Reaping the benefits of intrateam conflict : the role of transactive memory systems

(1)

Reaping the benefits of intrateam conflict:

The role of transactive memory

systems

Charity Miller

10112707

MSc. Business Administration - Strategy Track

Amsterdam Business School

University of Amsterdam

Thesis Coach: Pepijn van Neerijnen

August 31

st

, 2015

Final version

(2)

Statement of originality

This document is written by Student Charity Miller who declares to take

full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original

and that no sources other than those mentioned in the text and its

references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the

supervision of completion of the work, not for the contents.

(3)

Overview of figures and tables

Figure 1. Conceptual Model.

Figure 2. Moderating Effect of Role Clarity on the Relationship Between Cognitive Conflict and TMS

Figure 3. Standardized Regression Coefficients for the Relationship between Cognitive Conflict & Team Performance, as Mediated by TMS.

Table 1. Explanatory factor analysis for items in the TMS scale (Pattern Matrix) Table 2. Comparison of Measurement Models

Table 3. Results of Confirmatory Factor Analysis and Measurement Properties Table 4. Descriptive Statistics and Correlations

Table 5. Multiple Regression Results

(5)

0 Abstract

Conflict in teams is a frequently occurring phenomenon, which can be both beneficial as well as detrimental for team performance. The positive effects are argued to come forth out of an increase in information exchange leading to less information processing failures. Yet empirical evidence explaining how teams are able to overcome

inefficiencies in information processing, caused by increased information sharing, is absent in current literature. One of the ways information can be processed effectively in a team is through the team‟s transactive memory system (TMS), which describes how knowledge within the team is encoded, stored, and retrieved. However, little is known about what role a team‟s TMS plays in the presence of cognitive conflict. Furthermore, effect of role clarity on TMS development has not been fully explored. In order to close this gap, 96 teams of in total 445 individuals were studied, based on a three-wave lagged design. Results of this research show that cognitive conflict has a positive impact on the strength of a team‟s TMS, which in turn has a positive effect on team performance. Role conflict was found to positively moderate the influence of cognitive conflict on TMS development in teams with an average or high level of cognitive conflict. Furthermore, the effect of cognitive conflict on performance was mediated by TMS. This implies that TMS is a crucial factor in reaping the benefits of cognitive conflict and that defining clear roles can help teams to prosper when cognitive conflict is present.

1 Introduction

Many organizations have switched from formal bureaucratic structures to organizing work though team based designs (Marks, DeChurch, Mathieu, Panzer & Alonso, 2005).

(6)

Teams are responsible for developing products, improving services and managing operations (Cummings, 2004). As a result, uncovering how team performance can be influenced has been the focus of numerous studies (Mathieu, Maynard, Rapp & Gilson, 2008). Processes within teams are a key element of this research, as they play a large role in determining how effective a team functions (Ilgen, Hollenbeck, Johnson & Jundt, 2005).

Due to its common occurrence in teams, the process of intragroup conflict is of particular interest (Kozlowski & Bell, 2003). A lively debate is now taking place because cognitive conflict has been found to have positive as well as negative effects on desired outcomes. This specific type of conflict occurs when team members have opposing views on matters related to the task the team has to perform (Amason & Sapienza, 1997). Research has shown that cognitive conflict can increase decision quality (Amason, 1996; Olson, Parayitam & Bao, 2007), innovation (De Dreu, 2006; De Dreu & West, 2001) and performance in various settings (Jehn, 1995; Jehn & Mannix, 2001; Bradley, Klotz, Baur, & Banford, 2013). However, these results are contrasted by a number of studies claiming conflict is inherently detrimental for numerous aspects of team performance (Dreu & Weingart, 2003; Jehn, Northcraft, & Neale; 1999).

The positive effects of conflict are thought to come forth from an increased information exchange between team members. This leads to a deeper, more thorough examination of the subject matter, which can help the team achieve higher performance (De Wit, Greer & Jehn, 2012; Greer, Jehn & Mannix, 2008).

The counter argument is that cognitive conflict may evolve into relationship conflict when disagreements are taken personally (Simons & Peterson, 2000). This type

(7)

of conflict relates to members feeling a sense of interpersonal incompatibility (Jehn & Mannix, 2001), resulting in team members feeling less satisfied in their role as a group member (Jehn, 1995) and less willing to collaborate (De Dreu, 2006). These negative affective reactions can in turn lead to lower team performance (Mathieu et al., 2008). Overall, the majority of current conflict literature argues that benefiting from cognitive conflict depends on the team‟s ability to avoid relationship conflict by

maintaining the quality of the intra-team dynamics. Teams that have a strong trust base, resolve conflicts, and communicate in a collaborative manner are more likely to benefit from cognitive conflict (Behfar, Peterson, Mannix & Trochim, 2008; Bradley, Klotz, Postlethwaite & Brown, 2013; Simons & Peterson, 2000). While avoiding relationship conflict helps teams to prevent negative affective reactions from hurting team

performance, it does not explain how teams are able to turn cognitive conflict into an advantage or how these positive effects can be enhanced.

To address this gap in the literature this thesis argues that teams that benefit from conflict do so by overcoming biases in the way they process information. This helps the team overcome information-processing failures, thereby increasing performance

(Shippers, Edmondson & West, 2014). A new perspective on team conflict is presented by viewing the effects of conflict though an information processing lens. This allows the following contributions to be made.

First, empirical evidence is provided that cognitive conflict can influence the development of information processing systems within teams. Most studies focus on a negative relationship between cognitive conflict and information processing due to an increases the cognitive load inhibiting creativity and taking resources and focus away

(8)

from the main task (Carnevale & Probst, 1998; De Wit et al., 2012). However, this study taps into the way information processing could be enhanced through cognitive conflict. Results show that cognitive conflict positively impacts the strength of a team‟s

transactive memory system (TMS). A TMS describes how teams divide responsibilities related to the encoding, storage, retrieval, and communication of knowledge from various domains by relying on each other for information (Hollingshead, 2001; Hollingshead & Brandon, 2003). Examining the relationship between conflict and TMS contributes to the understanding how conflict affects team members‟ willingness to work together to divide cognitive labor. Furthermore, it emphasizes the importance of conflicting points of view within the team for the development of a TMS.

Second, the presence of a well-developed information-processing system helps explain how teams are able to efficiently integrate different perspectives into a collective outcome. Accordingly, a team‟s TMS may act as a mediator between cognitive conflict and performance. This explains how teams engaged in cognitive conflict are able to reach better decisions, without the increase of shared information impairing effective

information processing. Consequently, a contribution is made towards resolving the question how teams are able to benefit from conflict.

Third, this study examines how team members‟ perception of their role within the team affects the strength of the team‟s TMS. Role clarity is predicted to help teams use cognitive conflict to build their TMS. This highlights the importance of defining clear roles for the development of information processing systems.

