Miami University The Graduate School Certificate for Approving the Dissertation We hereby approve the Dissertation of Matthew Elliot Groebe Candidate for the Degree: Doctor of Philosophy ______________________________ Director Dr. Garold Stasser ______________________________ Reader Dr. Susanne Abele ______________________________ Reader Dr. Carrie Hall ______________________________ Graduate School Representative Dr. Monica Schneider ABSTRACT BEHAVIORAL MIMICRY IN THE COURTROOM: PREDICTING JURORS’ VERDICT PRFERENCE FROM NONCONSCIOUS MIMICRY OF ATTORNEYS by Matthew Elliot Groebe Mimicry is an unconscious reaction of imitating other people’s behaviors, postures, and facial expressions (Chartrand & Bargh, 1999). It has been shown to lead to a host of positive outcomes, such as increased liking and persuasiveness. Mimicry has not yet received any empirical attention in the courtroom. This research examines behavioral mimicry as a predictor of verdict preference. Specifically, the primary research question was whether a juror’s mimicry of the plaintiff’s attorney and defense attorney predicts verdict preference. Six mock trial videotapes were used (43 jurors in total). Jurors’ mimicry behaviors, as well as commonly held nonverbal indicators of agreement and disagreement, were coded. It was hypothesized that overall mimicry would predict final predeliberation verdict preference, and that mimicry would be a stronger predictor of verdict preference than nonverbal agreement behaviors or disagreement behaviors. The hypotheses were partially supported. Although overall mimicry did not predict final predeliberation verdict, mimicry did predict verdict preference on a segment-by-segment basis. Furthermore, mimicry was a stronger predictor of verdict preference than nonverbal agreement behaviors or disagreement behaviors. These results suggest that mimicry may be a subtle means of communicating agreement with an attorney as she presents her argument. Attorneys can focus on jurors’ mimicry as a tool for deselecting unfavorable jurors during jury selection and for assessing jurors’ temporary preferences and reactions during evidence presentation. BEHAVIORAL MIMICRY IN THE COURTROOM: PREDICTING JURORS’ VERDICT PRFERENCE FROM NONCONSCIOUS MIMICRY OF ATTORNEYS A Dissertation Submitted to the Faculty of Miami University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Psychology by Matthew Elliot Groebe Miami University Oxford, Ohio 2013 Dissertation Director: Dr. Garold Stasser Table of Contents List of Tables............................................................................................................................ iii List of Figures...........................................................................................................................iv Acknowledgments......................................................................................................................v Introduction....................................................................................................................................1 An overview of mimicry............................................................................................................1 The evolutionary basis of mimicry............................................................................................ 1 Theories behind mimicry........................................................................................................... 3 Consequences of mimicry..........................................................................................................4 Moderators for amount of mimicry............................................................................................5 Legal application........................................................................................................................6 Other nonverbal behaviors.........................................................................................................7 Overview of research................................................................................................................. 8 Methods...........................................................................................................................................9 Materials.....................................................................................................................................9 Procedure....................................................................................................................................9 Results........................................................................................................................................... 10 Interrater reliability.................................................................................................................. 10 Descriptive statistics................................................................................................................ 11 Predicting final verdict preference from overall mimicry....................................................... 12 Segment-by-segment analyses: Predicting subsequent preference..........................................12 Segment-by-segment analyses: Predicting mimicry................................................................14 Discussion..................................................................................................................................... 14 Strengths.................................................................................................................................. 16 Limitations............................................................................................................................... 16 Implications and recommendations......................................................................................... 17 Future directions...................................................................................................................... 18 Conclusion............................................................................................................................... 19 Literature Cited............................................................................................................................25 Appendices....................................................................................................................................27 Appendix A. Sample verdict preference item..........................................................................27 Appendix B. Nonverbal behavior list...................................................................................... 28 ii List of Tables Table 1. Interrater reliabilities, means, and standard deviations (by segment) for coded nonverbal behaviors. __________________________________________________________20 Table 2. Descriptive statistics for nonverbal agreement and nonverbal disagreement behaviors for plaintiff’s attorney and defense attorney across mock trials. ________________________21 iii List of Figures Figure 1. The main regression analysis, in which total mimicry of the plaintiff’s attorney and total mimicry of the defense attorney across all four segments were used to predict final predeliberation preference. _____________________________________________________22 Figure 2. A secondary analysis conducted to determine the relationship between prior preference, mimicry behavior, nonverbal agreement behaviors, nonverbal disagreement behaviors, and subsequent preference. Analyses were conducted on a segment-by-segment basis. The exception was that prior verdict preference for segment 1 could not be included because there was no prior preference for this first segment. Note that there are six predictors (Mn, SMn, FLn, HNn, ACn, and HSn) included in each segment, Sn. Significant regression coefficients are marked with an asterisk.__________________________________23 Figure 3. The significant predictors from the segment-by-segment analyses. ______________24 iv Acknowledgments I would like to thank Gary Stasser for his continued help and direction with this dissertation, as well as for providing invaluable guidance throughout the graduate program. I would also like to thank Susanne Abele, Carrie Hall, and Monica Schneider for their participation on the dissertation committee. I would like to thank Gregory Tortoriello and Melissa Zbacnik for their tireless devotion to coding. And I would like to give a special thank you to Kevin CosgriffHernandez and all of Tara Trask and Associates for their guidance and for allowing access to their mock trial materials. I would also like to thank my parents, Beth and Keith Groebe, for their love and support throughout my time at Miami. I would also like to thank my brother and sister, Brad and Dana Groebe, for always being there for me. And finally, I would like to thank Alexandra Hummel for her love and support. All of these people helped me along the way and made it possible for me to complete my degree. I thank you all. v Understanding how a juror is responding at the moment towards a particular argument can confer a large advantage to a litigator at trial. Not only can this information help attorneys successfully strike unfavorable jurors during jury selection, but it can also provide an indicator to an attorney as to how her argument (and the argument of her adversary) is resonating with a particular juror during trial. If a juror is not responding favorably to one’s message, this can foretell the need to alter one’s message to make it more favorable to that juror. While jurors do not announce their preferences out loud during trial, attorneys may be able to rely on certain nonverbal behaviors to help illuminate a juror’s current preference. One such nonverbal behavior that has not received empirical attention in the context of the courtroom is behavioral mimicry. The aim of this paper is to determine whether jurors’ mimicry behavior can foretell agreement with a particular argument. First, a general overview of mimicry will be presented, followed by a discussion of the evolutionary basis and theories behind of mimicry. Then, the paper will present the consequences of mimicry and some moderators for the amount of mimicry displayed. This will be followed by the legal application of mimicry and a discussion of other relevant nonverbal behaviors. Finally, an overview of the current research will be outlined. An Overview of Mimicry Mimicry is broadly defined as an unconscious reaction of imitating other peoples’ behaviors, movements, postures and facial expressions (Chartrand & Bargh, 1999). Expanding upon this definition, the first component of mimicry is that it normally occurs outside of conscious awareness. Thus, mimickers are unaware that they are imitating the mimickee. Note that mimicry can also occur as a result of deliberate intention (Yabar et al., 2006), but such mimicry is not the focus of the present research. The second component of the definition is that there are a host of possible interpersonal stimuli one can mimic. For one, individuals can mimic the facial expressions of others. Dimberg et al. (2000) found that people smile the most when exposed to happy faces, and show the most facial sadness when exposed to sad faces. Facial mimicry can also facilitate emotional mimicry via emotional contagion. Thus, when we mimic others’ facial expressions, we in a sense “feel” these same emotions felt by the mimickee (Lundquist & Dimberg, 1995). Verbal mimicry is another type of mimicry in which people adopt each other’s accents, speech rate, utterance