What You Do and How You Speak Matter

Abstract

The primary goal of the present study is to investigate complex relationships among interpersonal behaviors, language use, and group performance in short-term virtual teams. Thirty-four, four-person groups completed a decision-making task in real time using an online chat program. The findings suggest that having a negatively communicating collaborator in the group is associated with higher group performance compared with having a positively communicating collaborator. Also, linguistic style matching is a stronger predictor of group performance for groups with a positively communicating confederate compared with groups with a negatively communicating confederate. The findings are discussed within the theoretical framework of shared mental models, minority influence, and communication accommodation theory.

Keywords

virtual teams interpersonal behaviors linguistic style matching performance

Virtual teams are ubiquitous in today’s organizational, educational, and research landscapes. These teams are implemented in global work collaborations (Edwards & Sridhar, 2005), online learning contexts (Loh & Smyth, 2010), and in the field of scientific research and development (Schunn, Crowley, & Okada, 2002). The integration of virtual teams has created new research opportunities for communication scholars. For instance, several studies examine perceptual, behavioral, and cognitive processes embedded in virtual team interactions (Bazarova & Walther, 2009; Walther, Bunz, & Bazarova, 2005; Yilmaz & Peña, 2014). Similarly, a group of studies focus on language use and group outcomes in the context of virtual teams (Gonzales, Hancock, & Pennebaker, 2010; Leshed, Hancock, Cosley, McLeod, & Gay, 2007; Leshed et al., 2009; Tausczik, 2013).

Recent studies show that providing verbal feedback on virtual group communication may enhance the quality of interactions among collaborators, leading to improved group performance (Tausczik, 2013). Several other studies shed light on the links between language use and group dynamics (Gonzales et al., 2010; Leshed et al., 2007, 2009). While there are studies examining the language-performance link in virtual teams, there are currently no studies looking at the combined effects of interpersonal behaviors and language use on group performance. In general, most studies look at the direct effects of language use on group outcomes, ignoring the possible effects of interpersonal behaviors. For instance, Gonzales et al. (2010) look at the relationship between linguistic style matching (LSM) and performance in face-to-face and computer-mediated groups, yet do not directly investigate how positive or negative behaviors affect performance in computer-mediated groups. Also, the same study reports a positive link between LSM and performance in face-to-face groups, but not in virtual groups. Similarly, Tausczik (2013) examines the relationship between verbal feedback, LSM, and group performance, without a specific focus on the role of interpersonal behaviors on these processes.

Virtual team literature is rich with evidence that interpersonal behaviors play an important role in attribution (Bazarova & Walther, 2009), social attraction (Wang, Walther, & Hancock, 2009), interpersonal liking (Walther, Bunz, & Bazarova, 2005), and subgroup formation (Yilmaz & Peña, 2014). In addition to these perceptual and behavioral processes, it is also important to examine how interpersonal behaviors affect language use and reflect group performance. Deeper understanding of the influence of interpersonal demeanor on language use, and the ways it influences performance would help virtual members and managers have a greater control over group processes. Thus, the main goal of this study is to further our understanding of the complex relationships among interpersonal demeanor, language use, and group performance. More specifically, this study will examine whether teams with a positively communicating collaborator perform better than teams with a negatively communicating collaborator. Also, this study will explore the relationship between LSM and group performance. The following sections will discuss theoretical frameworks and empirical findings informing the effects of interpersonal behaviors and language use on performance in virtual teams and advance relevant hypotheses.

Interpersonal Behaviors and Team Performance

What does make a work group complete their tasks effectively, and thus, perform well? In a recent study, positive communication among group members has been linked to high performance (Tausczik, 2013). In this study, participants take an online quiz responding to a 10-page reading assignment. Following this quiz, the participants are randomly assigned to groups and instructed to discuss a different set of psychology theories and concepts with their group members before taking the second quiz. Group members interact via instant messenger and are given feedback on how much they are contributing to the group discussion. Overall, the goal of this study is to look at how the language use of group members, as well as the verbal feedback on this language use affects certain group dynamics and group outcomes. Individuals take the quizzes alone; however, the improvement in their quiz scores is related to which group they have been assigned.

One experiment in this study shows that positivity in group communication is related to increased group performance. The main linguistic marker for positivity is the use of assents (e.g., yes, yeah, OK). The findings suggest that when group members display positive communication behaviors in the form of increased assent use, they perform better compared with groups that do not (Tausczik, 2013). The main rationale for this finding is that having positive communication in group discussions (i.e., agreeing with one another, not challenging competing opinions) facilitates information sharing. That is, group members actively share and process information during group discussions, and thereby, perform better compared with groups that do not. In other words, this rationale equates positive communication with effective information sharing and superior performance.

Providing further evidence for this rationale, another study shows that successful negotiation tasks are related to using assent words representing positive and cooperative communication behaviors (Huffaker, Swaab, & Diermeier, 2011). Similarly, one study examining linguistic correlates of team performance in six computer-simulated search and rescue missions reports that team success is positively correlated with the number of positive emotion words (Fischer, McDonnell, & Orasanu, 2007).

These findings make sense in light of shared mental model theory of group performance. Mental models are knowledge structures that help individuals describe, explain, and predict events in their immediate environments (Cannon-Bowers, Salas, & Converse, 1993). The main function of shared mental models in team interactions is “to allow team members to draw on their own well-structured knowledge as a basis for selecting actions that are consistent and coordinated with those of their teammates” (Mathieu, Goodwin, Heffner, & Cannon-Bowers, 2000, p. 274). Accordingly, group members who share similar mental models in terms of how to complete the task, order of task, timeline, and so on, perform better compared with groups that lack shared mental models (Mathieu et al., 2000). Complex tasks require team members to share multiple mental models. First, team members must understand the medium through which they communicate, as the control over their interaction affects team processes. Second, team members should have a mutual understanding of task procedures, strategies, or problems. Third, team members must have shared mental models that can be used to understand communication patterns, flow of information, and role interdependencies among group members (Cannon-Bowers et al., 1993). For this third type of shared mental model to emerge, team members should get along interpersonally and cooperate with one another. As interpersonal liking is part of this type of shared mental model, positive behaviors are more likely to trigger shared mental models compared with negative behaviors. As such, it is possible that high agreeableness in group discussions is a reflection of shared mental models, which accounts for a complete understanding of group tasks and roles, leading to increased performance (Cannon-Bowers et al., 1993).