The conceptual model is presented in figure 1. To test this model data was used from 445 individuals in 96 teams, which participated in a simulation called the “Business

(9)

Strategy Game” (Thompson & Stappenbeck, 2001) over the course of six weeks. This game mimics the complexity and ambiguity of the actual business environment, creating a realistic setting in which to analyze teams (Tompson & Dass, 2000). The development of the teams was studied using a three-wave survey, with time lags between each moment of measurement. Time lag methodology provides stronger evidence for the causal

relationship between the level of cognitive conflict in a team, the strength of their TMS and their team performance. With a few exceptions (Jehn & Mannix, 2001; Behfar et al., 2008), most studies related to intra-team conflict are cross-sectional. However, the disadvantage of a cross-sectional design is that it does not allow the researcher to distinguish a causal explanation for an association from other possible explanations (Ployhart & Vandenberg, 2010). Introducing time lags also helps to minimize common method bias by reducing participant‟s memory of their answers related to a construct measured at an earlier time (Podsakoff, MacKenzie, Lee & Podsakoff, 2003).

In the following section the theoretical model is developed through the examination of team performance, cognitive conflict and TMS literature. Then, the research method is explained, followed by the results of the testing of the hypotheses. Finally, the implications for research and management are discussed.

(10)

Figure 1.

Conceptual Model.

2 Literature

2.1 Information processing and team performance

Within organizations, work can be organized based on individual jobs or based on teams of two or more employees. While these work teams can be created for various reasons, they are all characterized by individuals that perform organizationally relevant tasks, have shared goals, interact socially, exhibit task interdependencies and maintain and manage boundaries (Kozlowski and Bell, 2003, p. 334).

Having employees work in teams can have various benefits for organizations. First, teams are a solution to handling knowledge-intensive tasks that are too complex and novel to be handled by one person alone (Cummings & Haas, 2011). As a result, these teams of knowledge workers exist in various industries such as consulting, product development, engineering, and software services (Gardner, Gino & Staats, 2012). The

(11)

value of teams further increases in dynamic, competitive environments. Teams can provide the fast, adaptive responses needed to prosper in his type of climate by

combining the knowledge and skills of the different members (Kozlowski and Bell, 2003). Both advantages of teamwork come forth out of the team‟s ability to use input from multiple individuals to make decisions and devise solutions. This knowledge must be transformed and integrated through mutual adjustment for the team to be successful (Gardner at al, 2012; Majchrzak, More & Faraj, 2012). The ability to combine the

expertise of team members ultimately has a large influence in determining how profitable and innovative a firm can be (Gardner et al., 2012).

The importance of this aspect of team functioning has led to a large amount of related research. One popular approach to the study of knowledge integration in teams focuses on information processing. The term information processing relates to various reactions to informational input and objectives. Once information is perceived, it can be encoded through the process of structuring, evaluation, interpretation, and transformation. Information is then stored in the memory and can be brought back into awareness by the retrieval process (Hinsz, Tindale, and Vollrath, 1997).

The information processing perspective examines how teams engage in cycles of information processing at the individual level and at the team level, until the team reaches a decision (De Dreu, Nijstad, Bechtoldt & Baas, 2011). Thus, encoding, storage and retrieval takes place both within individuals as well as within teams. During these cycles, information is shared among the members of the team. This information can directly refer to the task at hand, but it can also be related to group interactions, other members within the group, or the larger context in which the group exists (Hinsz et al., 1997).

(12)

During information processing information is not always analyzed correctly. The process of storing and retrieving information can result in distortions and a mix up of evidence, causing biases in decision-making (Hilbert, 2012). These errors tend to be amplified in teams because members start to process information in more homogenous ways, accentuating similarities (Hinsz et al., 1997). When adopting a shared perspective steers the team away from the optimal decision, the quality of this decision is recued (Brodbeck, Kerschreiter, Mojzisch & Schulz-Hardt, 2007).

Another limitation teams have regarding information processing is a finite amount of information processing capacity. How well the team is able to perform depends on the team‟s capability to put their cognitive resources to optimal use (Hackman & Katz, 2010). Similarly, time is usually also a resource constraint as many teams face the pressure of having to fulfill their objective within a limited period (Majchrzak et al., 2012).

Consequently, teams have to choose how to divide their attention. While putting more time and effort into a task is likely to lead to better and more accurate decisions (De Dreu, 2007), timeliness is crucial for the effectiveness of strategic decisions (Schweiger, Sandberg & Rechner, 1989). This tradeoff is especially challenging for teams engaged in complex tasks, due to the overload of information in knowledge-intensive environments (Cummings & Haas, 2011). While information processing is inherently time-consuming, certain processes can either increase efficiency (Mathieu et al., 2008) or hinder it

(Steiner, 1972). Therefore, achieving high performance depends on the way the team divides their time and cognitive resources, as well as which processes they engage in during the processing of information.

(13)

2.2 The influence of intrateam conflict on team performance

Conflict is said to be present within a team when one or more members take actions that are conflicting with the interests of other group members (Baron, 1990, p. 199). These actions can be either verbal or nonverbal expressions of opposition (Weingart, Behfar, Bendersky, Todorova & Jehn, 2015). Furthermore, the sources of disagreement can be related to resources, beliefs, values or practices that matter to the team members (De Dreu & Gelfand, 2008, p. 6).

The fundament of the positive effects of conflict is an increased amount of information-exchange between members of the team. This allows the team to overcome biases and errors in decision-making. Shippers, Edmondson & West (2014) identified three team information-processing failures. Cognitive conflict can help teams to overcome these three facets of suboptimal information processing.

The first failure involves team members refraining from sharing relevant information when they believe other members are already aware of this knowledge, or when they deem it unimportant for the issue being considered. Cognitive conflict increases members‟ understanding of the subject matter, as a result of it being discussed more thoroughly (Amason, 1996; De Wit et al., 2012; Greer et al., 2008; Olson et al., 2007). This deeper understanding can illuminate the importance of the unshared information while a more elaborate discussion can highlight which members are not aware of it. Therefore, cognitive conflict can stimulate members to share previously unshared information and reduce this information processing failure.

Second, team members can fail to analyze shared information to the extent necessary to identify critical relationships and implications related to this information.

(14)

Biased information processing is the cause of this failure, which can be the result of members having individual goals that differ from the team goals, being influenced by the way the problem is framed or relying on heuristics and subjective preferences. Cognitive conflict can counter this failure by ensuring critical evaluation of other member‟s

knowledge and ideas (Amason, Thompson, Hochwater, & Harrison, 1995; De Wit et al., 2012; Nemeth, 1995). Identifying gaps or errors in argumentations helps teams to filter out any subpar ideas or improve the soundness of them. Furthermore, conflict allows teams to combine the input from multiple points of view (De Dreu & Weingart, 2003; Greer et al., 2008; Schweiger et al., 1989). Without different opinions being expressed, decisions are based on the ideas of the most influential or vocal group member (Amason et al., 1995). This causes the quality of the decisions to rely wholly on the expertise of one member. When members show differences in preference, more alternative solutions are considered (Schulz-Hardt, Jochims & Frey, 2002), and teams will be more able to spot critical implications and relationships.

