The influence of communication frequency on TMS

(1)

23 June 2017 - Final

Extensive research on transactive memory system (TMS) and its influence on team outcomes demonstrates a positive relation between TMS and multiple team outcomes such as performance, creativity, and innovation. Communication frequency is also linked to positive team outcomes; however, the findings concerning the relationship between TMS and communication frequency are inconclusive. This thesis attempts to clarify the possible interaction effect between communication frequency and face-to-face communication on TMS. As such, this study adds to the current debate on the importance of face-to-face communication. The interaction effect is investigated through a hierarchical regression analysis, based on the data of 96 teams participating in a business strategy game. The results indicate that communication frequency and TMS have a significantly positive correlation in the first week of the game. This positive correlation means the higher the communication frequency, the higher the level of TMS. Nevertheless, the results of the fourth week of the game reveal no significant relationship between TMS and communication frequency. Furthermore, in both weeks, no evidence is found to support the moderating role of face-to-face communication. The post-hoc analyses investigate the possible mediating role of trust. When comparing the importance of trust in week 1 and week 4, it appears that trust may play a crucial role in the relationship between communication frequency and TMS. This effect is found to be non-significant. Moreover, this thesis sets a clear direction for future research by indicating that face-to-face communication may have lost its value in student interaction due to the high-quality alternative ways of communicating.

The Influence of Communication Frequency on TMS

Student: Lotte Vingerhoets

Student number: 11356693

Supervisor: Pepijn van Neerijnen MSc. in Business Administration – Strategy Track

(2)

Statement of originality

This document is written by student Lotte Vingerhoets 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)

1 Introduction

Teamwork makes the dream work is a common expression; however, it is not always clear what makes teams and individuals within teams achieve more and perform better. Discerning what makes teams successful could potentially change the way teamwork is structured. To start at the beginning, performance is defined as “How well a person, machine, etc. does a piece of work or an activity” (Cambridge Dictionary, 2017). Starting from this explanation, team performance can be described as how well the team delivers the work it is assigned to. Previous research has pointed out team performance can be investigated at the following three sub-levels: the organizational level, the team behaviors and outcomes, and the performance based on roles (Mathieu, Maynard, Rapp, & Gilson, 2008). The first dimension can be described as the alignment between characteristics and performance outcomes at the organizational level. In this case, the results at the level of the organization, such as finical ratios, are considered to measure team performance (Barrick, Bradley, Kristof-Brown, & Colbert, 2007). Secondly, at the team level, the behavior and outcome performance can be examined. The behavior performance aspect of the team can be described as all the action contributing to the achievement of the team goals. The results of these actions are the team performance outcomes (Mathieu et al., 2008). Finally, at the individual level, performance can be measured based on the role assigned to the individual. The question here is how well the individual performs, compared to the expectations of his or her role (Welbourne, Johnson, & Erez, 1998). All these aspects contribute to the final outcome of the individual’s effort combined in the team performance.

Nonetheless, the above-mentioned dimensions are not the only key aspects of team performance as multiple external factors also influence the outcome. Throughout the years, numerous studies have examined teamwork and factors that drive the team performance. Communication and, in particular, the frequency of communication are often regarded as

(5)

examples of these factors (Barrick et al., 2007). As pervious research has stated the importance of information sharing to increase team performance, communication is linked to team performance (Janhonen, Johanson, & Gurteen, 2012). New knowledge sharing benefits the team performance since more knowledge is made available. Nevertheless, also communication about already known knowledge can enhance team outcomes because openness aids teammates to establish higher level of trusts and a strong relationship (Mesmer-Magnus & DeChurch, 2012).

Therefore, it is likely that the more teammates communicate, the better teams perform. Yet, previous research has revealed contradictory results about the extent of the effect communication frequency has on team performance (Hollingshead, 1998; Jackson & Moreland, 2009). Furthermore, maintaining relationship with external sources of knowledge, such as other teams, can be time consuming. This time can be better invested in internal communication as this is beneficial for the team performance (Janhonen et al., 2012). However, even within teams, creating strong relationships has bilateral effects. On the one hand it, increases the commitment to the team through alignment of goals; on the other hand, it weakens the impulse to search outside the team for new knowledge (Mom, van Neerijnen, Reinmoeller, & Verwaal, 2015). In the end, this will negatively influence the transactive memory system (TMS) and, consequently, the team performance. Therefore, it is essential to investigate the relation between communication and TMS in more detail.

As previously mentioned, communication serves as a medium to transfer skills and knowledge from one person to another. When this process occurs repeatedly, a routine of sharing information is established. According to Lewis (2003), sharing of information leads to development of a TMS. To elucidate, TMS is the way two or more individuals store, retrieve, and communicate information. Hence, to be able to manage TMS, the individual’s memory must consist of two aspects. First, the individual memory that represents the knowledge one

(6)

has. Second, the transactive memory, involves the knowledge of the other team members. Therefore, TMS allows the individual to access the information of others when needed (Wegner, Giuliano, & Hertel, 1985a). Hence, it is all about “who knows what” (Lee, Bachrach, & Lewis, 2014a).

Furthermore, TMS isa key research topic because sharing information among team

members could result in realizing full potential based on the collective team knowledge (Lewis, 2003). There are several studies on the influence of TMS regarding various team processes and outcomes. These studies have indicated a positive influence of a well-developed TMS on inter alia performance (Austin, 2003; Jackson & Moreland, 2009; Lee et al., 2014a; Rau, 2005), creativity (Ren & Argote, 2011), decision-making quality (Ren, Carley, & Argote, 2006), and innovation (Buyl, Boone, Hendriks, & Matthyssens, 2011). Because the positive effects on team outcomes and team performance of TMS is clear, this is not the subject of this thesis. However, the factors that influence TMS require more investigation.

As addressed, communication frequency is related to team performance. Yet, the results are not aligned. Moreover, previous studies have not taken TMS into account. In fact, the frequency of communication could possibly influence TMS, as sharing information is key aspect of TMS (Lewis, 2003). This sharing can occur through various communication media at different moments in time. Furthermore, previous research has concluded that the frequency of the communication is a significant indicator of the strength of the developed TMS. Here, a distinction is being made between face-to-face communication and non-face-to-face communication. The results concerning this distinction are, however, inconclusive. Especially the importance of face-to-face communication is questioned. More specifically, the significance of this type of communication after the team members have become familiar with each other’s expertise is unclear (Jackson & Moreland, 2009; Lewis, 2004). Therefore, it is

(7)

relevant to compare the influence of face-to-face communication and non-face-to-face communication on TMS at multiple moments in time.

