University of Southern California Dissertations and Theses

We're all in this (game) together: transactive memory systems, social presence, and social information processing in video game teams

(USC Thesis Other)

We're all in this (game) together: transactive memory systems, social presence, and social information processing in video game teams

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

WE’RE ALL IN THIS (GAME) TOGETHER:

TRANSACTIVE MEMORY SYSTEMS, SOCIAL PRESENCE, AND SOCIAL
INFORMATION PROCESSING IN VIDEO GAME TEAMS

by

Adam Scott Kahn

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(COMMUNICATION)

December 2013

Copyright 2013 Adam Scott Kahn
ii

Dedication
To Hillary…
To my parents…
To all the great teachers and mentors over the years…

iii

Acknowledgments
First and foremost, I must thank my advisor, Dmitri Williams, for all of the
mentorship he has given me over the last five years. Through encouragement and tough
love, I have grown so much as a scholar because of him. I must also thank my other two
dissertation committee members, Janet Fulk and Steve Read for their valuable feedback
along the way. An additional thanks goes to Andrea Hollingshead and Kwan Min Lee,
who were members of my qualifying exam committee. Their research greatly influenced
this dissertation. Thank you to Henry Jenkins and Ken Sereno for their collaborations
over the last five years as well.
This dissertation would not have been possible without a great team of
undergraduates assisting me with Studies 2 and 3. For an entire year, Ly Dinh was my
lead research assistant. Her commitment to the project and her countless hours spent
running participants and entering data were invaluable. I would not have been able to
complete this dissertation on time without her. Keika Stevenson and Kevin Barth also
devoted a semester to running participants and entering data.
I must also thank my friends at Annenberg. The biggest thanks goes to Robby
Ratan. Robby and I have worked together since our days at Stanford and then at USC.
We joked I would follow him to Michigan State, and we were only off by 85 miles. I
must thank my cohort, especially Laurel Felt and Julien Mailland for their friendship, and
the others who have worked in Dmitri’s lab: Cindy Shen, Leo Xiong, Li Lu, Alex Leavitt,
and Josh Clark. I would like to thank my future colleagues at Western Michigan
iv

University. Looking forward to joining you in the Fall helped move this dissertation
along.
Among the people I dedicate this dissertation to are the great teachers and
mentors over the years. In addition to Dmitri, this includes Cliff Nass and Fred Turner at
Stanford, and going even farther back, John Leistler and Barbara Gaims-Spiegel. I must
thank my parents for their support, both emotional and financial, as I navigated grad
school. Finally, the most important person in this process has been my fiancée, Hillary
Howard. The support I received from my qualifying exams through this dissertation has
been immense, and I couldn’t have done it without you.
v

Table of Contents
Dedication ii
Acknowledgments iii
List of Tables vii
List of Figures viii
Abstract ix
Introduction 1
Study 1: Group Composition, Transactive Memory Systems, and Social Presence 18
Method 27
Participants and Procedure 29
Measures 30
Data Analysis Procedures 32
Results 32
Discussion 42
Study 2: Communication Channel, Transactive Memory Systems, and Social Presence 46
Method 51
Participants 51
Procedure 54
Measures 57
Results 60
Discussion 75
Study 3: Social Information Processing and Transactive Memory Systems 79
Method 86
Participants 86
Procedure 86
Measures 86
Results 88
Discussion 102
Conclusion 107
References 114
vi

Appendix A: Measures From Study 1 131
Transactive Memory System Scale (adapted from Lewis, 2003) 131
Social Presence Scale (adapted from Biocca & Harms, 2003) 132
Appendix B: Measures From Study 2 133
Transactive Memory System Scale (adapted from Lewis, 2003) 133
Social Presence Scale (adapted from Harms & Biocca, 2004) 136
Appendix C: Measures From Study 3 141
Transactive Memory System Scale (adapted from Lewis, 2003) 141
Relational Communication Scale (adapted from Burgoon & Hale, 1987) 144

vii

List of Tables
Table 1. Odds Ratios and Fit of the Model Predicting Whether or not a LoL team Wins a
Match 35
Table 2. Means, Standard Deviations, and Intercorrelations of Team Formation and
Dimensions of Social Presence 38
Table 3. Means, Standard Deviations, and Differences in Social Presence Based on Team
Size 39
Table 4. Tests for Social Presence as a Mediator between Team Formation and
Transactive Memory Systems 41
Table 5. Cronbach's α for Study 2 Questionnaire Measures 58
Table 6. Means and Frequencies of r
wg(j)
for Study 2 Questionnaire Measures 60
Table 7. Correlation Among Study 2 Repeated Measures 62
Table 8. Power Analysis for Study 2 ANOVAs 63
Table 9. Correlation between TMS and Dimensions of Social Presence across Time 74
Table 10. Cronbach's α for Study 3 Questionnaire Measures 87
Table 11. Means and Frequencies of r
wg(j)
for Study 3 Questionnaire Measures 88
Table 12. Correlation Among Study 3 Repeated Measures 90
Table 13. Power Analysis for Study 2 ANOVAs 91
Table 14. Correlation between TMS and Dimensions of Relational Communication
across Time 101

viii

List of Figures
Figure 1. McGrath's Circumplex Model of Group Tasks 4
Figure 2. Study 1 Hypotheses and Research Questions 27
Figure 3. Results of Study 1 42
Figure 4. Mean Levels of Transactive Memory System over Time by Condition 64
Figure 5. Mean Levels of Performance over Time by Condition 67
Figure 6. Mean Levels of Copresence over Time by Condition 68
Figure 7. Mean Levels of Attentional Allocation over Time by Condition 69
Figure 8. Mean Levels of Perceived Message Understanding over Time by Condition 70
Figure 9. Mean Levels of Perceived Affective Understanding over Time by Condition 71
Figure 10. Mean Levels of Perceived Affective Interdependence over Time by Condition
72
Figure 11. Mean Levels of Perceived Behavioral Interdependence over Time by
Condition 73
Figure 12. Mean Levels of Immediacy/Affection over Time by Condition 92
Figure 13. Mean Levels of Similarity/Depth over Time by Condition 93
Figure 14. Mean Levels of Receptivity/Trust over Time by Condition 94
Figure 15. Mean Levels of Composure over Time by Condition 95
Figure 16. Mean Levels of Formality over Time by Condition 96
Figure 17. Mean Levels of Dominance over Time by Condition 97
Figure 18. Mean Levels of Equality over Time by Condition 98
Figure 19. Mean Levels of Task Orientation over Time by Condition 99
ix

Abstract
This dissertation examines transactive memory systems (TMS) in video game teams.
Study 1 is a field study of 16,499 players of the video game League of Legends and uses
survey and server data to explore the relationship between group composition (member
self-selection and team size) and TMS. It also explores the mediating role of social
presence. Study 2 is a League of Legends lab experiment comparing the effects of
communication channel (text-chat vs. voice-chat) on the development of TMS and social
presence in teams. Study 3, conducted concurrently with Study 2, tests Social
Information Processing theory in the context of video game teams and examined the role
of relational communication in the development of TMS.
1

Introduction
Video games are a ubiquitous medium. According to the Entertainment Software
Association (ESA; 2012), in the United States, the average household owns at least one
dedicated game console, PC, or smartphone. Similarly, video game play stretches across
all segments of the population, with approximately 47% of all game players being female
and 37% of all game players being older than 36 years of age. Part of the reason why
games are popular is because the there are many types of games ranging from virtual
worlds to casual games to children’s spaces to serious games (Williams & Kahn, 2013).
In addition to their importance in popular culture, video games provide a fertile
ground for social science research. Games and virtual worlds can serve as “petri dishes”
for the study of human behavior (Bainbridge, 2007; Castronova & Falk, 2009; Williams,
2010). It is possible that parallels exist between real-world behavior and behavior in
virtual worlds; with the right methodology, virtual worlds can provide better
experimental control and less expensive ways to test theories of human behavior than
would a face-to-face lab experiment. The validity of an experiment requires the
researcher to hope that everything is held constant at across participants. On the other
hand, an experiment in a virtual world can ensure everything (other than individual
differences in participants) is held constant by programming the world such that the only
difference between conditions is the experimental manipulation.
Furthermore, many video games (especially online games) require players to
create an account that a player must sign into every time. Therefore, player statistics (and
sometimes second-by-second player actions) are stored somewhere on a server
2

(Williams, 2008; Williams, Contractor, Poole, Srivastava, & Cai, 2011). These server
logs, assuming researchers are able to gain access, provide objective, unobtrusive
measures of player behavior (as opposed to requiring human coders, which is often
necessary when lab experiments are videotaped). This provides researchers with the
advantage of not having to rely on self-report data, which are often imperfect (Kahn,
Ratan, & Williams, 2010; Webb, Campbell, Schwartz, & Sechrest, 2000). Why ask
players what they do while playing a game when you have precise data that says what
they actually do in the game?
Video Games Teams and Small Group Communication
Games as a social and recreational context. Games are a very social medium.
The ESA reports that 62% of all game players play games with others, either online or in
person (Entertainment Software Association, 2012). For many, playing with others is one
of the biggest appeals of some games. One survey of players of the massively multiplayer
online game (MMO) EverQuest found that the most popular feature of the game was its
social components (Griffiths, Davies, & Chappell, 2004). Given the fact that playing
games is often such a social experience, they have the potential to answer many questions
of interest to group researchers (Wirth, Feldberg, Schouten, van den Hooff, & Williams,
2012). Given the methodological promise of video games, not only can questions be
asked about virtual groups, questions about real world groups can be tested in these
environments.
Yet relative to the importance of social motives of video game players and the
promise for small group scholars, research on video game teams is an understudied area
3

by communication researchers. Peña and Hancock (2006) posit two reasons for this. First,
video game scholars tend to come out of the media processing and effects tradition. This
tradition has a tendency to look at the individual as the unit of analysis. Second, group
scholars have a tendency to focus on instrumental and task-oriented communication,
seeing relational and socioemotional communication as secondary processes (Keyton,
2000; Rice & Love, 1987). And since the recreational nature of video games makes
relational and socioemotional communication the dominant forms of communication
during gameplay (Klimmt & Hartmann, 2008; Peña & Hancock, 2006), group researchers
tend to show little interest in the subject.
Yet relational and socioemotional communication are integral components of all
human interactions. Video games provide a space where these interactions are prevalent.
Peña and Hancock (2006) conducted an Interaction Process Analysis (Bales, 1950) of
transcripts of the text communications between members of a gaming clan
1
of the video
game Jedi Knight II: Jedi Outcast, a game set in the universe of Star Wars. The game
involves players engaging in lightsaber duels.
2
Despite the fact that the primary function
of the game was a task involving fighting other players, the analysis showed that the
majority of communication between players was socioemotional and of positive valence.

1
A clan is a group of “players organized as a stable and task-oriented team” (Klimmt &
Hartmann, 2008, p. 310)
2
A lightsaber is a “laser sword” in the Star Wars franchise, so a lightsaber duel would be
the equivalent of a sword fight in the real world.
4

Games as a contest task. However, not only does studying group processes in
video games add to the minimal research on mediated communication in socioemotional
and recreational contexts, it fills an overlooked deficiency in group research: groups
engaging in what are known as “Contest” tasks. After surveying and synthesizing the
literature on types of tasks groups engage in, McGrath (1984) proposed a circumplex
model of group task types (see Figure 1). The term circumplex was used, as the model
varied along two dimensions: conceptual/behavioral and cooperation/conflict. Together,
the different combinations of the possible values of the two dimensions could be
encompassed in a circle as they varied along the two dimensions.
Figure 1. McGrath's Circumplex Model of Group Tasks

5

The circumplex model resulted in eight mutually exclusive, collectively
exhaustive, logically related task types that fell into one of four “Quadrants”: tasks that
require groups to “Generate” an idea or a plan, tasks that require groups to “Choose” a
solution to a problem or a decision to an issue, tasks that require groups to “Negotiate” a
conflict of viewpoint or interest, and tasks that require groups to physically “Execute”
either a psychomotor or competitive task.
Video games tend to fall under this physical execution of a competitive task, or
what McGrath (1984) called “Type 7. Contest tasks”.
3
He defined contests as “tasks for
which the unit of focus, the group, is in competition with an opponent, an enemy, and
performance results will be interpreted in terms of a winner and a loser, with pay-offs in
those terms as well” (1984, p. 65). These tasks have no intragroup conflict or mixed-
motives relationships with other groups; these tasks involve groups “conquering” other
groups. He gave such examples as wars, winner-take-all conflicts, and competitive sports,
but such a definition would extend to contemporary multiplayer games.
It should be noted that in a Type 7 task, unlike the other tasks types, two metrics
have to be considered: performance and outcome (McGrath, 1984). There are multiple
possible metrics of performance, how well a team does during the course of a task. On
the other hand, outcome is binary: a group either wins or loses. A strong performance
does not guarantee a win, and a weak performance does not guarantee a loss. An athletic

3
Sometimes McGrath also used the terminology of “Contests/battles” or
“Contests/battles/competitive tasks” to describe Type 7 tasks.
6

team can score many points but still lose a game, and an army can incur many casualties
but still win a battle.
When he first proposed the circumplex model almost thirty years ago, McGrath
noted about Type 7 tasks that “such tasks are very heavily represented in the workaday
world and, against that baserate, are quite underrepresented in research on groups” and
“performance tasks of this type…constitute a large part of the time and energies of a
large number of actual groups. Considering that, the amount of attention given to such
performance in small group research is quite limited” (1984, pp. 65, 115). This is still
very true today. Despite the prevalence of contest tasks in the real world, in an informal
analysis conducted for the purposes of this dissertation, only four of the 92 articles
published in Small Group Research (the preeminent journal on the subject) between 2010
and 2012 dealt with any form of team engaged in a Type 7 task.
4

As McGrath notes, Type 7 tasks are central to many real world groups. These
tasks are the primary tasks of sports teams and military units. They are a component in
parliamentary groups in which opponents must debate and will ultimately pass or fail to
pass a law or win over public opinion. In order to land clients, companies must outbid
their competitors. Even a family arguing over where they should go on vacation may

4
This count includes any article that focuses solely on a Type 7 task (either empirically
or theoretically), surveys at least one group that primarily focuses on Type 7 tasks, or has
at least one experimental condition where a group engages in a Type 7 task. Simulations
(e.g. computerized simulations of military exercises) of Type 7 tasks were also included.
7

involve a Type 7 tasks. Thus, while Type 7 tasks have received minimal research, they
are an area of study ripe for investigation.
When McGrath (1984) proposed his eight types of group tasks, the distinguishing
element of Execute tasks (Contest tasks, along with Type 6, Performance tasks) is the
engagement in psychomotor activity. While video games may not involve the same type
of psychomotor activity that sports teams or real-world combat teams engage in, video
games that require teams to defeat other teams in real-time combat are engaging in the
psychomotor activity of using a keyboard and mouse and requires a great deal of hand-
eye coordination. In addition, these games are behavioral, inter-group conflict tasks,
making them most similar to Contest tasks when measured along the two dimensions
(conceptual/behavioral and cooperation/conflict). Thus, team vs. team, combat video
games provide a way to research the understudied Contest tasks while at the same time
not placing those being studied at physical risk (although the lack of physical risk must
be taken into consideration when interpreting the external validity of results).
Some Type 7 tasks such as a sports competition or military battles are very fast-
paced and require diverse expertise: sports teams have positions and military groups have
sub-units. With such speed, there may not be much time to decide what to do next, so it is
important for team members to be able to recognize the expertise of others and have a
mechanism to coordinate said expertise. This leads to the importance of looking at the
role of transactive memory systems in Type 7 tasks.
8

Transactive Memory Systems
A transactive memory system (TMS) is “the shared division of cognitive labor
with respect to the encoding, storage, retrieval, and the communication of information
from different knowledge domains, which often develops in groups and can lead to
greater efficiency and effectiveness” (Brandon & Hollingshead, 2004, p. 633). First
proposed by Wegner (1987) almost thirty years ago, the study of TMS has been of great
interests to scholars in social psychology, communication, and organizational science—
especially in recent years. Lewis and Herndon (2011) found 198 articles published
between 2000 and 2010 with the keyword “transactive memory,” with 77% of these
articles published after 2005. TMS has been studied in dyadic (e.g. Hollingshead, 1998a,
2000, 2001; Hollingshead & Fraidin, 2003; Wegner, 1987), small group (e.g. Jackson &
Moreland, 2009; Lewis, Belliveau, Herndon, & Keller, 2007; Lewis, Lange, & Gillis,
2005; Liang, Moreland, & Argote, 1995; Moreland, 1999; Moreland & Myaskovsky,
2000), and organizational (e.g. Austin, 2003; Lewis, 2003, 2004; Yuan, Fulk, & Monge,
2007; Yuan, Fulk, Monge, & Contractor, 2010) contexts.
There are two components to a TMS: the TMS structure and transactive processes
(Lewis & Herndon, 2011; Wegner, 1987). The TMS structure is a group’s shared mental
model of the expertise of group members. It is sometimes referred to as “a shared
understanding of who knows what” (Brandon & Hollingshead, 2004; Lewis & Herndon,
2011). Individuals may learn the expertise of partners/teammates through close
interpersonal relationships (Hollingshead, 1998a, 1998b; Wegner, 1987), prior experience
working together (Liang et al., 1995; Moreland, 1999; Moreland & Myaskovsky, 2000),
9

explicit feedback about the strengths of team members (Hollingshead, 2000; Moreland &
Myaskovsky, 2000), or even based on societal stereotypes (Hollingshead & Fraidin,
2003). The group shared mental model will only form when group members are
cognitively interdependent on one another (Hollingshead, 2001; Wegner, 1987). By
having a shared mental model, individuals in a group can both explicitly and implicitly
create a division of labor best suited to both the unique and shared expertise of the group
members.
Brandon and Hollingshead (2004) note that TMS “will be most effective when
knowledge assignments are based on group members' actual abilities, when all group
members have similar representations of the system, and when members fulfill
expectations” (2004, p. 639). Thus, the quality of the shared mental model depends on
three things: accuracy, sharedness, and validation. Group members must have an accurate
mental representation of the expertise of their teammates. Otherwise, tasks may be
delegated incorrectly. Group members must share a similar mental model. Otherwise,
different team members’ strategies for dividing labor may differ. Finally, group members
must validate the expectations of their group mates. Otherwise, it may lead to individuals
incorrectly modifying their mental model.
Lewis and Herndon (2011) note that TMS research tends to focus on TMS
structure, giving little attention to transactive processes. This is despite the fact that both
structure and process were equally important in the original (Wegner, 1987)
conceptualization of TMS. Transactive processes are “the mechanisms by which the
group coordinates members’ learning and retrieval of knowledge, so that the knowledge
10

can be applied to group tasks” (Lewis & Herndon, 2011, p. 1256). The three processes
originally proposed by Wegner were: a) directory updating, the process where teammates
learn the expertise of their teammates and form their shared mental model; b) information
allocation, where new information is assigned to the teammate with the most relevant
expertise; and c) retrieval coordination, where the group gets the information it needs
from the individual it judges most likely to have it. Transactive processes most frequently
occur in the form of verbal and nonverbal communication (Hollingshead, 1998a;
Hollingshead & Brandon, 2003; Hollingshead, Brandon, Yoon, & Gupta, 2011;
Palazzolo, Serb, She, Su, & Contractor, 2006; Yuan et al., 2007; Yuan et al., 2010).
There is a strong correlation between the amount of communication within the group and
the strength of a TMS (Jackson & Moreland, 2009; Kanawattanachai & Yoo, 2007;
Lewis, 2004; Palazzolo et al., 2006; Yoo & Kanawattanachai, 2001; Yuan et al., 2007;
Yuan et al., 2010).
There are three latent indicators of an established TMS: specialization, credibility,
and coordination (Lewis, 2003). If group members do not have specialized knowledge or
abilities, labor might as well be divided up at random. If group members do not trust one
another’s knowledge or abilities, individuals may not want to delegate tasks to somebody
who claims expertise in an area. Finally, even if there are specialization and credibility, if
there is no mechanism to coordinate the knowledge and abilities, this specialization is of
little use.
11