The third information processing failure is refraining from updating prior

conclusions and behaviors when the situation changes. This can be due to team members being stuck in routines or being overly committed or biased to a chosen course. Cognitive conflict on the other hand reduces biases during group discussions (De Wit et al., 2012). Members discuss information that is new to them, or inconsistent with their preferences (Schulz-Hardt, Brodbeck, Mojzisch, Kerschreiter, & Frey, 2006). This can lead to the reevaluation of member‟s own assumptions and recommendations (Schweiger et al, 1989). In this way cognitive conflict can help teams unearth the best solution, as opposed to the current one.

(15)

In conclusion, by reducing team information-processing failures cognitive conflict can lead to better decisions and solutions, which help teams to accomplishing or exceed their goals. Therefore the following hypothesis is proposed:

H1: Cognitive conflict positively impacts team performance.

2.3 The influence of cognitive conflict on Transactive Memory Systems

The increase of information-exchange with a team can also effect members‟ perceptions of the knowledge that other members have. Information about which people have useful knowledge, and what type of knowledge that is, is stored in an individual‟s transactive memory (Ren & Argote, 2011). When transactive memory is used by a group of people to store, retrieve and communicate information it is called a transactive memory system (Lewis, 2003).

Teams with a well-developed TMS have three characteristics: specialization, credibility and coordination (Lewis, 2003). These dimensions are indicators how well underlying latent construct is functioning (Pearsall, Ellis & Bell, 2010). Specialization entails that the knowledge within the team is differentiated, with each member having expertise in a specific area relevant to the execution of the task. Credibility refers to team members trusting each other to have reliable information. Finally, coordination is the effective and orchestrated retrieval and application of knowledge to complete a task (Lewis, 2003).

Cognitive conflict can help teams build a strong TMS by providing the

opportunity for teams to develop these aspects. Specialization can increase when team members become more aware of how the roles are divided in the group and more unique

(16)

information is shared. Role identification behavior is a process that occurs during team interaction, when team members discuss their own and their team members‟ capabilities and responsibilities within the team (Kozlowski, Gully, Nason & Smith, 1999). This behavior helps a team to development a TMS by increasing team members‟ awareness of each other‟s expertise (Pearsall et al., 2010). Cognitive conflict can involve the team disagreeing about the division and delegation of responsibility (Jehn, 1997, p. 540), thereby stimulating role identification behavior and ultimately specialization. Furthermore, the awareness of specialization among members may increase when cognitive conflict stimulates the sharing of information uniquely held by one individual. When this knowledge is integrated into the team‟s output, it could trigger members to further develop their expertise in a specific knowledge area and to rely on other members for information they do not possess themselves.

Credibility can benefit when team members argue their point of view and improve the accuracy of their knowledge in expectation of critique. During cognitive conflict team members elaborate their positions and the ideas and information that support them

(Tjosvold, 1991, p. 15), thereby arguing their point of view. Critical discussion of this shared knowledge offers the opportunity for this knowledge to be validated by other members (Brandon & Hollingshead, 2004). This helps the team assess the accuracy of the knowledge held by various members. Furthermore, when members are aware that their view might not automatically be accepted they could be triggered to make sure the information they share is correct.

Coordination within the team can improve when members increase their

(17)

facilitate coordination in two ways. First, increased understanding of the subject matter can help avoid misunderstandings between team members about what needs to be

accomplished. Second, a deeper understanding combined with differentiated opinions can help reveal new insights into how member‟s knowledge and skills can be combined to fulfill the task. Fewer misunderstandings and new combinations of expertise can in turn enhance effectiveness in team knowledge processing, which leads to increased

coordination (Lewis, 2003).

Concluding, cognitive conflict can help a team to identify the knowledge each member possesses, develop member expertise, assess and increase the credibility of this knowledge and increase coordination. This argument leads to hypothesis 2:

H2: Cognitive conflict positively impacts TMS.

2.4 The moderating effect of role clarity on TMS development

A role is a pattern of behaviors perceived by a member of an organization as behaviors that are expected, including behaviors that are not defined in terms of specific job tasks (Tubre & Collins, 2000). While a role is linked to a particular position in a specific social system (Polzer, 1995, p. 495), individuals in that position don‟t always know what their precise role is. This role ambiguity can be defined as “the level of uncertainty or lack of clarity surrounding expectations about a certain role” (Ilgen & Hollenbeck, 1991, p. 191). Following this definition, role clarity on the other hand can be explained as having a clear sense of the expectations associated with your role. Overall, members seem to prefer having clearly defined roles within a team. It makes them more satisfied and willing to remain working in the team (Jackson & Schuler, 1985; Lyons, 1971).

(18)

In relation to cognitive conflict, role clarity has only been examined for the effect it has on cognitive conflict turning into relationship conflict (Tidd, McIntyre & Friedman, 2004). Furthermore, no studies have examined the effect of role clarity on information processing systems within teams. However, there are a number of reasons why role clarity is likely to facilitate the development of a team‟s TMS.

First, team members might be more willing to invest time and energy towards increasing their knowledge when they are clear on the behavior that are expected of them. When members are not sure about which decisions fall under their responsibility, they can refrain from making any decisions by themselves out of uncertainty of other members‟ reactions (Rizzo, House & Lirtzman, 1970). This could cause members to retain from seeking information needed to make decisions, and prevent them from developing role specific expertise. Similarly, a lack of role clarity could lead members to feel effort towards specialization might be redundant, constraining motivation to start to develop it. Thus, role clarity is likely to positively influence members‟ perceptions of themselves as experts in role-related knowledge areas.

Second, like cognitive conflict, role clarity is likely to influence the role

identification behaviors within a team. When every member knows what his or her role is, it is easier to discuss and confirm everyone‟s capabilities and responsibilities within the team. Therefore, role clarity is likely to facilitate effective role identification behaviors occurring during conflict.

Once team members start seeing themselves, as well as each other, as specialists in a certain subject areas, a TMS can start to develop (Pearsall et al., 2010). This leads to the following hypothesis:

(19)

H3: The positive effect of cognitive conflict on TMS is positively moderated by role clarity.

2.5 The influence of transactive memory systems on team performance

A team‟s TMS plays a large role is determining how efficient a team can process and apply knowledge needed to complete their task. This can be explained by the

characteristics of teams with a strongly developed TMS leading to various efficiency gains. Because knowledge is divided between team members, it reduces individual cognitive load and prevents members wasting effort on overlapping knowledge

(Hollingshead, 1998b). This division in combination with increased coordination helps teams to allocate new information swiftly to the expert in that area, to be retrieved or shared when necessary (Brandon & Hollingshead, 2004).