duration, and latency to speak (Cappella & Planalp, 1981; Giles & Coupland, 1991; Gregory et al., 1997). The final type of mimicry is behavioral mimicry, which refers to the adoption of the mannerisms, posture, gestures, and motor movements of one’s interaction partner. For example, shaking one’s foot when one’s interaction partner is shaking her foot or leaning forward when one’s interaction partner is leaning forward. Behavioral mimicry, along with a lesser emphasis on facial mimicry, will be the main focus of the current research. Chartrand & Bargh (1999) coined the term “chameleon effect” to refer to the passive and unintentional tendency to adapt to our social surroundings and match the behaviors of others in our social environment. As will be discussed below, we are quite adept at altering our behavior to blend in with our surroundings. And, as recent work suggests, we are not even aware of our tendency or of others’ tendencies to engage in this behavior (Chartrand & Bargh, 1999; Chartrand & van Baaren, 2009). Taken as a whole, the literature suggests that humans routinely engage in mimicry throughout their daily lives. The Evolutionary Basis of Mimicry The discovery of mirror neurons in macaque monkeys and a similar mirror system in humans suggests a close link between perceiving an action and performing that same action (Iacoboni et al., 1999). Mirror neurons are neurons that fire both upon perceiving someone else 1 engage in an action, and upon oneself engaging in that same action (Chartrand & van Baaren, 2009). For instance, perceiving a certain behavior such as leg crossing automatically activates our own motor representation of crossing our own legs (Iacoboni et al., 1999). The presence of these mirror neurons suggests that this architecture that allows us to mimic is innate. However, the actual act of mimicry may be a learned action, as neurons that wire together fire together (Chartrand & van Baaren, 2009). In other words, repeated pairings or associations of perceptions and actions may help mimicry become automatized in an individual. Thus, although mimicry itself may not be innate, we appear to come pre-equipped with the mirror neurons that make such mimicry possible. The mirror neuron system supporting mimicry is so hard-wired that one-month-old infants have been shown to smile, open their mouths, and stick out their tongues when they see someone else doing the same (Meltzoff & Moore, 1977). And this tendency to mimic progresses rapidly. By nine months of age, infants can mimic more abstract emotional expressions such as joy, sadness, and anger (Termine & Izard). In addition to the biological evidence on mirror neurons, evolutionary psychology can also help explain the purpose of mimicry. Human beings are social animals (Aronson, 1999). Our daily lives are filled with social interactions. We talk to significant others as we prepare for our day, chat about current events with our co-workers over lunch, and spend our precious hours of recreation time with family and friends (Lakin et al., 2003). Given the importance of other people in our daily lives, we are strongly motivated to ensure that our social interactions are successful. In our less predictable and more dangerous evolutionary past, our ancestors lived in a harsh and unforgiving environment where individuals who were alone were at a survival and reproductive disadvantage (Buss & Kenrick, 1998). In order to survive and reproduce, individuals were forced to rely on others to complete necessary survival activities (e.g. hunting, gathering, protecting against predators, and raising offspring) (Lakin et al., 2003). Thus, harmonious group living was essential to early human survival. Individuals who were able to maintain cordial group relationships were more likely to be included in the group and thereby survive (de Waal, 1989). On the contrary, individuals who were unsuccessful at maintaining good standing with the group were unlikely to survive. Baumeister and Leary (1995) argue that humans have developed a fundamental need to belong. Since group living improved the odds of survival and transmission of genes to future generations, humans are deeply motivated to ensure that they remain in good standing with their social groups. Any behavior that increased the odds of remaining in one’s social group would be selected for, whereas any behavior that decreased those odds would be selected against. Thus, individuals who were able to maintain successful group relations were at an evolutionary advantage, and consequently their social strategies and techniques were passed on to future generations. Before the advent of language, nonverbal behaviors would have carried significant weight in portraying our inner states to others. In a sense, mimicry helps convey the message “I am like you” to the mimickee (Lakin et al., 2003). Given the significance of nonverbal behaviors in communicating important messages to others (DePaulo & Friedman, 1998), these adaptive behaviors would have become automatized over time, or able to occur without conscious awareness or intention (Bargh, 1990). The automatic nature of such nonverbal behaviors would make it much easier to affiliate with group members without excessive thought (Lakin et al., 2003). As will be discussed later, mimicry is one such nonverbal behavior that helps maintain positive interpersonal relationships. 2 Thus, while initially having survival value by facilitating communication, mimicry may have evolved to serve a “social glue” function, helping to increase affiliation, bind people together and create harmonious group relationships (Lakin et al., 2003). Automatically engaging in mimicry would increase the odds that the individual be included in the group and acquire her share of resources. The automatic nature of mimicry makes this behavior all the more adaptive. If one can engage in mimicry automatically and without intention, other cognitive resources are freed up to help navigate complex social situations. Mimicry plays such a prominent role in our daily interactions with others, such that even though participants who are not mimicked are not consciously aware of the lack of mimicry, they still show enhanced salivary cortisol levels and hence become stressed over the absence of mimicry (Kouzakova et al., 2010). This suggests that on an unconscious level we expect to be mimicked in our social interactions, and when that mimicry is lacking, we become distressed. Such distress, in turn, can lead to steps to attempt to reaffiliate to restore our belongingness needs (Baumeister & Leary, 1995). Theories Behind Mimicry William James’s principle of ideomotor action holds that merely thinking about a behavior increases the tendency to engage in that behavior (James, 1890). Chartrand and Bargh (1999) draw on this principle of ideomotor action in explaining their proposed perceptionbehavior link. They argue that the coding system for perceiving behaviors in others is the same as for performing those behaviors. In other words, there is a great deal of overlap between perceiving an action and performing that same action. So when we see someone touch their face, brain areas associated with touching our own face are activated as well. In their seminal study, Chartrand and Bargh (1999, study 1) had participants interact with two other participants (actually confederates) when doing a photo description task. The participant’s duty was to describe a series of photos to the confederate. As the participant described the photos, the confederate either shook her foot or rubbed her face. A hidden video camera surreptitiously recorded the participant’s behavior for evidence of behavioral mimicry. After the participant had finished the photo description task with the first confederate, the participant did a second photo description task with the second confederate. This second confederate performed whichever behavior the first confederate did not. Thus, if the first confederate shook her foot, the second confederate rubbed her face. If the first confederate rubbed her face, the second confederate shook her foot. Again, participants were video recorded for evidence of behavioral mimicry during the task. It was found that participants tended to shake their foot more when in the presence of the foot-shaking confederate, and they tended to rub their face more when in the presence of the face-rubbing confederate, showing evidence for behavioral mimicry. Two aspects about this study are noteworthy. First, participants mimicked a complete stranger with whom they had no expectation of future interaction. Second, when funnel debriefed at the end of the study, participants were completely unaware of their mimicry behavior. The authors interpreted these results as an indication of a perception-behavior link, or a direct expression account, which views mimicry as a passive, nonconscious, and unintentional behavior that occurs in the absence of the motivation to affiliate. This automatic nature of mimicry makes it quite effective, as it occurs without intent and allows the mimicker to direct cognitive resources elsewhere. Although early evidence (e.g. Chartrand & Bargh, 1999) suggested that mimicry occurs via a passive perception-behavior link, more recent evidence suggests that mimicry can be situationally moderated. Thus, instead of viewing mimicry as an indiscriminate behavior, later research views mimicry as a motivated process in which perceivers are more likely to mimic in 3 schema-consistent situations as opposed to schema-inconsistent situations (Dalton et al., 2010). If mimicry is indeed a motivated process, it would follow that we would be more likely to mimic certain people because we are more motivated to do so. Recent research supports this assertion. Recent research suggests that we are more likely to mimic those with whom we share similar opinions (van Swol & Drury, unpublished), same-race targets compared to cross-race targets (Dalton et al., 2010), politicians we support (McHugo et al., 1991; Bourgeois & Hess, 2008), ingroup members compared to outgroup members (Yabar et al., 2006), when we have an affiliation goal (Lakin & Chartrand, 2003), and those with whom we share the same name and academic major (Gueguen & Martin, 2009). Thus, mimicry is applied selectively and is contextdependent (Lakin & Chartrand, 2003). It is important to note that mimicry still occurs in the absence of a reason to affiliate (Lakin & Chartrand, 2003; Stel et al., 2010). Thus, the passive, automatic perception-behavior link still occurs even without the desire to affiliate. However, the motivation to affiliate, or a situation in which mimicry would be schema-consistent, can increase the tendency to mimic because it may strengthen the perception-behavior link. In other words, when we are highly motivated to affiliate, we tend to perceive more in our environment, hence increasing our tendency to mimic. Consequences of Mimicry Mimicry has been shown to have numerous positive effects, both upon the mimicker and the mimickee. Chartrand and Bargh (1999, study 2) were the first to empirically examine whether mimicry increases liking. Similar to their first study, participants took part in a photo description task with a confederate. In this second study, however, the confederate either mimicked or did not mimic the participant during the photo description task instead of viceversa. Subsequent to the interaction, participants were asked to rate their liking of the confederate as well as how smoothly the interaction went. Participants who had been mimicked by the confederate reported greater liking of