Computer-mediated communication and transactive memory systems. Given
the elements of specialization, credibility, and coordination, Hollingshead (2010) sees
multiplayer online games as a fruitful area of study in the area of TMS:
There are many related topics ripe for future research [on communication and
coordinated action in groups]. Some of the most exciting are in the area of
technology and coordination. For example, there are many online contexts that
relate to work, political action, social networking, and entertainment where
members must coordinate to achieve individual and collective goals. One example
is the popular multiplayer online game World of Warcraft. Players self-organize
into guilds, in which players complete missions as a team. Players are able to
choose the gender and race of their avatar. Different races have different attributes
and abilities. This is a fascinating context in which to study the interplay of
stereotypes and conventions in the creation of focal points by groups (pp. 401-
403).
In fact, one study (Richter & Lechner, 2009) looked at the formation of TMS in zero-
acquaintance World of Warcraft teams. However, it simply looked at the communication
transcripts of five teams and subjectively judged whether transactive communication
processes occurred. It concluded that some of the five teams seemed to have a TMS, and
the teams that did have TMS included both teams that communicated solely by voice and
teams that communicated solely by text.
Yet TMS research involving computer-mediated communication (CMC) has been
inconclusive and some might say even understudied relative to the importance of virtual
12

teams in today’s society (Moreland, Swanenburg, Flagg, & Fetterman, 2010). Some of
the earliest research (Hollingshead, 1998b, Experiment 1) of TMS compared retrieval
processes in intimate couples and strangers in CMC and face-to-face (FtF) contexts. This
research found that CMC inhibited TMS processes. A follow-up study (Hollingshead,
1998b, Experiment 2) concluded this was due to the fact that CMC eliminated nonverbal
and paralinguistic cues. While this is the only known study to compare TMS in CMC and
FtF contexts, it is hard to draw generalizable conclusions from this due to the newness of
CMC at the time the study was conducted. In addition, while the study did not draw on
any particular CMC theories, it implicitly took a “cues-filtered-out” perspective, which
sees CMC as inferior for complex tasks due to the lack of nonverbal cues. These theories
have since fallen out of favor, replaced by a different perspective: adaptation theories that
state that over time, CMC users have strategies to adapt to the lack of nonverbal cues (cf.
Walther, 2006, 2011; Walther & Parks, 2002).
Some longitudinal, field studies of TMS have found that communication volume
is positively related to TMS (Jackson & Moreland, 2009; Lewis, 2004; Yuan et al., 2007;
Yuan et al., 2010). One of these studies measured FtF and non-FtF communication
separately and found that this positive relationship between communication volume and
TMS only held for FtF communication volume; non-FtF communication volume had no
relationship with TMS (Lewis, 2004). Another study asked about overall communication
volume and FtF communication volume. It found that overall volume had a positive
relationship with TMS at both Time 1 and Time 2 (the study was a two-wave panel
13

study), but the amount of FtF communication only had a positive relationship with TMS
at Time 1, with no relationship at Time 2 (Jackson & Moreland, 2009).
Studies (Kanawattanachai & Yoo, 2007; Yoo & Kanawattanachai, 2001) that
have looked at TMS in virtual teams utilizing only text-based CMC have found that TMS
can develop in virtual teams. The teams were MBA students working together on a
business simulation over the span of a semester. Time played a crucial role in TMS
development, as intra-group communication was higher at the beginning of the semester
than later in the semester, once TMS had stabilized. This earlier communication volume
best predicted TMS throughout the process, more so than communication volume at later
time points. In addition, earlier in the semester, the proportion of task (as opposed to
socioemotional) communication predicted TMS, but later in the semester, the proportion
of task communication had no relationship with TMS.
Transactive Memory Systems in Video Game Teams
Only two known studies looked at TMS in video game teams. The first (Richter &
Lechner, 2009), as already mentioned, had little generalizability given that it was a
descriptive, qualitative study with a very small sample size. The second (Riedl,
Gallenkamp, Picot, & Welpe, 2012), involved a video game, Travian, which involved
mixed-motive tasks
5
as opposed to purely competitive tasks; thus, it would not
necessarily have been classified by McGrath (1984) as a Type 7: Contest/battle task.

5
Travian begins with players initially working individually. Along the way, players can
form and break alliances. While members of an alliance were interdependent, players
chose their alliances based on self-interest.
14

A popular new genre of video game, the multiplayer online battle arena (MOBA),
would be a classic example of a contest/battle task: ad hoc teams compete against other
ad hoc teams in fast-paced, head-to-head combat. Currently, the most popular video game
in the world is a MOBA: League of Legends (LoL). LoL has an active player based of
over 32 million players as of Fall 2012 (Riot Games, 2012).
MOBAs are a hybrid of the MMO and real-time strategy (RTS) genres; they
center around two small teams competing against one another. MOBAs may in fact be
especially appropriate to study group process given that research has shown that many
MMO players play predominantly on their own (Ducheneaut, Yee, Nickell, & Moore,
2006; Shen, 2010). The primary distinguishing feature of a MOBA as compared to
MMOs like World of Warcraft or Travian
6
is that a MOBA is a non-persistent world.
This means that whereas in a persistent world, one logs off and logs back on to where he
or she left off, in a non-persistent world, once a game has been completed (either a team
wins or loses), it has been completed. When one logs off and logs back on, he or she
starts a new game (although identity and friends lists are persistent). On the other hand,
many MOBAs resemble many MMOs in that both genres involve characters and stories
that would be considered of the fantasy genre. In addition, both MOBAs and MMOs have
players assume the identity of an avatar.

6
Travian is also hybrid of an MMO and RTS as well, but like an MMO, it is a persistent
world and player resources and not real-time combat determine the winners of battles
(e.g. the alliance with the most soldiers, weapons, etc… would win a battle).
15

The aforementioned TMS studies that used text-based CMC (Hollingshead,
1998b; Kanawattanachai & Yoo, 2007; Yoo & Kanawattanachai, 2001) may only say so
much about video game teams, as many video game teams use a combination voice and
text-chat when they play (Williams, Caplan, & Xiong, 2007). In fact, no research since
Hollingshead (1998b) has compared the role of communication modalities on TMS in an
experimental fashion, and that research was conducted at a time when most people had
little to no experience with any form of CMC. Riedl et al. (2012) found the perceived
communication richness of a medium was positively related to TMS, but contemporary
theories such as Social Information Processing (SIP) Theory (Walther, 1992) and the
Hyperpersonal Model of CMC (Walther, 1996) suggest that over time, media with fewer
paralinguistic and nonverbal cues may become just as, if not more, rich than FtF (or
media with more paralinguistic and/or nonverbal cues than text). In addition, all of these
studies looked at tasks very different from the Type 7, Contest/battle tasks one finds in a
MOBA.
Thus, in addition to being an example of two areas understudied by
communication and small group researchers (recreational/socioemotional CMC and Type
7, Contest/battle tasks), MOBAs also highlight several areas where scholars have
suggested future TMS research investigate. Lewis and Herndon (2011) believe
researchers should explore TMS in a wider range of task types and in dynamic contexts.
Similarly, Hollingshead (2010) believes that video game teams can provide a deeper
understanding of the coordination process of teams where some abilities lie in the avatar
(and thus stereotyping may occur) and some abilities lie in the ability of the player
16

(where individual expertise must be learned). Finally, video game teams can provide a
deeper understanding of the inconsistent findings on TMS in CMC contexts.
This Dissertation
This dissertation contains three studies looking at TMS in the video game League
of Legends, a MOBA and currently the most popular video game in the world. Study 1 is
a field study involving 16,499 players of the game and combines a survey administered to
the players and server logs for the players provided by the makers of the game. This
study look at the relationship between TMS and game outcome (as opposed to
performance), the relationship of team size and self-selection of teammates with TMS,
and the mediating role of social presence. Study 2 is a lab experiment comparing the
effects of text-chat versus voice-chat on TMS development, game performance, game
outcome, and social presence. It also looks at how the relationship between social
presence and TMS changes over time. Study 3, which is conducted concurrently with
Study 2, looks at SIP theory in video game teams and examines the role between
relational communication and TMS. Social presence and SIP theory will be talked about
in more detail in the portion of the dissertation presenting the actual studies.
Study 1 finds that TMS strongly predicts the outcome (win/loss) of a League of
Legends match. It also finds that self-selection of teammates is predictive of TMS but
team size is not. Finally, it finds that two dimensions of social presence, copresence and
comprehension, mediate the relationship between self-selection of teammates and social
presence.
17

Study 2 finds no differences between text-chat and voice-chat with respect to
TMS, game performance, and the individual dimensions of social presence (and is
inconclusive about game outcome due to a lack of variance). TMS again is positively
related to game performance, albeit more strongly at some time points than others. The
dimensions of social presence are also positively related to TMS, but the strength of this
relationship changes over time.
Study 3 finds no differences between text-chat and voice-chat in relational
communication. Some of the dimensions of relational communication are positively
related to TMS, but again, more strongly at some times than others. The results of Study
3 call SIP into question. Implications for this and the other two studies will be discussed
in the conclusion.

18

Study 1: Group Composition, Transactive Memory Systems, and Social Presence
Because of the mixed findings regarding the role of CMC in TMS development,
to assume that TMS can develop in all contexts should not be a given. Similarly, TMS is
more relevant and more likely to develop in some task types than others (Lewis &
Herndon, 2011). Finally, one should not assume that phenomena that occur in real-world
contexts will always map to a virtual context (Williams, 2010). Among other things,
different social architectures within virtual worlds may produce very different results.
Richter and Lechner (2009) found that some of the observed World of Warcraft
teams developed a TMS during the course of a one-time activity (specifically a multi-part
raid), but TMS was measured using subjective observer judgments. Riedl et al. (2012)
found that TMS developed in the RTS Travian, but this was a yearlong, mixed-motive
task. When teams engaged in battle, the winner was determined not by avatar combat but
by which team had statistical advantages going into a battle. This is very different from
the fast paced, psychomotor battle task in a MOBA. Yet Lewis and Herndon (2011) posit
that a TMS is not only relevant for Execute tasks but that Execute tasks are the most
relevant type of task for a TMS. Seeing as contest/battles tasks are of one of the two types
of Execute tasks, it is hypothesized:
H1: TMS develops in MOBA teams.
At the same time, while Riedl et al. (2012) found that TMS was predictive of
performance in Travian, McGrath (1984) notes that in the context of contest/battle tasks,
task performance and task outcome, while related, are not the same thing. In the case of
Travian, not only did Riedl et al. (2012) only look at team performance, only one team
19

per server could be declared the winner (after a year), thus giving a sample size of one in
the context of looking at the relationship between TMS and outcome. Just as Lewis and
Herndon (2011) posited that TMS was most relevant for Execute tasks, they also posited
that tasks with an objective (as opposed to subjective) outcome were more likely to
develop a TMS. Thus, it is hypothesized:
H2: In a MOBA, a stronger TMS is positively related to the outcome of a match.
Team Composition
Self-selective team formation. Much of the early research on TMS in small groups
focused on the effect of different ways of training group members (Liang et al., 1995;
Moreland, 1999; Moreland & Myaskovsky, 2000). This research found that group
members who trained together had greater TMS and performed better than group
members who had trained individually or with a different set of group members.
Research has suggested that teams composed solely of people who train together had
greater performance than those where only a subset of a team trained together (Lewis et
al., 2007). Other TMS research has found that prior familiarity with team members is
positively related to TMS development (Akgün, Byrne, Keskin, Lynn, & Imamoglu,
2005; Lewis, 2004). In some MOBAs, players can choose to play with specific
teammates or they can be randomly assigned to a team (or have a subset of teammates
self-selected and a subset randomly assigned). In order to play with specific people, one
must have prior familiarity, which is positively related to TMS, but at the same time, if
teams only partially trained together, only a partial TMS will be developed. Thus, it is
hypothesized:
20

H3: In a MOBA, there is a positive relationship between the percentage of team
members who self-selected their formation (as opposed to players randomly
assigned team members) and TMS.
Team size. Among the most frequently asked questions in small group research is,
“what is the ideal team size?” (Moreland, Levine, & Wingert, 1996). The answer to this
question is “it depends.” In the case of TMS research, some research suggests a positive
relationship between team size and TMS, and other research suggests a negative
relationship. Jackson and Moreland (2009) found that smaller student project teams
developed a stronger TMS than larger projects teams. They suggested this was because
smaller teams were able to communicate with one another better. Similarly, because
research has shown that social loafing and an overall decrease in productivity are more
likely to occur as group size increases (Steiner, 1972). Hollingshead et al. (2011) suggests
that increasing group size will lead to individual members to act in such a way that does
not actually reflect their full abilities, thus leading to problems with the validation
necessary for a TMS. In the case of online games, Xiong, Poole, Williams, and Ahmad
(2009) found that in the MMO, EverQuest II (EQII), group size had a negative total
effect on group performance, while not specifically addressing TMS. In another CMC
context, Lowry, Roberts, Romano Jr., Cheney, and Hightower (2006) found three person
CMC teams had better communication than six person CMC teams, and effective
communication is essential to the development of a TMS (Hollingshead, 2010;
Hollingshead & Brandon, 2003; Hollingshead et al., 2011; Jackson & Moreland, 2009).
21

On the other hand, some aspects of increased group size suggest a stronger TMS
in the context of video game teams. Xiong et al. (2009) found that in EQII that larger
groups had a greater aggregated expertise, which is a key element of a developed TMS
(Lewis & Herndon, 2011; Wegner, 1987). Riedl et al. (2012) had hypothesized that team
size would moderate the relationship between communication patterns and TMS in such
that these relationships would be stronger in smaller teams, but instead found the opposite
(albeit with a negligible effect size). Given the conflicting research, it is asked:
RQ1: Does team size affect the strength of a TMS in a MOBA team?
Social Presence
Lewis and Herndon (2011) note that when Wegner (1987) first proposed the
concept of transactive memory, TMS was composed of both knowledge structure and
knowledge-relevant transactive processes. Yet, they go on to note that most TMS
research focuses on the structure and not the process. The bulk of the research looks at
the cognitive antecedents and outcomes of TMS, but not cognitive processes. For this
reason, even longitudinal research fails to examine the dynamic nature of TMS
development. In the case of video game teams, there is an additional component to
cognitive processes not found in most other TMS reseach: interaction is often mediated
through virtual social actors (avatars and agents). One construct coming from the
mediated communication literature that addresses such psychological processes is that of
social presence. Not only should this psychological variable be explored in the context of
TMS given the focus on TMS structure over process, but it may also serve as explanation
of why team formation and team size may affect TMS in mediated contexts.
22

Lee (2004a) defines social presence as “a psychological state in which virtual
(para-authentic or artificial) social actors are experienced as actual social actors in a
sensory or non-sensory way” (p. 45). Biocca and colleagues (Biocca, Harms, & Burgoon,
2003; Biocca, Harms, & Gregg, 2001; Harms & Biocca, 2004) further explicate social
presence as consisting of three parts: copresence, psychological involvement and
behavioral engagement. Copresence is “the degree to which the observer believes he/she
is not alone and secluded, their level of peripheral or focal awareness of the other, and
their sense of the degree to which the other is peripherally or focally aware of them”
(Harms & Biocca, 2004, p. 247). Psychological involvement is “the degree to which the
observer allocates focal attention to the other, empathically senses or responds to the
emotional states of the other, and believes that he/she has insight into the intentions,
motivation, and thoughts of the other” (Biocca et al., 2001, p. 2). Behavioral engagement
is “the degree to which the observer believes his/her actions are interdependent,
connected to, or responsive to the other and the perceived responsiveness of the other to
the observer’s actions” (Biocca et al., 2001, p. 2).
It should be noted that this is a more contemporary definition of social presence
and slightly different from the original definition (Short, Williams, & Christie, 1976),
which was in the context of comparing different telecommunication media with each
other and with face-to-face interaction. The original definition referred to “the degree of
salience of the other person in the interaction and the consequent salience of the
interpersonal relationship” (Short et al., 1976, p. 65). Whereas the original Short et al.
definition was concerned only with the interpersonal aspect of mediated
23

telecommunication, the Lee (2004a) conceptualization of social presence comes from a
research community that concerns itself with cognition, affect, and behavior in virtual
environments (such as virtual reality, virtual world, and video games). Cognition has
always been the central component of TMS research; recent TMS research has suggested
affect plays an important role in the establishment of TMS (Gould, 2011); and McGrath
(1984) contrasted the behavioral product of Execute tasks with the conceptual product of
Choose and some Generate tasks. While the conceptualization of social presence this
study is using does assume that media that afford more communication cues should
exhibit greater levels of social presence (like the original conceptualization), the medium
is one of many possible variables that can affect feelings of social presence.
The dimensions of social presence are very related to antecedents of TMS.
Copresence would be a necessary condition: teammates must be mutually aware of one
another if they are to learn each other’s expertise and coordinate their actions based on
individual expertises. More generally, copresence is a defining feature of a small group
even in FtF contexts (McGrath, 1984). For psychological involvement, it is necessary to
devote attention to one’s teammates in order to either establish expertise or determine
characteristics of each other that will allow for using stereotypes, as they too can be used
(albeit not always correctly) to establish expertise (Hollingshead & Fraidin, 2003). Given
that communication is essential to the development of TMS (Hollingshead, 1998a;
Hollingshead & Brandon, 2003; Hollingshead et al., 2011; Yuan et al., 2010), it is
necessary for players to properly comprehend their teammates’ comunications and
intentions. They must mutually understand each other in order to engage in transactive
24

processes that allow for the creation of a shared, accurate, and validated group mental
model. For behavioral engagement, if cognitive interdependence is a necessary condition
for a purely cognitive task (Brandon & Hollingshead, 2004; Hollingshead, 2001),
behavioral interdependence should be necessary for a behavioral task such as video game
play. In addition, Gould (2011) found that the homegeneity of affect on the team had an
indirect effect on the development of the development of TMS, and thus affective
interdependence should also predict TMS. Thus, it is hypothesized:
H4: In a MOBA, there will be a positive relationship between social presence and
TMS.
Lee (2004b) proposes that social presence is a result of individuals being able to
attribute a theory of mind to virtual social actors. Social presence occurs when
individuals are able to simulate the mental states of avatars and agents. It would follow
then, if a game player already knows his or her teammates in advance, it would be easier
to simulate their mental states. Examples coming from video games and the more general
area of presence research would support such a view. Players exhibit greater levels of
presence and engagement when they are simply told they are playing against a human
opponent as opposed to a computer opponent (Lim & Reeves, 2010; Ravaja et al., 2006;
Weibel, Wissmath, Habegger, Steiner, & Groner, 2008). This is true not only for self-
reported measures of presence, but psychophysiological responses previously shown to
correlate with feelings of presence (such as increased skin conductance indicating
attentional processes). For example, Ravaja et al. (2006) found similar results, but more
importantly found that players showed increased self-reported and physiological
25

measures of presence when that human was a friend than when the human was a stranger.
Having a pre-established relationship with co-players created greater feelings of presence
than when playing with strangers. Taking Lee’s theory of mind explanation and the
presence research on knowing one’s coplayers, it is hypothesized that:
H5: In a MOBA, there is a positive relationship between the percentage of team
members who self-selected their formation (as opposed to players randomly
assigned team members) and social presence.
One could reason that as team size increases, it should become more difficult to
be aware all of one’s teammates at the same time and allocate attention to any individual.
At the same time, if motivation loss increases as team size increases (Steiner, 1972),
one’s teammates may be doing fewer things individually to be aware of and to allocate
attention to. Thus, increasing team size should lead to a decrease in copresence and
attentional allocation (a component of psychological involvement). Similarly,
miscommunications are more likely to occur as team size increases (Hollingshead et al.,
2011; Moreland et al., 1996), decreasing a player’s ability to gain insight into the
intentions, motivations, and thoughts of teammates (another component of psychological
involvement). Finally, if production loss increases as team size increases, individual
teammates may be less likely to respond to the actions of a player and the player may be
less likely to respond to the actions of his or her teammates with an increase in team size.
This would indicate a decrease in behavioral engagement.
On the other hand, Lee and Nass (2004) that when one heard a series of
persuasive statements in computer-synthesized voices, they felt greater levels of social
26

presence and were influenced more when different statements came from different voices
as compared to when the statements came from a single voice. The logic was, the more
voices one heard, the more virtual social actors one psychologically experienced, and
therefore the greater level of social presence. Thus, it is asked:
RQ2: Does team size affect the level of social presence players experience in a
MOBA team?
The rationale for establishing H4, H5, and asking RQ2 is that social presence has
been shown to mediate the relationship between variables in human-computer interaction
and computer-mediated communication. Many studies of avatar-mediated
communication and human-machine communication that concern the replication of
findings from FtF research have found social presence as a mediating variable.
Examples include similarity-attraction in voice interfaces (Lee & Nass, 2005) and
human-robot interaction (Lee, Peng, Jin, & Yan, 2006), the multiple source effect in
computer-synthesized speech (Lee & Nass, 2004), and perceptions of robots as
developing creatures (Lee et al., 2006). Social presence has also been shown to mediate
effects in contexts such as persuasive agents (Jin, 2011; Skalski & Tamborini, 2007), and
parasocial interaction with avatars (Jin, 2010). This is explained in Lee’s (2004b) theory
of mind explanation of social presence: when one can simulate the mental state of a
virtual social actor, he or she will respond and behave in the way he or she would with an
actual social actor. Thus it is hypothesized that:
27