Non-face-to-face communication is especially significant due to the growing number of virtual teams. In these virtual teams, team members face difficulties in sharing experiences and knowledge, which affects the development of TMS. It also takes more time for these teams to create a shared language to effectively communicate (Ren & Argote, 2011).

Considering the above-mentioned information, this thesis aims to reach a comprehensive conclusion about the moderating role of face-to-face communication in the relationship between communication frequency and TMS. This thesis is structured as follows. Chapter 2 provides a literature review, presenting the theoretical framework. The most important literature and theories are used to explain and highlight the key concepts. As a result, the research gap and hypothesis become clear. Finally, the chapter ends with the conceptual model. Subsequently, chapter 3 describes the details regarding the data collection. This contains a brief introduction to the business strategy game. Chapter 3 also contains the basic descriptions of the data collected such as means, standard deviation, and correlation. The results of the data analyses are presented in more detail in chapter 4. After this, these results are interpreted in terms of academic relevance and possible managerial implication in chapter 5. This chapter also indicates a number of limitations to the current study. Finally, chapter 6 provides the conclusion. In this chapter, the study is briefly summarized, and a few suggestions for further research are presented.

(8)

2 Theoretical framework

After a brief introduction, this chapter aims to provide a comprehensive explanation of the following main concepts: TMS, communication frequency, and face-to-face communication. This chapter also offers an overview of the state of the art concerning these concepts and demonstrates which information is available and what is still missing. First, the depended variable TMS is described. Next, the independent variable communication frequency is defined. Finally, it is clarified why a distinction is made between face-to-face communication and non-face-to-face communication. After each paragraph, the related hypotheses are presented, and these hypotheses are combined in the conceptual model at the end of the chapter.

2.1 Transactive Memory System (TMS) as the dependent variable

Although TMS is not a new phenomenon, it becomes more and more valuable in today’s business world because of the, for example, rapidly changing business environment. On a global scale, CEOs are confronted with an environment that changes in an expeditious space (Silberberg & Lämmel, 2014). Since the business environment is described as forces beyond the management’s direct control (Ward, Duray, Keong Leong, & Sum, 1995), it makes sense the CEOs occasionally doubt their ability to deal with this growing complexity (Silberberg & Lämmel, 2014). In this turbulent environment, internal knowledge and the sharing processes of this knowledge within an organization is often a crucial element to create a sustainable competitive advantage. Yet, organizations do frequently not have sufficient time to rethink their entire process, as they need to be able to respond quickly to changes or their competitors will. Therefore, it is crucial to create the most efficient processes as possible (Maula, Keil, & Zahra, 2012). Having an efficient way of sharing knowledge can fasten this process of responding to the changing environment.

(9)

knowledge (Suddaby & Greenwood, 2001). The codification of knowledge divides knowledge into clear steps that make up the routine. As a result, this knowledge is more easily stored, transferred, and reused. A case in point is a car manufacturer. In this example the processes conducted by the most efficient business unit will be codification to copy these processes to the other business unites. However, it is unlikely this will lead to the creation of new knowledge and, therefore, may not help organizations to deal with the changing, complex business environment. Tacit knowledge, on the other hand, is less tangible. Tacit knowledge can be described as the knowledge embedded within the employees’ minds and interactions

(Osterloh & Frey, 2000). It represents the knowledge that is commonly known and used, but

cannot be made explicit and is, consequently, more difficult to transfer. Furthermore, tacit

knowledge can be valuable when confronted with a challenge. One advantage is it creates a

“natural protection” against competition because it is harder to copy (Teece, 2007). Previous research has also highlighted that individuals accomplish more when they also have tacit knowledge about the company and the business processes compared to when they only use explicit knowledge. Tacit knowledge can be regarded as explanation of “skills”, and it is learned by experience (Chukwujioke, Owutuamor, & Oriarewo, 2013). Investing in these skills is imperative for organizations (Peteraf, 2014)

Through these sharing and learning processes, dynamic capabilities are established. According to Helfat et al. (2009, p. 4), “A dynamic capability is the capacity of an organization to purposefully create, extend, or modify its resource base.” For this reason, it can be said these capabilities will help the organization to cope with changing business environments and to create a sustainable advantage in the long run (Teece, 2007). This explains why an existing group has an advantage over a newly formed group since tacit knowledge is knowledge that is commonly known and used; however, this knowledge cannot be made explicit and is, therefore, more difficult to transfer.

(10)

A way to transfer and store knowledge at the level of the team is TMS (Grant, 1996). Moreover, TMS is defined as a construct that is built from two other components – structural and individual. The structural component explains how individual memories are linked to the collective knowledge network. Additionally, the individual transactive process consists of three additional components– encoding, storing, and retrieving information of the group memory. As mentioned in the introduction, the individual memory consists of two factors that are able to manage transactive processes, which are the individual memory and the transactive memory (Ren & Argote, 2011; Wegner et al., 1985a).

Furthermore, as a concept, TMS can be divided into the following dimensions: specialization, credibility, and coordination. This distinction is made because TMS consists of the knowledge of individuals (specialization), the degree to which they belief the knowledge of others is reliable (credibility), and how effectively the knowledge is shared (coordination) (Lewis, 2004). To completely understand TMS and how these dimensions interact, it necessary to examine each dimension individually. First, specialization forms the base of the TMS, as it provides the team with the information and knowledge needed to accomplish the team goals. However, individual team members also benefit from TMS as it offers them the necessary space to specialize certain valuable skills. Therefore, specialization and TMS are closely related and cannot be separated. Second, credibility could be described as cognition-based trust or the trust level trust team members have in each other’s knowledge (Kanawattanachai & Yoo, 2007; Ren & Argote, 2011). Finally, coordination, for a team to successfully exploit the information available within the team, coordination is essential. This coordination ensures the right person is selected to perform the task and to reach the goal (Seong, Kristof-Brown, Park, Hong, & Shin, 2015). After the introduction of the individual dimensions, it is clear they all contribute to same cause: TMS. They work together through combining the individual’s expertise, trusting this expertise, and facilitating the process of

(11)

working together (Lewis, 2004)

As TMS has been explained, the next step is to create an overview of the group outcomes to which TMS is positively linked. In general, TMS is positively correlated with group performances. This correlation stems from the fact that team members get to know each other better as TMS develops. As such, the team members form a mental model of each other’s skills, which allows them to allocate tasks accordingly. This further improves the way of coordination, which positively influences the performance (Austin, 2003; Kozlowski & Ilgen, 2006; Seong et al., 2015, 2015; Zhang, Hempel, Han, & Tjosvold, 2007). This improved coordination is especially exhibited in larger teams with a more complex distribution of knowledge. When a large team can benefit from a developed TMS, the efficiency of the team is greater compared to a team with a lower level of TMS. In smaller teams, the clarified coordination increases the quality of decisions (Faraj & Sproull, 2000; Ren et al., 2006).