It is known that positive socioemotional messages in face-to-face teams are more frequently exchanged than negative socioemotional messages (Bales, 1953). Testing this prediction in a virtual context, one study uses socioemotional and task messages generated by members of an online video game community, applying Bales’ interaction process analysis (Bales, 1953). This study reports that members produce more positive socioemotional messages (e.g., “That was great,” “Thanks for your help”) compared with negative socioemotional messages (e.g., “I hate this,” “Shut up”; Peña & Hancock, 2006). The researchers rationalize this finding using Bales’s equilibrium theory (Bales, 1953). According to this theory, groups use expressive messages, particularly positive ones, to counteract tensions in the group, and thereby, successfully complete their task. As such, while positive messages lead to enhanced performance, negative messages are disruptive to the group interaction, and impede the successful completion of task goals (Bales, 1953). Similarly, several studies examining the relationship between communication behaviors and group dynamics suggest that positive interpersonal behaviors such as text-based or visual feedback (Leshed et al., 2007; Tausczik, 2013) and cohesion (Gonzales et al., 2010) can positively affect how effectively groups complete their tasks. As informed by these empirical studies and theoretical models, the following hypothesis is proposed:

Hypothesis 1A: Groups with a positively communicating collaborator will perform better than groups with a negatively communicating collaborator.

Although positive communication can enhance interpersonal relationships in the long run, virtual groups that spend too much time on emotional expression may spend less time on the task, and thus, do not successfully achieve their task goals (Tausczik, 2013). Likewise, there is research suggesting that when group members do not participate in task-focused discussions, and blindly agree with one another about the solutions to a problem, the group outcome is not satisfactory (Leshed et al., 2007). When the group discussion is less task-oriented, the language of virtual team members reflects an increased use of assents (e.g., agree, OK, yes). Though there are studies that report positive functions of assents such as fostering turn taking and positively influencing negotiation outcomes (Curhan & Pentland, 2007) or reaching an agreement (Huffaker et al., 2011), Leshed et al. (2007) report that increased use of assents are associated with low participation and task focus in virtual teams. This finding provides a rival explanation for the link between positive communication and group outcomes. For instance, when group members value group harmony and interpersonal attraction more than the quality of group outcomes, they may experience groupthink (Janis, 1982). Groupthink can manifest in the form of avoiding the discussion of alternatives, and thereby, making decisions that are not ideal. However, integrating a stimulant for divergent views in group discussions, a negatively communicating collaborator might prevent the consequences of groupthink (Nemeth, 1995). In other words, having dissidents in the group might positively affect group performance, overriding the effects of groupthink.

Similarly, inducing a critical communication norm rather than a consensual norm can motivate team members to more critically approach their decisions and the alternatives. One study examining the impact of critical versus consensual norms on group decisions provides evidence for the positive effect of dissent on decision quality. According to this study, employing a critical approach to the decision-making process improves the quality of decisions by 56%, whereas there is only an 11% improvement in groups characterized by a consensus norm (Postmes, Spears, & Cihangir, 2001). The content analysis of discussions in the critical groups shows that group members generally use more probing language and challenge other group members compared with groups with a consensual norm. Rather than exchanging niceties, group members abiding by critical norms use a great number of critical contributions toward each other during discussions (Postmes, Spears, & Cihangir, 2001). In anonymous virtual teams, group norms are inferred from interactional patterns (Postmes, Spears, Sakhel, & De Groot, 2001). That is, having a critical member may set the tone for discussions, and thus, influence others to adopt an equally critical approach. On the basis of this rationale, it is possible that negative communication behaviors (i.e., criticizing and challenging others) might lead to better decisions compared with positive communication behaviors (i.e., providing positive feedback and encouragement). The following hypothesis is proposed to test this prediction:

Hypothesis 1B: Groups with a negatively communicating collaborator will perform better than groups with a positively communicating collaborator.

Linguistic Style Matching and Team Performance

Interpersonal communication is not confined to the content of conversations. It also consists of the style communicators use to convey the content. Although the content of communication is represented by content words (e.g., nouns), communication style is based on how people string these content words together using function words (e.g., articles). LSM reflects the extent to which communication partners mimic, synchronize, and match the language of one another at the function word level. LSM is measured using a metric that calculates the proportions of nine function word categories (i.e., verbs, articles, common adverbs, personal pronouns, indefinite pronouns, prepositions, negations, conjunctions, and quantifiers) used by communication partners (Gonzales et al., 2010; Ireland et al., 2010; Taylor & Thomas, 2008).

In a virtual group context, LSM can be used to better understand interpersonal and group dynamics. A recent study examining the relationship between LSM and group cohesion reports that group cohesion is positively related to LSM in face-to-face and virtual groups (Gonzales et al., 2010). Also, the results of this study illustrate that LSM is positively associated with task performance in face-to-face teams, but not in virtual teams (Gonzales et al., 2010).