the confederate than participants who had not been mimicked. Mimicked participants also reported that the interaction with the confederate had gone more smoothly than participants who were not mimicked. The link from mimicry to liking is not a one-way street. Lakin et al. (2003) report a bidirectional relationship between mimicry and liking. Thus, we also tend to mimic people we like more. Stel et al. (2010) manipulated a priori liking by telling participants the target woman was on a television show because she was either advocating helping others in need or advocating that it was a waste of time. A control condition was also included in which participants were not given any background information about the target woman. Participants were videotaped for mimicry behavior as they watched the silent video. The authors found that participants mimicked the target more in the liked target condition than in the disliked target condition. There were no differences in mimicry behavior in the disliked target condition and in the control condition, suggesting that mimicry may be the default behavior, and we tend to increase our tendency to mimic when we wish to affiliate. Participants also reported feeling more similar to the liked target. Furthermore, liking and similarity were both highly related and both led to increased mimicry. And, as one might expect, the relationship between liking and mimicry is cyclical. We mimic those we like more, which in turn leads to greater liking. This greater liking leads to more mimicry. Increased liking then occurs for both the mimicker and the mimickee. Researchers have recently attempted to study how mimicry can be used in the more applied context of persuasion. Bailenson and Yee (2005) had computerized avatars present a counter-attitudinal message to participants. The avatar either subtly mimicked the participant’s head movements or followed a pre-recorded set of movements as it delivered the persuasive 4 message. Participants who had been mimicked reported more agreement with the persuasive message than participants who had not been mimicked. Maddux et al. (2008) studied mimicry in the context of a negotiation between business school students. Participants were told to either mimic or not mimic their negotiation partner. Mimicking participants were able to create a larger joint gain for both themselves and their negotiation partner, as well as receive a larger personal gain compared to nonmimicking participants. Finally, Tanner et al. (2008, studies 2 and 3) examined mimicry in a laboratory sales context. Salespeople advocating for a sports drink who mimicked the consumer were more successful at getting the consumer to enjoy their drink, intend to buy their drink, predict future success for their drink, and consume more of their drink than were nonmimicking salespeople. Moderators for Amount of Mimicry Although Chartrand and Bargh (1999) posit that mimicry occurs because of a passive process via the perception-behavior link, as noted earlier more recent research suggests that the amount of behavioral mimicry can be moderated by situational and personality variables. One major moderator variable involves the degree to which the mimicker likes the mimickee. As discussed above, we still engage in some baseline amount of mimicry with a disliked person, but behavioral mimicry increases when the target is liked (Stel et al., 2010). We are also more likely to mimic when we have an activated affiliation goal (Lakin & Chartrand, 2003). An activated affiliation goal would be more likely with a liked target than with a disliked target. Similarity between the mimicker and the mimickee also leads to increased behavioral mimicry. In an unpublished study, van Swol and Drury examined how attitudinal similarity impacts mimicry. In their study, participants chose one of two places to take a hypothetical vacation. They then met with two other participants (actually confederates) and discussed their vacation preferences. One confederate agreed with the participant’s choice, while the other confederate disagreed with the participant’s choice. Participants mimicked the agreeing confederate to a larger extent than the disagreeing confederate. McHugo et al. (1991) had participants watch videos of two politicians (President Ronald Reagan and Senator Gary Hart) during a debate. Reagan supporters tended to mimic Reagan more than Hart, whereas Hart supporters tended to mimic Hart more than Reagan. Bourgeois and Hess (2008) found similar results with a French Canadian sample watching a debate between two French Canadian politicians. Thus, we tend to mimic those with whom we share similar attitudes to a greater extent than those with whom we have dissimilar attitudes. Incidental similarity can also moderate behavioral mimicry. Gueguen and Martin (2009) manipulated the name of the target as well as the target’s academic major. They found that participants were more likely to mimic targets with the same name and academic major than targets with different names or academic majors. This effect of incidental similarity on mimicry was mediated by liking. In other words, we tend to like similar others. This increased liking then leads to greater mimicry. Ingroup/outgroup status is a construct relevant to both liking and similarity, in that we tend to express greater liking for ingroup members and we view ourselves as more similar to ingroup members than to outgroup members. Indeed, it has been found that we mimic ingroup members to a greater extent than outgroup members (Yabar et al., 2006). In addition to the more situationally related variables, there are also a few personality related variables that have been shown to moderate behavioral mimicry. Chartrand and Bargh (1999, study 3) found that those high in empathy, specifically the perspective-taking cognitive facet of empathy, engaged in greater levels of behavioral mimicry than did participants lower in empathy. While empathy can be situationally manipulated, there are general personality 5 differences in empathy as well. High self-monitors also display more behavioral mimicry than low self-monitors, especially when there is more perceived self-other similarity (Cheng & Chartrand, 2003). Another personality variable that moderates behavioral mimicry is selfconstrual. Van Baaren et al. (2003b) found that individuals with an interdependent self-construal – due either to priming or to chronic differences from being raised in an Eastern culture – mimicked a confederate more than individuals with an independent self-construal. The research reviewed above suggests that there is a motivational aspect to mimicry. Therefore, although the perception-behavior link may help explain the tendency to mimic, behavioral mimicry can be situationally moderated. This suggests that mimicry may be selectively applied to help one maximize the chances of positive outcomes from mimicry. Hence, despite the fact that mimicry is nonconscious and unintentional, it is still employed selectively toward certain targets. This underscores the evolutionary advantage behind behavioral mimicry and why it would have been selected for throughout our evolutionary history. Legal Application To our knowledge, no study has empirically examined behavioral mimicry in the courtroom. Besides mimicry among jurors likely to occur during deliberation, it is also likely that jurors will show a tendency to mimic the attorneys as they present their arguments to the court. Identifying such behavioral mimicry would be a very useful tool for attorneys and trial consultants as they evaluate potential jurors in voir dire as well as attempt to determine how well their case is resonating during trial. As mentioned above, van Swol and Drury, in an unpublished study, found that participants tended to mimic targets with whom they agreed more than targets with whom they disagreed. The topic of conversation was vacation destination preference, and thus did not have the decisional ramifications as a civil trial with millions of dollars at stake or a criminal trial with a life at stake. Nonetheless, the results imply that we mimic those with similar attitudes. Furthermore, as reviewed earlier, we tend to mimic politicians whom we agree with more than politicians with whom we disagree (McHugo et al., 1991; Bourgeois & Hess, 2008). In a courtroom, we typically do not have strong pre-existing attitudes towards the attorneys for each side to the same extent that we have strong pre-existing attitudes towards politicians, but again the evidence suggests that we show greater mimicry toward those whom we support. Not only do we tend to mimic those we support, but such behavioral mimicry can also further increase our preference for the targets of our mimicry. Tanner et al. (2008, study 1) found that engaging in mimicry can actually shift one’s preferences. In their study, participants mimicked the snacking behavior of a confederate in a video as the confederate chose between a bowl of goldfish and a bowl of animal crackers. Although both snacks were present, the confederate in the video ate one snack exclusively. Participants watching the video of the confederate eating goldfish exclusively tended to eat goldfish, whereas participants watching the video of the confederate eating animal crackers exclusively tended to eat animal crackers. Furthermore, these mimicking behaviors impacted their endorsed preferences, such that participants in the confederate goldfish consumption condition reported greater liking of goldfish than animal crackers. The opposite results were found for participants in the confederate animal cracker consumption condition. These results are interesting because they suggest that engaging in behavioral mimicry can actually shift our preferences to align with our behavior. Thus, if we mimic one attorney more than the other, this can further cement our agreement with that side. 6 Also, as reviewed above, we are more likely to be persuaded when we are mimicked than when we are not mimicked (Maddux et al., 2008; Bailenson & Yee, 2005). The reverse question seems logical. Are we more likely to mimic someone who is persuading us than someone who is not persuading us? Individuals are more persuaded by those whom they like, trust, and to whom they feel similar (Cialdini, 2001). Mimicry has been shown to foster feelings of liking, trust, and similarity in the mimicker and mimickee (Chartrand & van Baaren, 2009; Stel et al., 2010; Lakin et al., 2003). Thus, can engaging in mimicry make one more receptive to certain persuasive messages? The evidence reviewed above suggests that attorneys and trial consultants may be able to observe the amount of behavioral mimicry in jurors and use it as a predictor of verdict preference. If we tend to mimic those with similar attitudes to a greater extent than those with dissimilar attitudes, then behavioral mimicry may be a very subtle indicator of agreement. In voir dire, differential amounts of mimicry towards each side may indicate agreement with or at least greater receptivity to one side’s arguments. Thus, attorneys would want to be attentive to mimicry behavior among jurors during voir dire and use their peremptory challenges accordingly. Attorneys would also want to take note of behavioral mimicry during evidence presentation. Are jurors engaging in greater mimicry of one side over the other? If so, it could suggest that certain arguments are resonating well with jurors, whereas other arguments need to be adjusted. Knowledge of how individual jurors are responding to arguments during evidence presentation would be crucial. While deliberation