Not only does performance directly benefit from the sharing of information and skills, but also multiple other team outcomes; for example, creativity and innovation take advantage of an increased TMS. These outcomes are partly responsible for the team performance. Additionally, TMS directly influences the creativity and innovation of teams as it serves as a mediator between individual and the functional diversity within the team. This entails that it provides team members with the language to understand each other knowledge better, which also improves the collaboration. This collaboration leads to more creative and innovative outcomes (Buyl et al., 2011; Gino, Argote, Miron-Spektor, & Todorova, 2010; Seong et al., 2015).

Finally, TMS also positively influence the team members on a personal level. As proven, TMS increases the reported level of job satisfaction among employees. This relation is explained by the fact that team members feel they can rely more on each other due to their

(12)

better understanding and higher knowledge of each other’s capabilities (Michinov, Olivier-Chiron, Rusch, & Olivier-Chiron, 2008).

2.2 The frequency of communication among teammates as independent variable

As previously mentioned, TMS functions as a key predictor for multiple team outcomes. Nevertheless, clarification is needed to elucidate which factors influence TMS. Therefore, this is the main focus of this thesis. More specifically, this thesis explores how the process of TMS is influenced by the frequency of communication, as this has been proven to play a crucial role in sharing information. Sharing information is one of the three aspects that underlie TMS (Jackson & Moreland, 2009; Wegner et al., 1985a). Therefore functional communication, communication related to the tasks of the team, is positively correlated to TMS (Lewis, 2004).

Although we all communicate, when asked to offer a comprehensive definition of communication, this common, daily process suddenly becomes complex. A broad definition could describe communication as “the practice of producing and negotiating meanings, a practice which always takes place under specific social, cultural and political conditions” (Schirato & Yell 1997 in Burns, O’Connor, & Stocklmayer, 2003, p. 186). Yet, this definition is not specific enough; in Katriel and Philipsen’s words (1981, p. 301), “communications refers to close, supportive flexible speech, which functions as the ‘work’ necessary to self-definition and interpersonal bonding.” Furthermore, in the business environment, communications is perceived as a tool to interact and share knowledge (Kuan Yew Wong & Elaine Aspinwall, 2004). Communication frequency is then easily explained as the frequency with which this process occurs.

In addition, communication is essential to organizations because it contributes to the organization’s identity and personality. Together with other capabilities, such as collective

(13)

skills and investments in the way they compensate and train staff, it defines what the organization’s essence is, what it does, and what it stands for (Ulrich & Smallwood, 2004).

As explained before, interactions or communication between team members is imperative for TMS development because teammates need to create a mental model of how knowledge, expertise, and skills are divided within the team (Hollingshead, 1998; Lewis, 2004). Therefore, communication and individual knowledge are key elements for the development of TMS. It has been argued that communication frequency has a positive influence on TMS since it provides the team members with a medium (communication) and the opportunities (frequency) to share information among them (Jackson & Moreland, 2009; Kuan Yew Wong & Elaine Aspinwall, 2004; Wegner et al., 1985a).

Especially in newly formed teams, it is essential to learn about the expertise and experience of other team members. This knowledge could be relevant for future projects. It provides the team with the necessary information to create and maintain an up-to-date mental model of the information available within the team. This model will form the basis for each strategic plan on how to handle new information or challenges. Related to their unique expertise and experience, certain responsibilities will be assigned to each team member. Thus, communication is crucial in the process of coordination of tasks so that the team members know who is responsible to remember which information (Hollingshead, 1998; Lewis, 2004). Moreover, a team with a longer history of working together is expected to have a stronger TMS (Ren & Argote, 2011). This can be explained by the fact that the more individuals communicate and spend time working together, the more familiar they become with each other’s knowledge. Previous research has ascertained that familiarity with the expertise of other group members can be linked to a strong TMS and increased group performance (Moreland, 1999). In addition, TMS is stimulated by the greater access to and familiarity of knowledge of individual team members because of the greater interpersonal

(14)

communication. These interpersonal relationships strengthen team members in developing a much more refined understanding of the group members’ knowledge. This improves the efficient access to information of the individual team members and, thus, the efficiency of TMS (Baldwin, Bedell, & Johnson, 1997). This is supported by social capital theory. The main idea of this theory is that potential and actual resources can be found through ties within a social network (Nahapiet & Ghoshal, 1998). The stronger, more frequent and more reciprocal the ties, the larger the “closure” of the social network (Burt, 2009). Moreover, the stronger these ties developed by interpersonal communication, the better they are suited to transfer tacit knowledge (Peltokorpi, 2004).

To conclude, repeated interactions between teammates increase the cognitive capabilities, which results in higher performance because this allows the team to react better and faster to changes (Peteraf, 2014). In this process, TMS is regarded as an intermediator. This positive relationship between TMS and performance is already extensively investigated (Ren & Argote, 2011). Furthermore, as it is assumed communication frequency is related to TMS, the first hypothesis is formulated as follows:

H1a: The frequency of communication is positively related to the development of TMS in week 1.

H1b: The frequency of communication is positively related to the development of TMS in week 4.

2.3 The frequency of face-to-face interaction as moderate variable

Since the start of humanity, individuals have been communicating. First with gestures and simple sounds, which has evolved into a complex lingual system that distinguishes us from any other mammal. Our languages and ways of communicating is what makes us human and allows us to develop further. Face-to-face communication forms the base of this. It is the most basic way of communication between individuals and can be defined as communication

(15)

directly from person to person without mediators such as technological devices. This makes face-to-face communication bound to time and place, as it requires people to actually meet, offline, to discuss their thoughts (Crowley & Mitchell, 1994; Sternberg, 2012).

However, nowadays, individuals also have the possibility to communicate online. Still within the possibilities of communicating, face-to-face communication remains the richest medium for interaction, as no other medium offers so many nonverbal, communicational, and paralinguistic cues. Moreover, it is unique in providing the possibility for immediate feedback (Daft & Lengel, 1986; Peltokorpi, 2004).