Another study looks at the relationship between LSM and hostage negotiation outcomes. The researchers examine the dialogue of police negotiators and hostage takers. The findings show that successful negotiations are associated with higher levels of LSM compared with unsuccessful negotiations (Taylor & Thomas, 2008). According to this study, the relationship between LSM and successful negotiations illustrates a mutual coordination that consists of problem-solving style, interpersonal thoughts, and emotional expression. For instance, the researchers demonstrate that when police negotiators employ concrete thinking represented by the use of articles (e.g., a, an, the) in language, the hostage taker generally mimics the same use of articles. The researchers rationalize the positive link between LSM and successful negotiations by arguing that LSM is a reflection of common framing of the conflict, which enables police negotiator and hostage taker to develop interdependence (Taylor & Thomas, 2008).

As documented by previous empirical studies, language use is associated with group performance. To provide further evidence for the relationship between LSM and group performance, particularly in the context of virtual teams, the following hypothesis is advanced:

Hypothesis 2: LSM is a positive predictor of group performance.

Interpersonal Behaviors, LSM, and Team Performance

Current empirical studies suggest that when LSM is conceptualized as a reflection of positive interpersonal behaviors (e.g., attraction, trust), it is associated with perceptions of high task performance (Gonzales et al., 2010; Scissors, Gill, & Gergle, 2008). Yet, it is important to note that Gonzales and associates document a positive relationship between LSM and performance only in face-to-face groups, not in computer-mediated communication groups. It is important to examine the complex relationships among LSM, interpersonal behaviors, and task performance in virtual groups.

One theoretical framework that sheds light on previous findings on LSM, interpersonal behaviors, and performance is communication accommodation theory (CAT; Giles & Soliz, 2014). CAT explains cognitive and affective factors underlying speech convergence or divergence between communicators. According to CAT, individuals adapt to communicative behaviors of others that they perceive to be similar, as perceived similarity induces positive affect and interpersonal affinity (Giles & Soliz, 2014). As such, individuals converge their language patterns with those they perceive as similar, and thus, more attractive. When examined through the lens of CAT, LSM reflects an effort to develop a closer relationship or elicit approval or respect. Thus, it is possible that when LSM is conceptualized as a reflection of positive affect among group members, it also influences performance positively. Overall, studies at the intersection of language use, interpersonal behaviors and group outcomes suggest that when group members like one another, they perform better (Gonzales et al., 2010; Leshed et al., 2007; Tausczik, 2013). On the basis of current evidence, the following hypothesis explores the effects of positive behaviors on the relationship between LSM and task performance:

Hypothesis 3A: LSM will more strongly predict group performance in groups with a positively communicating confederate compared with groups with a negatively communicating confederate.

When LSM is seen as a reflection of mutual engagement and increased task focus, it is possible that what drives this mutual engagement, namely the interpersonal behaviors, also influences group performance. Mutual engagement refers to the notion that group members are engaged with the task at hand and with other collaborators (Bryan-Kinns, Healey, & Leach, 2007). Mutually engaged interaction is characterized by the investigation of ideas of more than one member and an effort to integrate those ideas (Miell & MacDonald, 2000). When group members are mutually engaged, they contribute to joint production, and acknowledge and reflect on each other’s contributions (Bryan-Kinns et al., 2007).

Most studies predict that LSM occurs on the basis of interpersonal liking. Yet LSM may also occur as a result of minority influence processes. One model that speaks to the effects of negative interpersonal behaviors on language use and performance is rooted in the theoretical assumptions of minority influence theory (Moscovici, Lage, & Naffrechoux, 1969). A minority member refers to a group member who persistently disagrees with the majority view in a decision-making group. Challenging the views of other members, a minority member forces others to investigate more than one view and to integrate those views through a divergent thinking process (Nemeth, 1995). A minority member promotes a discussion of alternative viewpoints that would otherwise not brought up in discussion (Nemeth, 1986). That is, when there is a nonconformative group member who induces dissidence or acts to criticize and challenge others’ ideas, the rest of the group members feel forced to engage in the task. For instance, a previous study shows that groups with a dissenting member use more words and negations, reflecting increased task engagement, whereas groups with an assenting member use more assents, reflecting reduced task engagement (Yilmaz & Peña, 2015). When interpersonal behaviors induce such cognitive processes, it is possible that group members are more engaged with the task, enhancing mutual engagement among group members regarding the task, and leading to better performance. Therefore, it is possible that negative behaviors will have a stronger effect on the relationship between LSM and group performance compared with positive behaviors. This prediction is summarized in the following hypothesis:

Hypothesis 3B: LSM will more strongly predict group performance in groups with a negatively communicating confederate compared with groups with a positively communicating confederate.

Method

Research Design

This study uses part of the data collected as a larger research project that employed a 2 (negative vs. positive behaviors) × 2 (shared vs. unshared social identity) factorial design to examine the effects of interpersonal behaviors and social categories on language use. Given the primary goal of this study, the data unique to the effects of interpersonal behaviors has been used for the analysis. A previous study by the same author used the rest of the data to examine a different set of dependent variables. This study examined the effects of social categories and interpersonal behaviors on LSM and other linguistic markers (Yilmaz & Peña, 2015). One hundred and thirty-six students, 34 four-person groups in total, at a large public university in the United States participated in the study. Participants ranged in age from 18 to 24 years (M = 20.54, SD = 1.83). Fifty-eight percent of participants were female and 42% were male.

Task

Participants were assigned to anonymous groups and instructed to use Windows Live Messenger to work on Straus and McGrath’s (1994) decision-making task as a group. This task asks group members to make a series of decisions in a fictitious case of a college teaching assistant accepting a bribe from the basketball team’s star player to change the student’s grade on an exam. The task requires addressing five issues regarding the status of the athlete and the teaching assistant once the bribing is discovered. The following are the five issues: the student’s grade on the exam or in the course (Issue 1), the student’s status on the athletic team (Issue 2), the student’s academic status (Issue 3), the teaching assistant’s work status as an instructor (Issue 4), and the teaching assistant’s academic status (Issue 5). An additional requirement for effective decision making is to satisfy both the academic and athletic department with the final decisions (Straus & McGrath, 1994).