A strong belief of credibility between team members reduces the need to double-check information that is exchanged within the team. As a result, a well-developed TMS facilitates fast access to a large pool of accurate, task relevant information (Heavey & Simsek, 2015), which provides a greater diversity and depth of knowledge than any individual member could maintain (Lewis & Herndon, 2011).

Furthermore, a team‟s TMS also impacts the amount of information being shared between members (Hollingshead 1998a; Moreland and Argote 2004). Knowledge sharing refers to members within a team sharing task-relevant ideas, information, and suggestions (Srivastava, Bartol & Locke, 2006). A well-developed TMS allows members to anticipate which other members could use their knowledge, facilitating information-exchange, and in turn leading to higher performance (Choi, Lee & Yoo, 2010).

(20)

Considering the efficiency gains and increased knowledge sharing, it is no surprise that earlier studies have linked TMS to team performance in various contexts. TMSs have been found to be beneficial in teams in consumer good firms (Austin, 2003) and technology firms (Zhang, Hempel, Han, & Tjosvold, 2007) as well as in consulting teams (Lewis, 2004), new product development teams (Akgün, Byrne, Keskin, Lynn & Imamoglu, 2005) and teams within a laboratory setting (Lewis, 2003; Moreland &

Myaskovsky, 2000). Based on these earlier studies, the following hypothesis is proposed:

H4: TMS positively impacts team performance.

2.6 The mediating effect of transactive memory systems

Teams engaged in cognitive conflict will have to work under conditions where multiple points of view are shared that advocate different outcomes (Weingart et al., 2015). In the end, these perspectives need to be synthesized and incorporated into a single outcome to complete the task at hand (Tjosvold, 1991). A team‟s TMS provides a way for teams to combine various points of view in an effective manner through integrations. Integrations are the result of “members discovering links between members‟ knowledge and creating new knowledge that no member had previously possessed” (Lewis, Lange & Gillis, 2005, pp. 583–584). In this manner, teams engaged in cognitive conflict can attain higher performance by integrating various perspectives through their TMS.

Furthermore, cognitive conflict poses a significant risk of limiting the team‟s information processing abilities. It can increase the cognitive load, leading to decreased cognitive flexibility and creative thinking, and ultimately decreasing members‟ problem solving ability (Carnevale & Probst, 1998). This in turn can make people more prone to

(21)

narrow, black-and-white thinking (De Dreu, 2008). When the information processing ability of a team is recued, it is likely to cause the overall performance of the team to suffer (De Dreu & Weingart, 2003). A strong TMS ensures the efficient processing of the increased amount of shared information. It allows team members to divide responsibilit y for different knowledge areas, thereby lowering the cognitive burden on each individual (Brandon & Hollingshead, 2004). This provides the necessary mental resources for the team to focus on resolving the differences in opinions and to come up with a joint solution or answer.

The process of cognitive conflict also consumes time (Simons & Peterson, 2000). During conflict team members elaborate on their own ideas, search for information to support their view and weaken opposing views, and integrate various perspectives to form new ideas (Tjosvold, 1991). These processes require a time investment, which reduces the time available for the execution and implementations of ideas and solutions (Behfar et al., 2008). When there are tight time constraints, time spent on conflict could possibly be spent more effectively on accomplishing the task itself (De Wit et al., 2012). However, when teams use cognitive conflict to build their TMS, the loss of time can be won back by processing information more efficiently. This is because teams engaged in cognitive conflict have faster access to a larger amount of expertise when they are able to develop a strong TMS and thus spend less time searching for information (Lewis, 2004). A strong TMS compensates for time consumed during cognitive conflict, making the sharing and integration of different ideas worthwhile.

In conclusion, cognitive conflict is likely to benefit performance outcomes

(22)

cognitive resources and process information faster. This leads to the fifth and final hypothesis:

H5: Cognitive conflict positively effects team performance trough TMS.

3 Method

3.1 Participants, tasks & procedures

3.1.1 Participants

Data for this study was collected from undergraduate students of a large European

business school in a second year „strategic management‟ course. There were 597 students enrolled in course, divided into 117 teams of five members. These teams engaged in a nine-week business simulation game where they competed in a virtual industry against 10 or 11 other teams. The average age of the students was 19 years old, with the youngest being 17 and the oldest 30. The gender division was 32,4 % female and 67,6 % male. In total 584 students chose to participate in this research by voluntarily filling in surveys throughout the game. Response remained relatively consistent over the entire period, ranging from 94% to 96%. Teams that did not have data from at least four members were removed, leading to a final sample of 445 students within 96 teams.

3.1.2 Tasks

The teams participated in the Business Strategy Game (Thompson & Stappenbeck, 2001) where they represented firms in a virtual athletic footwear industry. They started with a well functioning firm and had to defend their position and improve their performance

(23)

during a nine-week period, representing nine years. This was done by designing and implementing a strategy for the firm. The teams could choose geographical markets to be active in and distribution channels to use as well as make decisions related to functional departments, such as R&D, marketing and production. The virtual business environment adapted according to choices made. Teams were evaluated based a number of different performance objectives, which can be found in section 3.2.1. The Business Strategy Game is a close representation of the actual business environment as it mimics the complexity and ambiguity of industries (Tompson & Dass, 2000). Student teams have to make decisions that real life business teams also have to make, under similar

circumstances, supporting the external validity of the design. Furthermore, teams faced the same starting conditions and all had the same means in the game, supporting the internal validity.

3.1.3 Procedures

Students were allowed to compose their own teams. Students without a team were randomly assigned into teams by the professors. Before the game started, students

attended a lecture about the game‟s interface and functionalities, followed by a one-week practice round. Thereafter, the teams completed the game autonomously without further guidance. Data was collected through surveys given at five time points, with questions about interpersonal dimensions of teamwork. Dutch students received the survey in Dutch, international students in English. Team performance was measured at nine time points through metrics included in the game simulation. As an incentive, students who filled in the survey were provided with personalized feedback reports with detailed

(24)

information on the individual team‟s decision-making process and performance. Insights into team functioning provided teams with increased means to reflect and improve.

3.2 Measurement of variables

An overview of all the items used to construct the variables described below can be found in appendix A.

3.2.1 Dependent variable team performance

Performance was determined based on the average score of the team‟s rating on „investor expectations‟ and „best-in-industry standard‟, relative to the performance of competing teams. The rating investor expectations was based on the team‟s ability to meet or exceed each year‟s expected performance targets for the earnings per share, return on equity, credit rating, image rating, and stock price appreciation. The best-in-industry standard rating was constructed out of a weighted average of earnings per share, return on equity, stock price appreciation, image rating and credit rating. The score was compiled

automatically by the business game software and was measured in the ninth, and last, week of the game. A scale ranging from 0 to 100 points was used, where a higher score indicated higher performance. The measure is realistic compared with outcomes that are important for real management teams, which have to balance investor expectations while, making sure the firm remains competitive in the market (Mishina, Dykes, Block, &

Pollock, 2010). Overall, the measure complies with the criteria proposed by Mathieu et al. (2008), which states that analysis of team performance should be differentiated into

multiple parts, combined using a formal algorithm and related to the function and tasks of the teams being studied.