Nevertheless, contemporary teams are not particularly bound by face-to-face communication due to numerous others possibilities available (Baltes, Dickson, Sherman, Bauer, & LaGanke, 2002). The fact that teams may make use other communications modes highly affects the working environment (Colbert, Yee, & George, 2016). Yet, these electronic media are less adequate to transfer knowledge due to the limited communicational cues (Gajendran & Harrison, 2007; Powell, Piccoli, & Ives, 2004).

Moreover, most previous studies have demonstrated that most individuals and teams prefer face-to-face communication to non-face-to-face written communication for sharing information. Multiple researchers have agreed with the fact that face-to-face communication, which has the opportunity for instant feedback and additional questions, makes it most suitable to deal with complex, rapidly changing information (Baron & Kenny, 1986 in; Dabbish & Kraut, 2006). It also provides room for checking and re-checking whether everyone understood the new information (Louhiala-Salminen & Kankaanranta, 2011). Furthermore, it ensures all individuals taking part in the interaction are immediately up-to-date (Baron & Kenny, 1986 in; Dabbish & Kraut, 2006). This makes face-to-face meetings and personal interactions highly valued among team members. Regular face-to-face contact is, moreover, essential during the first stages of creating relationships. In addition, it appears

(16)

team members favor face-to-face communication to, for example, email. The use of and volume of email is related to the number of face-to-face interactions (Dabbish & Kraut, 2006). Therefore, one may expect that this interaction effect between face-to-face communication and communication could influence TMS.

Furthermore, a study conducted by Ulijn, Lincke, and Karakaya (2001 in; Louhiala-Salminen & Kankaanranta, 2011) has provided prove that, although it is possible, creating true involvement and feelings of empathy is much more difficult via email compared to face-to-face interaction. Przybylski and Weinstein(2013) have gone a step further and have investigated and demonstrated that mobile phones can interfere with the establishment of human relationships. To illustrate, in the situations where mobile phones were present, participants indicated lower levels of closeness, trust, empathy, and understanding. This indicates TMS may experience a similar negative impact when teams only communicate through such technological devices.

Additionally, the difficulties to create meaningful relations through technologically mediated communication may also negatively affect team performance outcomes. The following two causes can be ascribed to this phenomenon: the lower degree of simultaneity and the less presence of nonverbal and para-verbal cues. More specifically, the performance outcomes that are negatively influenced by computer-mediated communication are the effectiveness of the team, the time to complete tasks or make decisions, and the level of satisfactions of team members (Baltes et al., 2002; Bordia, 1997).

Regarding the topic of this thesis, it is expected that the frequency of communication and face-to-face communication exhibit an interaction effect. The reason for this expectation is that previous studies have revealed contradicting effects of importance of face-to-face communication once TMS is established. Previous studies have demonstrated that the amount of face-to-face communication between team members is related to the development of TMS

(17)

in a sense that more face-to-face communication results in stronger TMS (Lewis 2004). Furthermore, Lewis (2004) has argued that also in mature teams, face-to-face communication is required to exploit TMS-related benefits. Nevertheless, a different opinion is shared by Jackson and Moreland (2009) who have claimed that when TMS is developed, the way of communication is no longer essential. Hence, this may be explained by the strength of the ties established in the initial stage (Gajendran & Harrison, 2007), as individuals have a clear mental model of how the knowledge is distributed and they experience a strong sense of reciprocity. In further stages, the way teammates communicate may become less important compared to the frequency of the communication (Jackson & Moreland, 2009; Peltokorpi, 2004). For example, electronic communication has certain benefits, such as speed, time, and costs, and it allows individuals to be part of several TMSs at once (Baltes et al., 2002). For this reason, it is relevant to explore how the frequency of face-to-face communication interacts with the frequency of communication. As such, the following hypotheses are established:

H2: Face-to-face communication moderates the relationship between communication frequency and TMS

H3: The moderating effect of Face-to-face on the relationship between communication frequency and TMS is greater in week 1 compared to week 4

Figure 1 Conceptual model.

Frequency of

communication

TMS

Face-to-face

communication

(18)

3 Methodology

This chapter elucidates the procedures of this research. The main purpose is to explain the steps taken during data collection and analysis to make it easy to reproduce the research. First, this chapter describes the sample through presenting a comprehensive overview of the basic demographics of the participants. In addition, the first paragraph explains why the decision was made to exclude some of the data. Secondly, the role of the business strategy game (BSG) is clarified. Finally, the chapter ends with a description of the measurements that represent the concepts explained in the literature review.

3.1 Sample

A survey was conducted during a bachelor course in strategic management at a large business school in the Netherlands, leading to a sample of 566 business school students. Before participating in the survey, all students signed a file to grant permission to use the data from the survey and the results of the course for research purposes. Because the course was designed around the BSG, which is explained in the next paragraph, the students formed teams of five. In this process, 65% of these teams were chosen by the students and the remaining 35% where randomly assigned by the professor. The first survey was conducted in the first week of the BSG. After this, the other five surveys were conducted every two weeks throughout the BSG to measure differences between the teams.

In the final sample used for the analyses, the input of 81 participants was removed because of incomplete data. A reason for this could be not completing the questionnaire each week. In addition, 20 teams were excluded from the final sample because fewer than 4 members completed the survey. This decision was made because when only three or fewer members filled out the survey, it cannot be guaranteed the data are representative for the team as a whole. Therefore, the final dataset contained the answers of 480 participates divided over 96 teams. The students were aged between 18 and 27, with an average of 18.5 years and

(19)

standard deviation of almost 1 year. The final sample consisted of more men (68%) than women (32%).

3.2 Data collection

The Business Strategy Game (BSG)

The data used in this research was collected by the means of a survey, which was repeated periodically during the execution of the BSG at nine moments. The BSG is a business simulation game designed to encourage the participants to act as a top management team. This created a unique learning environment. Furthermore, all teams started from an identical situation. The starting situation was being in charge of an athletic footwear manufacturer that is doing well in terms of both revenues and profit; however, the organizational environment was changing. These changes were due to the competitive environment of the BSG. Each team was directly competing with 10 other teams, and they set the rules of the game together. Based on their decisions, the environment could become more competitive or less competitive. Since this game was designed to mimic real-world situations, information was not always complete and decision did not have to be 100% rational. During nine rounds, participants had to make several strategic decisions concerning, for instance, the geographical market, the sales channel, the marketing, and human resources. After each round, five components, which were earning per share, return on equity, stock price, credit rating, and image rating, were presented as feedback for the teams. As they all started from the same situation and were operating in the same fictional industry, performance differences between team could only stem from the decisions made by the teams. Thereby, it allowed us to isolate the effects of different group formations, communication methods, and frequency on the performance of teams.