Procedures

An undergraduate research assistant was recruited for the interpersonal behavior manipulation. The same confederate participated in each group discussion as one of the four group members and he was blind to the dependent measures of the experiment. The instructions were similar to those in Wang et al. (2009) study, yet there were some differences in terms of how the confederate acted out negative and positive communication behaviors. For instance, in the negative condition, the confederate was instructed to criticize the views of other group members, and to do so in a way that he did not care about how he was perceived. In the positive condition, the confederate acted like an easily agreeable and nice group member. At the linguistic level, when in the negative condition, the confederate was instructed to use direct negative statements (e.g., this does not work) or to draw attention to the requirements of the task (e.g., but you do not take into account all departments). In the positive condition, the confederate would use mostly agreement words (e.g., yeah, OK, agreed) or encouragement phrases (e.g., great job, this was easy). The confederate was closely monitored during the first experimental session and given feedback on how to better display the behavioral roles.

As participants entered the laboratory at 5-minute intervals, they were seated separately to avoid any interaction before the experiment. Computers were separated with dividers to ensure visual anonymity and to avoid the influence of individuating information on language use. Once the participants completed the decision-making task, they completed a survey and were separately debriefed. The real-time interactions of team members were saved for linguistic analysis once the participants left the laboratory.

Measures

Linguistic Style Matching

LSM measures the extent to which individuals mimic each other’s communication at the linguistic level, particularly in terms of function words. LSM was calculated by averaging the absolute difference scores for nine function word categories derived through Linguistic Inquiry and Word Count Program (see Tausczik & Pennebaker, 2010, for a detailed review). The following nine function word categories were used to calculate LSM: Verbs (e.g., to be, to have), articles (e.g., an, the), common adverbs (e.g., very, often, nearly), personal pronouns (e.g., I, you, we), impersonal pronouns (e.g., it, this, those), prepositions (e.g., on, after, with), negations¹ (e.g., not, never), conjunctions (e.g., and, but, because), and quantifiers (e.g., many, few, lots). Several studies examining cognitive, affective, and behavioral processes in virtual and face-to-face teams used LSM metric, validating its reliability (Gonzales et al., 2010; Tausczik, 2009). The following formula was used to calculate LSM.

LSMpronoun = 1 - [(│ pronoun 1 - pronoun 2 │) / (pronoun 1 + pronoun 2)]

In this formula, pronoun1 is the percentage of pronouns used by one member, and pronoun2 is the percentage used by another. The same formula was used for all nine linguistic categories and all possible pairs excluding the confederate (see Gonzales et al., 2010, for a detailed review on LSM calculations). A composite LSM score for each group was averaged and the LSM scores at the group level, for all possible pairs excluding the confederate, were analyzed. The scores ranged from 0 to 1. The higher numbers represented greater linguistic matching at the group level.

Group Performance

Performance was conceptualized as overall task effectiveness, a measure representing the quality of the decisions made by the groups (Straus & McGrath, 1994). To measure group performance, the original scoring system for the decision-making task (Straus & McGrath, 1994) was used (see the appendix). According to this system, the judgment task included two sets of point values for the issue alternatives. One set of points represented Straus and McGrath’s (1994) judgment of the extent to which each alternative supported a given division’s interests. For instance, the option of giving the basketball player a failing grade had a high point value for the academic faculty, for the sake of academic integrity, whereas low point value for the athletic staff, as it would make the student ineligible to play basketball. The second set of point values referred to the importance of an issue for a division. For example, the academic status of the player held greater importance to the athletic staff compared with the academic department. Each division was assigned three scores (from the point of view of each department’s position). As such, each department’s score included the weighted point values for that department, summed over five issues. Each group’s performance was calculated by multiplying these three scores. The multiplication generates the highest scores for decisions that have the best degree of balance among the three departments’ positions. For a more detailed explanation of the scoring system, see Straus and McGrath (1994).

Manipulation Checks

To confirm the effectiveness of the interpersonal behavior manipulation, McCroskey and McCain’s (1974) social attraction scale was used. This measurement examines the perceptions of interpersonal liking among group members (Wang et al., 2009). The scale includes items such as “I think she/he could be a friend of mine” and “I would like to have a friendly chat with him/her.” The items were arrayed in 1 (strongly disagree) to 7 (strongly agree) Likert-type scale. The reliability for the social attraction scale was acceptable (α = .89).

In addition to the social attraction scale, linguistic markers associated with negative and positive communication behaviors were analyzed to check whether the interpersonal manipulation was conveyed through the language use. Negative communication as represented by the use of negations (Tausczik, 2013), and assents as conveyed by positive behaviors (Leshed et al., 2007) were analyzed to confirm that the confederate used language that was consistent with the enacted interpersonal behaviors.

Results

Manipulation Checks

The ratings of social attraction for the positively and negatively behaving confederate were analyzed using SPSS mixed-model procedure, as multiple ratings of the same participants may lead to interdependence among observations (Kenny, 1995). The analysis showed that participants rated the confederate communicating negatively (M = 3.89, SE = .15) as less attractive than the confederate communicating positively (M = 5.39, SE = .15), F(1, 120) = 14.56, p < .001, η² = .48. The results showed that interpersonal behavior manipulation worked.