H6a: In a MOBA, social presence will mediate the relationship between the
percentage of team members who self-selected their formation (as opposed to
players randomly assigned team members) and TMS.
H6b: In a MOBA, social presence will mediate the relationship between the
percentage of team size and TMS (assuming one exists).
A full diagram of the hypotheses and research questions (with the exception of H1, which
does not concern the relationship between variables) can be found in Figure 2.
Figure 2. Study 1 Hypotheses and Research Questions

Method
The current study focuses on the game League of Legends (LoL), a multiplayer
battle online arena (MOBA). LoL was chosen for two reasons. First, as of the Fall of
2012, it is the most popular video game in the world (Riot Games, 2012). Second, its
operator, Riot Games Inc., agreed to assist the research team in both providing server-
side behavioral data and soliciting a random sample of its user base to participate in the
survey described below.
In LoL, players take on the role of summoners. For every individual “match,”
summoners select one “champion” to control. Based on the performance of the
28

champions in matches, players earn “influence points,” or IP. IP allow players to
purchase in-game items and more powerful champions for future matches. The structure
of a match involves two teams of champions competing against each other, and the match
ends when one team wins by destroying their opponents’ “nexus” (essentially, the home
base). LoL and the MOBA genre in general, is a “non-persistent world.” Unlike an MMO,
where a player usually logs into a game and returns to the spot he or she left off (or at
least resumes the same character), in LoL, once a match is over, a player can choose any
champion he or she wants and play with any team he or she wants in the next match.
There are a three general types of LoL matches: PvP (a team of human players
versus another team of the same size of human players)
7
, co-op vs. AI (a team of human
players versus a team of the same size of computerized “bots”), and custom games
(where teams could vary in size and contain humans and “bots” together). PvP though is
by far the most prevalent type of game play.
In PvP games, players can choose to play on a three-member team against another
three-member team (3v3) or a five-member team against another five-member team
(5v5). In PvP games, the tasks in a 3v3 game and a 5v5 game are the same in the sense
that the goal is to destroy the enemy’s nexus. However, the “maps” on which the games
are played differ slightly to reflect the fact that three players can cover less terrain than
five players can.
In PvP games, players can choose to play an “arranged match,” where they
choose their teammates, or a “solo match,” where they are randomly assigned teammates

7
PvP is an abbreviation for player versus player
29

by the game.
8
Opposing teams are matched by the game, trying to balance the skill levels
of the two teams. Together, this provides a framework for testing the specified
hypotheses and research questions.
Participants and Procedure
Using a simple-random sample of all LoL accounts that had played at least one
match in the preceding month, Riot Games invited 113,579 players to participate in a
web-based survey about their experiences with LoL. Selected players received the
invitation via email and would click on a link to be directed to the survey, which was
hosted by the researchers. In compensation, players would have their earned IP doubled
for the next four LoL matches that they won. In one week, 25,996 (22.9%) of the emails
were opened and 22,521 complete responses were collected. After eliminating duplicate
responses, responses coming from invalid links (likely non-invitees), and responses
completed “too fast” (pretesting indicated twelve minutes would be a reasonable
minimum completion time given the survey length), 18,627 valid responses were
collected. This was a response rate of 16.4% of all emails sent and 71.7% of all emails
opened.
In addition, for the purpose of this study, players were eliminated (as determined
by server-side data) if their last match was not an official PvP match (as the survey
measures would ask a player about their last match and the server data only provided
information about the last match and career statistics). Furthermore, players were also

8
The term random is used loosely, as the game does match players of similar career
performance.
30

asked if they used multiple accounts, and if so, whether or not they received the survey
invite from their main account. If the player responded that he or she had received the
survey from an alternate account, the response was not included in the present study. The
reason for this is that players’ alternate (or “smurf”) accounts may not reflect their level
of experience or playing style, and thus the server data associated with the survey data
may threaten internal validity. After filtering out those that did not meet these two
inclusion criteria, the final sample size was 16,499.
Measures
Server-side measures.
Team outcome. Riot provided the information as to whether or not a player won
his or her last PvP match.
Team formation. Team self-selection was operationalized as the percentage of
teammates a player prearranged to play with. This was calculated by taking server-side
measures of how many other players were prearranged to be on a person’s team in his or
her last game played (a number ranging from 0 to 2 for 3v3 games and a number ranging
from 0 to 4 for 5v5 games) and dividing this by three or five (depending on whether the
game was 3v3 or 5v5). Because of the sheer size of LoL’s player base, it is highly
unlikely that two players have ever played together before if it was not prearranged.
Team size. Server-side data identified whether a player’s last game played was a
3v3 game or a 5v5 game.
Controls. For the purpose of analyzing predictors of the likelihood a team wins a
match, two server-side measures were collected: number of PvP matches played and
31

career PvP winning percentage. Because the number of PvP matches played ranged
anywhere from 1 to 1,749, with a mean of 333.56 and standard deviation of 241.88, this
number was divided by 100 in order to better interpret odds ratios.
Survey measures. Survey measures can be found in Appendix A.
Transactive memory system. TMS was measured using nine questions adapted
from Lewis (2003) pertaining to the three dimensions TMS: specialization, credibility,
and coordination. Together, the nine questions were aggregated into a single measure (M
= 3.40, SD = 0.82, Cronbach’s α = .87). While this measure is normally a team level
aggregation of individual members’ responses, this study only had survey measures from
one member of a team. It was not possible to survey all team members, as the random
sample was of individuals and not teams). However, Lewis’s validation found a high r
wg

(a measure of within-group agreement), suggesting that an individual measure would be
relatively consistent with a team aggregated measure.
Social presence. Social presence was measured using five items from the
Networked Minds Social Presence Inventory v. 1.2 by Biocca and Harms (2003), which
is generally consistent with the dimensions of social presence they had explicated
previously (Biocca et al., 2003). Players were asked to think about the last game they had
played and to what extent they agreed or disagreed with a set of statements, each of
which corresponded to one of the five subscales from the inventory: copresence (M =
2.68, SD = 1.25), perceived attentional engagement (M = 4.08, SD = 0.86), perceived
emotional contagion (M = 3.26, SD = 1.17), perceived comprehension (M = 3.66, SD =
1.10), and perceived behavioral interdependence (M = 3.95, SD = 0.98). Asking multiple
32

questions from each subscale would have been preferred for purposes of reliability, but
survey space was limited because it was being used for multiple studies conducted by
other researchers. The specific questions for social presence were selected based on high
factor loadings in previous studies (Biocca et al., 2001; Harms & Biocca, 2004) and high
face validity and clarity as determined by seventeen video game scholars.
Data Analysis Procedures
With a sample size as large as the present study, the statistical power provided
was likely to produce highly significant statistics for most tests conducted, regardless of
the substantive meaning. Therefore, standardized measures were used to determine if the
significant effects were meaningfully significant. In order to consider a hypothesis
supported, the statistical analysis needed to not only be statistically significant but also
needed to meet the minimum criteria of a small effect size.
Cohen (1988) provided the following indexes and values for measuring effect size
in social science research: d indicates small, medium, and large effect sizes at values of
0.20, 0.50, and 0.80 (respectively) for t-tests, and r indicates small, medium, and large
effect sizes at values of 0.10, 0.30, and 0.50 (respectively) for correlation.
Results
Common Method Variance
Before analyzing data, in order to test for common method variance, a Harman’s
one-factor test (Podsakoff & Organ, 1986) was conducted. This test loads all survey
items onto a single, unrotated factor using exploratory factor analysis, and if this single
factor accounts for less than 50% of the total variance, common method variance should
33

not be a major concern of the researcher. For the survey measures of TMS and social
presence, only 32.87% of the variance was accounted for by the single factor.
The Harman one-factor test is not without its criticisms (Podsakoff, MacKenzie,
Lee, & Podsakoff, 2003). While not meeting the 50% threshold, some of the variance of
the single-factor may be common method variance, nor does the Harman one-factor test
partial out any common method variance. In addition, the more measurements there are,
the less likely a single factor will account for 50% of the variance. An alternate procedure
suggested by Lindell and Whitney (2001) is to either include a “marker variable” that
should be theoretically unrelated to at least one variable of interest and assume that any
correlation between the marker variable the unrelated variable is the common method
variance, which can then be partialled out. Alternatively, the lowest positive correlation
in the correlation matrix of all measurements taken during the survey administration can
be used in lieu of the marker variable. In the case of the survey administration for Study
1, there was a correlation of r = .001 between two variables unrelated to this study.
Between the results of the Harman one-factor test and the Lindell and Whitney technique,
common method variance was ruled out.
Data Analysis
To test H1 that TMS could develop in MOBA teams a one-sample t-test was
conducted. Results indicated that the average level of TMS (M = 3.40, SD = 0.82)
differed significantly from the scale midpoint of three (“neither agree nor disagree”). This
difference of 0.40, 95% BCa CI [0.390, 0.418] was significant, t(15767) = 61.60, p <
34

.001, d = 0.49, just missing the threshold for a medium size effect. Thus, H1 was
supported.
To test H2, that there would be a positive relationship between TMS and the
outcome of a match, a hierarchical binary logistic regression was conducted, where a
losing outcome was coded as 0 and a winning outcome was coded as 1. The number of
PvP matches played and career PvP winning percentage were entered into the regression
before entering TMS. The odds ratios and model fit can be found in Table 1. After
controlling for career PvP experience and winning percentage, it was found that for every
one-unit increase on the five-point TMS scale, a team was 2.23 times more likely to win a
match. TMS accounted for anywhere from 7% to 12% of the variance in whether or not a
team wins a match (depending on the R
2
statistic used). Thus, H2 was supported.
35

Table 1. Odds Ratios and Fit of the Model Predicting Whether or not a LoL team Wins a
Match

Odds Ratio and CI
Step 1 Step 2
Total career PvP matches played
(divided by 100)
1.05
[1.034, 1.066]
1.05
[1.038, 1.074]
Career PvP win percentage
(multiplied by 100)
1.05
[1.043, 1.055]
1.05
[1.047, 1.069]
TMS
2.23
[2.119, 2.349]

Model R
2
R
L
2
= .02
R
CS
2
= .03
R
N
2
= .03
χ
2
(2) = 333.44
R
L
2
= .09
R
CS
2
= .11
R
N
2
= .15
χ
2
(3) = 1525.71
ΔR
2

ΔR
L
2
= .07
ΔR
CS
2
= .08
ΔR
N
2
= .12
Δχ
2
(1) = 1192.27
All odds ratios and χ2 have p < .001. All confidence intervals are 95% BCa on 1000 bootstrap samples.
To test H3, that the percentage of self-selected teammates would be positively
related to TMS, a bivariate correlation was conducted. Results indicated a small-to-
medium size relationship between familiarity and TMS, r(15766) = .19, p < .001, 95%
BCa CI [.173, .201]. Thus, H3 was supported.
36

To answer RQ1, as to whether three person teams or five person teams would
have greater levels of transactive memory, an independent-samples t-test was conducted.
While TMS in three person teams (M = 3.56, SD = 0.83) was significantly different from
five person teams (M = 3.39, SD = 0.82), t(15147) = 7.91, p < .001, mean difference 95%
BCa CI [.125, .209], the effect size, d = .13, did not meet Cohen’s threshold for a small
effect. Thus, in response to RQ1, it is concluded that there is no meaningful difference in
TMS for three person teams as opposed to five person teams.
To test H4, that social presence would be positively related to TMS, a
simultaneous multiple regression was conducted with the five dimensions of social
presence as individual predictors. The full model was found to be significant with a large
effect-size, adjusted R
2
= .31, F(5, 15143) = 1360.95, p = .001. However, only
copresence (b = 0.19, 95% BCa CI [0.183, 0.204], β = 0.29, p = .001, r = .31) and
comprehension (b = 0.26, 95% BCa CI [0.242, 0.268], β = 0.34, p = .001, r = .35) met
any of Cohen’s thresholds for effect sizes (in these cases, medium effect sizes). The
remaining three dimensions of attentional engagement (b = 0.06, 95% BCa CI [0.039,
0.072], β = 0.06, p = .001, r = .06), emotional contagion (b = ˗0.01, 95% BCa CI [˗0.023,
˗0.003], β = -0.02, p = .014, r = ˗.02), and behavioral interdependence (b = 0.07, 95%
BCa CI [0.054, 0.080], β = 0.08, p = .001, r = .09) did not meet Cohen’s threshold for a
small effect size. Thus, H4 was partially supported.
To test H5, that the percentage of teammates self-selected would be positively
related to social presence, bivariate correlations were conducted for each of the individual
dimensions of social presence. An intercorrelation matrix of TMS and the individual
37

dimensions can be found in Table 2. Results indicate a small-to-medium size relationship
between self-selection and copresence and a small-to-medium size relationship between
self-selection and message understanding. The remaining three dimensions of social
presence did not meet Cohen’s threshold for a small effect. Thus, H5 was partially
supported.
38

Table 2. Means, Standard Deviations, and Intercorrelations of Team Formation and
Dimensions of Social Presence
M SD 1 2 3 4 5
1. Team formation
(percentage self-selected)
.30
[.297, .308]
.35
1

2. Copresence
2.68
[2.659,
2.697]
1.25
.19
*
[.171,
.203]

1
3. Attentional engagement
4.07
[4.062,
4.089]
0.86
.05
[.034,
.066]
.15
*
[.137,
.171]
1
4. Emotional contagion
3.27
[3.247,
3.286]
1.17
.10
[.080,
.111]
.23
*
[.216,
.251]
.12
*
[.099,
.133]
1
5. Comprehension
3.66
[3.636,
3.676]
1.10
.18
*
[.164,
.196]
.34
**
[.310,
.350]
.30
*
[.282,
.316]
.11
*
[.091,
.127]
1
6. Behavioral
interdependence
3.95
[3.937,
3.968]
0.97
.06
[.043,
.072]
.11
*
[.092,
.130]
.33
**
[.316,
.352]
.18
*
[.159,
.195]
.19
*
[.171,
.207]
All correlations have p < .001. All confidence intervals are 95% BCa on bootstrap 1000 samples.
*
Meets Cohen’s threshold for a small effect
**
Meets Cohen’s threshold for a medium effect
39

To answer RQ2, as to whether three person teams or five person teams would
have greater levels of social presence, a series of independent-samples t-tests were
conducted for each of the five dimensions of social presence. Statistics for each t-test can
be found in Table 3. None of the mean differences had a d that met Cohen’s threshold for
a small effect size. Thus, in response to RQ2, it is concluded that there are no differences
in social presence for three person teams as opposed to five person teams.
Table 3. Means, Standard Deviations, and Differences in Social Presence Based on Team
Size

3 person team 5 person team
t df d CI for mean differences
M SD M SD
Copresence 2.86 1.30 2.65 1.24 6.51 15147 0.11 [0.146, 0.277]
Attentional engagement 4.10 0.88 4.07 0.86 1.35 2103.52 0.06 [˗0.016, 0.074]
Emotional contagion 3.29 1.18 3.26 1.17 0.88 15147 0.01 [˗0.032, 0.087]
Comprehension 3.76 1.11 3.64 1.10 4.14 15147 0.07 [0.064, 0.169]
Behavioral interdependence 3.89 1.01 3.96 0.96 ˗2.63 2089.98 0.12 [˗0.119, ˗.017]
All t-tests have p < .001. All confidence intervals are 95% BCa on 1000 bootstrap samples.
To test H6a, that social presence would mediate the relationship between team
formation and TMS, Model 4
9
of the PROCESS custom dialog for SPSS was used

9
Model 4 tests for up to ten mediators running in parallel and operates based on the ideas
outlined byPreacher and Hayes (2004, 2008).
40

(Hayes, 2013).
10
The five dimensions of social presence were entered in simultaneously
as mediators with team familiarity as the predictor and TMS as the outcome. Preacher
and Kelley (2011) note that there are no interpretable measures of indirect effects when
measures come from arbitrary units (e.g. latent scales as opposed to observable
measurements). They propose an analogue to R
2
called κ
2
, but this only measures indirect
effects in simple single mediator models. Thus, in testing the mediation hypothesis of
H6a, to determine whether to consider a hypothesis supported, an r was calculated for t-
tests (for coefficients) and z-scores (for Sobel tests) when PROCESS did not explicitly
provide r.
Results of PROCESS’s mediation analysis can be found in Table 4. While there
are small indirect effects of self-selection on TMS through copresence and
comprehension, there are negligible indirect effects through attentional engagement,
emotional contagion, and behavioral interdependence. In addition, the direct effect of
familiarity on TMS did not meet Cohen’s criteria for a small effect size (as judged by the
r for the z score of the Sobel test), indicating total mediation. Thus, H6a was partially
supported. Because no relationship was found between group size and either TMS or
social presence, H6b was not tested, as two of the conditions of mediation are that the
predictor has a relationship with the outcome as well as the mediator (Baron & Kenny,
1986). A final diagram of the results can be found in Figure 3.