(20)

3.3 Measurements

To measure TMS, the validated 15 item 5-point scale of Lewis (2003) was used. The correspondents had to indicate how much they agreed or disagreed with the presented statement. The answers possibilities ranged from 1 = strongly disagree to 7 = strongly agree (Lee, Bachrach, & Lewis, 2014b). In the survey, four questions were counter-indicative. Therefore, these items were recoded before moving forward with the reliability tests

To test if a scale was consistent at measuring the variable it represented, a reliability test was performed. The Cronbach’s alpha had to be higher than the 0.7 for the scale to pass the test. This test was conducted for the TMS during week 1 and 4. As demonstrated in Table 1, the Cronbach’s alpha of TMS in week 1 was just below the threshold. Because four items expressed to have a low total correlation < 0.30, these items were deleted. This increased the Cronbach’s alpha from 0.696 to 0.819. In contrast, the TMS measured in week 4 did not indicate any problems. Therefore, all items could remain. Furthermore, the interclass correlation coefficient of the items was significant (p <0.001). This means that individual respondents expressed a significant consistency in their answers on the different questions.

Table 1: Results reliability test TMS

Variable Cronbach’s Alpha Adjusted Cronbach’s Alpha

TMS 1 0.691 0.817

TMS 2 0.733

After the reliability test, an exploratory factor analysis was conducted. This analysis went beyond the standard factor analysis where the observed variables were linked to the underlying latent variable based on patterns in the responses. This analysis is often regarded as a method of data reduction. Exploratory factor analysis has as main goal to identify underlying factor structures. In the current research, three groups of items could be distinguished as these items each have a high factor loading on only one factor. This analysis

(21)

supported the expectations based on the literature review and Lewis’ design (2003). Based on the theory, these three groups could be classified as the dimensions of TMS: specialization, credibility, and coordination.

The final step before starting the regression analysis was creating new variables based on the existing variables to test the hypotheses. The hypotheses were tested based on the means. In the case of TMS, the mean was calculated based on the selected items.

Communication frequency was measured through the following question: “How often do the members of your group communicate about the project?” Finally, to examine the frequency of face-to-face communication, the question “How much of your group’s communication about the project is face-to-face?” was asked. For the first question, the answer possibilities were “never”, “once a week”, “twice a week”, “three to four times a week”, “daily”, or “multiple times a day.” For the second question, a range was provided from 1 = never to 7 = always.

Since the goal of this study was to clarify the relation between TMS, communication frequency, and face-to-face communication, a few control variables were selected, namely closeness, intragroup trust, and behavioral integration. Based on the literature review, these variables were expected to influence this relationship. Therefore, they were kept constant to strengthen the internal validity of the results.

The first one, closeness, was measured by the question: “How close do you feel to your contacts in your working relation?” For this question, the answer possibilities were between 1 = distant/arm’s length and 5 = very close (Baer, 2012; Moran, 2005).

Second, trust was investigated through Simons and Peterson’s scale (2000). An example statement of this scale was “We expect every team member to show absolute integrity”. The answers ranched from 1 = never to 7 = always. The Cronbach’s alpha for this scale was α = 0.883.

(22)

Finally, behavioral integration was represented by the scale of Mooney, Holohan, and Amason (2007). One of the statements of this scale was “Team members help each other solve problems.” The answer scale for this was 1 = never to 7 = always. The Cronbach’s alpha for this scale was α = 0.845

To ensure that there are no problems regarding discriminant validity, a second explanatory factor analysis (EFA) was performed. This analysis is similar to the EFA conducted for TMS but this time, the three control variables were put into the same analysis to observe whether the questions only loaded high in one of the three constructs. In general, a factor loading above 0.4 or below -0.4 signals that the item could be important for the corresponding factor. In this study, this would mean that some of the questions corresponding to the measurements represent more than one factor. However, when a more stringent norm of 0.7 was used, this problem disappeared as the questions related to the factors all score above

0.7 on only one factor.As expected, based on the literature review, the items loaded on three

diverse factors. These factors represent the three variables trust, closeness, and behavioral integration. This is true for week 1; yet, in week 4, a problem arises concerning the question related to closeness, as it also has a high factor loading (> 0.7) on trust. This is considered in the discussion chapter as a limitation of the current study.

After this, a third EFA is performed, including the constructs representing TMS. Together with the control variables, this resulted in a model with four factors. Based on the literature and the measurements used, six factors were expected. Moreover, the option of forced extraction was used. This decision can be rationalized because the validity of the constructs was already proven by previous research. This forced extraction of week 1 revealed that one of the questions measuring behavioral integration does not load on the factor related to behavioral integrations. For this reason, this question was not included when computing the mean. The results of week 4 demonstrated that the question, which represented trust, loaded

(23)

high on two factors. These factors were associated with TMS and closeness. The impact of this is discussed in more detail in the discussion chapter, as this is a limitation of the current study.

3.4 Data aggregation

The next step in describing the measurements was testing the level of consistency of the team members’ answers. This step was necessary to justify the aggregation of the data to the team level. To test consistency, a multilevel model was used with the individual respondents as level 1 and the team as level 2. It was expected that the team, as contextual variable, affected the individual’s answers, and, therefore, the team member’s observations could be similar. This is statistically regarded as a problem since statistic models assume independency between results of different respondents (Field, 2013). For this reason, this should be taken into account when analyzing the data. To calculate the effect of team on the different outcomes, the interclass correlation coefficient was used and represented the part of the total variation explained by the team.

Since the aim of this thesis was to examine the effects of communication frequency on TMS, which is a construct measured at the team level, the group means of the variables were used in the analysis. According to Enders and Tofighi (2007), this practice is appropriate when the variable of level 2 is the most essential but level 1 should also be taken into consideration.

Because of the difference in -2LL between the model without team identified as class

and the model with team identified as class in the first week (Χ2 (1) = 0.359, p>0.05), the

critical value of the chi-square distribution test was not reached. This means that there was no significant improvement of the model by adding teams as class. These results were for the

first week; yet, the results of the fourth week differed substantially (Χ2 (1) = 139,099 p<0.01).

(24)

increased over time. Furthermore, it indicated that teammates provided similar answers to the question. This offered a reason to aggregate the data to the team level to be able test the hypotheses on the team level.