To confirm that the confederate modified the use of negative and positive communication in accordance with the assigned interpersonal behaviors, linguistic markers associated with negative behaviors (i.e., negations such as no, never) and positive behaviors (i.e., assent words such as yes, agree, yeah) were analyzed. The analysis showed that negatively communicating confederate used more negations (M = 2.26, SD = .12) compared with positively communicating confederate (M = 1.72, SD = .60), t(34) = 3.64, p = .001. On the other hand, positively communicating confederate used more assent words (M = 5.20, SD = .54) compared with negatively communicating confederate (M = 3.12, SD = .36), t(34) = −3.23, p = .004. These results corroborate that the confederate’s language reflected the assigned interpersonal communication roles as negatively communicating confederate used negations at higher rates than the positively communicating confederate. Similarly, positively communicating confederate used assents at higher rates than the negatively communicating confederate.

Hypotheses Tests

Hypothesis 1A predicted that groups with a positively communicating confederate would perform better than groups with a negatively communicating confederate. Hypothesis 1B, on the other hand, predicted that groups with a negatively communicating confederate would perform better than groups with a positively communicating confederate. To test for the mean differences in performance scores between two conditions, independent t test function of SPSS was used. The results showed that groups with a negatively communicating confederate performed better (M = 551.94, SD = 136.23) compared with groups with a confederate communicating positively (M = 308.31, SD = 75.63), t(32) = 5.39, p < .001, d = 1.90. These results supported Hypothesis 1B disconfirming Hypothesis 1A.

To test for the main effects of LSM scores on performance, we entered LSM and interpersonal behaviors into an ordinary least squares linear regression model predicting group performance. We expected that controlling for interpersonal behaviors, LSM would positively predict performance. Though there was a positive association between LSM and performance, this relationship was not statistically significant (b = 0.22, p = .10). As reported in the testing of Hypothesis 1A, interpersonal behaviors, on the other hand, significantly predicted group performance (b = −0.60, p < .001), F(2, 31) = 16.77, p < .001. These results disconfirmed the prediction in Hypothesis 2.

To examine whether interpersonal behaviors affected the relationship between LSM and group performance, a second model that included an interaction term for interpersonal behaviors and LSM was tested (see Table 1). The results revealed that LSM and interpersonal behaviors interacted to influence group performance (b = 3.40, p < .01). Accordingly, positively communicating groups produced a significant relationship between performance and LSM, F(2, 31) = 15.21, p < .001.

Table 1.

Standard Deviation Change in Group Performance due to Interpersonal Behaviors and LSM.

	Main effects model	Main effects and interpersonal behaviors × LSM model
Independent variables	β	β
Behaviors (0 = negative, 1 = positive)	−0.60**	−4.17*
LSM	0.22	−0.35
Behaviors × LSM		3.40**
R ²	0.52	0.60
Adjusted R	0.48	0.56

Note. LSM = linguistic style matching.

p ≤ .05. **p ≤ .01.

To interpret the interaction effects, we decomposed the effect of LSM on performance in each condition separately. For each standard deviation increase in LSM, positively communicating groups performed better by .67 standardized unit increase in performance scores (b = 0.67, p < .01). Performance in negatively communicating groups decreased by .28 standardized units for each standard deviation increase in LSM. Yet this value did not achieve significance (b = −0.28, p = .25). Although LSM accounted for 45% of variation in performance among positively communicating group members, it only accounted for 8% of variation in performance among negatively communicating groups. These results provided evidence for Hypothesis 3A, disconfirming Hypothesis 3B (see Figure 1).

Figure 1.

The interaction of interpersonal behaviors and LSM on performance.

Discussion

Interpersonal Effects on Group Performance

How do interpersonal communication behaviors manifested in language affect group performance? The findings of this study document that negative communication behaviors—in the form of challenging others’ opinions or pushing other collaborators to take into account alternative options—trigger higher group performance compared with positive communication behaviors.

In the presence of a negatively communicating collaborator, it is possible that other group members feel more accountable and conspicuous during group discussions. This explanation makes sense in light of the research providing evidence for the positive effects of critical communication norms on group performance (Postmes, Spears, & Cihangir, 2001). When applied to the context of this study, it is plausible to argue that a dissenting collaborator might trigger a critical communication norm in group interaction, which influences the way other group members communicate. Linguistic markers can provide evidence for the presence of criticality in discussions. For instance, a previous study by this author reported that in the presence of a dissenting group member, group discussions included more negations than those with an assenting group member. Similarly, positively communicating groups used more assents compared with negatively communicating groups (Yilmaz & Peña, 2015). This finding suggests that a negatively communicating member can affect whether other group members contribute to the group discussions in a critical or a consensual way. Future studies should test for the direct effects of dissenting behaviors on the formation of communication norms in group discussions.

By criticizing and challenging majority ideas, the negatively communicating member can elicit a minority influence (Moscovici et al., 1969). Past research showed that minority influence overrode the negative effects of groupthink by inducing a divergent thinking process (Nemeth, 1995). That is, if the group members are inclined to reach a consensus without discussing pros and cons of their ideas, minority influence can deter them from making hasty decisions. For instance, in the below conversation, the dissenter points out the importance of satisfying the needs of all departments before settling on a decision. By doing so, the dissenting member reminds other collaborators that their decisions are only valid if they follow the task instructions. Though the positive confederate attempts to draw attention to similar issues in group discussions, the negative confederate uses a more persistent and firm language compared with the positive confederate. When we look at this episode involving a dissenting member (i.e., negatively communicating confederate), we observe a critical communication pattern that forces other group members to take the task requirements into consideration, implying that the confederate is not mindlessly agreeing with the majority’s viewpoint:

Sphinx2:

Look whoever wants to give him an F is not following the instructions on the sheet of satisfying all depts. How does this satisfy athletics dept?

Pyramid1:

OK then you enforce the penalty the next season. Maybe you give him a B so he can finish and the team may win the conference championship. But then the next season he is suspended for a certain amount of time?