10
Structural equation modeling could not be used because only having a single item for
each dimension of social presence would result in a model being underidentified.
41

Table 4. Tests for Social Presence as a Mediator between Team Formation and
Transactive Memory Systems
X M M Y Indirect effect
M a r b r
t
ab CI ab
cs
P
M
R
M
z r
z
κ
2

Copresence 0.66 .19
*
0.19 .30
**
0.12
[0.110,
0.136]
0.05 0.28 0.75 20.00 .16
*
.08
Attentional
engagement
0.12 .05 0.06 .06 0.007
[0.004,
0.010]
0.003 0.02 0.04 4.88 .04 .01
Emotional contagion 0.32 .10 ˗0.02 ˗.03 ˗0.005
[˗0.008,
˗0.002]
˗0.002 ˗0.01 ˗0.03 ˗3.00 ˗.02 <.01
Comprehension 0.56 .18
*
0.25 .34
**
0.14
[0.125,
0.152]
0.06 0.32 0.85 20.02 .16
*
.08
Behavioral
interdependence
0.16 .06 0.07 .09 0.01
[0.008,
0.014]
0.005 0.02 0.06 5.98 .05 .01
Total indirect effect 0.27
[0.253,
0.295]
0.12 0.63 1.66
Total effect: b = 0.44, r = .19
*

Direct effect: b = 0.16, r
t
= .08
X = Team formation (percentage self-selected)
Y = TMS
All tests have p < .01. All confidence intervals are 95% BCa on 1000 bootstrap samples.
Note: κ
2
is calculated for a simple mediation model, as it is not calculable for multiple mediator models.
*
Meets Cohen’s threshold for a small effect size
**
Meets threshold for a medium effect size

42

Figure 3. Results of Study 1

After data analysis was complete, it was suggested to the researcher that one
explanation for the relationship between team formation and TMS was that teams that
self-selected would be more likely to use voice-chat, which might improve TMS. While
no data existed on the mode of communication used in the previous match, players had
also been asked about how frequently they used three different modes of communication
(for the purposes of a different study): in-game text-chat, voice-chat, and face-to-face (in
the same room) communication. However, none of the three modes of communication
had a bivariate correlation with TMS that reached Cohen’s threshold for a small effect.
For text-chat: r(15493) = .04, p < .001, 95% BCa CI [0.023, 0.056]. For voice-chat:
r(15493) = .09, p < .001, 95% BCa CI [0.073, 0.102]. For face-to-face communication:
r(15493) = .09, p < .001, 95% BCa CI [0.075, 0.108].
Discussion
This study established that, on average, TMS could develop in MOBA teams,
despite the fact that they are rich in socioemotional communication. Similarly, in a
43

Contest/battle task, such as a MOBA, TMS predicts outcome. In fact, it strongly predicts
outcome: with an odds ratio of 2.23 (after controlling for player experience and prior win
percentage), a team that would rate the highest on the 5-point TMS scale would be almost
five times as likely to win compared to a team that rated at the midpoint, and 25 times as
likely to win compared to a team that rated at the lowest point.
It was hypothesized that team formation would predict TMS and that this
relationship would be mediated by social presence. As predicted, the percentage of team
members who self-selected their formation (as opposed to randomly assigned team
members) was positively related TMS. For social presence, which was originally
hypothesized unidimensionally but due to measurement issues was addressed
multidimensionally, only the dimensions of copresence and perceived comprehension
mediated the relationship between team formation and TMS. These two dimensions
completely mediated the relationship to the extent that a direct effect remained but of
negligible effect size. The other three dimensions of social presence, attentional
engagement, emotional contagion, and behavioral interdependence, had no relationship
with team formation or TMS.
In regards to the question of the relationship between team size, TMS, and social
presence, there were no differences in TMS or social presence between 3v3 teams and
5v5 teams. In addition to the possibility that indeed no relationship exists, there are three
possible explanations. First, there may not be sufficient differences between 3 person and
5 person teams (although Lowry et al., 2006 found differences in communication quality
between 3 and 6 person teams). Second, in the case of LoL, because both 3v3 and 5v5
44

involved evenly matched teams as well as different game maps (albeit similar and with
the same goal), the tasks were sufficiently different that they were an inappropriate
comparison. Finally, group size may have indirect effects on social presence and TMS,
but the indirect effects cancel each other out. Indeed, Xiong et al. (2009) found that group
size had indirect effects on team performance, but since some of the effects were positive
and some were negative, the total effect was negligible in size.
Limitations
There were a handful of limitations for this study. As previously mentioned, the
differences between 3v3 and 5v5 matches may have either not differed enough (in the
sense that three person teams may be too similar to notice an effect) or differed too much
(in the sense that the tasks were too different to compare the two). Another limitation was
that while the survey measures described teams, the survey used measures taken from
individuals as a proxy for their teams. While the measures used have been shown to have
strong inter-rater agreement, variance may have been gained or lost due to the use of
individuals instead of the aggregation of individuals.
Finally, even Cohen (1988) concedes that using his thresholds for effect sizes is
almost as arbitrary as null hypothesis significance testing. However, some criteria were
needed to determine whether hypotheses should be supported. Bootstrapping showed that
on many occasions, when an effect size did not meet Cohen’s criteria, the 95% BCa
confidence interval did include 0, which is another way to conclude that the null
hypothesis is a better explanation than the alternative hypothesis. However, in some
45

cases, the negligible effect sizes did not include 0 in the 95% BCa confidence interval,
which would suggest an effect truly did exist (however small it may be).

46

Study 2: Communication Channel, Transactive Memory Systems, and Social
Presence
While Study 1 had the advantages of being a field study of actual League of
Legends players, because it used a one-time survey (combined with some quasi-
experimental design), it lacked the ability to a) demonstrate causality and b) demonstrate
change over time. Among the limitations of much of the early work on CMC channel
effects was a lack of looking at how differences between CMC and FtF may diminish
over time (Hollingshead, Mcgrath, & O'Connor, 1993; Walther, 1992). One of the key
differences between the two modes of communication is the rate at which information is
transmitted; with enough time, CMC groups can have group performance and
interpersonal interaction reaching (and possibly even exceeding) that of FtF groups.
Channel Effects and TMS
The earliest TMS experiments comparing the two channels (Hollingshead, 1998b)
suffered from looking at only one point in time and did not see if dyads would adapt to
limited non-verbal cues. Other longitudinal, correlational TMS research might suggest
that groups adapt to limited non-verbal cues, as the relationship between communication
channel and TMS only existed at earlier time points (Jackson & Moreland, 2009; Lewis,
2004) in the study. Riedl et al. (2012) found that perceived communication channel
richness was positively related to TMS, although perceptions of medium richness may in
fact be more of a social construction than an actual property of a medium (Fulk, Schmitz,
& Steinfeld, 1990; Fulk, Steinfeld, Schmitz, & Power, 1987).
47

There are two reasons why there may be an interaction between medium and time
in the development of TMS. First, the rate that information is transmitted is slower in
CMC. Therefore, it may take longer to learn the expertise of TMS and enact coordination
mechanisms. Second, Ashleigh and Prichard (2012) argue that trust is not only a
component of a TMS, but it is also an antecedent, and furthermore, this extends not only
to the cognitive trust in the abilities of one’s teammates but also the affective trust in the
benevolence and integrity of one’s teammates. According to SIP (Walther, 1992),
initially affective trust may be higher in FtF teams that CMC teams due to the
paralinguistic and nonverbal cues, but over time, CMC teams develop verbal strategies to
compensate for the lack of nonverbal cues and reach levels of affective trust that equal
(or possibly exceed) that of FtF teams. Empirical evidence supports this convergence for
affective trust (Walther, 1995; Walther & Burgoon, 1992; Wilson, Straus, & McEvily,
2006) and finds similar patterns with cognitive trust (Wilson et al., 2006). Thus, if trust is
slower to develop in CMC teams than FtF teams, TMS should also be slower to develop.
In the case of video game teams though, the two primary communication channels
are not CMC and FtF but text-chat and a combination of text- and voice-chat (Williams et
al., 2007). For the purposes of this study though, to isolate channel effects, text-chat and
voice-chat (without text-chat) will be compared. By the logic of SIP, that time is
necessary to adapt to the lack of nonverbal cues and the notion that text communication is
transmitted more slowly than oral communication, it should follow that for zero-
acquaintance groups, with no preexisting communication patterns, voice-chat should
48

produce higher levels of TMS than text-chat initially. However, over time, TMS levels
should converge between the two conditions. Thus, it is hypothesized:
H1: In newly formed MOBA teams, a) teams communicating via voice-chat will
have greater initial levels of TMS than teams communicating via text-chat, and b)
over time, the differences in TMS between voice-chat teams and text-chat teams
will diminish.
11

Task Outcome and Performance
While Study 1 had the strength of demonstrating that in the case of a Type 7 task,
TMS predicted outcome (win or loss), as opposed to past research that has only looked at
the relationship between TMS and performance (metrics of achievement that may be high
even in a loss or low even in a win), this study will attempt to test both simultaneously.
Thus, it is hypothesized:
H2: In MOBA teams, TMS will be positively related to both a) performance and
b) outcome.
As such, because TMS will be slower to develop in voice-chat teams than text-chat
teams,
H3: In newly formed MOBA teams, a) teams communicating via voice-chat will
have greater initial levels of performance than teams communicating via text-chat,
and b) over time, the differences in performance between voice-chat teams and
text-chat teams will diminish, and c) teams communicating via voice-chat will

11
A MOBA match (which upon completion will be considered a time point) tends to last
30-60 minutes.
49

have better initial outcomes than teams communicating via text-chat, and d) over
time, the differences in outcomes between voice-chat teams and text-chat teams
will diminish.
Social Presence
While SIP is a theory about interpersonal relations, and social presence addresses
cognition, affect, and behavior, the SIP should still provide some insight into social
presence. SIP has posited, and empirical evidence has supported, that while initially there
is greater impression development in FtF than CMC, these levels become equal over time
(Walther, 1992, 1993). Thus, one would expect a similar pattern to hold with social
presence. However, in Study 1, hypotheses about social presence were only supported for
some dimensions of social presence. The hypotheses assumed that social presence would
serve as a mediator, uniformly across all of its dimensions. It hypothesized about social
presence unidimensionally even though it measured it multidimensionally. It assumed
that relationships with social presence would be stable regardless as to how long a team
has played together. Thus, any further hypotheses and research questions should a priori
be set out separately for the six individual dimensions of social presence (copresence,
attentional allocation, perceived message understanding, perceived affective
understanding, perceived affective interdependence, and perceived behavioral
interdependence). Thus, it is hypothesized that:
H4: In newly formed MOBA teams, teams communicating via voice-chat will
have greater initial levels of a) copresence, b) attentional allocation, c) perceived
message understanding, d) perceived affective understanding, e) perceived
50

affective interdependence, and f) perceived behavioral interdependence than
teams communicating via text-chat, but over time, the differences in g)
copresence, h) attentional allocation, i) perceived message understanding, j)
perceived affective understanding, k) perceived affective interdependence, and l)
perceived behavioral interdependence between voice-chat teams and text-chat
teams will diminish.
In the case of TMS, the relationship strength of different dimensions may vary
over time. TMS development has been described a series of learning cycles, where over
time, as tasks are repeatedly performed with a group, groups first learn the basics of
everybody’s knowledge, skills, and abilities and then later learn generalizable
mechanisms for coordinating previously unencountered problems (Lewis et al., 2005).
Thus, it may be necessary to pay close attention to teammates when TMS is first being
developed, in order to learn expertise, but once more abstract understanding of teammates
has been developed, TMS makes it such that one does not need to pay as close attention
to teammates. Similarly, while behavioral interdependence might be necessary for TMS
to first develop, once there are coordination mechanisms in place, teammates may be able
to act more independently because of an inherent trust in other’s abilities to carry out
responsibilities. Therefore, while much research has shown that social presence serves as
a mediator as to when real world behaviors occur in response to virtual social actors, it
has failed to address whether mediation processes change over time. Thus, it is asked:
51

RQ1: In newly formed MOBA teams, how does the relationship between different
dimensions of social presence and TMS change as teams play more matches
together?
RQ2: In newly formed MOBA teams, which dimensions of social presence serve
as mediators between communication channel and TMS as teams play more
matches together?
Method
This study was 2 × 3 mixed design experiment, where communication channel
(text-chat versus voice-chat) was between-subjects and time (Match 1, Match 2, Match 3)
was within-subjects. This study also used the game League of Legends, but unlike the
first study, this study used custom games. This was chosen for two reasons. First, in
custom games, players can play against computerized “bots” set at a constant challenge
level, whereas there would be no way of knowing the experience level of opponents if
players played against humans. Second, in custom games, team sizes can vary anywhere
from a single players to five players, thereby allowing research sessions to run as long as
three participants (the minimum number of people to qualify as a small group) showed up
for a given session. Experimental research on small groups is plagued by the problem of
getting enough people to sign up for a given session and then having those signed up
actually come to the lab (Wittenbaum, 2012).
Participants
Potential participants were recruited from the University of Southern California
through announcements in communication classes and via snowball sampling. During
52

recruitment, the potential participants were informed that if they qualified for and
participated in the actual lab experiment, they would receive either extra credit or a $40
Amazon.com gift certificate if they were not enrolled in a class offering extra credit.
They were also informed that those who qualified for and participated in the actual lab
experiment would have their names entered in a drawing for an iPad and that based on
game performance participants would have the opportunity to have their names entered in
the drawing additional times and/or win up to $150.
Potential participants received a link, where they filled out an online
prequestionnaire to make sure participants qualified. There were three straightforward
qualification criteria: they must have been USC undergraduates (this facilitated
compensation logistics), they must have been at least 18 years of age (so minor assent
was not necessary), they must not have had the lead researcher as an instructor previously
(both to avoid coercion and possible contamination if he had spoken about it in other
classes). This was made clear to participants when the announcement was made.
A fourth qualification existed that was simply explained to potential participants
at the time of recruitment as “certain levels of video game experience.” In the
prequestionnaire, there was a question as whether or not they had ever played a series of
games. All of the games listed were MOBAs. If a respondent had indicated that he or she
had ever played one of these games, the person was informed that he or she was
ineligible to participate in the study. People with MOBA experience were excluded so as
to not give any team too much of an advantage, in addition to the fact that anecdotal
53

experience has suggested that more experienced players can become frustrated and harass
inexperienced players in LoL.
If a potential participant did not qualify due to any of the four criteria, he or she
was informed of this. If a potential participant did qualify, he or she would be directed to
a list of open three-hour time slots where he or she could sign up to participate. Up to five
people could sign up for a time slot, and participants were not able to see the names of
others who had already signed up for the time slot. A session would run if at least three of
the people signed up showed up to the lab. Thus, any session that ran would have three to
five people.
The final sample contained sixty-seven individuals (34% male, 66% female) in
groups ranging in size from 3 to 5 players, for a total of 20 teams (fourteen were of size
3, five were of size 4, and one was of size 5). The number of participants who signed up
for and showed up for a given slot determined the team size. Teams were randomly
assigned to either the text-chat or the voice-chat condition (10 per condition). All players
utilized the Summoner’s Rift map, which is typically used for 5v5 matches, but teams
only competed against the same number of bots as their own team size (a 3-person team
competed against three bots).
To test for equivalence between the randomly assigned conditions, the percentage
of females on a team were calculated. There was no significant difference between the
percentage of females in the text-chat condition (M = 75.12%, SD = 21.02%) and the
text-chat condition (M = 54.12%, SD = 32.69%), t(18) = 1.71, p = .11. In addition, at the
time participants signed up for the study, they completed the Game Playing Skill scale by
54

Bracken and Skalski (2006). Despite random assignment, there was a significant
difference in the average team member skill for the text-chat condition (M = 3.42, SD =
0.48) and the voice-chat condition (M = 4.01 SD = 0.62), t(18) = -2.40, p = .03, but post-
hoc analysis found no significant correlations between average team member skill and
any of the other variables in the study.
Procedure
After filling out the prequestionnaire and coming to the building where the
research lab was housed, the experimenter accompanied the participants to a room and
each sat at one of five computers. Dividers existed between the computers, and
participants would not be able to see one another while the session ran to isolate
communication to text or voice (and not any form of FtF communication). In addition,
participants were asked not to communicate with one another except when instructed to.
Once seated, participants were explained how the session would run. They were
told that they would be playing three consecutive matches of the game League of Legends
and that after each match, they would fill out questionnaires. They were informed that for
every match their team won, everybody on the team would have his or her name entered
in the drawing for the iPad an additional time, and that the team (out of all teams) that
had the best kill-to-death ratio
12
for each individual match would earn $50 per player (and
since there would be three matches, up to $150). The extrinsic incentives to winning,

12
The kill-to-death ratio is the number of times a member of one’s team kills an opposing
champions divided by the number of times one’s champion dies.
55

maximizing killing opposing champions, and minimizing the number of team deaths was
to create the interdependence that is a prerequisite for TMS.
After having providing an overview of how the next three hours would go,
participants were provided verbal instructions how to create a LoL account. After creating
the account, they would place on headsets and do the Basic Tutorial that is part of LoL so
they could get a sense of the rules of the game and how to control champions. After all
participants finished the tutorial, they took off the headsets for further verbal instructions
and were given five minutes to read up on the champion they would be assigned to play.
Before the players came into the lab, the experimenter had pulled up the profile of
a different champion on each of the five computer screens. Just as every session had been
randomly one of the two conditions (text or voice), every computer in a given session
was randomly assigned one of the five champion roles in LoL: Assassin, Carry, Fighter
Mage, or Tank. No computer was assigned the same champion type. Only ten LoL
champions are available as free-to-play in a given week (when a champion is not free-to-
play, it must be “bought” using either influence points accumulated during game play or
Riot points purchased using real-world money), and these champions rotate every week.
Therefore, during the span of the research study, the specific champion corresponding to
a given role varied. When multiple champions of a given role were available during a
given week, the champion that was rated the least difficult to play was chosen.
Whereas the free-to-play champions rotate every week, the possible champions
that could be assigned to a bot in a custom game stay the same. Thus, players would
compete against a bot team consisting of the same size as the number of players in the lab
56

and of the same champion roles (e.g. a participant team consisting of a Carry, Fighter,
and Tank would play against a team with a Carry bot, a Fighter bot, and a Tank bot). The
opposing champions corresponding to a given champion role remained the same
throughout the course of the study.
In the text-chat condition, players were told they would be playing the game and
were encouraged to communicate via text and work together with their teammates to
destroy their enemy’s nexus. They were given explicit instructions how to use LoL’s in-
game text-chat. The incentives of winning and kill-to-death ratio were reiterated. They
then went about playing their first match. The volume on the headsets was at such a level
that they could not hear anything other than game play and the microphones on the
headsets remained off at all times.
In the voice-chat condition, players were told they would be playing the game and
were encouraged to communicate via voice and work together with their teammates to
destroy their enemy’s nexus. The microphones on the headsets were then unmuted, and a
sound check was done to ensure that the microphones picked up participants’ voices. The
voices of the players were picked up and transmitted to the headsets of other players
using a voice-over-IP (VOIP) software program called Ventrilo with an open
microphone. Due to the volume of the headsets, participants were only able to hear each
other through the mediated communication channel and not across the room.
After completing the first match, participants filled out a questionnaire containing
measures of TMS and social presence (as well as measures used in Study 3). Teams
would play two more matches together and fill out the same questionnaire after each
57

match. After the third questionnaire was filled out, the experimenter thanked the
participants and dismissed them. Later, the experimenter would use LoL’s “Match
History” feature (which provides detailed statistics about the last fifteen matches a
summoner has participated in) to look up the statistics of the participants (whose
summoner names had been recorded) for the matches that occurred in the lab.
In the span of time between when a session ran and when the statistics were
retrieved, one participant played more than fifteen matches, and thus his or her statistics
were irretrievable. Because this study is interested in group performance, and because the
Match History feature only provides individual statistics, the group performance
measures for this participant’s team were excluded (though survey measures were
retained).
Measures
Questionnaire measures can be found in Appendix B. Reliability for each
questionnaire measure at each administration can be found in Table 5.
Questionnaire measures.
Transactive memory system. TMS was measured using Lewis’s (Lewis) fifteen
item scale adapted to describe video game teams.
Social presence. The six dimensions of social presence were measured using
Harms and Biocca’s (2004) most recent version of the Networked Minds Social Presence
Inventory. Each dimension had six questions.
58

