the current data, we also found that a prosecutor’s expression of anger in a murder trial pushed jury members to render a guilty verdict. In particular, the perceptions that a prosecutor was showing an authentic expression of anger improved perceptions of the prosecutor’s competence and at the same time provoked mock juror anger. In turn, both mock juror anger and prosecutor competence increased the odds of a guilty verdict. Furthermore, we found that perceptions of the prosecutor’s competence mattered more than participant anger in guiding mock juror verdict. However, we did not replicate Côté et al.’s (2013) finding that inauthentic anger would violate trust and consequently backfire on the attorneys. Therefore, we cannot conclude that failing to appear genuinely angry would pose a risk to a practicing prosecutor. Nonetheless, this does not mean that attorneys should not be cautious when expressing emotions in court. In general, the literature suggests that an individual’s knowledge of fake emotions seems to generate inauspicious consequences (Tng & Au, 2014). Until this area of research is further developed, this point is moot.

The practical implications cited here might be limited to cultures like that in the United States where anger displays are more overt. As previously mentioned, Swedish researchers have found that legal professionals in their country are cautioned and taught to manage their emotions, and that displays of anger are often subtle (Bergman Blix & Wettergren, 2016; Flower, 2014). Anger displayed in courtrooms in countries such as Sweden might yield different reactions from jurors, whether they are authentic or not.

Future directions and limitations

In most real-world cases, jury members would hear direct examinations, cross examinations and arguments from both the prosecution and defense. The defense would present a series of exculpatory facts in favor of the defendant, while the prosecution would present a series of inculpatory facts against the defendant. The present study only included the prosecutor’s closing statement. Thus, ecological validity may have been diminished. Nevertheless, the purpose of this experiment was to determine whether the expression of anger from a prosecutor, not the actual case, would impact jury verdict. Moreover, even with a burden of proof statement and a dearth of actual incriminating evidence against the defendant, a significant percentage of mock jurors rendered guilty verdicts. Thus, our manipulation did have an impact on decisions. Future studies should build on the current study by incorporating more trial evidence along with the attorney’s expression of anger. These studies could elucidate whether the effect we found generalizes to a more extended trial context and identify the conditions in which the persuasiveness of trial evidence would be enhanced or hindered by anger. Research suggests that anger is less effective at persuading individuals under certain conditions. For instance, one study found that in the context of charitable giving, conveying anger is significantly less effective in soliciting donations than conveying disappointment (Van Doorn et al., 2015). This finding implies that the instrumentality of anger may depend on the conditions in which it is presented. Given that anger tends to elicit higher conviction rates (Georges et al., 2013; Nuñez et al., 2017; Nuñez et al., 2015; Semmler & Hurst, 2017), perhaps anger expression is less effective for defense attorneys whose job is to get their client acquitted of a crime. Future studies could test this idea.

Another limitation of the current study was the lack of stimulus sampling (Wells & Windschitl, 1999). Stimulus sampling involves employing multiple examples of a stimulus for an experimental manipulation. Having multiple stimulus samples averages out any peculiarities due to a singular instance of a stimulus sample. If these peculiarities are not averaged out across multiple stimuli, the odds of spurious effects (Type I error) or false negatives (Type II error) may increase (Wells & Windschitl, 1999). In the present study, we used only one male and one female actor to deliver the closing statement. There could have been peculiarities of our single male or female actor, such as the attractiveness of vocal tone or dynamics, that could have explained the non-significant effects. Future studies examining the interaction between attorney gender and anger should employ various stimulus samples to reduce this potential error.

The manipulation could have limited internal validity because we gave participants information about the authenticity of the attorney’s anger in the prompt and instructed the actors to express anger with varying degrees of authenticity. We did both to strengthen the effect of the manipulation and reduce Type II error. However, it is unclear whether the prompt, instructions or both drove the effects. Also, in a real-world scenario, jurors would not be told of the prosecutor’s authenticity through a written prompt. They would have to discern deceit on their own. To address these issues, we recommend that in future studies examining attorney emotions, researchers only use vocal stimuli in order to increase internal and ecological validity. Regardless of these limitations, the current study demonstrates what would happen if mock jurors knew the prosecutor was being fake or legitimate about their anger. That is, mock jurors would favor the authentically angry prosecutor. Moreover, the current study opens up an avenue for further investigation on the interpersonal effects of attorney emotions in the courtroom.

Finally, it was awkward to implement the closing statement in the no-anger condition. Throughout the script for the closing statement, there were anger-related themes and statements, such as, ‘ . . . they found a horrific murder scene . . . ’. Therefore, while the actors were tasked with delivering the closing statement with no anger (i.e. reading it with a neutral tone and making no arm movements), the contents of the script were not neutral. Nevertheless, given the nature of the case being a murder trial and the need for the actors to express anger in the authentic condition, the angry themes in the script were unavoidable. Further, the results of our study showed that participants assigned to the authentic anger condition were significantly angrier than those assigned to the no-anger and inauthentic condition, which suggests that it was the manipulation that made the participants angry, not necessarily the content.