As illustrated in the preceding example, when there is a dissenting member in the group (i.e., Sphinx2), the discussion is not characterized by the exchange of agreement words. Instead, group members are forced to provide a rationale for their decisions or generate alternative solutions. A main implication of this finding is that the negatively communicating member sets a critical communication norm, holding group members accountable for evaluating the validity of their decisions in light of the task requirements. In this example, the confederate responds to the nonconfederate member by pointing out the importance of satisfying all departments while making a decision. This episode exemplifies one of many different ways the negatively communicating member engages the rest of the group members to make sure that the pros and cons of each decision are discussed. Yet it is important to note that the intention of the confederate is not to lead the group members toward the best decision by using this approach, as he does not have any information about the best or the worst decision in the task. The confederate simply forces other group members to revisit their decisions and revaluate them to confirm that they are aligned with the requirements of the task.

In the absence of a dissenting member, collaborators tend to arrive at a decision without analyzing the pros and cons of their choices critically. When group members are agreeing with one another, without comparing and contrasting different views, it is not possible to reach mutual engagement (Bryan-Kinns et al., 2007). Linguistically, idea generation is reflected through word counts. Previous research found that high-performing group members communicated more frequently and generated more words than low-performing members (Sexton & Helmreich, 2000). Similarly, certain linguistic markers such as the use of assents were found to be associated with low task focus and high groupthink (Leshed et al., 2007). When evaluated through the lens of the groupthink phenomenon, it is not surprising that groups with a negatively communicating collaborator performed better than those with a positively communicating collaborator, as the former forces group members to be engaged in the task. Research on creativity in groups also documented that reaching a premature group consensus was associated with low productivity and creativity (Chirumbolo, Mannetti, Pierro, Areni, & Kruglanski, 2005). Facilitating a divergent thinking process, dissenting communication can steer groups toward a greater mutual engagement.

Past research showed that when group members were primed with a critical group norm, they approached the task more critically, and their discussions included critical contributions, compared with a group where a consensual norm was salient (Postmes, Spears, & Cihangir, 2001). In the context of the present study, when there was not a dissenting group member in the group, group members attempted to reach decisions without paying much attention to whether they were considering all the requirements of the task. For instance, in the positively communicating groups, very few members mentioned the importance of satisfying the needs of each department, which was a crucial component of their performance results. As seen in the following example, Sphinx1 is the positively communicating confederate. He is initially trying to steer the direction of discussion toward the issue of satisfying the needs of all departments. Yet once the majority disagrees with him, he does not push others further, and complies with the majority view:

Sphinx1:

Yes the B could be, but it doesn’t satisfy everyone—the college needs to keep their rep while the basketball team needs Jack, so by giving him his initial grade, they are actually cutting him some slack

Pyramid2:

So . . . ur telling me you never cheated before. Everyone makes mistakes and should get a second chance

Sphinx2:

I agree with P2. 20 years down the line no one will remember this but if we ruin jacks bbal career

Pyramid1:

You may ruin part of his season, but at least he’ll learn something about cheating not being ok

Sphinx1:

Yes but life isn’t fair and we need to try to treat everyone the same.

Pyramid2:

It wasn’t cheating. It was bribery. It was the TAs fault for taking the money

Sphinx1:

OK. You are right.

This example illustrates that in the absence of a dissenting member, group members reach a consensus without “critically” analyzing the effects of their discussion on the quality of decisions. For instance, the above example illustrates that the group members do not take into consideration how their decision would satisfy the academic department of the TA or the academic department of the basketball player. Though the confederate tries to make group members look at the issue from multiple perspectives to address each department’s expectations, the majority view overrides the confederate’s efforts to further discuss this issue. Eventually, without further discussion, the confederate agrees with the majority. Thus, we can argue that the presence of a negatively communicating group member can make critical group norms more salient in the group discussions, and thereby, lead to improved decision-making processes and better group performance. After all, situational group norms are inferred through real-time interactions in anonymous virtual groups (Postmes, Spears, Sakhel, et al., 2001). That is, when group members are consistently exposed to a critical communication style, they may infer that this is the normative communication behavior in the group, and thus, adapt to it (Postmes, Spears, Sakhel, et al., 2001). As such, future studies should examine how dissenting behaviors lead to changes in the language use of nonconfederate members (e.g., increased use of cognitive words), representing critical processing of information.

Linguistic Style Matching, Interpersonal Behaviors, and Group Performance

The present study examined the relationship between how virtual team members accommodated their language toward one another and how well they performed in an interdependent task. Though LSM did not significantly predict performance, the interaction of LSM and interpersonal behaviors showed that when the group communication was characterized by positive behaviors, there was a positive relationship between LSM and performance. However, when the group was communicating negatively, LSM did not predict group performance. This finding shows that the predictive value of LSM for group performance should be examined in light of the interpersonal behaviors underlying the group discussions, particularly in the context of virtual teams.

The notion that LSM affects the association between positive behaviors and performance validates previous findings on the relationship between LSM, performance, and positive behaviors (Gonzales et al., 2010; Tausczik, 2013). This finding also provides evidence for CAT-based approaches to the relationship between LSM, interpersonal behaviors, and performance. That is, in positively communicating groups, LSM can be used as an indicator of group performance. Yet this assumption should be interpreted with a caveat, as in the present study negatively communicating groups performed better compared with positively communicating groups. Previous studies reporting evidence for LSM as a performance indicator did not tease apart unique effects of interpersonal behaviors (e.g., negative vs. positive communication) within group discussions. For instance, Gonzales et al. (2010) looked at the relationship between LSM and performance across face-to-face and computer-mediated groups. This same study reported that LSM was a positive predictor of performance only in face-to-face teams. That is, LSM did not predict performance in computer-mediated groups. Similarly, Tausczik (2013) examined how verbal feedback affected group processes, and thereby, the relationship between LSM and performance in virtual teams. Though this study found a significant association between LSM and performance, the researchers did not manipulate the interpersonal dynamics in the group discussions.