is a fundamental component of a trial, there is substantial evidence suggesting that it does little to change attitudes. Due to informational and normative social influence (Deutsch & Gerard, 1955), the majority opinion pre-deliberation is the final jury verdict 97% of the time (Kalven & Zeisel, 1966). Therefore, understanding how individual jurors are responding to a case can paint a very accurate picture of how a jury will decide. Other Nonverbal Behaviors While behavioral mimicry is the primary focus of this line of research, a secondary aim is to examine nonverbal behavior more generally. It has been estimated that 60-65% of people’s total communication occurs through nonverbal behaviors (Frederick, 2006). Although mimicry is one type of nonverbal behavior, it does not tell the whole picture. Therefore, this study also evaluates how different nonverbal behaviors may be predictive of verdict preference as well. Specifically, we focused on nonverbal behaviors commonly thought of as indicators of agreement and disagreement. Importantly, there is no single nonverbal indicator of agreement or disagreement (Frederick, 2006). But, when evaluating how jurors are receiving their case, attorneys have a tendency to focus on single nonverbal cues, such as smiling or head nodding. However, relying on single nonverbal cues is a dangerous mistake as a single cue can have different meanings. For instance, fixed eye gaze may indicate interest and agreement, but it may also indicate hostility to one’s argument (Frederick, 2006). Similarly, forward lean may indicate interest and receptivity, but may also indicate hostility. Conversely, arm folding may indicate disagreement with the speaker, but could also suggest that the juror is cold. Although certain cues are fairly reliable indicators of agreement (e.g. head nod, smile) and disagreement (e.g. head shake, eye roll, arm folding), they are not always infallible (Bousmalis et al., 2009). Thus, relying on single nonverbal cues as indicators of agreement or disagreement has its drawbacks. Putting too much stock in interpreting jurors’ attitudes through their nonverbal behaviors can backfire because people are fairly well versed at controlling their behaviors. Especially in a 7 formal courtroom setting, jurors may attempt to conceal outward signs of disagreement. Mimicry, on the other hand, is a behavior that occurs outside of conscious awareness, and as such, should be less amenable to conscious control. Nonetheless, a secondary purpose of this study will be to examine the predictive ability of certain select nonverbal behaviors that attorneys perceive to be indicators of agreement and disagreement. For instance, is a juror who nods her head often when the defense attorney is presenting likely to favor the defense? Such an examination can potentially allow attorneys to compare the predictive ability of behavioral mimicry against other nonverbal behaviors. Overview of Research The purpose of the current research is to evaluate a potential trial strategy that attorneys can use in court to help maximize success. The main question is whether jurors’ mimicry behavior predicts their verdict preference. A secondary question is whether other nonverbal behaviors commonly thought to indicate agreement or disagreement hold any predictive validity for verdict preference. These are pivotal questions that can help attorneys and trial consultants organize their trial strategy. Specifically, attorneys can take notice of jurors’ mimicry behavior during voir dire when deciding how best to use their peremptory challenges. Attorneys can also observe jurors’ mimicry behavior during evidence presentation to provide a quick snapshot of how well their arguments are resonating with the jury. To answer these questions, we coded mock trial videotapes taken from recent civil mock trials conducted by a national trial consulting firm. These videos feature a split-screen, in which one side of the screen focuses on the mock jury, and the other side is split between the attorney and the attorney’s visual presentation. Trained assistants focused on the mock jurors and coded for predetermined target nonverbal behaviors as well as the time at which they occurred. The first author performed the same task while focusing solely on the attorneys. Mimicry indices were calculated by matching up the attorney behaviors with the juror behaviors. Verdict preferences, as expressed on Likert scales, were assessed after each side presented a set of arguments. A primary aim of this research was to examine the predictive ability of mimicry behavior (and other select nonverbal behaviors) as a predictor of verdict preference. Thus, it is hypothesized that behavioral mimicry will predict subsequent agreement with a message. However, the reverse direction can also be the case. Given its bidirectional nature, it may also be that agreement with a message predicts behavioral mimicry. As noted earlier, single nonverbal behaviors are not always the best predictors of agreement or disagreement. They nonetheless might have some predictive validity. It is hypothesized, however, that mimicry will be a better predictor of verdict preference than the other nonverbal behaviors. This study adds to the existing literature by taking the field of mimicry to a new, applied domain. This research adds to the existing literature because it uses a community sample of mock in a relatively naturalistic context making an important decision that, at least for real jurors, would have lasting consequences. Furthermore, unlike other research involving a clear persuasion context (McHugo et al., 1991; Bourgeois & Hess, 2008), this research involves mimickees (the attorneys) about whom little is known. In other words, there should not be strongly pre-defined attitudes towards the mimickees in the current study. Lastly, in these mock trial videos, the mock jurors knew that they would be deliberating after evidence presentation. Thus, this would lead to engagement with the trial and motivation to pay attention. Together, all of these factors help increase the ecological validity of the research. The goal is to be able to identify the predictive ability of behavioral mimicry in a very applied context with big decisional ramifications. 8 Methods Materials Mock trial videotapes were provided by Tara Trask LLC. The videos were posted to a server to which the authors had been granted access. Six mock trial videos fit the necessary requirements for the research study. These requirements included that the defense attorney and plaintiff’s attorney remain constant throughout all portions of the trial, at least four mock jurors remain in the video frame throughout the mock trial, and the attorneys remain in the video frame when presenting their arguments. All mock trial videos involved civil trials and varied in length from two hours to eight hours. The videos were always presented in a split-screen format, where the mock jurors were on one side of the screen and the other side was split between the presenting attorney and the attorney’s presentation. Mock jurors wore a number tag to identify their juror number. To evaluate juror verdict preferences, the authors obtained entrenchment charts from Tara Trask LLC. These charts display jurors’ verdict preferences and are broken down by each segment in the mock trial. For instance, jurors are asked their preference after the plaintiff’s opening statement. They are then asked their preference after the defense opening statement, their preference after the plaintiff evidence presentation, and so on. Each juror’s preference on the entrenchment chart is delineated by her juror number (as seen in the videos). Preferences ranged from 1-4 for each side. Thus, a preference of 1 for the plaintiff designates a strong proplaintiff “opinion that cannot be changed”, whereas a preference of 4 for the plaintiff means a pro-plaintiff “opinion that would be easy to change”. Similarly, a preference of 1 for the defense means a strong pro-defense “opinion that cannot be changed”, whereas a preference of 4 for the defense means a pro-defense “opinion that would be easy to change”. See Appendix A for a sample preference scale. (Note that the scale for some of the older mock trial videos was 1-6; in these instances the authors converted these preferences to a 1-4 scale.) From this measure we created a Likert scale from -4 to +4 assessing verdict preference. A preference of +4 signaled a very strong pro-defense opinion, whereas a preference of -4 signaled a very strong pro-plaintiff opinion. As mentioned above, juror preferences were assessed after each segment in the mock trial. Procedure Six mock trial videos fit the necessary criteria outlined above. In the videos, mock jurors were normally seated in three rows of eight jurors, although the exact seating arrangements varied from trial to trial. We only included mock jurors seated in the front row who were fully visible. The rationale behind these criteria was that we did not want anyone blocking the juror and the attorney. This could create a sense of interpersonal distance, and more practically, it would be more difficult to code behavioral mimicry in jurors in the second and third rows. Depending on the mock trial, we coded anywhere between four and eight mock jurors. We coded the nonverbal behavior for 43 mock jurors throughout the six mock trials. We used both attorneys as targets of mimicry. To achieve this, we coded four segments per mock trial video, two involving the plaintiff’s attorney and two involving the defense attorney. We attempted to code a plaintiff and defense segment early in the mock trial (e.g. plaintiff opening statement, defense opening statement) and a plaintiff and defense segment towards the end of the mock trial (e.g. plaintiff closing statement, defense closing statement). Therefore, we followed the general outline of plaintiff segment, defense segment, plaintiff segment, and then defense segment. In essence, we were testing whether jurors who mimic the 9 defense attorney to a greater extent than the plaintiff’s attorney were more likely to side with the defense attorney, and vice versa. Two undergraduate research assistants were trained by the authors on how to code the videos. They learned how to recognize and identify various nonverbal behaviors. A coding protocol was created to outline the target nonverbal behaviors that were focused on in the research (see Appendix B). Both coders were blind to the specific research hypotheses. The research assistants’ duty was to independently view the four predetermined segments for each mock trial, focusing on each eligible juror and noting each instance of the target nonverbal behaviors as well as the time at which they occurred. The first author separately viewed the four segments for each mock trial, focusing on the attorneys as they presented their arguments. The first author noted each instance of the target nonverbal behaviors and the time at which they occurred. Thus, there was a list of behaviors performed both by each of the jurors and a list of behaviors performed by the attorneys. These lists also contained the times at which these behaviors occurred. Juror behaviors were classified as instances of mimicry if they occurred within 10 seconds of the attorney’s behavior. For instance, if the attorney touched his face at 14:35, a juror face touch at 14:40 was classified as mimicry. However, if the juror touched her face at 14:50, it was not considered an instance of mimicry. Note that several of the nonverbal behaviors listed in Appendix B are indicators of agreement