(25)

3.2.2 Independent variable cognitive conflict

The item „Team members share and accept constructive criticisms without making it personal‟ measured the amount of cognitive conflict within a team. This statement was rated on a seven-point Likert scale ranging from “strongly disagree” (1) to “strongly agree” (7). This item taps into various aspects of cognitive conflict.

First, cognitive conflict is most beneficial for teams when the norms within the team are accepting of conflict being shared and received (Jehn, 1995). Secondly, it is necessary to distinguish cognitive conflict from relationship conflict. Even when team members critique each other on subjects related to the task, it can still be perceived and interpreted as relationship conflict (Simons & Peterson, 2000). Therefore it is crucial that the

statement specifically emphasized constructive, as opposed to hurtful, criticism, which does not lead to a personal conflict.

3.2.3 Mediating variable transactive memory system (=0,726)

TMS was measured by six items that were derived from a 15-item scale designed by Lewis (2003). It includes three dimensions with each two questions, namely

specialization, credibility and coordination. These dimensions are manifestations of the underlying latent construct. An example of an item in the specialization dimension is: „Each team member has specialized knowledge of some aspect of our project.‟ „I trusted that other members‟ knowledge about the project was credible.‟ is an item included in the credibility dimension. Finally, an example of a coordination item is: „We accomplished the task smoothly and efficiently.‟ All items have been rated on a seven-point Likert scale ranging from “strongly disagree” (1) to “strongly agree” (7).

(26)

3.2.4 Moderating variable ‘Role clarity’ (= 0,715)

Role ambiguity was measured with a five-item scale developed by Rizzo et al. (1970). Responses were given on a 7-point Likert scale which ranged from 1 = „totally not applicable‟ to 7 = „very applicable‟. To increase the reliability of the scale to an acceptable level, the second and the fifth item were dropped. An example of an item included in the final measure is: „I have clear, planned tasks within my group.‟ An average of the three remaining items was taken to create a composite score. A higher score relates to more clear perception the individual‟s role in the group.

3.3 Control variables

Four control variables were incorporated to minimize the effect of variables external to the model on the results.

3.3.1 Team member familiarity

Team member familiarity was controlled for as this might influence the team‟s TMS and conflict within the team (Jehn & Mannix, 2011; Ren & Argote, 2011). This construct was measured by taking the average of the two items of the scale developed by Jackson & Moreland (2009). Both items were rated on a seven-point Likert scale ranging from 1 = „strongly disagree‟ to 7 = „strongly agree‟.

3.3.2 Prior individual performance

Prior individual performance was included as control variable to account for the effects of individual members‟ ability to produce a high quality outcome of tasks. This has been shown to influence the performance of the entire team (Schippers, Homan &

(27)

Knippenberg, 2013). This construct was measured by self-reported GPA, with possible values ranging from 1.0 to the maximum grade of 10.0.

3.3.3 Gender diversity

Gender diversity was also controlled for as this could influence both cognitive conflict as well as team performance (Jehn, Northcraft & Neale, 1999). This variable was measured by the percentage of females on the team.

3.3.4 Age

Finally, age was also accounted for, indicated by the self-reported age of the participants at the beginning of the business simulation game.

3.4 Timing of measurements

To increase the reliability of statements about cause and effect, the variables in this study were each measured at multiple time points. Prior research within conflict literature has largely been cross-sectional, leading to an inability to predict causality (Jehn, 1995; Lovelace, Shapiro & Weingart, 2001; Olson et al., 2007). Ployhart & Vandenberg (2010) propose that a powerful way to test change is to measure the independent variable at time t, the mediator variable at time t+1 and the dependent variable at time t+2. However, these authors also emphasize that constructs should be measured at moments when they are likely to be of importance to the team. Consequently, the independent variable cognitive conflict is measured during the third (middle) survey wave (t=1). Cognitive conflict is most likely to benefit team performance at the midpoint of group interaction as cognitive conflict in earlier stages might interfere with creating an action plan and

(28)

Mannix, 2001). Mediating variable TMS was measured during wave 4 (t=2) to introduce a time lag between the variables. Following this reasoning, team performance was measured after the fifth, and final wave (t=3). The moderating variable role clarity was measured at the start of the project, wave 1. This was chosen because role identification behaviors occur during the initial stages of team development (Pearsall et al., 2010).

3.5 Measurement model

In the first part of this section the measurement model will be tested for discriminate and convergent validity of the constructs, as well as for their reliability. Furthermore the appropriateness of aggregating the data to team level will be analyzed. In the second part, the structural model is examined to determine if the hypotheses are supported.

3.5.1 Exploratory factor analysis

An exploratory factor analysis (EFA) was performed to validate the discriminant validity of the constructs and as well as the underlying dimensions of TMS as reflective measure. It is expected that the underlying factors of this construct covary because they have a common cause (Lewis, 2003). Therefore, an oblique rotation was used as this allows the underlying factors to be correlated (Pedhazur & Schmelkin, 1991).

The KMO measure of sampling adequacy was .707, which is a good value according to Hutcheson & Sofroniou (1999). Bartlett‟s test for sphericity, which is significant at p<.001, provides evidence that the correlation matrix is not an identity matrix and that the variables are sufficiently related to perform an EFA. Together with the KMO measure, this confirms the appropriateness of the data.

(29)

Table 1.

Explanatory Factor Analysis for Items in the TMS Scale (Pattern Matrix)

Rotated factor loadings

Item SPC CRD COR CON RC Total

TMS1 Each team member has specialized

knowledge of some aspect of our project.  

.95

TMS2 I know which team members have expertise

in specific areas.  

.85

TMS3 I trusted that other members‟ knowledge

about the project was credible.

.95

TMS4 I was confident relying on the information

that other team members brought to the discussion.

.96

TMS5 Our team needed to backtrack and start over

a lot.

.96

TMS6 We accomplished the task smoothly and

efficiently.

.67

Conflict1 Team members share and accept

constructive criticisms without making it personal.

.98

RC1 I have clear, planned goals and objectives for

my job.

.44 .59

RC3 I can divide my time properly. .87 RC4 I know what my responsibilities are. .87

% of variance 16.0 33.8 9.7 7.7 13.8 81.0

α .83 .92 . 57 - .72 .76

Note. SPC = specialization; CRD = credibility; COR = coordination, CON = conflict, RC = role clarity.