The aggregation to the level of the team meant that all the values for the selected variables were gathered in groups. After this, an aggregation or combination was made for each group to summarize the data (Yu, Gunda, & Isard, 2009). In this thesis, the team was used as a group to combine the data.

Table 2 presents the results of the normality check for the aggregated data. To check normality, the skewness and kurtosis of the variables were calculated. The skewness represents the symmetry of the distribution of values for the selected variable. Ideally, the histogram representing values follows a normal distribution, meaning the values are equally distributed around the center. Kurtosis, on the other hand, involves whether the peak of the distribution is sharper or flatter compared to the normal. Based on the generally accepted rule of thumb, the divided the skewness or kurtosis score by its standard error has to be smaller than ±1.96. In the current study, multiple variables did not have a normal distribution. To resolve this issue related to the data distribution, the bootstrapping technique provided by SPSS was employed in the additional analysis. This technique resamples the original data sample to provide a better estimate of the distribution of the population, which results in more robust estimates (Field, 2013).

(25)

Table 2: normality checks team level: skewness and kurtosis of variables Skewness SD Kurtosis SD Week 1 TMS -.381 .246 1,479 .488 Communication frequency -.026 .246 -.096 .488 Face-to-face communication -.884 .246 1.118 .488 Trust -.623 .246 .084 .488 Closeness -.348 .246 -.369 .488 Behavioral integration -.521 .246 .403 .488 Week 4 TMS -.249 .246 .328 .488 Communication frequency -1.438 .246 3.053 .488 Face-to-face communication -.800 .246 1.258 .488 Trust -1.339 .246 2.994 .488 Closeness -.755 .246 .674 .488 Behavioral integration -1.297 .246 2.461 .488

This chapter focused on describing the variables through reliability and factor analysis. Furthermore, it pointed out that the data could be aggregated to the team level because this plays a significant role in week 4. The following chapter has the goal to provide the necessary information to support or reject the hypotheses presented in chapter 2. For this, the correlation between the dependent, independent, and control variables are discussed, followed by a hierarchical regression analysis.

(26)

4 Results

The previous chapter focused on describing the collection and characteristics of the data. The main goal of this chapter is to present the correlations between the variables and the results of the regression analysis. This is done to provide support to accept or reject the hypotheses stated in chapter 2. The type of analysis that was used is a hierarchical regression analysis with three models for each week. The first model contained only the control, namely variables closeness, trust, and behavioral integration. This model served as a baseline model. The baseline model was employed to assess whether communication frequency and face-to-face communication have an added value in explaining variance of TMS among teams.

4.1 Correlation among the variables

In Table 3, the descriptive of variables and correlation coefficients based on the aggregated data are provided. As can be observed, in week 1, TMS did not significantly correlate with face-to-face communication. Yet, in week 4, face-to-face communication was significantly correlated with TMS. This indicated that face-to-face communication became more important over time. Moreover, the frequency of communication in week 1 positively correlated with TMS in week 4; this revealed that communication frequency in the initial state may have a lasting effect on TMS. Regarding the control variables, a high correlation (r > 0,6) (Field, 2013) was found in week 1 between communication frequency and trust (r = .666, p < .01), and between communication frequency and behavioral integration (r = .757, p < 0.01). Furthermore, a high correlation was found between trust and behavioral integration (r = .701, p < 0.01). The correlations of week 4 reveal comparable results, respectively r =.877, r = .798 and r = .817, all with p < 0.01. In week 4, closeness also illustrated a high correlation with communication frequency (r = .735, p < 0.01), trust (r = .793, p < 0.01), and behavioral integration (r = .712, p < 0.01). This can be explained by the fact that trust, behavioral integration, and closeness are all concepts that are part of the group dynamic, which can only

(27)

occur when individuals communicate and get to know each other better.

As was pointed out in the EFA, there was no risk of an overlap in most of these correlations. Only the relationship between trust and closeness and trust and TMS in week 4 could be partly explained by the question related to trust loading high on these factors.

(28)

28 Ta b le 3 : de sc ri pt iv es o f va ri abl es a nd co rr el at io n at t he t ea m le ve l M SD 1 2 3 4 5 6 7 8 9 10 11 W ee k 1 1. TMS 4.89 .36 2. Com m uni ca ti on f re que nc y 5.87 .42 .594** 3. Face -to -f ac e c om m uni ca ti on 4.99 .75 .101 .195 4. T rus t 6.06 .43 .454** .666** .178 5. Cl os ene ss 5.33 .73 .462** .498** .507** .524** 6. Be ha vi ora l i nt egra ti on 5.78 .41 .520** .757** .188 .701** .595** W ee k 4 7. TMS 5.08 .53 .335** .388** -.020 .323** .121 .400** 8. Com m uni ca ti on f re que nc y 5.62 .69 .369** .462** .105 .435** .319** .512** .690** 9. Face -to -f ac e c om m uni ca ti on 4.68 .93 .075 .183 .636** .259* .343** .268** .237* .403** 10. T rus t 5.64 .67 .360** .515** .145 .534** .428** .565** .709** .877** .410** 11. Cl os ene ss 5.21 .95 .365** .412** .330** .423** .607** .502** .513** .735** .519** .793** 12. Be ha vi ora l i nt egra ti on 5.6 7 .62 .419** .577** .160 .523** .403** .641** .622** .798** .354** .817** .712** ** Corre la ti on i s s igni fi ca nt a t t he 0.01 l eve l (2 -t ai le d) * Corre la ti on i s s igni fi ca nt a t t he 0.05 l eve l (2 -t ai le d)

(29)

4.2 Hierarchical regression

As stated in Table 3, there was correlation between most of the variables; therefore, it is essential to first test for multicollinearity before conducting any regressions to check whether a strong linear relationship exists between two independent variables. The data did not reveal any problems related to multicollinearity. For this, the variance inflation factors (VIF) was computed. The scores of week 1 are all within the accepted ranch (Miles & Shevlin, 2000): communication frequency (VIF=1.049, Tolerance =.953) and face-to-face communication (VIF=1.049, Tolerance =.953). The same is true for the scores of week 4: communication frequency (VIF=1.136, Tolerance =.880) and face-to-face communication (VIF=1.136, Tolerance =.880). This indicates there were no problems related to multicollinearity.