and disagreement. If these behaviors occurred absent similar behavior from the attorney or outside the 10-second window, they were counted as instances of agreement or disagreement and not as instances of mimicry. Mimicry indices were also calculated based on a 5-second window; however there were no material differences between the results for the shorter and longer time window. Thus, only analyses using the 10-second window will be presented. The setup of these videos is slightly different than most mimicry studies. First, the attorney is standing (sometimes behind a podium) while the mock jurors are sitting. This precludes any sort of mimicry in the lower half of the body. Second, the attorneys are exclusively presenting while the mock jurors are exclusively listening. Thus, many hand gestures that the attorneys make would be unlikely to be mimicked by the mock jurors. To address this issue, the list of codable mimicry behaviors had to include behaviors that would be plausible for the jurors to engage in. Behaviors that both the attorney and mock jurors could plausibly engage in (e.g. face touching, forward lean, and head nodding) are outlined in Appendix B. After the jurors’ nonverbal behaviors were coded in each of the four trial segments, preferences from the entrenchment charts after each of the four relevant trial segments were matched up with each qualifying juror. As discussed above, verdict preferences were converted to a -4 to +4 Likert scale, with -4 indicating a very strong pro-plaintiff preference, and +4 indicating a very strong pro-defense preference. Results Interrater Reliability Two trained coders viewed all four segments of each of the six mock trial videos. They focused on the mock jurors and noted specific instances of target nonverbal behaviors. Interrater reliability analyses were conducted to determine the amount of agreement between the two coders. Reliability estimates for the frequencies of each of the target nonverbal behaviors generated by the two coders are listed in Table 1. As all mock jurors were double-coded, reliability estimates were adjusted using the Spearman Brown Prophecy Formula: 10 Spearman Brown correlation coefficient = 2(r)/1+r Target nonverbal behaviors with a Spearman Brown correlation coefficient of 0.6 and above were included in the mimicry indices. Nonverbal behaviors included in the mimicry indices were: face touching (r = 0.994); hair touching (r = 0.983); neck touching (r = 0.797); general body touching (r = 0.917); forward lean (r = 0.908); posture shift (r = 0.689); cross arms (r = 0.908); adjust clothes (r = 0.929); adjust accessories (r = 0.821); smile (r = 0.759); yawn (r = 0.912); head nod (r = 0.966); and head shake (r = 0.625). Nonverbal behaviors excluded from the mimicry indices and further analyses due to unacceptable interrater reliability included: arm touching (r = 0.573); backward lean (r = 0.587); uncross arms (r = 0.431); frown (r = -0.017); and gaze aversion (r = 0.152). Descriptive Statistics Combining across the six mock trials, final juror predeliberation preferences slightly favored the plaintiff (M = -1.16, SD = 3.08), wherein negative preferences signify a preference in favor of the plaintiff. Juror preferences combined across all mock trials after each of the four segments were as follows: segment one preference (M = -2.02, SD = 1.65); segment two preference (M = -0.18, SD = 2.82); segment three preference (M = -1.46, SD = 2.75); and segment four preference (M = -1.06, SD = 3.03). Segments one and three were after plaintiff presentations, whereas segments two and four were after defense presentations. The general proplaintiff bias appeared after each segment, although it was lessened following defense segments. Mimicry was defined as any behavior performed by a juror within 10 seconds of that same behavior being performed by the attorney. A mimicry index representing the summation of instances of mimicry for each of the target nonverbal behaviors with adequate interrater reliability was calculated for each juror after each segment. An overall plaintiff mimicry index was calculated for each juror by totaling the mimicry indices for each juror after the plaintiff segments (one and three). An overall defense mimicry index was calculated for each juror by totaling the mimicry indices for each juror after the defense segments (two and four). Consistent with the pro-plaintiff trend in juror preferences, in comparing the overall plaintiff mimicry index to the overall defense mimicry index, jurors tended to show greater behavioral mimicry of the plaintiff’s attorney (M = 1.70, SD = 2.80) than of the defense attorney (M = 0.80, SD = 1.12), t(42) = 2.18, p<.05. Nonverbal behavior counts for each juror and segment were averaged across the two coders. Thus, if one coder counted six instances of face touching by a juror in a segment while the other coder counted four instances of face touching by the juror in that segment, it was counted as five face touches. Means and standard deviations for the target nonverbal behaviors are presented in Table 1. The three commonly held nonverbal indicators of agreement (smile, forward lean, and head nod) were not correlated with each other (Cronbach’s alpha = -.042). Thus, these three nonverbal behaviors were not combined to form a nonverbal index of agreement, and instead were treated as three separate variables. Each of the three variables for segments one and three were combined to form an index for the plaintiff segments. For instance, a juror’s smiling behavior in segments one and three was combined to get an overall count of smiling during the plaintiff segments. Similarly, each of the three variables for segments two and four were combined to form an index for the defense segments. For instance, a juror’s smiling behavior in segments two and four was combined to get an overall count of smiling during the defense 11 segments. Table 2 displays the descriptive statistics for the nonverbal agreement behaviors for the plaintiff’s attorney and defense attorney combined across mock trials. As with the commonly held nonverbal indicators of agreement, the commonly held nonverbal indicators of disagreement (arm cross and head shake; frown and gaze aversion were not included as the interrater reliabilities were too low) were not correlated with each other (Cronbach’s alpha = 0.02). Thus, arm cross and head shake were not combined to form a nonverbal index of disagreement, and instead were treated as two separate variables. Each of the two variables for segments one and three were combined to form an index for the plaintiff segments. For instance, a juror’s head shaking behavior in segments one and three was combined to get an overall count of head shaking during the plaintiff segments. Similarly, each of the two variables for segments two and four were combined to form an index for the defense segments. For instance, a juror’s head shaking behavior in segments two and four was combined to get an overall count of head shaking during the defense segments. Table 2 also displays the descriptive statistics for the nonverbal disagreement behaviors for the plaintiff’s attorney and defense attorney combined across mock trials. Predicting Final Verdict Preference from Overall Mimicry The primary goal of the research was to determine whether overall mimicry predicts final predeliberation verdict preference. The regression model is outlined in Figure 1. The primary predictor variables are overall mimicry of the plaintiff’s attorney and overall mimicry of the defense attorney. The outcome variable is final predeliberation verdict preference. Since the research used a nested design with the level of juror nested within the level of case, five dummy variables were created for the six different cases to ensure there were no case-specific effects. When both mimicry indices and all five dummy variables were entered into the model, the model was not significant, F(7,35) = 0.423, ns. While this demonstrates that there were no effects of case, it also shows that overall mimicry of the plaintiff’s attorney and overall mimicry of the defense attorney do not predict final verdict preference. Overall mimicry of the plaintiff’s attorney did not significantly predict final verdict preference, = -0.242, t(35) = -0.833, ns. Similarly, overall mimicry of the defense attorney did not significantly predict final verdict preference, β = 0.453, t(35) = 0.789, ns. As there were no case effects, additional analyses were not conducted as a nested design with the level of juror nested in the level of case. Even if the case-specific dummy codes are left out of the regression model, the model is still not significant and overall mimicry of the plaintiff’s attorney and overall mimicry of the defense attorney are not significant predictors of final verdict preference, F(2,40) = 0.660, ns. Mimicry of the plaintiff’s attorney did not significantly predict final verdict preference, β = 0.176, t(40) = -0.945, ns. Mimicry of the defense attorney also did not significantly predict final verdict preference, β = .394, t(40) = 0.843, ns. However, they both were in the predicted direction. Therefore, the main hypothesis that overall mimicry of the plaintiff’s attorney and overall mimicry of the defense attorney would predict final verdict preference was not supported. Segment-by-Segment Analysis – Predicting Subsequent Preference Figure 2 details a more nuanced segment-by-segment analysis, where the regression models focus on the relationship between prior preference, mimicry, nonverbal agreement behaviors (smile, head nod, forward lean), nonverbal disagreement behaviors (arm cross and head shake), and subsequent preference at the segment level and not at an aggregate level across all four segments. When attempting to predict subsequent verdict preference, analyses were first conducted using a full model containing prior preference, mimicry, the three nonverbal 12 agreement behaviors, and the two nonverbal disagreement behaviors as predictors.1 To demonstrate that mimicry has more predictive validity than the three nonverbal agreement behaviors or both nonverbal disagreement behaviors, the final five variables were dropped from the full model and only prior preference and mimicry were entered into the nested model. A change in R2 was then calculated to determine whether prior preference and mimicry were indeed stronger predictors than the three nonverbal agreement behaviors or both nonverbal disagreement behaviors. All of the significant findings from the segment-by-segment analyses are presented in Figure 3. In the first step of this set of analyses, segment one mimicry, nonverbal agreement behaviors2, and nonverbal disagreement behaviors3 were tested as predictors of preference after the first segment. The overall regression model was not significant, F(6,35) = 0.673, ns. R2 for the model was .103. As this was the first segment for each mock trial, there was no prior preference included in the model. The next step in this set of analyses tested segment one preference and segment two mimicry, nonverbal agreement behaviors, and nonverbal disagreement behaviors as predictors for segment two preference. When all predictors were added in the full model, the model was significant, F(7,35) = 3.47, p<.01. R2 for the full model was 0.410. To obtain a nested model with only segment one preference and segment two mimicry as predictors, segment two nonverbal agreement behaviors and nonverbal disagreement behaviors were dropped from the full model. The nested model was significant, F(2,40) = 9.150, p<.01. R2 for the nested