Previous findings highlight the relationship between LSM and positive communication behaviors, yet do not point out the importance of negative interpersonal behaviors in the examination of how LSM can reflect performance in virtual teams. This study shows that the effects of interpersonal behaviors, particularly the negative ones, override the influence of LSM on performance in virtual teams. Though negative behaviors may induce critical communication norms (Postmes, Spears, & Cihangir, 2001) and minority influence (Nemeth, 1995), leading to high performance in negatively communicating groups, it is possible that the LSM metric only captures performance among positively communicating members.

The fact that LSM predicted performance only in positively communicating groups implies that LSM did not capture mutual engagement processes among negatively communicating members. LSM has been traditionally associated with positive behaviors reflecting mutual liking among communication partners (Gonzales et al., 2010; Ireland et al., 2010). As such, it is not surprising that LSM did not predict performance in the negative condition. However, this finding should not preclude LSM as a performance indicator in negatively communicating virtual groups. It is possible that groups with dissidents go through more turbulent communication episodes, and their synchronization at the linguistic level may have more ups and downs compared with positively communicating collaborators. In this study, LSM was calculated as a one-time construct. Future studies should measure LSM during different points of the decision-making process (e.g., beginning, middle, end). It would be helpful to see how the fluctuations at the linguistic level in different time intervals affect LSM, and thereby, its association with group performance, particularly in negatively communicating groups.

One limitation of this study is measuring group performance using the quality of decisions and how well each decision met the expectations of the departments involved. Though this is a validated and reliable measurement of group performance (Straus & McGrath, 1994), there are other operationalizations of performance such as the level of agreement among group members, satisfaction, and speed of decision making (Hirokawa & Salazar, 1999). Future studies could employ different performance measures to examine how interpersonal behaviors and LSM affect team success.

Also, dissenting communication (i.e., minority influence) was manipulated through the confederate’s behaviors. Although this poses an important limitation, the linguistic analysis documented that the confederate conveyed dissenting and assenting communication through assigned behavioral roles. For instance, when the confederate was the minority member, he played the devil’s advocate and used negations more so than the positively communicating confederate. Similarly, while acting out the positive behaviors, the confederate used assents, confirming the majority views. Future studies could manipulate minority influence by making majority views salient to validate current findings.

Conclusion

This study found that negatively communicating groups performed better than positively communicating groups. Also, LSM was more strongly associated with group performance in positively communicating groups compared with negatively communicating groups. The complex relationship between interpersonal behaviors, LSM, and performance suggests that while using LSM as a predictor of group performance, the valence of interpersonal behaviors should be taken into consideration. In a work group where all members get along and like each other, the discussion might not necessarily lead to optimal decision making if group members focus too much on socioemotional aspects of group interactions, neglecting information sharing for task completion (Lebie, Rhoades, & McGrath, 1996). However, when there is a dissenting member who induces a critical communication norm (Postmes, Spears, & Cihangir, 2001) or functions as a minority member (Nemeth, 1995; Nemeth & Rogers, 1996), groups perform better. Though LSM is associated with group performance, this association is only in effect when group members are communicating positively. As such, it is important to analyze how this relational aspect interacts with LSM to influence overall performance. After all, as documented by the effects of negative behaviors on performance, it is rarely the sweet-talking that gets the job done in virtual teams.

Footnotes

Appendix

Acknowledgements

The author would like to thank Howard Giles and the two anonymous reviewers for their valuable feedback on this article. Special thanks to Chris Parlato for his helpful comments and suggestions on previous drafts of the article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

Gamze Yilmaz, (PhD, University of Texas at Austin) is an assistant professor in the Communication Department at University of Massachusetts Boston. She studies social, cognitive and behavioral processes in computer-mediated communication with a particular focus on language use.

References

Bales

R. F.

(1953). The equilibrium problem in small groups. In Parsons

Bales

R. F.

Shils

E. A.

(Eds.), Working papers in the theory of action (pp. 111-161). Glencoe, IL: Free Press.

Bazarova

N. N.

Walther

J. B.

(2009). Attributions in virtual groups: Distances and behavioral variations in computer-mediated discussions. Small Group Research, 40, 138-162.

Bryan-Kinns

Healey

Leach

(2007). Exploring mutual engagement in creative collaborations. In Proceedings of the 6th Conference on Creativity and Cognition (pp. 223-232). New York, NY: Association for Computing Machinery.

Cannon-Bowers

J. A.

Salas

Converse

S. A.

(1993). Shared mental models in expert team decision making. In Castellan

N. J.

Jr. (Ed.), Individual and group decision making: Current issues (pp. 221-246). Hillsdale, NJ: Lawrence Erlbaum.

Chirumbolo

Mannetti

Pierro

Areni

Kruglanski

A. W.

(2005). Motivated closed-mindedness and creativity in small groups. Small Group Research, 36, 59-82.

Curhan

J. R.

Pentland

(2007). Thin slices of negotiation: Predicting outcomes from conversational dynamics within the first five minutes. Journal of Applied Psychology, 92, 802-811.

Edwards

H. K.

Sridhar

(2005). Analysis of software requirements engineering exercises in a global virtual team setup. Journal of Global Information Management, 13, 21-41.

Fischer

McDonnell

Orasanu

(2007). Linguistic correlates of team performance: Toward a tool for monitoring team functioning during space missions. Aviation, Space, and Environmental Medicine, 78, 86-95.

Giles

Soliz

(2014). Communication accommodation theory: A situated framework for interpersonal, family, and intergroup dynamics. In Braithewaite

Schrodt

(Eds.), Engaging interpersonal theories (2nd ed., pp. 159-167). Thousand Oaks, CA: Sage.

10.