Analysis week 1

As mentioned in the introduction, the first step in a hierarchical regression analysis was choosing the baseline model. The baseline model can be found in Table 4 and is referred to as Model 1. In this model, only the control variables were used. The second model builds on the first model by adding the communication frequency as independent variable. This model was used to test H1a. Finally, in the third model the face-to-face communication and the interaction term of communication frequency and face-to-face communication were added to test H2

In the second model, communication frequency was introduced as independent

variable. After communication frequency was added, the R2 _{changed from .315 to .391, which}

is a significant change. Furthermore, the standardized beta value that corresponds with communication frequency was positive. This finding confirmed the first hypothesis, H1a: “The frequency of communication is positively related to the TMS of teams in week one.”

The next step in the hierarchical regression analysis involved adding face-to-face communication and the interaction term of face-to-face communication and communication

(30)

frequency. This resulted in the third model. This model was created to investigate the possible moderating effect of face-to-face communication on the relationship between communication frequency and TMS. However, the interaction term did not significantly increase the explained variance. Thus, in week 1, face-to-face communication did not serve as a moderator variable. Therefore, the second hypothesis, “Face-to-face communication moderates the relationship between communication frequency and TMS,” was rejected.

Analysis week 4

The hierarchical regression analysis of the data of week 4 followed the same steps as the analysis of the first week. The results in Table 5 demonstrate that in the fourth week of the business strategy game, communication frequency did not cause a significant increase in the

explained variance of TMS. Moreover, R2 did not change significantly when communication

frequency was added in Model 2 compared to Model 1. Consequently, H1b: “The frequency of communication is positively related to the TMS of teams in week 4” was not supported. Additionally, after introducing face-to-face communication and the interaction term, there

was no significant increase in R2. Therefore, the third hypothesis, “The moderating effect of

Face-to-face on the relationship between communication frequency and TMS is greater in week 1 compared to week 4,” was rejected. Although it is not related to any of the hypotheses, it is noteworthy to mention that only trust had a significantly positive relation with TMS in the fourth week of the project. The high-standardized beta indicates a strong relationship. Therefore trust may mediate the relation between communication frequency and TMS. This would explain why communication frequency does no longer have a significant impact on TMS in week 4.

(31)

31 b le 4 : H ie ra rc h ic a l r e g re s s io n w e e k 1 T M S w ee k 1 M ode l 1 M ode l 2 M ode l 3 B S td.E rror B S td.E rror B S td.E rror d ep en d en t var iab le s m uni ca ti on f re que nc y .3 77 ** .1 32 .3 83 ** .1 39 -to -f ac e c om m uni ca ti on -.0 69 .0 44 te rac ti on m uni ca ti on f re que nc y & -to -f ac e c om m uni ca ti on -.0 09 .0 34 on tr ol s rus t .110 .0 96 .0 13 .0 97 .0 03 .1 00 os ene ss .106 .0 62 .0 99 .0 57 .1 40 * .0 63 ha vi ora l i nt egra ti on .258 .1 46 .0 47 .1 30 .0 29 .1 27 2 .315 .391 .406 R 2 .076** .015 = 96 ll va lue s re fl ec t s ta nda rdi ze d be ta c oe ff ic ie nt s nda rd e rrors a re ba se d on boot st ra ppe d da ta

(32)

32 Ta b le 5 : H ie ra rc h ic a l r e g re s s io n w e e k 4 T M S w ee k 4 M ode l 1 M ode l 2 M ode l 3 B S td.E rror B S td.E rror B S td.E rror In d ep en d en t var iab le s Com m uni ca ti on f re que nc y .2 25 .1 37 .2 12 .1 35 Face -to -f ac e c om m uni ca ti on -.0 21 .0 52 In te rac ti on Com m uni ca ti on f re que nc y & Face -to -f ac e c om m uni ca ti on -.0 45 .0 59 C on tr ol s T rus t .5 51 ** .1 34 .4 05 ** .1 43 .3 91 ** .1 41 Cl os ene ss -.0 89 .0 78 -.1 00 .0 78 -.0 84 .0 84 Be ha vi ora l i nt egra ti on .1 34 .1 31 .0 77 .1 35 .0 64 .1 37 R 2 .518 .536 .543 Δ R 2 .018 .007 N = 96 A ll va lue s re fl ec t s ta nda rdi ze d be ta c oe ff ic ie nt s S ta nda rd e rrors a re ba se d on boot st ra ppe d da ta *p<0.05, **p<0.0

(33)

Post-hoc analysis

The results of the correlation matrix and the two previous hierarchical regressions led to a number of additional questions concerning these data and the correlations between these particular variables. The communication frequency of week 1 was correlated with the TMS in week 4. Therefore, it is relevant to test this by means of a hierarchical regression (see the Appendix for the results). The models are similarly constructed as the models of the hierarchical regression performed for week 1 and 4; meaning face-to-face communication and the control variables of week 1 were included. The difference with the previous models is that in this table, the influences of communication frequency, face-to-face communication, and the control variables of week 1 on TMS in week 4 were investigated. This created a lag, “a delay in the period of time in which events happen”(Cambridge Dictionary, 2017). The results of these analyses were not significant; hence, the frequency of communication or face-to-face communication in week 1 did not have a direct effect on TMS in week 4.

Furthermore, when the regressions of week 1 and week 4 were compared, a high increase of the importance of trust was evident. This may indicate that in the initial state, communication frequency was important but once a certain level of trust was reached, this role shifted to trust. This was verified through a mediation model (see Figure 2). For this, Model 4 of the program process was used as an addition to SPSS (Hayes, 2013).

Figure 2: Trust as mediator between communication frequency and TMS

Communication frequency week 1

Trust week 4

(34)

Table 6 and Table 7 present the results of this analysis. The effect of communication

frequency in week 1 on trust in week 4 a1 = 0.677. As such, two teams that differ by one unit

on communication frequency differ by 0.677 units on trust. The sign of a1 was positive, which

entails that these teams with a higher communication frequency were estimated to have a higher trust level. This effect was statistically different from zero, t= 8.663, p= .000.

The effect b1= 0.144 indicated that two teams with the same communication frequency

but differ by one unit in their trust level are expected to differ by 0.144 units in TMS. However, this effect was not significant, t = .912, p = .364.

The indirect effect of a1b1 = 0.097 meant that two teams that differ by one unit in

communication frequency are expected to differ by 0.097 units in TMS due to the difference in the trust levels between the teams. This indirect effect was not significant, as revealed by a 95% BC bootstrap confidence interval that is not entirely above zero (-.086 to .298).

The direct effect of communication frequency, c’= .390, was the estimated difference in TMS between two teams when the level of trust is the same in both teams but differ by one unit in communication frequency, meaning that the team that communicated more but had an equal level of trust that was estimated to be .390 units higher in TMS. This direct effect was statistically significant, t = 2.434, p=0.017. The total effect of communication frequency on TMS was c = .390, revealing two teams that diverged by one unit in communication frequency are estimated to differ .390 units in their TMS. The positive sign designates that the team, which communicates more, has a higher TMS. This effect was significant, t = 2.434, p=0.017.