model was 0.314. The change in R2 was not significant, F(5,35) = 1.139, ns. Since the change in R2 was not significant, this implies that the dropped predictors (segment two nonverbal agreement behaviors and nonverbal disagreement behaviors) are not needed. Thus, segment one preference (β = 0.933, t(40) = 4.029, p<.001, partial η2 = .289) and segment two mimicry (β = 2.037, t(40) = 2.484, p<.01, partial η2 = .134) explain most of the variance in the model, and segment two nonverbal agreement behaviors and nonverbal disagreement behaviors do not add any predictive validity to the model. The next step in this set of analyses tested segment two preference and segment three mimicry, nonverbal agreement behaviors, and nonverbal disagreement behaviors as predictors for segment three preference. When all predictors were added in the full model, the model was significant, F(7,35) = 2.400, p<.05. R2 for the full model was 0.324. To obtain a nested model with only segment two preference and segment three mimicry as predictors, segment three nonverbal agreement behaviors and nonverbal disagreement behaviors were dropped from the full model. The nested model was significant, F(2,40) = 8.992, p<.01. R2 for the nested model was 0.310. The change in R2 was not significant, F(5,35) = 0.145, ns. Since the change in R2 was not significant, this implies that the dropped predictors (segment three nonverbal agreement behaviors and nonverbal disagreement behaviors) are not needed. Thus, segment two preference 1 To control for segment length, we included the total number of behavior codes in a particular segment for each juror as a predictor. This variable was used as a proxy for segment length because longer segments would generally be associated with a higher number of behavior codes in that segment. The total number of behavior codes was never a significant predictor and did not change any of the findings. Thus, it is omitted from the reported analyses. 2 The nonverbal agreement behaviors included in each model were smile, head nod, and forward lean. 3 The nonverbal disagreement behaviors included in each model were arm cross and head shake. 13 (β = 0.481, t(40) = 3.737, p<.001, partial η2 = .259) and segment three mimicry (β = -0.220, t(40) = -1.688, p<.10, partial η2 = .066)4 explain most of the variance in the model, and segment three nonverbal agreement behaviors and nonverbal disagreement behaviors do not add any predictive validity to the model. The final step in this set of analyses tested segment three preference and segment four mimicry, nonverbal agreement behaviors, and nonverbal disagreement behaviors as predictors for segment four preference. When all predictors were added in the full model, the model was significant, F(7,35) = 6.621, p<.001. R2 for the full model was 0.577. To obtain a nested model with only segment three preference and segment four mimicry as predictors, segment four nonverbal agreement behaviors and nonverbal disagreement behaviors were dropped from the full model. The nested model was significant, F(2,40) = 20.510, p<.001. R2 for the nested model was 0.513. The change in R2 was not significant, F(5,35) = 1.059, ns. Since the change in R2 was not significant, this implies that the dropped predictors (segment four nonverbal agreement behaviors and nonverbal disagreement behaviors) are not needed. Thus, segment three preference (β = 0.789, t(40) = 6.330, p<.001, partial η2 = .550) and segment four mimicry (β = 0.585, t(40) = 1.895, p<.10, partial η2 = .082)5 explain most of the variance in the model, and segment four nonverbal agreement behaviors and nonverbal disagreement behaviors do not add any predictive validity to the model. Segment-by-Segment Analysis – Predicting Mimicry The first step in this set of analyses used segment one preference to predict mimicry, nonverbal agreement behaviors, and nonverbal disagreement behaviors in segment two (see Figure 2). While segment one preference did not significantly predict the three nonverbal agreement behaviors or either nonverbal disagreement behavior, segment one preference did significantly predict mimicry, β = -.077, t(41) = -1.817, p<.106. The next step in this set of analyses was to test segment two preference as a predictor of mimicry, nonverbal agreement behaviors, and nonverbal disagreement behaviors in segment three. Segment two preference did not significantly predict segment three mimicry, nonverbal agreement behaviors, or nonverbal disagreement behaviors. The final step in this set of analyses was to test segment three preference as a predictor for mimicry, nonverbal agreement behaviors, and nonverbal disagreement behaviors in segment four. Segment three preference did not significantly predict segment four mimicry, nonverbal agreement behaviors, or nonverbal disagreement behaviors. Discussion The main hypothesis was not supported. Overall mimicry of the plaintiff’s attorney and defense attorney did not predict final predeliberation verdict. There are two possible explanations for why mimicry was not a significant predictor of final predeliberation verdict. First, unlike in three of the four segment-by-segment analyses, there was no prior preference variable in the main regression model. Without prior preference in the main regression model, all of its explained variance was in the error term and mimicry was not a significant predictor of 4 While segment three mimicry was not a significant predictor with a two-tailed test, it was marginally significant and would have been significant with a one-tailed test. As a priori hypotheses predicted mimicry and preference within a given segment to be positively correlated, significance under a one-tailed test is meaningful in this analysis. 5 See footnote 4 6 See footnote 4 14 final predeliberation preference. A second reason the primary model may not have been significant is because there were several segments in each mock trial that were not coded. Since only an early plaintiff segment and early defense segment and a late plaintiff segment and late defense segment were coded, important information in the middle of each mock trial was not examined. Without mimicry and other nonverbal counts from these middle segments, the strength of the primary model was reduced, further explaining why the primary model was not significant. While the main regression model was not significant, a more nuanced analysis revealed that behavioral mimicry was a significant predictor of verdict preference at the more micro level of case segment. Specifically, mimicry of the defense attorney in segments two and four predicted juror preferences after those defense segments. Similarly, mimicry of the plaintiff’s attorney in segment three predicted juror preferences after that plaintiff segment. In all three instances, preferences changed in a manner consistent with behavioral mimicry. Thus, jurors who engaged in more mimicry of the defense attorney during the defense segments showed stronger pro-defense preferences after the segment than jurors who mimicked the defense attorney to a lesser extent. In a similar manner, jurors who engaged in more mimicry of the plaintiff’s attorney during the second plaintiff segment showed stronger pro-plaintiff preferences after the segment than jurors who mimicked the plaintiff’s attorney to a lesser extent. However, mimicry was not a significant predictor of juror preference after the first plaintiff segment. This is likely because there was no prior juror preference before the first segment. Thus, without prior preference as a second predictor variable entered in to the regression model alongside mimicry, mimicry was not a significant predictor of preference. Prior verdict preference was available for the other three segments analyzed, potentially explaining why mimicry was not a significant predictor in the first plaintiff segment. In other words, prior verdict preference helped explain a large amount of the variance in the nested model. When prior preference was not a predictor in the nested model along with mimicry, the model was not significant because mimicry alone was not a strong enough predictor and the variance explained by prior preference went in to the error term, lowering mimicry’s test statistic. In line with the secondary hypothesis, mimicry was a better predictor of verdict preference than nonverbal behaviors thought to signal agreement (e.g. smiling, forward lean, head nod) and nonverbal behaviors thought to signal disagreement (e.g. arm cross, head shake). Behavioral mimicry was a stronger predictor than the three nonverbal agreement behaviors and both nonverbal disagreement behaviors in all segments analyzed. In segment two, mimicry accounted for 13.4% of the variance, certainly not an insignificant amount. These findings suggest that attorneys would be wise to weigh mimicry more heavily than commonly held nonverbal indicators of agreement and disagreement. Although only speculation, one potential reason that mimicry is a better predictor than nonverbal agreement behaviors and nonverbal disagreement behaviors is because it may be more outside of conscious control. Especially in a formal courtroom setting that values impartiality, people are well-versed at controlling outward signs of agreement and disagreement. Mimicry, on the other hand, may be less amenable to conscious control. Another potential explanation for mimicry’s greater predictive ability may lie in the nature of the different constructs. It has been well-established that mimicry signals liking, agreement, and an intent to affiliate (Chartrand & van Baaren, 2009). Commonly held nonverbal agreement and disagreement behaviors, on the other hand, may have multiple meanings. For instance, crossing one’s arms may indicate disagreement, but it may also indicate the person is cold (Frederick, 2006). Regardless of the explanation, the results suggest that behavioral 15 mimicry is a better predictor of juror verdict preference than either nonverbal agreement behaviors or disagreement behaviors. Strengths This study had several strengths that help bolster the findings. First, all mock jurors’ nonverbal behaviors were coded by two hypothesis-blind coders. Thus, jurors’ nonverbal behaviors were thoroughly coded. Double coding also allowed for an average count for each of the nonverbal behaviors that is more reliable than either of the behavioral counts from each individual coder going into the average. Interrater reliabilities were also quite high for most of the nonverbal behaviors; nine of the thirteen nonverbal behaviors included in the mimicry indices had a Spearman Brown correlation coefficient of 0.8 or above. The benefit of keeping the coders blind to the research hypotheses was that they were not aware that mimicry was the main focus of the research. Second, the duty of the coders was to focus on the jurors, whereas the first author’s duty was to focus on the attorney. Mimicry was calculated by identifying post hoc any behaviors the jurors performed that the attorney had just performed within ten seconds or less. Thus, at no point was mimicry even a focus of the coding process. Instead, the focus was on the nonverbal behaviors performed by two separate groups (the attorney and the jurors). As no single coder focused on both the attorney and the jurors, mimicry did not enter into the analyses until after all data had been collected. Therefore, the potential for coder bias was minimized.7 Third, there were no case effects. None of the dummy variables included in the primary regression model were significant predictors of predeliberation verdict preference. Thus, it