Conclusion

The impact of attorney emotion on jury decision-making is understudied. The current experiment helped to fill the gap in this literature by investigating the effect of prosecutor anger. It was found that the prosecutor’s authentic expression of anger increased the odds of a guilty verdict through mock juror anger and perceptions of the prosecutor’s competence – which is consistent with contagion and social information, two paths proposed by the EASI model (Van Kleef, 2009; Van Kleef et al., 2010). In other words, the prosecutor gained an advantage by delivering their closing statement in (authentic) anger. Thus, the current evidence bolsters the claim of trial advocacy books – that attorneys can gain an edge by expressing certain emotions (e.g. anger) in court.

Appendix.

No anger and angry closing statement, respectively

Tiffany was a young single mom of a one-year-old son named Kendall. She had a family who cared about her, including an aunt who rented an apartment to try and help Tiffany get started in life. Tiffany was doing the best that she could for her and her son. Tiffany also had a circle of friends who cared about her. They called her Little Tiff. Tiffany’s best friend was a girl named Tammi Brown, and Tammi Brown had a boyfriend. Tammi Brown’s boyfriend was this man, the defendant, Steven Dixon. On October 12, 2015, the defendant went to a party, and then after that he went to Tiffany’s apartment. And sometime early in the morning of October 13^th, something set the defendant off. He turned on Tiffany! He beat her in the face, knocking out several of her teeth and fracturing her skull. He slashed at her face, at her neck, and at her head. He cut her with such force that he broke not just one knife, but two knife blades! The first knife landed next to Tiffany’s body on the floor. The second knife blade stayed lodged in her neck from where he stabbed her. And when he stabbed her he cut her carotid artery. When the defendant was done with Tiffany, when he was satisfied that she was dead, he tried to cover what he had done with a blanket. And then he ran from that apartment, and he left behind one-year-old Kendall alone with his mother’s body. When Tiffany’s friends showed up at the apartment a little while later, they found a horrific murder scene. They found Tiffany on the floor, and they found one-year-old Kendall crying. . . . Covered in blood from crawling around his mother’s body. [PAUSE FOR DRAMATIC EFFECT, CUT-OFF POINT]

Tiffany was a young single mom of a one-year-old son named Kendall. She had a family who cared about her, including an aunt who rented an apartment to try and help Tiffany get started in her life. Tiffany was doing the best that she could for her and her son. Tiffany also had a circle of friends who [RAISE VOICE] CARED about her. They called her Little Tiff. Tiffany’s best friend was a girl named Tammi Brown, and Tammi Brown had a boyfriend. Tammi Brown’s boyfriend was [RAISE VOICE – CONTEMPT] THIS man [POINT ANGRILY], the DEFENDANT, Steven Dixon. On October 12, 2015, the defendant went to a party, and then after that he went to Tiffany’s apartment. And sometime early in the morning of October 13^th, something set the defendant off. [RAISE VOICE MORE] He TURNED on Tiffany! He BEAT [HIT TABLE] her in the FACE, knocking out several of her teeth and FRACTURING HER SKULL. He SLASHED [SLASH HAND MOTION] at her face, at her neck, and at her head. He CUT her with SUCH FORCE that he broke not just one knife, but TWO knife blades! The first knife landed next to Tiffany’s body on the floor. The second knife blade stayed LODGED in her NECK from where he stabbed her. And when he stabbed her he CUT her carotid artery. [LOUD, ANGRY INCREDULOUSNESS] When the defendant was DONE with Tiffany, when he was SATISFIED that she was dead, he tried to cover what he had done WITH A BLANKET. And then he ran from that apartment, and he left behind ONE [SLAM]-YEAR-[[SLAM]-] OLD [[SLAM]-] Kendall alone with his mother’s body. [STILL SOMEWHAT LOUD/ANGRY EMPHASIS] When Tiffany’s friends showed up at the apartment a little while later, they found a HORRIFIC murder scene. They found Tiffany on the floor, and they found one-year-old Kendall CRYING . . . COVERED IN BLOOD from crawling around his mother’s body. [PAUSE FOR DRAMATIC EFFECT, CUT-OFF POINT]

Ethical standards

Declaration of conflicts of interest

Samuel Choi has declared no conflicts of interest

Narina Nuñez has declared no conflicts of interest

Benjamin M. Wilkowski has declared no conflicts of interest

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee [Institutional Review Board at the University of Wyoming; Protocol #20181116SC02192] and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study