Table 5. Cronbach's α for Study 2 Questionnaire Measures
Match 1 Match 2 Match 3
TMS .76 .77 .88
Copresence .93 .95 .98
Attentional allocation .66 .76 .85
Perceived message understanding .88 .93 .95
Perceived affective understanding .90 .92 .92
Perceived affective interdependence .95 .96 .95
Perceived behavioral interdependence .92 .93 .95
Game measures.
Performance. The intended measure of performance was the kill-to-death (KTD)
ratio, which is calculated by taking the number of times a team kills an enemy champion
divided by the number of champion deaths a team suffers. This is a common measure of
performance by video game players in games like LoL, where players can die multiple
times since they regenerate after they die. It also would work well for this study, as its
ratio nature makes it independent of team size and match length (the longer the match,
the more opportunity both kill and die). However many teams made no kills during
matches, leaving many teams with a KTD of 0 (yet in terms of measuring performance, a
KTD of 0 with three deaths is different from a KTD of 0 with twenty deaths). Thus, an
alternative measure similar to KTD, the amount of damage inflicted by a team divided by
the amount of damage taken by the team was calculated. However, this measure was not
normally distributed. However, the inverse of this measure (the amount of damage taken
by a team divided by the damage inflicted by a team) was normally distributed. Because
59

higher values of this would indicate worse performance, this metric was subtracted from
1 to change its direction. Thus, a positive value would indicate that more damage was
inflicted than taken and a negative value would indicate more damage was taken than
inflicted (a value of 0 would indicate an equal amount of damage was inflicted and
taken).
Outcome. Outcome was measured by recording whether a team won a match.
Aggregation. For each scale, at each time point, r
wg(j)
, a measure of within-group
interrater agreement for multi-item measures, was calculated for each team (James,
Demaree, & Wolf, 1984, 1993). An r
wg(j)
greater than .90 represents very strong
agreement, greater than .70 indicates strong agreement, and greater than .50 indicates
moderate agreement (LeBreton & Senter, 2008). Means and frequencies of r
wg(j)
for each
measure at each questionnaire administration can be found in Table 6. These values
would indicate that individual scores for all of the scales could be aggregated for a team.
60

Table 6. Means and Frequencies of r
wg(j)
for Study 2 Questionnaire Measures
Match 1 Match 2 Match 3
M > .90 > .70 > .50 M > .90 > .70 > .50 M > .90 > .70 > .50
TMS .78 35% 80% 90% .79 55% 75% 90% .68 40% 75% 75%
Copresence .70 30% 75% 80% .85 60% 85% 90% .86 65% 85% 90%
Attentional allocation .80 40% 85% 95% .59 40% 75% 75% .74 25% 80% 85%
Perceived message understanding .57 35% 55% 65% .82 40% 85% 95% .74 50% 75% 75%
Perceived affective understanding .56 35% 55% 65% .79 40% 75% 90% .82 30% 80% 80%
Perceived affective interdependence .75 25% 55% 70% .77 55% 75% 85% .73 50% 80% 85%
Perceived behavioral interdependence .67 30% 70% 70% .82 35% 90% 90% .77 35% 75% 80%
.90 = very strong agreement
.70 = strong agreement
.50 = moderate agreement
Results
To test for common method variance, a Harman one factor test was conducted for
each of the three questionnaire administrations as well as all three administrations
combined (for the individual measurements, not the group aggregation). After the first
match, a single factor explained 30.46% of the variance. After the second match, a single
factor explained 30.60% of the variance. After the third match, a single factor explained
43.52% of the variance. Across all three administrations, a single factor explained
31.76% of the variance. Using the Lindell and Whitney (2001) technique of finding the
smallest positive correlation of all measures taken during a single questionnaire
administration, for Match 1 there was a correlation of r = .009, for Match 2 there was a
correlation of r =.008, and for Match 3 there was a correlation of r = .022. Between the
61

Harman one-factor tests and the Lindell and Whitney technique, common method
variance was ruled out.
Power Analysis
Post-hoc power analysis was conducted for α = .05. For the individual time points
(n = 20), with a bivariate correlation, the power to detect a large effect (r = .10) was .11,
the power to detect a medium effect (r = .30) was .37, and the power to detect a large
effect (r = .50) was .76. Aggregated across all three time points (n = 60), with a bivariate
correlation, the power to detect a large effect (r = .10) was .19, the power to detect a
medium effect (r = .30) was .76, and the power to detect a large effect (r = .50) was .99.
For repeated measures ANOVA, power depends on the correlations among
repeated measures and thus will differ across individual tests. A table of repeated
measures correlation can be found in Table 7. Power analysis for the different measures
can be found in Table 8.
62

Table 7. Correlation Among Study 2 Repeated Measures
r
1,2
r
1,3
r
2,3
r
average

TMS
.79
***

[.621, .915]
.71
***

[.417, .873]
.87
***

[.718, .967]
.79
Copresence
.75
***

[.411, .907]
.71
***

[.367, .877]
.81
***

[.596, .944]
.75
Attentional allocation
.42
*

[.035, .688]
.42
*

[-.251, .732]
.64
**

[.206, .892]
.49
Perceived message understanding
.73
***

[.456, .876]
.67
***

[.316, .857]
.71
***

[.456, .876]
.70
Perceived affective understanding
.68
***

[.414, .880]
.60
**

[.335, .822]
.93
***

[.829, .972]
.74
Perceived affective interdependence
.84
***

[.651, .946]
.71
***

[.500, .871]
.84
***

[.656, .951]
.80
Perceived behavioral interdependence
.75
***

[.380, .918]
.64
**

[..286, .855]
.81
***

[.562, .946]
.74
Performance
.65
**

[.240, .898]
.53
**

[-.074, .896]
.75
***

[.423, .900]
.64
r
i,j
= bivariate correlation between measurements after Matches i and j
*
p < .05,
**
p < .01,
***
p < .001
All confidence intervals are 95% BCa on 1000 bootstrap samples.

63

Table 8. Power Analysis for Study 2 ANOVAs
Channel Time Channel × Time
S M L S M L S M L
TMS .07 .21 .45 .29 .96 >.99 .29 .96 >.99
Copresence .07 .21 .46 .25 .93 >.99 .25 .93 >.99
Attentional allocation .08 .26 .55 .14 .64 .97 .14 .64 .97
Perceived message understanding .08 .22 .47 .21 .87 >.99 .21 .87 >.99
Perceived affective understanding .08 .21 .46 .24 .91 >.99 .24 .91 >.99
Perceived affective interdependence .07 .21 .45 .29 .96 >.99 .29 .96 >.99
Perceived behavioral interdependence .08 .21 .46 .24 .91 >.99 .24 .91 >.99
Performance .08 .22 .47 .17 .74 .99 .17 .74 .99
S = power to detect small effect (f = .10)
M = power to detect medium effect (f = .25)
L = power to detect large effect (f = .40)
Data Analysis
To test H1, that a) teams communicating via voice-chat would have greater initial
levels of TMS than teams communicating via text-chat, and b) over time, the differences
in TMS between voice-chat teams and text-chat teams would diminish, a 2 × 3 mixed
factorial ANOVA was conducted. No interaction effect was found for time and condition,
F(2, 36) = 0.06, p = .94, partial η
2
= .004. There was also no main effect for condition,
F(1, 18) = 0.18, p = .68, partial η
2
= .01. Thus, H1b was supported but H1a was not.
There was however a main effect for time, F(2, 36) = 25.48, p < .001, partial η
2
= .59.
Post hoc pairwise comparisons of simple effects with Sidak correction indicated that
TMS was significantly less after Match 1 than after Match 2 or 3 (p < .001 for both
64

pairwise comparisons), but there was no significant difference (p = .36) between TMS
after Match 2 and Match 3. Mean levels of TMS can be found in Figure 4. One-sample t-
tests indicated that the mean level of TMS (aggregated across the two conditions) after
Match 1, was significantly less than the scale midpoint of three, t(19) = -4.37, p < .001,
mean difference = -0.47, 95% BCa CI [-0.684, -0.252]. Match 2 did not differ significant
from the midpoint, t(19) = -0.01, p =.99, mean difference = -0.001, 95% BCa CI [-0.237,
0.232], nor did Match 3, t(19) = 0.77, p =.45, mean difference = 0.12, 95% BCa CI [-
0.192, 0.395].
Figure 4. Mean Levels of Transactive Memory System over Time by Condition

65

To test H2a, that TMS would be positively related to performance, bivariate
correlations were conducted. After Match 1, a large size correlation was found, r(17) =
.53, p = .01, 95% BCa CI [.135, .767]. Nonsignificant small size correlations were found
for the other two time points. After Match 2, r(17) = .22, p = .18, 95% BCa CI [-.234,
.615], and after Match 3, r(17) = .37, p =.06, 95% BCa CI [-.087, .731]. Aggregated
across all three matches, a medium-to-large relationship was found, r(57) = .43, p < .001,
95% BCa CI [.253, .606]. While the correlations after Matches 2 and 3 were
nonsignificant, the confidence intervals would indicate that none of the four correlations
differed significantly between time points nor differed from the aggregated point
estimate. Thus, H2a was supported for some time points and would have likely been
supported at all times had there been enough power to detect an effect.
Only one team won Match 1 (a text-chat condition), no teams won Match 2, and
only one team won Match 3 (a voice-chat condition). Thus, H2b, that TMS would be
positively related to outcome, could only be tested for Match 1 and Match 3. Binary
logistic regressions were conducted individually for Match 1 and Match 3. At neither
time individual time point were the regressions significant. After Match 1, R
L
2
= .43,
R
CS
2
= .16, R
N
2
= .48, χ
2
(1, N = 20) = 3.41, p = .07, and after Match 3, R
L
2
= .43, R
CS
2
=
.16, R
N
2
= .48, χ
2
(1, N = 20) = 3.43, p = .06. Aggregated across all three matches, the
regression was significant, R
L
2
= .25, R
CS
2
= .07, R
N
2
= .28, χ
2
(1, N = 60) = 4.04, p = .04.
However, while the constant of the regression was significant (p < .02), the odds ratio for
66

TMS, 14.03, 95% CI [0.833, 236.27]
13
was just short of significant, p = .07. Because of a
lack of variance in outcome, it was not possible to support H2b, as the probability of a
team losing was so high that a regression equation would predict the losing outcome
regardless of TMS.
14

To test H3a and H3c, that a) teams communicating via voice-chat would have
greater initial levels of performance than teams communicating via text-chat, and c) over
time, the differences in performance between voice-chat teams and text-chat teams would
diminish, a 2 × 3 mixed factorial ANOVA was conducted. No interaction effect was
found for time and condition, F(2, 34) = 0.13, p = .87, partial η
2
= .008. There was also
no main effect for condition, F(1, 17) = 0.14, p = .25, partial η
2
= .08. Thus, H3c was
supported but H3a was not. There was however a main effect for time, F(2, 36) = 25.48,
p < .001, partial η
2
= .59. Post hoc pairwise comparisons of simple effects with Sidak
correction indicated that performance differed between Match 1 and Match 3 (p = .04),
but Match 2 was neither significantly different from Match 1 (p = .13) or Match 3 (p =
.35). Mean levels of performance can be found in Figure 5.

13
This confidence interval was not bootstrapped, as many bootstrapped samples would
have no teams winning a match, leading to a variance of 0 for the outcome.
14
A small pilot test found greater variance in outcome than in the actual study. This may
have been because the pilot test had no incentive other than a desire to play the game.
Thus pilot participants may have had more game experience than the participants in the
main study.
67

Figure 5. Mean Levels of Performance over Time by Condition

To test H4, that for each dimension of social presence, communicating by voice-
chat would have greater levels of social presence initially as compared to text-chat but
that over time the differences would converge, a 2 × 3 mixed factorial ANOVAs were
conducted for each dimension. For copresence, there was no interaction effect, F(2, 36) =
1.97, p = .15, partial η
2
= .10, nor main effect for condition, F(1, 18) = 0.35, p = .56,
partial η
2
= .02. There was however a main effect for time, F(2, 36) = 11.97, p < .001,
partial η
2
= .40. Post hoc pairwise comparisons of simple effects with Sidak correction
indicated that copresence differed between Match 1 and Match 3 (p = .001) and Match 2
68

and Match 3 (p = .03), but Match 1 and Match 2 did not differ significantly (p = .11).
Mean levels of copresence can be found in Figure 6.
Figure 6. Mean Levels of Copresence over Time by Condition

For attentional allocation, there was no interaction effect, F(2, 36) = 2.44, p = .10,
partial η
2
= .12, nor main effect for condition, F(1, 18) = 0.71, p = .41, partial η
2
= .04.
There was however a main effect for time, F(2, 36) = 7.05, p = .003, partial η
2
= .28. Post
hoc pairwise comparisons of simple effects with Sidak correction indicated that
attentional allocation differed between Match 1 and Match 3 (p = .01), but Match 2 did
69

not differ from Match 1 (p = .21) or Match 3 (p = .12). Mean levels of attentional
allocation can be found in Figure 7.
Figure 7. Mean Levels of Attentional Allocation over Time by Condition

For perceived message understanding, there was no interaction effect, F(2, 36) =
0.98, p = .39, partial η
2
= .05, nor main effect for condition, F(1, 18) = 0.57, p = .46,
partial η
2
= .03. There was however a main effect for time, F(2, 36) = 21.75, p < .001,
partial η
2
= .55. Post hoc pairwise comparisons of simple effects with Sidak correction
indicated that perceived message understanding differed between Match 1 and Match 2 (p
< .001) and Match 1 and Match 3 (p < .001), but Match 2 and Match 3 did not differ (p =
.23). Mean levels of perceived message understanding can be found in Figure 8.
70

Figure 8. Mean Levels of Perceived Message Understanding over Time by Condition

For perceived affective understanding, there was no interaction effect, F(1.31, 36)
= 0.79 (Greenhouse-Geisser corrected), p = .42, partial η
2
= .04, nor main effect for
condition, F(1, 18) = 0.04, p = .84, partial η
2
= .002. There was however a main effect for
time, F(1.31, 36) = 7.96 (Greenhouse-Geisser corrected), p = .006, partial η
2
= .31. Post
hoc pairwise comparisons of simple effects with Sidak correction indicated that perceived
affective understanding differed between Match 1 and Match 3 (p = .02), but Match 2 did
not differ from Match 1 (p = .07) or Match 3 (p = .12). Mean levels of perceived affective
understanding can be found in Figure 9.
71

Figure 9. Mean Levels of Perceived Affective Understanding over Time by Condition

For perceived affective interdependence, there was no interaction effect, F(2, 36)
= 1.24, p = .30, partial η
2
= .06, nor main effect for condition, F(1, 18) = 2.10, p = .16,
partial η
2
= .11. There was however a main effect for time, F(2, 36) = 5.39, p < .001,
partial η
2
= .41. Post hoc pairwise comparisons of simple effects with Sidak correction
indicated that perceived affective interdependence differed between Match 1 and Match 2
(p = .047), Match 1 and Match 3 (p = .02), and Match 2 and Match 3 (p = .03). Mean
levels of perceived affective interdependence can be found in Figure 10.
72

Figure 10. Mean Levels of Perceived Affective Interdependence over Time by Condition

For perceived behavioral interdependence, there was no interaction effect, F(2,
36) = 1.10, p = .35, partial η
2
= .06, nor main effect for condition, F(1, 18) = 0.23, p =
.64, partial η
2
= .01. There was however a main effect for time, F(2, 36) = 11.63, p <
.001, partial η
2
= .39. Post hoc pairwise comparisons of simple effects with Sidak
correction indicated that perceived behavioral interdependence differed between Match 1
and Match 2 (p = .03), Match 1 and Match 3 (p = .003), and Match 2 and Match 3 (p =
.04). Mean levels of perceived behavioral interdependence can be found in Figure 11.
73

Figure 11. Mean Levels of Perceived Behavioral Interdependence over Time by
Condition

To address RQ1, as to how would the relationship between different dimensions
of social presence and TMS change as teams play more matches together, bivariate
correlations were conducted for each dimension and TMS for each individual time point
as well as aggregated across all three time points. These can be found in Table 9. Results
indicate that all dimensions are positively correlated with TMS and all correlations are
significant except for Match 1’s perceived affective understanding, perceived affective
interdependence, and perceived behavioral interdependence. In terms of correlations
differing, for copresence, the correlations for Match 1 and Match 2 fell outside of each
74

other’s confidence intervals. For perceived affective interdependence, the correlation for
Match 1 fell outside the confidence interval for Match 3. For perceived behavioral
interdependence, the correlation for Match 1 fell outside the confidence intervals for
Match 2 and Match 3.
Table 9. Correlation between TMS and Dimensions of Social Presence across Time
Match 1 Match 2 Match 3 Aggregated
Copresence
.47
*
[.183, .716]
.78
***

[.550, .898]
.59
**

[.231, .859]
.65
***

[.476, .778]
Attentional allocation
.69
***

[.306, .857]
.77
***

[.469, .923]
.81
***

[.604, .916]
.79
***

[.693, .867]
Perceived message understanding
.85
***

[.645, .943]
.84
***

[.632, .949]
.86
***

[.703, .941]
.88
***

[.801, .925]
Perceived affective understanding
.32
[-.121, .715]
.60
**

[.055, .905]
.62
**

[.239, .842]
.58
***

[.341, .763]
Perceived affective interdependence
.32
[-.287, .806]
.38
*

[-.218, .828]
.68
**

[.330, .868]
.52
***

[.253, .742]
Perceived behavioral interdependence
.31
[-.133, .674]
.66
**

[.361, .834]
.75
*

[.489, .907]
.64
***

[.482, .777]
*
p < .05,
**
p < .01,
***
p < .001
All confidence intervals are 95% BCa on 1000 bootstrap samples.
In regards to RQ2, because there was no main effect for communication channel
on TMS or any individual dimension of social presence, social presence does not mediate
a relationship between communication channel and TMS (as none exists).
75