can be inferred that any results found are specific to mimicry and other nonverbal behavior, and not to the idiosyncrasies of any particular case or attorney. Fourth, this research had an atypical design and paradigm for mimicry research. For one, it included a community sample of jurors. Most mimicry research uses a convenience sample of college-age participants. Using a more diverse community sample of jurors increases the external validity of the results. On a related note, mock jurors were put in a more naturalistic situation than participants in most mimicry research. Instead of watching a mundane video of a stranger (e.g. Stel et al., 2010), this research focused on compensated mock jurors in a legal setting with potentially high stakes. Thus, these mock jurors should have been invested in their duty, potentially strengthening the conclusions drawn from the results. Fifth, although it has been suggested that mimicry is a useful tool to examine in the courtroom, no empirical research to date has focused on the predictive ability of jurors’ mimicry of attorneys in the courtroom. Therefore, this study takes mimicry to a new applied domain, further demonstrating the ubiquity and predictive validity of behavioral mimicry. As will be discussed later, this research has much to offer trial attorneys who might want an extra tool to use at trial. Limitations One of the primary limitations of this research was the relatively low number of mock jurors in the sample. Although six different cases were included in the analyses, the total number of mock jurors coded was 43. The rationale behind this was that jurors in the front row would 7 Although not reported in the Results section, the two coders made guesses as to the preferences of each juror after each segment. Preferences were made on the same -4 to +4 scale. The correlation between the preference estimates of the two coders was very low, offering further support of a lack of coder bias. 16 have less interpersonal distance between themselves and attorneys, thereby precluding jurors in the second and third rows from analysis. More practically, it would have been more difficult to code the nonverbal behaviors of jurors in the second and third rows as they were more difficult to fully see. Some of the jurors in the front row were also either not fully in the camera frame or else blocked from view by other front row jurors, further reducing the sample of codable jurors. Nonetheless, the results suggest that even with a relatively small sample, mimicry is still a fairly strong predictor of verdict preference. Another limitation is that the mimicry counts were quite low. As stated earlier, the average mimicry count of the plaintiff’s attorney across both plaintiff segments was 1.70, whereas the average mimicry count of the defense attorney across both defense segments was 0.80. Although average mimicry counts were quite low, some jurors engaged in much greater amounts of mimicry than other jurors. Mimicry indices ranged from 0-14.50 for the plaintiff’s attorney and 0-4.50 for the defense attorney. Despite the relatively low mimicry counts across segments, mimicry was still a significant predictor of verdict preference in three of the four segments analyzed. Another limitation of the study was that all six cases analyzed were civil cases. While no case-specific effects were found, it is an open question whether the same results would apply in a criminal case. It would seem unlikely that a juror’s mimicry in a criminal trial would not be as predictive of verdict as in a civil trial, but it is an open empirical question. Further, the mock jurors in this study were not real jurors in a real case in a real courtroom. Each segment lasted anywhere from several minutes to an hour and a half, far shorter than many presentations at trial. Therefore, any effects of mimicry may be further muted (or enhanced) by a longer trial. While ecological validity is higher in this research than in many other mimicry studies, results should still be interpreted with caution. As mentioned earlier, another limitation was that there were intermediate segments in each mock trial that were not coded. Given the abundance of information, we were only able to code an early plaintiff segment, early defense segment, late plaintiff segment, and late defense segment from each mock trial. This left out important mimicry and other nonverbal behavior information from segments in the middle of the mock trials. Thus, without this data, the strength of mimicry as a predictor in the overall regression model was reduced. This helps explain why mimicry was not a significant predictor in the overall regression model while mimicry was a significant predictor in the segment-by-segment analyses. Implications and Recommendations As a whole, the results suggest that behavioral mimicry is a moderately strong predictor of verdict preference. While it is not surprising that prior preference is be a very strong predictor of subsequent preference, it is noteworthy that mimicry accounts for a significant amount of additional variance in predicting subsequent preference. In segments two, three, and four, mimicry explained anywhere between 6% and 13% of the variance. In a long and complex trial, that is not an insignificant amount. Attorneys and trial consultants can use the findings from this research as a tool during jury selection. Since people are quite adept at monitoring excessive displays of agreement or disagreement in a formal courtroom setting, behavioral mimicry may be a promising way to ascertain a juror’s current leanings completely irrespective of the juror’s verbalized responses. Jurors who engage in greater mimicry of the other side’s attorney during jury selection may be worthy candidates for peremptory challenges. These results can also be useful during evidence presentation. As the jury’s decision is not known until after deliberation (when the case is over), there is really no way to judge how 17 one’s case is resonating with the jury. However, the current findings suggest that mimicry may be a subtle indicator of agreement, at least on a temporary basis. Thus, if jurors engage in a large amount of mimicry as an attorney presents arguments about topic A but less mimicry when she presents a separate line of arguments about topic B, this could be an indication that the arguments for topic A are more persuasive than the arguments for topic B. The attorney can then either continue to reinforce the topic A arguments and/or go back and refine the topic B arguments. Another way mimicry might be used during evidence presentation is as a quick indicator of jury preferences. If jurors are mimicking one attorney more than the other attorney throughout the trial, it may indicate that the former side is currently “in the lead”. Although this research examined the effects of jurors mimicking the attorneys, the direction can be reversed. Previous mimicry findings show that mimicry results in increased liking of the mimicker (e.g. Lakin & Chartrand, 2003) as well as increased persuasive ability for the mimicker (e.g. Bailenson & Yee, 2005; Maddux et al., 2008; Tanner et al., 2008). To this end, attorneys can use these findings to their advantage. Starting with jury selection, attorneys can attempt to subtly mimic the jury. This would have to be done in such a manner that it is undetectable to jurors, as obvious imitation can backfire (Chartrand & van Baaren, 2009). If an attorney subtly mimics different jurors throughout the trial, the increased liking and persuasiveness engendered from being mimicked might push several on-the-fence jurors over onto the mimicking attorney’s side. Finally, the results suggest that mimicry is a stronger predictor of verdict preference than commonly held nonverbal agreement or nonverbal disagreement behaviors. Attorneys often rely on one or two indicators of nonverbal agreement (e.g. head nod, smile) or disagreement (e.g. cross arms, head shake) as indicative of a juror’s reaction to certain arguments (Frederick, 2006). The results of this study suggest that these commonly held nonverbal indicators should be interpreted with caution. Mimicry, given its nonconscious nature and strong relationship with liking, attitudinal similarity, and persuasiveness, may be a better indicator of a juror’s current preference. If one is going to monitor the jury’s nonverbal behaviors during trial, a greater focus on behavioral mimicry and a de-emphasis on commonly held nonverbal agreement and nonverbal disagreement behaviors may be a promising strategy. Future Directions This work presents several intriguing future directions. One question is whether certain types of mimicry are more predictive than others. Thirteen different nonverbal behaviors had high enough interrater reliability to be included in the mimicry indices. All of these variables were combined together in the mimicry index, so it is not clear whether certain mimicry behaviors are more predictive of verdict preference than others. Future research can attempt to tease out this issue of which specific mimicry behaviors attorneys should focus on in the courtroom. A second possible future direction is to try to replicate the results using different types of cases. As noted earlier, all six mock trials involved different civil cases. While there is no reason to suggest the type of case has an impact on the predictive validity of mimicry, it nonetheless remains an open question. Third, this research only focused on jurors in the front row. Jurors in the second and third rows were excluded due to concerns with coding difficulty. The question, then, is whether jurors in the second or third rows would show a similar pattern of findings as the jurors in the front row. On the one hand, someone else is between them and the attorney presenting, which might create a sense of interpersonal distance and thereby inhibit mimicry. On the other hand, 18 being towards the back might make these jurors more comfortable, resulting in slightly less constrained behavior and a greater propensity to mimic. Thus, expanding the focus beyond jurors in the front row would be an interesting future direction. Fourth, on a note related to the third point above, most of the mock trials used in this research had 24 mock jurors. The typical jury is either six or twelve individuals. Thus, jurors in the present research may have felt more anonymous than jurors on an actual jury. This could potentially lead to disengagement with the task (e.g. social loafing), or it could reduce pressure on the jurors, enabling them to invest more of their attention on the attorneys’ presentations. How this might impact mimicry is unclear. Perhaps the large number of other mock jurors reduces engagement with the task and thus potentially mimicry. Or conversely a dimmer spotlight might lead to greater mimicry. Again, it is an empirical question deserving of future research. Conclusion The findings from this research suggest that behavioral mimicry is an important factor to consider at trial. Despite the robustness of prior preference as a predictor of subsequent preference, mimicry still accounts for between 6%-13% of the variance. This can make a world of difference at trial. The evidence suggests that mimicry is a better predictor of verdict preference than commonly held nonverbal indicators of agreement and disagreement. This research provides attorneys and trial consultants with a tool they can use in court to help them deselect unfavorable jurors and to assess the effectiveness of their case on a moment-to-moment basis. Instead of waiting until the