Factor loadings with a value <.3 have been suppressed. Extraction method: principal axis factoring. Rotation Method: promax with Kaiser Normalization. Missing values deleted listwise.

The factor loadings and cross-loadings can be found in table 1. No cross-loadings higher than .5 were observed, which is well below the generally accepted limit of .7. Therefore discriminant validity can be assumed. All factors loadings are higher than .55, the minimum factor loading required for significance at a sample size of 100 (Hair, Anderson, Tatham & Black, 1998, p.112), indicating that the items belonging to one factor are highly correlated. This gives support for convergent validity. Collectively, these factors account for 81% of total variance. Chronbach‟s alpha was used to assess reliability. The proposed threshold for this measure is .6 (Hair, Black, Babin & Anderson, 2010, p. 125). Three out of four factors score higher than this cut point, as well as for the

(30)

full scale (α= .76). This supports the expectation that the construct TMS consist of three underlying factors, namely specialization, credibility and coordination.

3.5.2 Confirmatory factor analysis

To validate the results of the EFA, a confirmatory factor analysis (CFA) was performed using Amos Graphics. First, a complete dataset was created by removing the data of 33 respondents that did not fill in all questions related to the constructs. This left a final sample of 531. This eliminated the need to estimate means and intercepts and allowed for the calculation of SRMR.

Comparing the variance-extracted estimates with the squared correlation between variables tested for discriminant validity. Role clarity was compared with the three factors that made up the latent variable TMS. Table 2 shows that squared correlation between role clarity and specialization was lower than the variance-extracted estimates of the two variables (Φ21 = .292, Φ221 = .085; SE of Φ21 = .067; p < .001). That was also the

case for role clarity and credibility (Φ21 = .214, Φ221 = .046; SE of Φ21 = .048; p < .001)

and role clarity and coordination ((Φ21 = .288, Φ221 = .082; SE of Φ21 = .061; p < .001).

All variance-extracted estimates were higher than the squared correlation between factors, as advocated by Netemeyer, Johnston & Burton (1990, p.152). Furthermore, all variance-extracted estimates were higher than the proposed minimum of .50 (Fornell & Larcker, 1981, p. 46).

(31)

Table 2.

Comparison of Measurement Models

Model Factors 2 df ∆2 RMSEA SRMR CFI TLI GFI

Null 1733.432 45 Baseline model Five-factor model: Cognitive conflict (1), role clarity (2), specialization (3), credibility (4), coordination (5) 90.307 26 .072 .0527 .962 .934 .963

Model 1 Four-factor model: Cognitive conflict & credibility (1), role clarity (2), specialization (3), coordination (4)

105.904 29 15.597 .0604 .102 .954 .929 .919

Model 2 Four-factor model: Cognitive conflict (1), role clarity &

specialization (2), credibility (3), coordination (4)

332.840 30 226.936 .1002 .135 .821 .731 .873

Model 2 Three factor model: Cognitive conflict & credibility (1), role clarity & specialization (2), coordination (4)

345.507 32 12.667 .1040 .139 .814 .739 .870

Note. **p<.001

The results of the CFA can be found in Table 3. The baseline five-factor model had a comparative fit index (CFI) of .96, the Tucker-Lewis fit index (TLI) was .93. These values are in line with the recommended minimum .9 proposed by Hu and Bentler (1999). The SRMR was .0527, falls below the cutoff value of .8 they propose for this fit index. The RMSEA was .07, indicative of a fair fit according to Browne and Cudeck, (1993). This indicates that the proposed model is a good fit with the data. Furthermore, the baseline model was compared with model with two or one factors. Table 3 indicates that the three-factor model fits the data best.

(32)

Table 3.

Results of Confirmatory Factor Analysis and Measurement Properties

Variable

Standardized

loadings (λyi) Reliability (λ2yi) Variance (Var(εi))

Variance-Extracted Estimate (AVE) Role clarity .50 Item 1 .577 .333 1.220 Item 3 .596 .355 .640 Item 4 .892 .687 .251 SPEC .75 Item 1 .696 .484 1.220 Item 2 1.009 1.018 .030 CRED .85 Item 1 .936 .876 .111 Item 2 .907 .823 .172 COOR .57 Item 1 .397 .158 .008 Item 2 .997 .994 1.797 3.5.3 Aggregation analysis

To determine if the constructs can be aggregated to team level, inter-rater reliability and inter-rater agreement were measured using the intraclass correlation coefficients ICC(1) and ICC(2). This shows the consistency of the measures, giving an indication if the there is agreement between members of the same team if the mean of all the team members‟ rating differ between teams.

First, the data was restructured using LeBreton & Senter‟s (2007) syntax. ICC(1) and ICC (2) values were calculated in SPSS. ICC(1) and ICC(2) values for cognitive conflict were .16 and .46 respectively (F=1.536, p<.2). For role clarity they were .12 and .67 (F=7.981, p<.001), and for TMS they were .10 and .62 (F=16.542, p<.001). All (?) ICC(2) values approach the benchmark value of 0.70, which is common practice (Klein et al., 2000, p. 518). However, ICC(1) values between .01 and 0.10 are considered a small to medium effect by LeBreton & Senter‟s (2007), who advise addition investigation when these values are measured.

(33)

Therefore, the rwg(j) was computed. This statistic is used to establish the degree of

similarity between the responses of members in a team, thus determining the interrater agreement. A uniform distribution was assumed for all measures (σ2E = 4.00 using a 7-point likert-scale). The average rwg(j) was .79 for cognitive conflict, .80 for role clarity

and .81 for TMS. All values are above the acceptable level of 0.70 proposed by George (1990) and James, Demaree and Wolf (1984). Combined, the values indicate that aggregation is justified.

4 Results

In this chapter the descriptive statistics are analyzed and the structural model is examined to determine if the hypotheses are supported.

4.1 Descriptive statistics

Descriptives of all the constructs are shown in table 4, together with their zero-order correlations. The highest correlation observed was .62. This is well below 0.85, the level that might lead to multicollinearity problems (Kenny, 1998). The variance inflation factors (VIFs) were of all the independent variables were calculated. The highest value found was 1.8, which was well under the limit of 3.3 proposed by Petter, Straub and Rai (2007). Together, the acceptable level of the correlations and VIF‟s provide evidence that multicollinearity is not an issue. An overview of all VIF values can be found in appendix D.

(34)

Table 4.