Discussion
The results of Study 2 only partially supported the hypotheses. Hypotheses about
initial differences in text- and voice-chat were not supported but hypotheses about
convergence between text- and voice-chat were supported because no differences existed
in the first place and both text-and voice-chat increased in similar patterns. In the case of
TMS, there was no significant difference between groups at any time point. In fact, text-
chat groups had ever so slightly more TMS than the voice-chat condition. While the
study’s statistical power was low, increasing power would still be unlikely to result in a
statistically significant result. The best predictor of TMS was time itself; TMS drastically
increased between Match 1 and Match 2. There was a non-significant increase between
Match 2 and Match 3 as well (this would have likely been significant with greater
statistical power). It would appear that TMS could develop at a similar rate regardless of
the two communication channels. However, it would appear that after Match 1, on
average, there was no TMS, and after Matches 2 and 3, on average, there may or may not
have been TMS.
As for the hypothesis about TMS predicting performance and outcome, this too
was only partially supported. Across all time points, a medium-to-large size correlation
was found between TMS and performance, but looking at individual time points, only a
significant correlation was found after Match 1. However, the lack of significance after
Match 2 and Match 3 may have been due to lack of power, as the correlation at any given
time point was within the confidence interval of the correlation at other time points.
Because the measure of performance (1 minus the damage taken to inflicted ratio) was
76

not a statistic that appeared on screen to participants while playing, it would suggest that
evaluations of TMS were less likely to be a retrospective evaluation of performance and
more likely to actually be an evaluation of TMS.
On the other hand, TMS did not predict winning a match. However, this may have
been because very few matches were won by any team. Thus, the probability of winning
was so low, very little could predict winning. At individual time points, TMS was not a
predictor of winning a match (in the case of Match 2, it was because no teams won a
match), but aggregated across all three time points, TMS was just short of being a
significant predictor of winning (though the actual odds ratio was quite large).
Much like the hypothesis that TMS would initially differ between text- and voice-
chat and later converge, the hypothesis that performance would initially differ and then
converge was also only partially supported. There were no significant differences
between groups at any time point. Thus, there was convergence from the start. Unlike
TMS though, performance was slightly better for the voice-chat condition than the text-
chat condition, and with greater power, there might have been a main effect (in which
case, the hypothesis about initial differences would have been supported and the
hypothesis about convergence would not have been supported). Again, the best predictor
of performance was time itself. Between Match 1 and Match 3, performance increased.
Match 2 was not significantly different from either, but with greater power, it may have
been.
Similarly, the hypotheses about the initial difference and later convergence of
text-chat and voice-chat for the individual dimensions of social presence also found no
77

differences between conditions at any time point. Both groups had increases of all of the
dimensions of social presence across time. Without greater power, it is hard to
differentiate which specific time by condition interactions might actually be occurring.
More data need to be collected in order to interpret the results of these hypotheses.
Finally, in terms of the open research questions, it would appear that all of the
dimensions of social presence had strong positive relationships with TMS (unlike in
Study 1). Affective understanding, affective interdependence, and behavioral
interdependence were not significant predictors for Match 1, but this may have been due
to a lack of statistical power, as the actual correlation was of medium effect size. It does
seem though that at least for copresence, affective interdependence, and behavioral
interdependence, the correlations with TMS increase in strength after Match 1. This
suggests that the role each dimension plays in the development of TMS is dynamic. Thus,
if the experiment had been extended from three matches to one hundred matches, some
dimensions may have lost their predictive power. This might explain why in Study 1, the
dimensions of social presence were weaker predictors of TMS (and in some cases did not
predict TMS at all). Because there were no differences in social presence or TMS
between conditions, asking about mediation became an irrelevant research question.
However, this question will be addressed in the future if greater statistical power finds
some differences between groups (either with a main effect or interaction effect).
In addition to the lack of statistical power, one limitation of this study was that it
heavily relied on SIP to formulate its hypotheses. While a highly cited theory, there are
two shortcomings in using this. SIP is about relational communication, while TMS is
78

very focused on instrumental communication. Thus, SIP was applied implicitly to TMS.
Second, many of the propositions Walther made when formulating SIP (Walther, 1992)
were not supported by empirical evidence (Walther, 1995; Walther & Burgoon, 1992).
For many dimensions of relational communication, Walther found convergence between
CMC and FtF because there were no initial differences. Follow-up research discovered
that longitudinal work led to an anticipation of future interaction, which resulted in no
initial differences in relational communication between channels (Walther, 1994).
However, the present study was longitudinal in the sense that multiple measurements
were taken, but not in the sense that there was an anticipation of future interaction over
the course of the semester (like the original SIP empirical studies). Because of the
indirect application of SIP for Study 2 and because of the inconsistency between SIP
theory and empirical findings, an exploratory study was conducted simultaneous with
Study 2 and may be used to understand better the lack of channel differences in TMS.
79

Study 3: Social Information Processing and Transactive Memory Systems
The logic of Study 2 was based on SIP theory, which states that initially there will
be greater positive relational communication in FtF interaction than CMC interaction due
to the rate at which information can be transmitted, but over time, the amount of
relational communication should converge, as CMC users will develop verbal strategies
to compensate for the lack of nonverbal cues (Walther, 1992). While the “initial
differences with eventual convergence” was shown unequivocally to be true for
impression development (Walther, 1993), it was only partially true for some dimensions
of relational communication (Walther, 1995; Walther & Burgoon, 1992): some
dimensions converged from the start, whereas other dimensions had more positive
relational communication in CMC than FtF; only a few dimensions had “initial
differences with eventual convergence.” The hypothesis generation in Study 2 also
focused on a specific dimension of relational communication, receptivity/trust, in its logic
that there would be “initial differences with eventual convergence” for levels of TMS, yet
it did not address the other dimensions of relational communication. However, if some
dimensions of relational communication are negatively related to TMS (which has yet to
be explored for many dimensions of TMS), if there are initial differences between text-
chat and voice-chat, these dimensions of relational communication might actually
suppress the relationship between communication channel and TMS. Thus, Study 3 is an
exploratory study of three general questions: does the original premise of SIP hold for
text-chat and voice-chat in newly formed MOBA teams; what is the relationship between
80

the different dimensions of relational communication and TMS; and does relational
communication suppress TMS, thereby explaining the results of Study 2?
Relational Communication and Social Information Processing Theory
Burgoon and Hale (1984, 1987), using theoretical and factor analytic insight,
identified eight dimensions of verbal and nonverbal communication central to
interpersonal relationships: a) immediacy/affection, b) similarity/depth, c)
receptivity/trust, d) composure, e) formality, f) dominance, g) equality, and h) task
orientation. Immediacy/affection pertains to communication that expresses liking,
affection, inclusion, and involvement. Similarity/depth pertains to communication that
expresses that individuals are alike one another and desire a non-superficial relationship.
Receptivity/trust pertains to an open rapport and a desire to be trusted. These first three
all pertain to the intimacy of an interpersonal interaction. Composure is the extent to
which communication is calm as opposed to tense. Formality is the extent to which the
communication is formal as opposed to informal. Dominance is the extent to which
individuals try to persuade or control others. Equality is the expression of cooperation
and respect towards others. Task orientation is the extent that a group focuses its
communication on the task as opposed to on the interpersonal relationships between
members. For each of these dimensions, the original SIP hypotheses (Walther, 1992) will
be tested against the empirical findings (Walther & Burgoon, 1992), replacing FtF with
voice-chat and CMC with text-chat.
Immediacy/affection. While SIP originally hypothesized that there would be
greater initial immediacy/affection in FtF than CMC but both would increase over time
81

and converge, empirical results found no initial differences and that both increased in
tandem. Thus, the following hypothesis will be tested:
H1: In newly formed MOBA teams, there is an interaction between time and
channel such that initial levels of immediacy/affection will be greater in voice-
chat teams than text-chat teams, and immediacy/affection increases and converges
in voice-chat and text-chat teams as more matches are played.
Similarity/depth. SIP hypothesized that similarity/depth would have no initial
differences between conditions but that both would increase over time.
H2: In newly formed MOBA teams, there is only a main effect for time such that
similarity/depth increases in voice-chat and text-chat teams as more matches are
played and both conditions are converged from the start.
Receptivity/trust. While SIP originally hypothesized that there would be greater
initial receptivity/trust in FtF than CMC but both would increase over time and converge,
empirical results found no initial differences and that both increased in tandem. Thus, the
following hypothesis will be tested:
H3: In newly formed MOBA teams, there is an interaction between time and
channel such that initial levels of receptivity/trust will be greater in voice-chat
teams than text-chat teams, and receptivity/trust increases and converges in voice-
chat and text-chat teams as more matches are played.
Composure. While SIP originally hypothesized that there would be greater initial
composure in FtF than CMC but both would increase over time and converge, empirical
82

results found no initial differences and that both increased in tandem. Thus, the following
hypothesis will be tested:
H4: In newly formed MOBA teams, there is an interaction between time and
channel such that initial levels of composure will be greater in voice-chat teams
than text-chat teams, and composure increases and converges in voice-chat and
text-chat teams as more matches are played.
Formality. While SIP originally hypothesized that there would be different initial
levels of formality in FtF than CMC but both would decrease over time and converge,
empirical results found no initial differences and that both decreased in tandem. Thus, the
following hypothesis will be tested:
H5: In newly formed MOBA teams, there is an interaction between time and
channel such that initial levels of formality will differ in voice-chat teams than
text-chat teams, and formality decreases and converges in voice-chat and text-chat
teams as more matches are played.
Dominance. While SIP originally hypothesized that there would be greater levels
of dominance in CMC than FtF and that CMC, which would decrease, would converge
with FTF, which would have an inverted U-shape, empirical results found no initial
differences but the converging decrease of CMC with the inverted U-shape of FtF was
found. Thus, the following hypothesis will be tested:
H6: In newly formed MOBA teams, there is an interaction between time and
channel such that initial levels of dominance will be greater in text-chat teams
83

than voice-chat teams, and dominance decreases in text-chat team and converges
with voice-chat teams, which has an inverted U-shape.
Equality. While SIP originally hypothesized that there would be greater levels of
equality in FtF than CMC and that CMC, which would increase, would converge with
FTF, which would have a U-shape, empirical results found no initial differences but the
converging increase of CMC with the U-shape of FtF was found. Thus, the following
hypothesis will be tested:
H7: In newly formed MOBA teams, there is an interaction between time and
channel such that initial levels of equality will be greater in voice-chat teams than
text-chat teams, and equality increases in text-chat team and converges with
voice-chat teams, which has a U-shape.
Task orientation. While SIP originally hypothesized that there would be greater
initial task orientation in CMC than in FtF but both would decrease over time and
converge, empirical results found that while both decreased over time, CMC was less
task oriented than FtF. Thus, the following hypothesis will be tested:
H8: In newly formed MOBA teams, there is an interaction between time and
channel such that initial levels of task orientation will be greater in text-chat
teams than voice-chat teams, and task orientation decreases and converges in
voice-chat and text-chat teams as more matches are played.
Relational Communication and TMS
Very little research has been done pertaining to links between relational
communication and TMS. Much of the relevant research is only pertinent to some of the
84

dimensions of relational communication or does not address the full concept of TMS. For
example, the credibility dimension of TMS deals with cognitive trust, which is how much
one trusts the knowledge and abilities of others. What the credibility dimension does not
deal with is affective trust, which is an interpersonal trust in the benevolence and
integrity of others, which is more relevant to the receptivity/trust dimension of relational
communication. While TMS research has dealt with cognitive trust, Ashleigh and
Prichard (2012) have proposed an integrative model of TMS development that proposes
that both cognitive and affective trust are necessary before the specialization and
coordination dimensions of TMS cane develop (even if only cognitive and not affective
trust is a dimensions of TMS). While the model proposed by Ashleigh and Prichard has
yet to be tested, if it is in fact true, one would hypothesize:
H9: Receptivity/trust is positively related to TMS in MOBA teams.
The aspects of affective trust that Ashleigh and Pritchard directly address are
benevolence and integrity. One might expect that these two things would like lead to a
greater positive attitude towards another person. Thus, it is hypothesized:
H10: Immediacy/affection is positively related to TMS in MOBA teams.
While not dealing explicitly with TMS, Littlepage and colleagues (Littlepage &
Mueller, 1997; Littlepage, Schmidt, Whisler, & Frost, 1995) found four variables that
predicted expertise recognition in groups, two of which are also dimensions of relational
communication—dominance and task-oriented communication. While these studies dealt
with homogenous culture teams (all were American), Yuan, Bazarova, Fulk, and Zhang
(2013) found that only task-oriented communication is positively related to expertise
85

recognition but dominance was not. Similarly, Kanawattanachai and Yoo (2007) found
initial levels of task oriented communication (as measured using IPA) are predictive of
TMS. All of this being said, the tasks used in these studies may have lacked the
socioemotional communication that dominates video games. Furthermore, in the studies
that focused solely on expertise recognition, the tasks used were on the opposite end of
the circumplex model of group tasks (McGrath, 1984). For the time being though, it is
hypothesized that:
H11: Task orientation is positively related to TMS in MOBA teams.
As for the dominance dimension, when using multicultural teams, Yuan et al.
(2013) were unable to replicate Littlepage and colleagues’ (Littlepage & Mueller, 1997;
Littlepage et al., 1995) finding that dominance was positively related to expertise
recognition. Given the inconsistent results, and given these studies used a task type very
different from a Type 7 task, hypothesis formation is not clear. Another dimension that
does not have a clear direction to hypothesize is similarity/depth. It may be positively or
negatively related, as diversity has both positive and negative indirect effects on the
performance of video game teams (Xiong et al., 2009). As for the other dimensions, no
relevant research could be found
Thus, it is asked:
RQ1: What is the relationship between similarity/depth, composure, formality,
dominance, and equality with TMS development?
RQ2: For all eight dimensions of relational communication, how do the
relationships with TMS change over time?
86

If some dimensions of relational communication are positively related to TMS
development and others negatively related to TMS development, this may lead to
suppression effects if relational communication differs between text-chat and voice-chat
teams. If SIP theoretically holds true for the game teams, it is possible that the fact that
the greater initial task orientation of text-chat teams cancels out the greater initial
receptivity/trust of voice-chat teams. Thus, it is asked:
RQ3: Can the lack of interaction effect between time and communication channel
on TMS be explained by positive and negative indirect effects canceling each
other out?
Method
Participants
Study 3 was conducted simultaneously with the same participants from Study 2.
Procedure
Study 3 drew questionnaire responses from the same questionnaire administration
as Study 2.
Measures
Questionnaire measures can be found in Appendix C. Reliability analysis can be
found in Table 10. Aggregation analysis can be found in Table 11.
Transactive memory system. TMS was measure concurrently with Study 2.
Relational communication. The eight dimensions of relational communication,
immediacy/affection, similarity/depth, receptivity/trust, composure, formality,
dominance, equality, and task orientation were measured using a version of the 41-item
87

scale developed by Burgoon and Hale (1987) that was adapted to reflect a team rather
than a dyad.
Table 10. Cronbach's α for Study 3 Questionnaire Measures
Match 1 Match 2 Match 3
TMS .76 .77 .88
Immediacy/affection .86 .89 .88
Similarity/depth .88 .89 .86
Receptivity/trust .92 .93 .93
Composure .76 .74 .79
Formality .39 .54 .66
Dominance .63 .78 .72
Equality .69 .66 .69
Task orientation .74 .62 .73

88

Table 11. Means and Frequencies of r
wg(j)
for Study 3 Questionnaire Measures
Match 1 Match 2 Match 3
M > .90 > .70 > .50 M > .90 > .70 > .50 M > .90 > .70 > .50
TMS .78 35% 80% 90% .79 55% 75% 90% .68 40% 75% 75%
Immediacy/affection .69 20% 80% 80% .85 55% 90% 95% .80 35% 90% 95%
Similarity/depth .74 35% 80% 85% .76 15% 70% 85% .76 25% 90% 90%
Receptivity/trust .76 55% 80% 85% .83 60% 85% 85% .84 55% 95% 95%
Composure .69 5% 70% 80% .71 25% 65% 80% .58 40% 60% 65%
Formality .73 40% 75% 85% .57 20% 60% 65% .57 30% 60% 60%
Dominance .75 35% 75% 90% .72 40% 55% 85% .72 40% 75% 75%
Equality .55 15% 45% 60% .51 30% 55% 65% .63 20% 55% 70%
Task orientation .51 5% 30% 60% .66 15% 50% 75% .69 30% 55% 80%
.90 = very strong agreement
.70 = strong agreement
.50 = moderate agreement
Results
Common Method Variance
To test for common method variance, a Harman one factor test was conducted for
each of the three questionnaire administrations as well as all three administrations
combined. After the first match, a single factor explained 25.23% of the variance. After
the second match, a single factor explained 31.40% of the variance. After the third match,
a single factor explained 34.23% of the variance. Across all three administrations, a
single factor explained 26.76% of the variance. Between the Harman one-factor tests and
89

the Lindell and Whitney technique (which would be the same as Study 2), common
method variance was ruled out.
Power Analysis
Post-hoc power analysis was conducted for α = .05. Just as in Study 2, for the
individual time points (n = 20), with a bivariate correlation, the power to detect a large
effect (r = .10) was .11, the power to detect a medium effect (r = .30) was .37, and the
power to detect a large effect (r = .50) was .76. Aggregated across all three time points (n
= 60), with a bivariate correlation, the power to detect a large effect (r = .10) was .19, the
power to detect a medium effect (r = .30) was .76, and the power to detect a large effect
(r = .50) was .99.
For repeated measures ANOVA, power depends on the correlations among
repeated measures and thus will differ among individual tests. A table of repeated
measures correlation can be found in Table 12. Power analysis for the different measures
can be found in Table 13.
90

Table 12. Correlation Among Study 3 Repeated Measures
r
1,2
r
1,3
r
2,3
r
average

TMS
.79
***

[.621, .915]
.71
***

[.417, .873]
.87
***

[.718, .967]
.79
Immediacy/affection
.79
***

[.594, .920]
.82
***

[.662, .911]
.91
***

[.798, .958]
.84
Similarity/depth
.92
***

[.842, .969]
.86
***

[.626, .970]
.91
***

[.789, .975]
.90
Receptivity/trust
.91
***

[.834, .965]
.84
***

[.695, .923]
.96
***

[.901, .988]
.90
Composure
.66
**

[.342, .849]
.69
***

[.399, .842]
.87
***

[.735, .950]
.74
Formality
.39
*

[-.049, .730]
.62
**

[.262, .839]
.60
**

[.192, .853]
.54
Dominance
.67
**

[.345, .897]
.55
**

[.159, .833]
.89
***

[.776, .952]
.70
Equality
.89
***

[.723, .985]
.81
***

[.603, .932]
.85
***

[.681, .933]
.85
Task orientation
.59
**

[.209, .831]
.56
**

[.191, .816]
.74
***

[.536, .887]
.63
r
i,j
= bivariate correlation between measurements after Matches i and j
*
p < .05,
**
p < .01,
***
p < .001
All confidence intervals are 95% BCa on 1000 bootstrap samples.