verdict has already been announced to learn of the jury’s opinion, this research points to behavioral mimicry as a means of ascertaining the current mood of the jury before it is too late. 19 Tables Table 1. Interrater reliabilities, means, and standard deviations (by segment) for coded nonverbal behaviors Face Touching* Hair Touching* Arm Touching Neck Touching* General Body Touching* Forward Lean* Backward Lean Posture Shift* Uncross Arms Cross Arms* Adjust Clothes* Adjust Accessories* Smile* Frown Yawn* Head Nod* Head Shake* Gaze Aversion Total Behaviors Performed by Mock Jurors Total Possible Mimicable Behaviors Performed by Attorneys Reliability Estimates 0.994 0.983 0.573 0.797 0.917 Mean 13.63 3.55 1.21 1.35 1.02 Standard Deviation 18.19 5.19 1.73 1.83 3.10 0.908 0.587 0.689 0.431 0.908 0.929 0.821 0.759 -0.017 0.912 0.966 0.625 0.152 0.983 0.59 0.54 1.80 0.22 0.92 1.40 1.06 0.26 0.01 1.39 1.71 0.04 0.49 31.65 1.50 0.92 2.51 0.54 1.84 3.96 2.08 0.82 0.13 2.91 7.50 0.21 1.79 33.63 - 28.91 37.08 * Target behaviors included in the mimicry indices Note that behaviors not included in the mimicry indices do not have means and standard deviations due to poor interrater reliability 20 Table 2. Descriptive statistics for nonverbal agreement and nonverbal disagreement behaviors for plaintiff’s attorney and defense attorney across mock trials. Nonverbal Behavior Agreement/ Disagreement Plaintiff Mean Defense Mean 0.74 1.15 Plaintiff Standard Deviation 1.67 2.42 Significant Difference 0.33 1.28 Defense Standard Deviation 0.70 3.08 Smiling Forward Lean Head Nodding Arm Crossing Head Shaking Agreement Agreement Agreement 4.30 16.00 2.78 10.46 No Disagreement 1.79 2.85 2.01 3.63 No Disagreement 0.15 0.39 0.02 0.11 No 21 Yes No Figures Figure 1. The main regression analysis, in which total mimicry of the plaintiff’s attorney and total mimicry of the defense attorney across all four segments were used to predict final predeliberation preference. PFinal = f(MP, MD) MP = M1 + M3 Total mimicry of plaintiff’s attorney across segments 1 and 3 MD = M2 + M4 Total mimicry of defense attorney across segments 2 and 4 PFinal = Final predeliberation preference 22 Figure 2. A secondary analysis conducted to determine the relationship between prior preference, mimicry behavior, nonverbal agreement behaviors, nonverbal disagreement behaviors, and subsequent preference. Analyses were conducted on a segment-by-segment basis. The exception was that prior verdict preference for segment 1 could not be included because there was no prior preference for this first segment. Note that there are six predictors (Mn, SMn, FLn, HNn, ACn, and HSn) included in each segment, Sn. Significant regression coefficients are marked with an asterisk. **** S1 M1 SM1 FL1 HN1 AC1 HS1 P1 S2 **** P2 S3 M2*** SM2 FL2 HN2 AC2**8 HS2 M3** SM3 FL3 HN3 AC3 HS3 **** P3 S4 P4 M4** SM4 FL4 HN4 AC4 HS4 Sn = Segment (segments 1 and 3 for plaintiff presentations; segments 2 and 4 for defense presentations) Pn = Preferences (Preferences 1 and 3 after plaintiff presentations; preferences 2 and 4 after defense presentations) Mn = Total mimicry behavior for a given segment SMn = Total smiling behavior for a given segment FLn = Total forward lean behavior for a given segment HNn = Total head nodding behavior for a given segment ACn = Total arm crossing behavior for a given segment HSn = Total head shaking behavior for a given segment ** p<.05 ***p <.01 ****p<.001 8 This is likely the result of a type I error. By chance alone we would expect 5% of the predictors to be significant. As there were 27 individual predictors throughout the segment-bysegment analyses, a type I error is not unlikely. 23 Figure 3. The significant predictors from the segment-by-segment analyses. **** P1 M2*** AC2**9 **** P2 M3** **** P3 M4** P4 Pn = Preferences (Preferences 1 and 3 after plaintiff presentations; preferences 2 and 4 after defense presentations) Mn = Total mimicry behavior for a given segment ACn = Total arm crossing behavior for a given segment ** p<.05 ***p <.01 ****p<.001 9 See footnote 8 24 Literature Cited Aronson, E. (1999). The social animal. (8th ed.). New York: W.H. Freeman and Company. Bailenson, J.N. & Yee, N. (2005). Digital chameleons: Automatic assimilation of nonverbal gestures in immersive virtual environments. Psychological Science, 16(10), 814-819. Bargh, J.A. (1990). Auto-motives: Preconscious determinants of social interaction. In E. Higgins & R. Sorrentino (Eds.), Handbook of motivation and cognition (Vol. 2, pp. 93-130). New York: Guilford Press. Baumeister, R.F. & Leary, M.R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497-529. Bourgeois, P. & Hess, U. (2008). The impact of social context on mimicry. Biological Psychology, 77, 343-352. Bousmalis, K., Mehu, M. & Pantic, M. (2009). Spotting agreement and disagreement: A survey of nonverbal audiovisual cues and tools. Proc. IEEE Int'l Conf. Affective Computing and Intelligent Interfaces. Buss, D.M. & Kenrick, D.T. (1998). Evolutionary social psychology. In D.T. Gilbert, S.T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 982-1026). New York: Oxford University Press. Cappella, J. N., & Planalp, S. (1981). Talk and silence sequences in informal conversations III: Interspeaker influence. Human Communication Research, 7, 117–132. Chartrand, T.L. & van Baaren, R. (2009). Human mimicry. Advances in Experimental Social Psychology, 41, 219-274. Chartrand, T.L. & Bargh, J.A. (1999). The chameleon effect: The perception-behavior link and social interaction. Journal of Personality and Social Psychology, 76(6), 893-910. Cheng, C.M. & Chartrand, T.L. (2003). Self-monitoring without awareness: Using mimicry as a nonconscious affiliation strategy. Journal of Personality and Social Psychology, 85, 1170-1179. Cialdini, R. (2001). Influence: Science and practice. Needham Heights, MA: Allyn and Bacon. Dalton, A.N., Chartrand, T.L., & Finkel, E.J. (2010). The schema-driven chameleon: How mimicry affects executive and self-regulatory resources. Journal of Personality and Social Psychology, 98(4), 605-617. DePaulo, B.M. & Friedman, H.S. (1998). Nonverbal communication. In D.T. Gilbert, S.T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 3-40). New York: Oxford University Press. Deutsch, M. & Gerard, H.G. (1955). A study of normative and informational social influence upon individual judgment. Journal of Abnormal and Social Psychology, 51, 629-636. de Waal, F. (1989). Peacemaking among primates. Cambridge, MA: Harvard University Press. Dimberg, U., Thunberg, M., & Elmehed, K. (2000). Unconscious facial reactions to emotional facial expressions. Psychological Science, 11, 86–89. Frederick, J. (2006). Understanding juror’s nonverbal communication. The Jury Expert, 18(3), 18. Giles, H., & Coupland, N. (1991). Language contexts and consequences. Milton Keynes: Open University Press. Gregory, S. W. Jr., Dagan, K., & Webster, S. (1997). Evaluating the relation of vocal accommodation in conversation partners’ fundamental frequencies to perceptions of communication quality. Journal of Nonverbal Behavior, 21, 23–43. 25 Gueguen, N. & Martin, A. (2009). Incidental similarity facilitates behavioral mimicry. Social Psychology, 40(2), 88-92. Iacoboni, M., Woods, R. P., Brass, M., Bekkering, H., Mazziotta, J. C., & Rizzolatti, G. (1999). Cortical mechanisms of human imitation. Science, 286, 2526–2528. James, W. (1890). Principles of psychology. New York: Holt. Kalven, H., Jr. & Zeisel, H. (1966). The American jury. Boston: Little, Brown. Kouzakova, M., van Baaren, R., & van Knippenberg, A. (2010). Lack of behavioral imitation in human interactions enhances salivary cortisol levels. Hormones and Behavior, 57, 421426. Lakin, J.L. & Chartrand, T.L. (2003). Using nonconscious behavioral mimicry to create affiliation and rapport. Psychological Science, 14(4), 334-339. Lakin, J.L., Jefferis, V.E., Cheng, C.M., & Chartrand, T.L. (2003). The chameleon effect as social glue: Evidence for the evolutionary significance of nonconscious mimicry. Journal of Nonverbal Behavior, 27(3), 145-162. Lundquist, L. O., & Dimberg, U. (1995). Facial expressions are contagious. Journal of Psychophysiology, 9, 203–211. Maddux, W.W., Mullen, E., & Galinsky, A.D. (2008). Chameleons bake bigger pies and take bigger pieces: Strategic behavioral mimicry facilitates negotiation outcomes. Journal of Experimental Social Psychology, 44, 461-468. McHugo, G., Lanzetta, J., & Bush, L. (1991). The effect of attitudes on emotional reactions to expressive displays of political leaders. Journal of Nonverbal Behavior, 15(1), 19-41. Meltzoff, A.N. & Moore, M.K. (1977). Imitation of facial and manual gestures by human neonates. Science, 198, 75-78. Stel, M., van Baaren, R., Blascovich, J., van Dijk, E., McCall, C., Pollmann, M., van Leeuwen, M., Mastop, J., & Vonk, R. (2010). Effects of a priori liking on the elicitation of mimicry. Experimental Psychology, 57(6), 412-418. Tanner, R.J., Ferraro, R., Chartrand, T.L., Bettman, J.R., & van Baaren, R. (2008). Of chameleons and consumption: The impact of mimicry on choice and preferences. Journal of Consumer Research, 34, 754-766. Termine, N.T. & Izard, C.E. (1988). Infants’ response to their mothers’ expressions of joy and sadness. Developmental Psychology, 24, 223-229. van Baaren, R. B., Maddux, W. W., Chartrand, T. L., De Bouter, C., & van Knippenberg, A. (2003b). It takes two to mimic: Behavioral consequences of self- construals. Journal of Personality and Social Psychology, 84, 1093–1102. Van Swol, L. M., & Drury, M. (2008). The effects of shared opinions on nonverbal mimicry. University of Wisconsin-Madison, submitted for publication. Yabar, Y., Johnston, L., Miles, L., & Peace, V. (2006). Implicit behavioral mimicry: Investigating the impact of group membership. Journal of Nonverbal Behavior, 30, 97113. 26 Appendix A – Sample Verdict Preference Item Which side do you currently favor? Plaintiff Defense How easy would it be to change your opinion? 1 My opinion cannot be changed 2 My opinion would be slightly difficult to change 27 3 My opinion would be slightly easy to change 4 My opinion would be very easy to change Appendix B– Nonverbal Behavior List The following is a list of nonverbal behaviors that were coded. The research assistants independently focused on the eligible jurors and noted the occurrence of each nonverbal behavior and the times at which they occurred. The first author focused on the attorneys and noted the occurrence of each nonverbal behavior and the times at which they occurred. Mimicry was defined as the juror performing the same behavior as the attorney within a 10-second time frame. If a juror performed a behavior absent similar earlier behavior by the attorney, or if the behavior occurred after 10 seconds, it was not counted as mimicry. Note that there are some behaviors on this list that are commonly held indicators of agreement or disagreement. The occurrence of these was not treated as mimicry, unless the attorney performed these behaviors less than 10 seconds before the juror performed one of these behaviors. 11 - Face touching 12 – Hair touching 13 – Arm touching 14 – Shoulder/neck touching 15 – General body touching 21 – Forward lean 22 – Backward lean 23 – Posture shift 24 – Uncross/open arms 25 – Cross/close arms 31 – Adjust clothes 32 – Adjust accessories (e.g. watch, necklace, glasses) 41 – Smile 42 – Frown 43 – Yawn 51 – Head nodding 52 – Head shaking 53 – Gaze aversion 61 – Juror obstructed by attorney 62 – Juror reappears after attorney moves Note: Codes 61 and 62 were used if a juror was obstructed for four or more seconds 28
© Copyright 2026 Paperzz