Descriptive Statistics and Correlations

Variables Descriptives Correlations

Mean SD 1 2 3 4 5 6 7 8 1 GPA 7.01 .359 - 2 Prior relationships 4.71 1.33 .49** - 3 Age 18.5 .697 -.06 -.079 - 4 Gender ₃₇a 37 .39 -.030 -.062 - 5 Cognitive conflict 5.69 .630 .21 .24* -.012 -.087 - 6 Role clarity 5.27 .386 .32** .17 -.11 .022 .22* - 7 TMS 5.08 .526 .16 .07 .15 -.15 .62** .35** - 8 Performance 74.9 21.6 .024 -.02 .07 -.17 .23* .091 .47** -

Note. a% percentage females in team *p<.05, **p<.01

4.2 Hypothesis test

The results of the testing of all five hypotheses are presented in table 5. Hypothesis 1 stated that cognitive conflict positively impacts team performance. To test this hypothesis, the control variables and cognitive conflict were included in the model, with team

performance as dependent variable. This is shown in model 4 of table 5, which confirms that cognitive conflict significantly predicts team performance (=.238, p<.05). Therefore, hypotheses 1 is supported, confirming that cognitive conflict positively impacts team performance.

Hypothesis 2 stated that cognitive conflict positively impacts TMS. This hypothesis was tested with the model including the control variables as well as

independent variable cognitive conflict. The dependent variable was TMS, as can be seen in model 2. Results confirm that cognitive conflict has a positive effect on TMS (=.464, p<.001). Furthermore, the R-squared statistic indicated that this model explains 47.1% of the variability of different teams‟ TMS. Interestingly enough, this model also shows that age has a significant effect on TMS development (=.135, p<.05). However, this

(35)

significant effect disappears when only the control variables are included in the model, as can be seen in model 1.

Table 5.

Multiple Regression Resultsa

TMS _Performance

Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Control variables GPA .149 .028 .032 -.001 -.043 -.042 Prior relationships -.055 -.042 -.040 -.033 -.004 .006 Age .157 .135* .119 .059 -.010 -.015 Gender -.146 .147 -.141 -.15 -.104 -.104 Main effects Cognitive conflict .464*** .458*** .238* -.052 Role clarity .389** .310* TMS .452*** .482*** Interactions Cognitive conflict * .348 Role clarity Mediator TMS R2 .077 .471 .487 .086 .223 .225 Adjusted R2 .031 .431 .441 .029 .175 .166 F value (sig?) 1.679 11.721*** 10.582*** 1.507 4.594** 3.816** ∆ R2 .016 ∆F value (sig?) 2.456 df 4, 81 6, 79 7, 78 5, 80

Note. a Standardized regression coefficients are reported, *p<.05, **p<.01, ***p<.001

To test for hypothesis 3, which stated that the positive effect of cognitive conflict on TMS is positively moderated by role clarity, cognitive conflict was multiplied with role clarity created an interaction variable. The variables were mean centered to avoid issues with multicollinearity, as recommended by Aiken and West (1991). The main effects of cognitive conflict and role clarity were also included, as can be seen in model 3. The model including the interaction effect accounted for slightly more, variance, but at a weak significance level (∆R2=

.016, p=.121). However, an examination of the interaction plot (figure 2) shows that the impact of role clarity seems to depend on whether the team

(36)

has low, average or high levels of conflict. As the level of cognitive conflict increased, the amount of role conflict had a larger effect on the strength of the team‟s TMS.

Figure 2.

Moderating Effect of Role Clarity on the Relationship Between Cognitive Conflict and TMS

Note. Low = mean minus one SD, average = mean, high = mean plus one SD.

Therefore, the variables were further analyzed using the Johnson-Neyman technique, as described by Preacher, Curran and Bauer (2006). These results, which can be found in table 6, indicate that the coefficient of the interaction is statistically significant (p.05) and positive when the level of cognitive conflict is equal or larger than the cutoff value. It follows that role clarity was positive associated with TMS when teams had an average or high level of cognitive conflict. Hence, hypothesis 3 is partially supported.

These results imply that the coefficient in each column is significant and negative when uncertainty is less than the lower bound, not statistically significant when

uncertainty has values between the lower and upper bound, and significant and positive

Low conflict Average conflict High conflict 4,2 4,4 4,6 4,8 5,0 5,2 5,4 5,6 5,8

Low Average High

T M S Role clarity Low conflict Average conflict High conflict

(37)

However, the model including the interaction effect did not explain a significant larger percentage of variance compared with model 2 (∆R2=.016, p=.121).

Table 6.

Exploring Role Clarity Interaction Coefficients a

Variables Cognitive Conflict x Role Clarity 95%- significance region

Cutoff value -.085

Percentage of teams .

Below cutoff value 47.7% Above cutoff value 52.3% Simple slopes

Low cognitive conflict Not significant, .24<p<.49 Average cognitive

conflict Significant, p<.05

High cognitive conflict Significant, p<.001

Note. a Values calculated using the Johnson-Neyman technique, based on the

approach of Preacher et al. (2006). Low = mean minus one SD, average = mean, high = mean plus one SD.

Hypothesis 4 stated that TMS positively impacts team performance. To test for this hypothesis, TMS was regressed on dependent variable team performance, while controlling for the control variables. Model 5 shows TMS has both a positive and significant effect on team performance, thus supporting hypothesis 4 (=.452, p<.001).

Hypothesis 5 stated that cognitive conflict positively effects team performance trough TMS. Testing this hypothesis requires that three conditions must hold (Baron & Kenny, 1986). First, independent variable cognitive conflict must affect mediating variable TMS. This condition is confirmed by hypothesis 2. Secondly, cognitive conflict must affect the dependent variable, team performance. This condition is met through hypothesis 1. The final condition relates to TMS as mediator affecting dependent variable team performance. Hypothesis 4 is the final confirmation that all conditions hold.

Mediation occurs when the effect of the impendent variable on the dependent is less when the mediator is included in the model. In table 5, model 6 shows that the direct

(38)

effect of cognitive conflict has been reduced to a non-significant level when TMS is included in the model (=-.052, p=.687). A visual representation of these effects can be seen in figure 3. Preacher, Rucker and Hayes (2007) argue that the indirect effect of cognitive conflict on team performance, through TMS, should also be significantly different from zero. This was tested using the bootstrapping technique. The 95%-CI bootstrap distribution of the indirect effect with a sample of 10,000 did not contain zero (5.5505, 16.7383), providing evidence for its significance (p<.05). Together, these results provide support for hypothesis 5.

Figure 3.

Standardized Regression Coefficients for the Relationship Between Cognitive Conflict & Team Performance, as Mediated by TMS.

Note. The standardized regression coefficient between conflict and performance, controlling for TMS, is between parentheses. *p<.05, **p<.01, ***p<.001.

5 Discussion

This thesis examined the effect of cognitive conflict on TMS development within teams. Role clarity was explored as moderator of this relationship. Furthermore, both cognitive conflict and TMS were also hypothesized to influence team performance. The findings

(39)

contribute to the understanding of the complex relationship between cognitive conflict, information processing and team performance. This in turn has implications for research and practice.