91

Table 13. Power Analysis for Study 2 ANOVAs
Channel Time Channel × Time
S M L S M L S M L
TMS .07 .21 .45 .29 .96 >.99 .29 .96 >.99
Immediacy/affection .07 .20 .43 .37 .99 >.99 .37 .99 >.99
Similarity/depth .07 .20 .42 .54 >.99 >.99 .54 >.99 >.99
Receptivity/trust .07 .19 .42 .55 >.99 >.99 .55 >.99 >.99
Composure .08 .21 .46 .24 .91 >.99 .24 .91 >.99
Formality .08 .25 .53 .15 .68 .98 .15 .68 .98
Dominance .08 .22 .42 .21 .87 >.99 .21 .87 >.99
Equality .07 .20 .43 .38 .99 >.99 .38 .99 >.99
Task orientation .08 .23 .50 .18 .79 >.99 .18 .79 >.99
S = power to detect small effect (f = .10)
M = power to detect medium effect (f = .25)
L = power to detect large effect (f = .40)
Data Analysis
To test the SIP hypotheses, a 2 × 3 mixed factorial ANOVA was conducted for
each of the individual dimensions of relational communication. In testing H1 about
immediacy/affection, there was no interaction effect, F(2, 36) = 0.69, p = .51, partial η
2
=
.04, nor main effect for condition, F(1, 18) = 0.02, p = .89, partial η
2
= .001. There was
however a main effect for time, F(2, 36) = 17.99, p < .001, partial η
2
= .50. Post hoc
pairwise comparisons of simple effects with Sidak correction indicated that
immediacy/affection differed between Match 1 and Match 2 (p = .003), Match 1 and
92

Match 3 (p = .001), but Match 2 and Match 3 did not differ significantly (p = .17). Mean
levels of immediacy/affection can be found in Figure 12. H1 was not supported.
Figure 12. Mean Levels of Immediacy/Affection over Time by Condition

In testing H2 about similarity/depth, there was no interaction effect, F(2, 36) =
1.30, p = .28, partial η
2
= .07, nor main effect for condition, F(1, 18) = 0.31, p = .59,
partial η
2
= .02. There was however a main effect for time, F(2, 36) = 14.36, p < .001,
partial η
2
= .44. Post hoc pairwise comparisons of simple effects with Sidak correction
indicated that similarity/depth differed between Match 1 and Match 3 (p = .001) and
Match 2 and Match 3 (p = .006), but Match 1 and Match 2 did not differ significantly (p
= .14). Mean levels of similarity/depth can be found in Figure 13. H2 was supported.
93

Figure 13. Mean Levels of Similarity/Depth over Time by Condition

In testing the H3 about receptivity/trust, there was no interaction effect, F(1.35,
24.35) = 1.00 (Greenhouse-Geisser corrected), p = .35, partial η
2
= .05, nor main effect
for condition, F(1, 18) = 0.08, p = .78, partial η
2
= .005. There was however a main effect
for time, F(1.35, 24.35) = 14.45 (Greenhouse-Geisser corrected), p < .001, partial η
2
=
.45. Post hoc pairwise comparisons of simple effects with Sidak correction indicated that
receptivity/trust differed between Match 1 and Match 3 (p = .002), Match 2 and Match 3
(p = .001), but Match 1 and Match 2 did not differ significantly (p = .05). Mean levels of
receptivity trust can be found in Figure 14. Thus, H3 was not supported
94

Figure 14. Mean Levels of Receptivity/Trust over Time by Condition

In testing the competing H4 about composure, there was an interaction effect,
F(1.70, 30.64) = 4.16 (Huynh-Felt corrected), p = .03, partial η
2
= .19, and a main effect
for time, F(1.70, 30.64) = 5.85 (Huynh-Felt corrected), p = .01, partial η
2
= .19, but no
main effect for condition, F(1, 18) = 0.05, p = .83, partial η
2
= .003. Post hoc pairwise
comparisons of simple effects with Sidak correction indicated that the interaction was
that composure only increased for the text-chat condition between Match 2 and Match 3
(p = .006) and only increased for the voice-chat condition between Match 1 and Match 2
(p = .02). At no individual point in time did the two conditions differ. Thus, while
composure increased over time in both conditions, this increase happened earlier for the
95

voice-chat condition than text-chat condition. Mean levels of composure can be found in
Figure 15. Thus, H4 was not supported.
Figure 15. Mean Levels of Composure over Time by Condition

In testing H5 about formality, there was no interaction effect, F(2, 36) = 0.01, p =
.99, partial η
2
= .001; or main effect for condition, F(1, 18) = 0.17, p = .69, partial η
2
=
.009; and no main effect for time, F(2, 36) = 1.00, p = .38, partial η
2
= .05. Mean levels
formality can be found in Figure 16. Thus, H5 was not supported.
96

Figure 16. Mean Levels of Formality over Time by Condition

In testing H6 about dominance, there was no interaction effect, F(1.43, 25.75) =
0.02 (Greenhouse-Geisser corrected), p = .94, partial η
2
= .001; no main effect for
condition, F(1, 18) = 0.02, p = . 88, partial η
2
= .001; and no main effect for time, F(1.43,
25.75) = 3.11 (Greenhouse-Geisser corrected), p = .08, partial η
2
= .15. Mean levels of
dominance can be found in Figure 17. Thus, H6 was not supported.
97

Figure 17. Mean Levels of Dominance over Time by Condition

In testing H7 about equality, there was no interaction effect, F(2, 36) = 0.65, p =
.53, partial η
2
= .04, nor main effect for condition, F(1, 18) = 0.07, p = .80, partial η
2
=
.004. However, while the ANOVA indicated a main effect for time, F(2, 36) = 3.50, p =
.04, partial η
2
= .16, post hoc pairwise comparisons of simple effects with Sidak
correction indicated no significant differences in equality between the time points. Mean
levels of equality can be found in Figure 18. Thus, H7 was not supported.
98

Figure 18. Mean Levels of Equality over Time by Condition

In testing H8 about task orientation, there was no interaction effect, F(2, 36) =
0.83, p = .44, partial η
2
= .04, nor main effect for condition, F(1, 18) = 0.17, p = .68,
partial η
2
= .01. There was however a main effect for time, F(2, 36) = 5.65, p = .007,
partial η
2
= .24. Post hoc pairwise comparisons of simple effects with Sidak correction
indicated that task orientation increased between Match 1 and Match 2 (p = .008), but did
not differ significantly between Match 1 and Match 3 (p = .12) or Match 2 and Match 3
(p = .85). Mean levels of task orientation can be found in Figure 19. H8 was not
supported.
99

Figure 19. Mean Levels of Task Orientation over Time by Condition

To test H9, H10, H11, RQ1, and RQ2, bivariate correlations were conducted for
each dimension and TMS for each individual time point as well as aggregated across all
three time points. These can be found in Table 14. Receptivity/trust had a significant
large effect size positive correlation after Matches 2 and 3 but a nonsignificant medium
effect size positive correlation after Match 1, which fell outside the confidence interval of
the other two matches. Thus, H9 was partially supported. Likewise, there was a
significant, large effect size positive correlation between immediacy/affection after
Matches 2 and 3, but there was a nonsignificant medium effect size positive correlation
after Match 1. The correlation after Match 1 fell outside of the confidence intervals for
100

Matches 2 and 3. H10 was therefore also partially supported. For task orientation, there
was a significant large effect size positive correlation after Match 2 and a just short of
significant medium-to-large effect size positive correlation after Match 3. Match 1 had a
nonsignificant small-to-medium effect size after Match 1. Match 1 fell outside the
confidence interval of Match 2 but not Match 3. Matches 2 and 3 were within each
other’s confidence intervals. H11 too was partially supported.
As for the other dimensions of relational communication, after the individual
matches, there were nonsignificant medium effect size positive correlations for
similarity/depth with TMS, but once aggregated across all three matches, the correlation
was significant. Like immediacy/affection, with composure, there was a nonsignificant
small-to-medium effect size positive correlation after Match 1 and a significant medium-
to-large effect size positive correlation after Matches 2 and 3. All three matches fell
within the confidence intervals of the others. Formality had a significant large effect size
negative correlation with TMS after Match 3 only. After Matches 1 and 2, the correlation
was nonsignificant and of negligible effect size. The correlation after Matches 1 and 2
fell outside the confidence interval of the correlation after Match 3. For dominance, the
correlations were nonsignificant of small effect size. For equality, there was a significant
large effect size positive correlation after Matches 2 and 3. Match 1 was a nonsignificant
small effect size positive correlation. This fell within the confidence interval for Match 2
but not Match 3.
101

Table 14. Correlation between TMS and Dimensions of Relational Communication
across Time
Match 1 Match 2 Match 3 Aggregated
Immediacy/affection
.33
[.076, .555]
.65
**

[.366, .826]
.62
**

[.341, .805]
.61
***

[.432, .744]
Similarity/depth
.30
[-.158, .642]
.30
[-.319, .678]
.34
[-.068, .681]
.36
**

[.099, .063]
Receptivity/trust
.31
[-.001, .584]
.62
**

[.260, .825]
.61
**

[.326, .790]
.56
***

[.369, .713]
Composure
.29
[-.092, .666]
.47
*

[.113, .711]
.45
*

[.059, .714]
.48
***

[.228, .640]
Formality
-.12
[-.402, .133]
.07
[-.602, .518]
-.57
**

[-.818, -.089]
-.28
*

[-.526, .004]
Dominance
.25
[-.250, .623]
.01
[-.588, .490]
.17
[-.257, .508]
.20
[-.051, .500]
Equality
.16
[-.242, .537]
.50
*

[.140, .734]
.54
*

[.206, .752]
.43
**

[.188, .621]
Task orientation
.10
[-.408, .571]
.57
**

[.190, .849]
.30
[-.058, 591]
.39
**

[.167, .580]
*
p < .05,
**
p < .01,
***
p < .001
All confidence intervals are 95% BCa on 1000 bootstrap samples.
In terms of RQ3, because there was no main effect for condition on any of the
dimension of relational communication and the only interaction effect did not find any
between condition differences at an individual time point, it would be impossible for
there to be any indirect effects of channel on TMS through relational communication. To
102

answer this research question, greater statistical power would be necessary in order to
determine which dimensions of relational communication should be tested as possible
intermediaries for indirect effects.
Discussion
The results of Study 3 do not support the theoretical assertions of SIP. At the
same time, they do not support all of the previous empirical findings of Walther and
Burgoon (1992). For immediacy/affection, there appeared to be a main effect only for
time, similar to the previous empirical findings. For similarity/depth, one of the few
dimensions that Walther and Burgoon (1992) hypothesized to not differ between
conditions, from a significance testing perspective, the results of the present study were
consistent. However, the pattern of means actually demonstrated a pattern similar to the
theoretical assertions for the other dimensions: there was greater similarity/depth initially
for voice-chat than text-chat but the two conditions converged over time. With greater
statistical power, there may indeed be a detectable interaction effect. Receptivity/trust
was consistent with the previous empirical findings but the pattern of means was
consistent with the theoretical assertions. Thus, to interpret the results properly, greater
statistical power will be necessary. Composure was neither consistent with the theoretical
nor the empirical findings. The interaction effect demonstrated that composure increased
between different time intervals for the different conditions. However, there was not
enough statistical power to determine at which individual time points the two conditions
significantly differed. Formality did not differ between the two conditions or over time.
Similarly, neither the theoretical assertions of SIP nor the previous empirical findings
103

were consistent with the results for dominance. No effects, main or interaction, were
found (although greater statistical power may be able to detect main effects for condition
and time). There do not appear to be interaction effects from the pattern of means.
Equality also did not demonstrate results consistent with the theoretical assertions of SIP
or previous empirical findings. However, the pattern of means suggest an “initial
difference with convergence” pattern, despite the fact that this is not what SIP
hypothesizes for this particular dimension. Finally, task orientation was inconsistent with
theoretical predictions of SIP or previous empirical findings. Task orientation actually
increased across both conditions. Similar to previous findings, although not significant in
the case of the present study, the pattern of means indicated that voice-chat teams were
more task oriented that text-chat teams (in contrast to the theoretical predictions of SIP).
One possible explanation for the increase in task orientation is that the game was quite a
novel task for participants so their task orientation may have increased across matches as
they understood the task better.
As for the relationship between the individual dimensions of relational
communication and TMS, no dimensions were significant correlated after Match 1
(although some may have been with greater statistical power. What is clear is
immediacy/affection, receptivity/trust, composure, equality, and task orientation are
positively related to TMS, although this relationship becomes stronger after multiple
matches have been played. Similarity/depth is positively related but consistent across
time. Formality is relatively independent of TMS initially but becomes negatively
correlated after three matches. Dominance appears to be independent of TMS.
104

It would appear that as players get more experience, relational communication
becomes more important in the formation of TMS. It may be that early on, as players are
learning the game, they are more focused on the game than their teammates. The results
of Study 2 indicated a steep jump in TMS between Matches 1 and 2. Thus, because
players do not understand the game mechanics, in their first match, they lack the
interdependence that defines a team. This is either because they are learning the game
instead of learning about their teammates or it takes time for them to develop a sense of
interdependence. This might explain why, on average there was no TMS after Match 1,
and why after Match 2, on average, the TMS level was not significantly different from
the midpoint of the scale. Thus, the reason why TMS may not have developed swiftly
was because, based on the results of Study 2, the team had yet to be interdependent, an
antecedent of TMS.
As for whether relational communication suppresses TMS, because there were no
meaningful main or interaction effects for condition on relational communication, this
could not be tested. However, given that only one dimension, formality, was negatively
related to TMS, it is unlikely that relational communication would be a suppressor, even
if greater statistical power found main or interaction effects for condition.
Obviously, the results of the Study 3 call SIP into question. Walther (Walther,
1994; Walther & Burgoon, 1992) explained the original findings of CMC and FtF not
differing from the start as being due to the fact that the longitudinal nature made
participants anticipate future interaction, which led to there being relational
communication in the CMC condition from the start. However, the present study, while
105

longitudinal in the sense that repeated measures was taken, lacked an anticipation of
future interaction as the three matches were all played in one sitting (had players been
asked to come in three consecutive weeks, there may have been different results).
Limitations
Studies 2 and 3 were limited due to the small sample size. With greater statistical
power, more hypotheses may have been supported (or refuted in a more interpretable way
than null differences). The patterns of means for several dimensions of relational
communication suggested (in sometimes unexpected ways) interaction and main effects.
Another major limitation in the interpretation of results is a lack of metric for
communication volume. Unfortunately, the transcripts of text-chat sessions were lost
when the software recording the game became incompatible with a LoL software update.
During the course of the lab studies, the experimenters noticed that in both the text and
voice conditions, players would go extended periods not communicating with their
teammates. Therefore, text-chat players may have had the time to communicate just as
much information as the voice-chat players if all teams had extended periods of silence.
The transmission rate may have differed but the volume may have been constant across
conditions. This metric may have served as an important covariate.
Finally, the fact that players were inexperienced with LoL may have slowed down
the development of TMS. While using inexperienced players was an intentional choice to
understand how groups learned novel tasks together, because of the novelty, it may have
required more than three matches to get a better sense of TMS development. Players may
106

have spent a good portion of the time learning game mechanics instead of learning about
their teammates (or even realizing that they needed to work as a team to be successful).
107

Conclusion
Study 1 demonstrated that TMS could develop in video game teams engaged in a
Type 7, battle/contest, task. Because such a task has two metrics in evaluating a group,
performance and outcome, the study was the first to demonstrate that TMS was positively
related to outcome. It found that the self-selection of teammates was positively related to
TMS and that three person teams and five person teams did not differ in TMS. Finally, it
found that two dimensions of social presence, copresence and message comprehension
mediated the relationship between self-selection and TMS.
Study 2 was less conclusive, in part because of a lack of statistical power, about
the relationship between communication channel, TMS, and social presence. It would
appear that the affordances for nonverbal cues do not affect TMS development in this
study, in contrast to some previous findings. Juxtaposed with the results of Riedl et al.
(2012), it would appear that it is perceived channel richness and not a technological
deterministic interpretation of channel richness that leads to greater TMS. Unlike Study
1, which examined the relationship between TMS and outcome, Study 2 looked at TMS
and both performance and outcome. It found that TMS predicted a metric performance
that participants did not see while playing, but it did not predict outcome (although the
latter may be because few teams won a match), which would suggest that self-reports of
TMS after the fact are valid and not necessarily retrospective evaluations of observed
performance or outcome. Finally, it found that all dimensions of social presence were
positively related to TMS, though some dimensions had a stronger relationship in later
matches than earlier matches did. Thus, the relationship is dynamic and may explain why
108

not all dimensions demonstrated a relationship in Study 1, which looked at players of
differing levels of experience.
Study 3 looked at the role of relational communication and SIP in the
development of TMS in video game teams. The results did not support the theoretical
predictions of SIP nor were they consistent with many of the previous empirical findings
of SIP, although some of the results were hard to interpret in light of the low statistical
power. What it did find though was most dimensions of relational communication were
positively related to TMS, with the exception of dominance, which appeared to be
independent of TMS, and formality, which initially appeared to be independent of TMS
but later negatively related to TMS.
Theoretical Implications
The results of Study 2 along with the post-hoc analysis in Study 1 would suggest
that the early findings of channel effects on TMS might not hold true today. The early
findings may have been a result of the fact that most individuals were not familiar with
CMC at the time. The lack of channel effects is important, as TMS research on emergent
response teams (Majchrzak, Jarvenpaa, & Hollingshead, 2007) has suggested that such
teams often faced limited communication channels in emergency situations. If groups are
able to develop strategies to maximize information exchange when nonverbal cues are
limited, they may in fact be able to develop TMS at rates similar to groups with more
communication channels available.
These results also indicate the robustness of TMS development applying in Type
7, battle/contents, tasks and tasks filled with relational communication. These two
109

important aspects of human communication are often overlooked by researchers due to
the predisposition to study task-oriented work teams. Whereas early TMS research was
conducted in the context of dyads and small groups, much research has transitioned into
the study organizational behavior.
These results also call into question SIP. It is very possible that the reason why
early results were inconsistent with theoretical predictions was not the anticipation of
future interaction but because the theory was not entirely correct. One also has to wonder
to what extent SIP holds today with college-age students. Whereas SIP posited that
individuals develop strategies over time to compensate for the lack of nonverbal cues,
given the fact that today’s college-age population grew up with all forms of mediated
communication, it may be that some people don’t need time to develop strategies, as
compensatory conventions are already established. In the white paper that laid out many
technologies commonplace today in human-computer interaction, Douglas Engelbart
(1962) proposed what he called a “Neo-Whorfian Hypothesis.” Extending the Sapir-
Whorf Hypothesis (Whorf, 1956), which states that one’s native language affects one’s
cognitive processes and conceptualization of the world, Engelbart’s Neo-Whorfian
Hypothesis posited that the way one manipulates symbols affects cognitive processes and
conceptualizations. Given that people, especially young people, are used to manipulating
mediated symbols, CMC strategies may be very different than when Walther and
Burgoon (1992) was conducted.
Overall, the results suggest that task type is very important when drawing
conclusions about both virtual groups and groups in general. This of course is not a novel
110

concept. In terms of virtual groups, Williams (2010) cautions about assuming outcomes
in real-world groups map to virtual worlds (and mapping may differ between different
virtual worlds). Futhermore, the reason McGrath (1984) proposed the circumplex model
was because as tasks differed on the conceptual/behavioral and the cooperation/conflict
dimensions, group processes would be very different.
When thinking about the issues of mapping and the circumplex model, MOBAs
may most resemble a short-term sports team such as a pickup game of basketball or
multiple military units that come together right as they are about to go into battle.
Furthermore, the findings of this study may hold true for Type 7 tasks in zero-
acquaintance groups even if they will result in long-term groups, as the MOBA teams
should have had no anticipation of ongoing interaction, which Walther (1994) said would
affect interpersonal relationship development in groups.
By communication channel having no relationship with relational communication,
one must think about the overall effects of relational communication on TMS
development. While task-orientation is positively related to TMS, other dimensions of
relational communication are as well. Developing positive interpersonal relationships is
central to the ability to learn and coordinate expertise. Many smart people with different
expertise may be put in a room together, but if they do not get along, expertise
recognition and knowledge sharing may suffer.
Emergent groups responding to disasters has been proposed as an area to rethink
TMS theory (Majchrzak et al., 2007). While these teams are not engaging in a Type 7
tasks, they are engaging in the other type of Execute tasks. Among the reasons why these
111

groups have forced a rethinking of TMS theory are: they often have a sense of great
urgency, volatility, little prior experience with teammates, limited communication
channels, geographic dispersion, and possible self-interests on top of group-interests. For
this reasoning, Majchrzak et al. have proposed replacing specialization as expertise being
tailored to a specific situation, credibility as trust through action, and coordination with
coordinating knowledge processes without a knowledge structure. However, given the
danger involved in emergency responses, these new conceptualizations are hard to test.
Jarvenpaa and Majchrzak (2008) were able to examine some of these things in regards to
national security teams, but they focused mostly on the egocentric aspect characterizing
emergent teams responding to a disaster.
All of the aforementioned reasons as to why Majchrzak et al. (2007) said that
TMS development may differ than past teams used in the study of TMS, MOBAs share
many of these properties, even if emergency response teams do not engage in Type 7
tasks (unless one would consider the disaster an opponent). Study 1 demonstrated TMS
could develop in MOBA teams, although the teams studied may have been more long
term. Study 2 though demonstrated that players engaged in an unfamiliar task with
unfamiliar teammates can build TMS in a relatively short period of time (in this case
between one and a half and two and a half hours, depending on match length).
Future Directions
First, it is necessary to run more participants for this study to increase statistical
power for main effects and small size interaction effects. Other studies of TMS (Jackson
& Moreland, 2009; Lewis et al., 2007; Lewis et al., 2005; Liang et al., 1995; Moreland,
112