In this study, the total effect of communication frequency was the same as the direct effect since the effect of trust as mediator was not significant.

(35)

Table 6: regression model based on process

Consequent

Trust week 4 TMS week 4

Antecedent Coeff. SE P Coeff. SE P

Communication frequency week 1 a1 .679 .078 <.001 c1’ .390 .160 <.05 Trust week 4 --- --- --- b1 .144 .158 = .364 Constant i1 2.094 .459 <.001 i2 1.921 .777 <.05 R2 = .443 R2 = .158 F(1,94) = 75.042, p <.001 F(2,93) = 8.735, p <.001

Table 7: mediated model based on process

Effect SE P LLCI ULCI

Direct effect c1’ .390 .160 <.05 .072 .708

Total effect c1 .390 .160 <.05 .072 .708

Boot SE Boot LLCI Boot ULCI

Indirect effect a1b1 .097 .096 -.086 .298

In chapter 4, the results were presented, which led to the acceptation of H1a and the rejection of the H1b, H2 and H3. Because a number of findings triggered further questions concerning the variables used in the models, a post-hoc analysis was performed to test the role of trust. The findings indicated that trust did not have a significant mediating role.

The next chapter interprets these results in more detail and explain possible implications. At the end, the chapter provides the limitations of the current research and offers suggestions for further research.

(36)

5 Discussion

It is commonly known that many hands make light work. Everyone agrees that when individuals join forces in a team project, they will achieve more than they would have been able to achieve on their own. Within the research field, it is generally accepted that the TMS strength is a good predictor of team performance. In turn, TMS is influenced by multiple factors such as trust, closeness, behavioral integration, and communication. Previous researchers have mostly recognized the value of frequent communication in the development process of TMS. For example, they have argued that more frequent communication results in a stronger TMS (Baldwin et al., 1997; Ren & Argote, 2011). However, the importance of face-to-face communication has often been questioned. Moreover, it is difficult to find a consensus on this factor and its influence on TMS (Louhiala-Salminen & Kankaanranta, 2011; Ulijn, Lincke, & Karakaya, 2001). Some researchers have claimed face-to-face communication to be crucial to establish a certain level of personal relationship, which is required for effective team work (Burt, 2009; Nahapiet & Ghoshal, 1998; Peltokorpi, 2004), whereas others have stated that face-to-face communication remains essential. To effectively perform the teamwork and reach the team’s goals, teammates need to meet frequently (Baltes et al., 2002; Bordia, 1997; Lewis, 2004). Therefore, to clarify this contradiction, this thesis investigates the relationship between the frequency of communication, TMS, and the moderating role of face-to-face communication.

The results in chapter 4 demonstrate that, as expected, communication frequency is positively related to TMS both in the beginning (week 1) and during the group project (week 4). Based on the previously mentioned results, H1a: “The frequency of communication is positively related to the TMS of teams in week 1” can be supported. In week 1, most of the variety of the levels of TMS can be explained based on the frequency with which the teams communicated. Therefore, it can be said that the frequency of communication plays a role in

(37)

the development of TMS among student teams during group projects in the initial state of the project. This observation, which indicates that communication frequency positively influences TMS, is not new (Hollingshead, 1998; Lewis, 2004). Nevertheless, this effect was not found in week 4 and, thus, H1b “The frequency of communication is positively related to TMS in week 4” is rejected.

Moreover, for the second hypothesis, “Face-to-face communication moderates the relationship between communication frequency and TMS”, no proof was found. The interaction term does not reveal a significant moderating effect on TMS. Furthermore, the overall model (Model 3) has no improvement in explained variance of the TMS among teams. Finally, based on the data, there is no evidence to support H3 “The moderating effect of face-to-face on the relationship between communication frequency and TMS is greater in week 1 compared to week 4”. This implies that after the team members are familiar with each other, the way of communication no longer influences the development of TMS. These results on the moderating role of face-to-face communication contradict the expectations.

Although the focus was on the hypotheses, the two hierarchical regressions demonstrate unexpected findings. To illustrate, the importance of trust in explaining TMS increased between week 1 and week 4. This suggests that the relationship between communication frequency and TMS is more dynamic than has been argued in previous research. A post-hoc analysis was performed to answer the question: “Does trust mediate the direct effect of communication on TMS in week 4?” It could be that the main role of communication frequency in the development of TMS is to establish trust among team members, but once a certain level of trust is created, it loses its direct effect on TMS. However, when this was tested through a mediation model, the effect was not significant.

The following paragraphs discuss three implications of these findings regarding the context of the research field and a more practical context.

(38)

Different relational antecedents, different times: toward an understanding of the social dynamics of antecedents of TMS

The results in chapter 4 illustrate that communication frequency is significantly and positively related to TMS in the first week of the business strategy game. This is in accordance with previous research findings that the frequency with which teammates communicate is a key factor in explaining the different levels of TMS among multiple teams (Jackson & Moreland, 2009; Lewis, 2004; Wegner, Giuliano, & Hertel, 1985b). Considering the results, the findings of week 4 are in contrast with findings of week 1. In the fourth week of the group project, the frequency of communication among group members did no longer have a significant relation with TMS. This contradicts most of the previous studies, as they have found that communication frequency remained a key factor in explaining the variance of TMS. When communication frequency is not related to TMS, this implies that team members do not have to communicate on a frequent basis to work efficiently together and exploit the benefits of a TMS.

The more practical implication of this finding is that it provides an argument in favor of frequent team meetings, following the formation of a new team. These frequent meetings will encourage the teammates to learn about each team member’s skills and expertise; this will result in a mental model about how the knowledge is distributed within the team. This model will then be used to divide roles and tasks, allowing the team to work more efficiently and to fully benefit from the available knowledge. Moreover, frequent communication provides the team members with the opportunity to build strong interpersonal relationships and a sense of reciprocity, which is essential to develop a strong TMS (Baldwin et al., 1997; Nahapiet & Ghoshal, 1998; Peltokorpi, 2004). As a result, a team with a strong TMS does not necessarily has to communicate as often as predicted by pervious research.

A reason for this could be that interpersonal relationships are developed and, hence, a feeling of trust and reciprocity (Gajendran & Harrison, 2007). If this is true, teammates

The influence of communication frequency on TMS