5.1 Implications for research

First, the results clearly demonstrate that cognitive conflict can facilitate the development of team processes related to effective information processing. This provides empirical evidence that cognitive conflict is not inherently detrimental for information processing. These findings go against current conflict literature, which claims that even small

amounts of cognitive conflict can significantly increase cognitive load, reducing creative thinking processes (De Dreu & Weingart, 2003; De Wit et al., 2012). The results of this study suggest the need for a more nuanced view. This opens the door for future research examining the various contexts that determine how cognitive conflict affects information processing.

In contrast, these findings are in line with previous research on TMS

development, which suggests that teams benefit from the team engaging in processes that allow members to critically evaluate each other‟s performance. This research has focused on processes where team members observe how other members perform (Moreland, Argote, & Krishnan, 1996; Prichard & Ashleigh, 2007) or received feedback about other members‟ performance (Moreland & Myaskovsky, 2000). While communication has also been identified as an essential antecedent of TMS (Ren & Argote, 2011), the positive effect of cognitive conflict on TMS emphasizes the importance of verbal argumentation and inquisition regarding the knowledge of team members.

(40)

Secondly, results show that cognitive conflict affects teams performance through TMS. In fact, when TMS was controlled for, the direct effect of cognitive conflict on team performance became insignificant. This suggests that information processing plays a crucial role in explaining why cognitive conflict can be beneficial for teams.

Furthermore, results indicate that information processing can be a useful lens through which the positive effects of cognitive conflict, such as understanding and increased decision quality (Olson et al., 2007), can be studied.

Third, findings indicate that role clarity complements the effect of cognitive conflict on TMS development in teams with an average or high level of cognitive

conflict. This was contrary to the expectation that role clarity would act as a moderator in all instances, including in teams with low levels of cognitive conflict. The assumption was made that role clarity would facilitate role identification behaviors because team members already have a good idea what their own role is. However, results suggest that teams that engage in less cognitive conflict also spend less time discussing the roles of different members in the team, irrespective of the amount of clarity the members have about their role. It would be interesting for future research to investigate how teams with lower levels of cognitive conflict divide their workload and coordinate activities. This could help explain why teams with less cognitive conflict have a less developed transitive memory systems.

5.2 Managerial implications

The findings of this thesis have various implications for the practical organization of teams. First of all, managers must be aware that when cognitive conflict arises in a team, it is not automatically a bad thing. However, while other studies have gone so far to

(41)

suggest in should be stimulated (Chen & Leung, 2010; Jehn & Mannix, 2011), this is likely to do more harm than good. There is a significant risk of cognitive conflict turning into relationship conflict (De Dreu & Weingart, 2003; Simons & Peterson, 2000), which is both bad for performance (Jehn, 1995) as well as for TMS development (Chen & Leung, 2010).

On the other hand, when a team experiences high levels of cognitive conflict this does not necessarily have to be reduced. As long as the conflict does not involve

relationship conflict, it might be worth focusing on how the team can use it to increase their performance. Findings from this study suggest teams can learn from conflict. Specifically, cognitive conflict can help members of a team figure out who holds knowledge in which domain, and assess the credibility of that knowledge. The resulting TMS can in turn help the team to process information more efficiently, and increase performance.

Managers must also be aware that a strong TMS plays an important role in reaping the benefits of cognitive conflict. Cognitive conflict alone is not likely to lead to increased performance. Certain teams may be more prone to cognitive conflict, for instance when members differ in education, experience, and expertise (Jehn et al., 1999) or when there is friendship among the team members (Shah & Jehn, 1993). When managers foresee that a team is likely to engage in moderate to high levels of cognitive conflict, stimulating a clear division of roles within the team can help the team to use this type of conflict to build a strong TMS, and ultimately achieve higher performance.

(42)

5.3 Limitations regarding research design

This study also has a number of limitations, which require future research to be resolved. First, this study has operationalized cognitive conflict as a single item construct, which limits the opportunity for extensive reliability checks. Furthermore, while the item was phrased to capture the cognitive dimension of conflict, as opposed to the relationship dimension, measuring the two dimensions separately would offer a higher level of construct validity. Measuring cognitive and relationship conflict separately would also offer the opportunity to see if the advantage of cognitive conflict related to information processing still holds when there are high levels of relationship conflict.

Second, future research should focus on extending the analysis of role clarity and its relationship with cognitive conflict and TMS. While findings of this study suggest that role clarity is likely to play a more pronounced role in teams engaged in high levels of conflict, additional research is needed to confirm and clarify this effect. Without these results, caution has to be exercised with regards to drawing conclusions about how to stimulate TMS development to reap the benefits of cognitive conflict. Looking at other moderators, such as group training (Liang, Moreland & Argote, 1995; Moreland et al, 1996) or the level of goal congruency between team members (Zhang et al., 2007) could also further increase understanding of the relationship between cognitive conflict and TMS.

Finally, there are a few limitations that are the result of studying teams in a simulation setting, as opposed to an organizational context. The participants in this study were all higher-educated individuals with an average age of 19 years. The task they completed was a cognitive, non-routine task. Therefore, it is hard to predict if these

(43)

results are also generalizable to teams with different demographics or teams performing manual labor or routine tasks. There is evidence that cognitive conflict is more likely to have a negative effect on team performance when tasks are routine (Jehn, 1995).

Therefore future research should ideally also look at the relationship between cognitive conflict, TMS and performance in teams performing routine tasks. Additionally, team members in this study had different performance incentives compared to those of most firms. The results of the business game counted for 20% of the grade for the course and no monetary incentives were involved. This could lead to a decrease in the external validity as organizations provide different incentives. This could also lead to unrealistic goal congruity in the teams. Team members most likely all had more or less the same goal, to do as well as possible in the game. In real life team members might have different agenda‟s such as steering the firm in a direction that facilitates their personal ambitions.

6 Conclusion

This study examined the role of TMSs within the relationship between cognitive conflict and team performance. Cognitive conflict was argued to benefit performance by

decreasing information-processing failures within teams. It was proposed that the

performance benefits could be achieved through a strong TMS, which provides a way for the team to integrate perspectives, focus cognitive resources and process information faster. In this manner, inefficiencies caused by increased information exchange could be overcome. Role conflict was thought to facilitate the development of TMS by stimulating role identification behaviors within a team and influencing members‟ view of themselves as specialists. The theorized model was analyzed using a three-wave lagged design with

(44)

data from 445 individuals within 96 teams. Results show that TMS mediates the

relationship between cognitive conflict and team performance. Furthermore, role conflict was found to positively moderate the effect of cognitive conflict on TMS in teams with an average or high level of cognitive conflict. The main takeaway of this study is that TMS plays an important role in teams that are able to use cognitive conflict to achieve higher performance, and that defining clear roles can increase the chance of cognitive conflict having this beneficial effect. These findings help explain how teams are able to turn cognitive conflict into an advantage and how these positive effects can be enhanced. However, future research is needed to verify these results by using a full-scale measure of cognitive conflict.

Reaping the benefits of intrateam conflict : the role of transactive memory systems