Gonzales

A. L.

Hancock

J. T.

Pennebaker

J. W.

(2010). Language style matching as a predictor of social dynamics in small groups. Communication Research, 37, 3-19.

11.

Hirokawa

R. Y.

Salazar

A. J.

(1999). Task-group communication and decision-making performance. In Frey

L. R.

(Ed.), The handbook of group communication theory and research (pp. 167-191). Thousand Oaks, CA: Sage.

12.

Huffaker

D. A.

Swaab

Diermeier

(2011). The language of coalition formation in online multiparty negotiations. Journal of Language and Social Psychology, 30, 68-81.

13.

Ireland

M. E.

Slatcher

R. B.

Eastwick

P. W.

Scissors

L. E.

Finkel

E. J.

Pennebaker

J. W.

(2010). Language style matching predicts relationship initiation and stability. Psychological Science, 22, 1-6.

14.

Janis

I. L.

(1982). Groupthink: Psychological studies of policy decisions and fiascoes. Boston, MA: Houghton Mifflin.

15.

Kenny

D. A.

(1995). The effect of nonindependence on significance testing in dyadic research. Personal Relationships, 2, 67-75.

16.

Lebie

Rhoades

J. A.

McGrath

J. E.

(1996). Interaction process in computer-mediated and face-to-face groups. Computer Supported Cooperative Work, 4, 127-152.

17.

Leshed

Hancock

J. T.

Cosley

McLeod

P. L.

Gay

(2007). Feedback for guiding reflection on teamwork practices. In Proceedings of the International ACM Conference on Supporting Group Work (pp. 217-220). New York, NY: Association for Computing Machinery.

18.

Leshed

Perez

Hancock

J. T.

Cosley

Birnholtz

Lee

. . . Gay

(2009). Visualizing real-time language-based feedback on teamwork behavior in computer-mediated groups. In Proceedings of ACM Conference on Human Factors in Computing Systems (pp. 537-546). New York, NY: Association for Computing Machinery.

19.

Loh

Smyth

(2010). Understanding students’ online learning experiences in virtual teams. Journal of Online Learning and Teaching, 6, 335-342.

20.

Mathieu

J. E.

Heffner

T. S.

Goodwin

G. F.

Salas

Cannon-Bowers

J. A.

(2000). The influence of shared mental models on team process and performance. Journal of Applied Psychology, 85, 273-283.

21.

McCroskey

McCain

(1974). The measurement of interpersonal attraction. Speech Monographs, 41, 261-266.

22.

Miell

MacDonald

(2000). Children’s creative collaborations: The importance of friendship when working together on a musical composition. Social Development, 36, 348-369.

23.

Moscovici

Lage

Naffrechoux

(1969). Influence of a consistent minority on the responses of a majority in a color perception task. Sociometry, 32, 365-380.

24.

Nemeth

(1986). Differential contributions of majority and minority influence. Psychological Review, 93, 23-32.

25.

Nemeth

(1995). Dissent as driving cognition, attitudes, and judgments. Social Cognition, 13, 273-291.

26.

Nemeth

Rogers

(1996). Dissent and the search for information. British Journal of Social Psychology, 35, 67-76.

27.

Peña

Hancock

J. T.

(2006). An analysis of socioemotional and task communication in online multiplayer video games. Communication Research, 33, 92-109.

28.

Postmes

Spears

Cihangir

(2001). Quality of decision making and group norms. Journal of Personality and Social Psychology, 80, 918-930.

29.

Postmes

Spears

Sakhel

De Groot

(2001). Social influence in computer-mediated communication: The effects of anonymity on group behavior. Personality and Social Psychology Bulletin, 27, 1243-1254.

30.

Schunn

Crowley

Okada

(2002). What makes collaborations across a distance succeed? The case of cognitive science community. In Hinds

Kiesler

(Eds.), Distributed work (pp. 407-430). Cambridge: MIT Press.

31.

Scissors

L. E.

Gill

A. J.

Gergle

(2008). Linguistic mimicry and trust in text-based CMC. In Proceedings of the ACM Conference on Computer Supported Cooperative Work (pp. 277-280). New York, NY: Association for Computing Machinery.

32.

Sexton

J. B.

Helmreich

R. L.

(2000). Analyzing cockpit communications: The links between language, performance, and workload. Human Performance in Extreme Environments, 5, 63-68.

33.

Straus

S. G.

McGrath

J. E.

(1994). Does the medium matter? The interaction of task type and technology on group performance and member reactions. Journal of Applied Psychology, 79, 87-97.

34.

Tausczik

(2009). Linguistic analysis of workplace computer-mediated communication (Unpublished master’s thesis). University of Texas, Austin.

35.

Tausczik

(2013). Changing group dynamics through computerized language feedback (Unpublished doctoral thesis). University of Texas, Austin.

36.

Tausczik

T. R.

Pennebaker

J. W.

(2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29, 24-54.

37.

Taylor

P. J.

Thomas

(2008). Linguistic style matching and negotiation outcome. Negotiation and Conflict Management Research, 1, 263-281.

38.

Walther

J. B.

Bunz

Bazarova

(2005). The rules of virtual groups. In Proceedings of the 38th Annual Hawaii International Conference on System Sciences. Washington, DC: Institute of Electrical and Electronics Engineers.

39.

Wang

Walther

J. B.

Hancock

(2009). Social identification and interpersonal communication in computer-mediated communication: What you do versus who you are in virtual groups. Human Communication Research, 35, 59-85.

40.

Yilmaz

Peña

(2014). The influence of social categories and interpersonal behaviors on future intentions and attitudes to form subgroups in virtual teams. Communication Research, 41, 333-352.

41.

Yilmaz

Peña

(2015). How do interpersonal behaviors and social categories affect language use? The case of virtual teams. Communication Quarterly, 63, 427-443