1999; Moreland & Myaskovsky, 2000) and relational communication (Walther, 1993,
1994, 1995; Walther & Burgoon, 1992) in small groups often used between 30 and 40
groups. Power analysis would suggest that even this many groups may only be able to
detect medium-to-large effects (Cohen, 1988).
While the choice to study players without MOBA experience was an intentional
choice, it would be interesting to compare them to zero-acquaintance teams that have
MOBA experience. Such teams would not need to learn the mechanics of the game and
would be focused solely on learning about their teammates. Such teams would also allow
the exploration of TMS in cases where there may be discrepancy between player
expertise and avatar expertise. More experienced teams would also know that constant
communication is necessary. More experienced players may also be willing to participate
in a more longitudinal study, as three matches may not have been enough to fully trace
the trajectory of the relationship between social presence and TMS.
Studying the actual communication that goes on in the game would also be
interesting using IPA. This has only been done in two studies of TMS (Kanawattanachai
& Yoo, 2007; Richter & Lechner, 2009). One was in a work-oriented team, as opposed to
a socioemotional team like a video game team, and the one that did deal with a video
game team only had a sample size of six and only reported descriptive statistics without
evaluating the relationship of different types of communication with TMS. Given the
relational nature of video game teams and given that relational communication in such
teams was positively related to TMS, it is possible that earlier findings about task-
oriented communication being positively related to TMS will not hold true.
113

Finally, SIP should be revisited. The original study should be replicated and
communication experience should be examined as a covariate. Do more experienced
CMC users not differ in relational communication as compared to FtF groups? Do
younger adults already have strategies in place that make the lack of nonverbal cues
irrelevant?
114

References
Akgün, A. E., Byrne, J., Keskin, H., Lynn, G. S., & Imamoglu, S. Z. (2005). Knowledge
networks in new product development projects: A transactive memory
perspective. Information & Management, 42(8), 1105-1120.
doi:10.1016/j.im.2005.01.001
Ashleigh, M., & Prichard, J. (2012). An integrative model of the role of trust in
transactive memory development. Group & Organization Management, 37(1), 5-
35. doi:10.1177/1059601111428449
Austin, J. R. (2003). Transactive memory in organizational groups: The effects of
content, consensus, specialization, and accuracy on group performance. Journal
of Applied Psychology, 88(5), 866-878.
Bainbridge, W. S. (2007). The scientific research potential of virtual worlds. Science,
317(5837), 472-476. doi:10.2307/20037445
Bales, R. F. (1950). Interaction process analysis: A method for the study of small groups.
Cambridge, MA: Addison-Wesley.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in
social psychological research: Conceptual, strategic, and statistical considerations.
Journal of Personality and Social Psychology, 51(6), 1173-1182.
doi:10.1037/0022-3514.51.6.1173
Biocca, F., & Harms, C. (2003). Guide to the networked minds social presence inventory
v. 1.2 Retrieved October 1, 2010, from
115

http://www.mindlab.msu.edu/Biocca/pubs/papers/2002_guide_netminds_measure
.pdf
Biocca, F., Harms, C., & Burgoon, J. K. (2003). Toward a more robust theory and
measure of social presence: Review and suggested criteria. Presence:
Teleoperators and Virtual Environments, 12(5), 456-480.
doi:10.1162/105474603322761270
Biocca, F., Harms, C., & Gregg, J. (2001). The networked minds measure of social
presence: Pilot test of the factor structure and concurrent validity. Paper
presented at the 4th annual International Workshop on Presence, Philadelphia,
PA.
Bracken, C. C., & Skalski, P. (2006). Presence and video games: The impact of image
quality and skill level. Paper presented at the 9th annual International Workshop
on Presence, Cleveland, OH.
Brandon, D. P., & Hollingshead, A. B. (2004). Transactive memory systems in
organizations: Matching tasks, expertise, and people. Organization Science,
15(6), 633-644. doi:10.1287/orsc.1040.0069
Burgoon, J. K., & Hale, J. L. (1984). The fundamental topoi of relational communication.
Communication Monographs, 51(3), 193.
Burgoon, J. K., & Hale, J. L. (1987). Validation and measurement of the fundamental
themes of relational communication. Communication Monographs, 54(1), 19-41.
doi:10.1080/03637758709390214
116

Castronova, E., & Falk, M. (2009). Virtual worlds: Petri dishes, rat mazes, and
supercolliders. Games and Culture, 4(4), 396-407.
doi:10.1177/1555412009343574
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.).
Hillsdale, NJ: Lawrence Erlbaum.
Ducheneaut, N., Yee, N., Nickell, E., & Moore, R. J. (2006). "Alone together?"
Exploring the social dynamics of massively multiplayer online games. Paper
presented at the CHI 2006, Montréal, Québec, Canada.
Engelbart, D. C. (1962). Augmenting human intellect: A conceptual framework. Menlo
Park, CA: Stanford Research Institute.
Entertainment Software Association. (2012). Essential facts about the computer and
video game industry Retrieved March 1, 2013, from
http://www.theesa.com/facts/pdfs/ESA_EF_2012.pdf
Fulk, J., Schmitz, J., & Steinfeld, C. W. (1990). A social influence model of technology
use. In J. Fulk & C. W. Steinfeld (Eds.), Organizations and communication
technology (pp. 117-140). Newbury Park, CA: Sage.
Fulk, J., Steinfeld, C. W., Schmitz, J., & Power, J. G. (1987). A social information
processing model of media use in organizations. Communication Research, 14(5),
529-552. doi:10.1177/009365087014005005
Gould, J. J. (2011). Examining affect and transactive communicative processes in
organizational teams. Unpublished doctoral dissertation, University of Southern
California, Los Angeles, CA.
117

Griffiths, M. D., Davies, M. N. O., & Chappell, D. (2004). Demographic factors and
playing variables in online computer gaming. CyberPsychology & Behavior, 7(4),
479-487. doi:10.1089/cpb.2004.7.479
Harms, C., & Biocca, F. (2004). Internal consistency and reliability of the networked
minds social presence measure. Paper presented at the 7th annual International
Workshop on Presence, Valencia, Spain.
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process
analysis: A regression-based approach. New York, NY: Guilford Press.
Hollingshead, A. B. (1998a). Communication, learning, and retrieval in transactive
memory systems. Journal of Experimental Social Psychology, 34(5), 423-442.
doi:10.1006/jesp.1998.1358
Hollingshead, A. B. (1998b). Retrieval processes in transactive memory systems. Journal
of Personality and Social Psychology, 74(3), 659-671. doi:10.1037/0022-
3514.74.3.659
Hollingshead, A. B. (2000). Perceptions of expertise and transactive memory in work
relationships. Group Processes & Intergroup Relations, 3(3), 257-267.
doi:10.1177/1368430200033002
Hollingshead, A. B. (2001). Cognitive interdependence and convergent expectations in
transactive memory. Journal of Personality and Social Psychology, 81(6), 1080-
1089. doi:10.1037/0022-3514.81.6.1080
Hollingshead, A. B. (2010). Communication, coordinated action, and focal points in
groups: From dating couples to emergency responders. In C. R. Agnew, D. E.
118

Carlson, W. G. Graziano & J. R. Kelly (Eds.), Then a miracle occurs: Focusing
on behavior in social psychological theory and research (pp. 391-410). New
York, NY: Oxford University Press.
Hollingshead, A. B., & Brandon, D. P. (2003). Potential benefits of communication in
transactive memory systems. Human Communication Research, 29(4), 607-615.
doi:10.1111/j.1468-2958.2003.tb00859.x
Hollingshead, A. B., Brandon, D. P., Yoon, K., & Gupta, N. (2011). Communication and
knowledge-sharing errors in groups: A transactive memory perspective. In H. E.
Canary & R. D. McPhee (Eds.), Communication and organizational knowledge:
Contemporary issues for theory and practice (pp. 133-150). New York, NY:
Routledge.
Hollingshead, A. B., & Fraidin, S. N. (2003). Gender stereotypes and assumptions about
expertise in transactive memory. Journal of Experimental Social Psychology,
39(4), 355-363. doi:10.1016/s0022-1031(02)00549-8
Hollingshead, A. B., Mcgrath, J. E., & O'Connor, K. M. (1993). Group task performance
and communication technology: A longitudinal study of computer-mediated
versus face-to-face work groups. Small Group Research, 24(3), 307-333.
doi:10.1177/1046496493243003
Jackson, M., & Moreland, R. L. (2009). Transactive memory in the classroom. Small
Group Research, 40(5), 508-534. doi:10.1177/1046496409340703
119

James, L. R., Demaree, R. G., & Wolf, G. (1984). Estimating within-group interrater
reliability with and without response bias. Journal of Applied Psychology, 69(1),
85-98. doi:10.1037/0021-9010.69.1.85
James, L. R., Demaree, R. G., & Wolf, G. (1993). r
wg
: An assessment of within-group
interrater agreement. Journal of Applied Psychology, 78(2), 306-309.
doi:10.1037/0021-9010.78.2.306
Jarvenpaa, S. L., & Majchrzak, A. (2008). Knowledge collaboration among professionals
protecting national security: Role of transactive memories in ego-centered
knowledge networks. Organization Science, 19(2), 260-276.
doi:10.1287/orsc.1070.0315
Jin, S.-A. A. (2010). Parasocial interaction with an avatar in Second Life: A typology of
the self and an empirical test of the mediating role of social presence. Presence:
Teleoperators and Virtual Environments, 19(4), 331-340.
doi:10.1162/PRES_a_00001
Jin, S.-A. A. (2011). "It feels right. Therefore, I feel present and enjoy": The effects of
regulatory fit and the mediating roles of social presence and self-presence in
avatar-based 3D virtual environments. Presence: Teleoperators and Virtual
Environments, 20(2), 105-116. doi:10.1162/pres_a_00038
Kahn, A. S., Ratan, R. A., & Williams, D. (2010). Why we distort in self-report: The
effects of cognitive dissonance and balance theory on self-report errors. Paper
presented at the 60th annual conference of the International Communication
Association, Singapore.
120

Kanawattanachai, P., & Yoo, Y. (2007). The impact of knowledge coordination on virtual
team performance over time. MIS Quarterly, 31(4), 783-808.
Keyton, J. (2000). The relational side of groups. Small Group Research, 31(4), 387-396.
doi:10.1177/104649640003100401
Klimmt, C., & Hartmann, T. (2008). Mediated interpersonal communication in
multiplayer video games: Implications for entertainment and relationship
management. In E. A. Konijn, S. Utz, M. Tanis & S. B. Barnes (Eds.), Mediated
interpersonal communication (pp. 309-330). New York, NY: Routledge.
LeBreton, J. M., & Senter, J. L. (2008). Answers to 20 questions about interrater
reliability and interrater agreement. Organizational Research Methods, 11(4),
815-852. doi:10.1177/1094428106296642
Lee, K. M. (2004a). Presence, explicated. Communication Theory, 14(1), 27-50.
doi:10.1111/j.1468-2885.2004.tb00302.x
Lee, K. M. (2004b). Why presence occurs: Evolutionary psychology, media equation,
and presence. Presence: Teleoperators and Virtual Environments, 13(4), 494-505.
doi:10.1162/1054746041944830
Lee, K. M., & Nass, C. (2004). The multiple source effect and synthesized speech:
Doubly-disembodied language as a conceptual framework. Human
Communication Research, 30(2), 182-207. doi:10.1111/j.1468-
2958.2004.tb00730.x
121

Lee, K. M., & Nass, C. (2005). Social-psychological origins of feelings of presence:
Creating social presence with machine-generated voices. Media Psychology, 7(1),
31-45. doi:10.1207/S1532785XMEP0701_2
Lee, K. M., Peng, W., Jin, S.-A., & Yan, C. (2006). Can robots manifest personality?: An
empirical test of personality recognition, social responses, and social presence in
human–robot interaction. Journal of Communication, 56(4), 754-772.
doi:10.1111/j.1460-2466.2006.00318.x
Lewis, K. (2003). Measuring transactive memory systems in the field: Scale development
and validation. Journal of Applied Psychology, 88(4), 587-604. doi:10.1037/0021-
9010.88.4.587
Lewis, K. (2004). Knowledge and performance in knowledge-worker teams: A
longitudinal study of transactive memory systems. Management Science, 50(11),
1519-1533. doi:10.1287/mnsc.1040.0257
Lewis, K., Belliveau, M., Herndon, B., & Keller, J. (2007). Group cognition, membership
change, and performance: Investigating the benefits and detriments of collective
knowledge. Organizational Behavior and Human Decision Processes, 103(2),
159-178. doi:10.1016/j.obhdp.2007.01.005
Lewis, K., & Herndon, B. (2011). Transactive memory systems: Current issues and future
research directions. Organization Science, 22(5), 1254-1265.
doi:10.1287/orsc.1110.0647
122

Lewis, K., Lange, D., & Gillis, L. (2005). Transactive memory systems, learning, and
learning transfer. Organization Science, 16(6), 581-598.
doi:10.1287/orsc.1050.0143
Liang, D. W., Moreland, R., & Argote, L. (1995). Group versus individual training and
group performance: The mediating role of transactive memory. Personality and
Social Psychology Bulletin, 21(4), 384-393. doi:10.1177/0146167295214009
Lim, S., & Reeves, B. (2010). Computer agents versus avatars: Responses to interactive
game characters controlled by a computer or other player. International Journal
of Human-Computer Studies, 68(1-2), 57-68. doi:10.1016/j.ijhcs.2009.09.008
Lindell, M. K., & Whitney, D. J. (2001). Accounting for common method variance in
cross-sectional research designs. Journal of Applied Psychology, 86(1), 114-121.
doi:10.1037/0021-9010.86.1.114
Littlepage, G. E., & Mueller, A. L. (1997). Recognition and utilization of expertise in
problem-solving groups: Expert characteristics and behavior. Group Dynamics:
Theory, Research, and Practice, 1(4), 324-328. doi:10.1037/1089-2699.1.4.324
Littlepage, G. E., Schmidt, G. W., Whisler, E. W., & Frost, A. G. (1995). An input-
process-output analysis of influence and performance in problem-solving groups.
Journal of Personality and Social Psychology, 69(5), 877-889.
doi:dx.doi.org/10.1037/0022-3514.69.5.877
Lowry, P. B., Roberts, T. L., Romano Jr., N. C., Cheney, P. D., & Hightower, R. T.
(2006). The impact of group size and social presence on small-group
123

communication: Does computer-mediated communication make a difference?
Small Group Research, 37(6), 631-661. doi:0.1177/1046496406294322
Majchrzak, A., Jarvenpaa, S., & Hollingshead, A. B. (2007). Coordinating expertise
among emergent groups responding to disasters. Organization Science, 18(1),
147-161. doi:10.1287/orsc.1060.0228
McGrath, J. E. (1984). Groups: Interaction and performance. Inglewood, NJ: Prentice
Hall.
Moreland, R. L. (1999). Transactive memory: Learning who knows what in work groups
and organizations. In L. L. Thompson, J. M. Levine & D. M. Messick (Eds.),
Shared cognition in organizations: The management of knowledge (pp. 3-31).
Mahwah, NJ: Lawrence Erlbaum.
Moreland, R. L., Levine, J. M., & Wingert, M. L. (1996). Creating the ideal group:
Composition effects at work. In E. Witte & J. H. Davis (Eds.), Understanding
group behavior: Small group processes and interpersonal relations (Vol. 2, pp.
11-35). Mahwah, NJ: Lawrence Erlbaum.
Moreland, R. L., & Myaskovsky, L. (2000). Exploring the performance benefits of group
training: Transactive memory or improved communication? Organizational
Behavior and Human Decision Processes, 82(1), 117-133.
doi:10.1006/obhd.2000.2891
Moreland, R. L., Swanenburg, K. L., Flagg, J. J., & Fetterman, J. D. (2010). Transactive
memory and technology in work groups and organizations. In B. Ertl (Ed.), E-
124

collaborative knowledge construction: Learning from computer-supported and
virtual environments (pp. 244-274). Geneva, Switzerland: IGI Global Press.
Palazzolo, E. T., Serb, D. A., She, Y., Su, C., & Contractor, N. S. (2006). Coevolution of
communication and knowledge networks in transactive memory systems: Using
computational models for theoretical development. Communication Theory,
16(2), 223-250. doi:10.1111/j.1468-2885.2006.00269.x
Peña, J., & Hancock, J. T. (2006). An analysis of socioemotional and task communication
in online multiplayer video games. Communication Research, 33(1), 92-109.
doi:10.1177/0093650205283103
Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. (2003). Common
method biases in behavioral research: A critical review of the literature and
recommended remedies. Journal of Applied Psychology, 88(5), 879-903.
doi:10.1037/0021-9010.88.5.879
Podsakoff, P. M., & Organ, D. W. (1986). Self-reports in organizational research:
Problems and prospects. Journal of Management, 12(4), 531-544.
doi:10.1177/014920638601200408
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect
effects in simple mediation models. Behavior Research Methods, Instruments, &
Computers, 36(4), 717-731. doi:10.3758/BF03206553
Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for
assessing and comparing indirect effects in multiple mediator models. Behavior
Research Methods, 40(3), 879-891. doi:10.3758/brm.40.3.879
125

Preacher, K. J., & Kelley, K. (2011). Effect size measures for mediation models:
Quantitative strategies for communicating indirect effects. Psychological
Methods, 16(2), 93-115. doi:10.1037/0022-3514.51.6.1173380635410.1037/0022-
3514.51.6.11731987-13085-001
Ravaja, N., Saair, T., Turpeinen, M., Laarni, J., Salminen, M., & Kivikangas, M. (2006).
Spatial presence and emotions during video game playing: Does it matter with
whom you play? Presence: Teleoperators and Virtual Environments, 15(4), 381-
392. doi:10.1162/pres.15.4.381
Rice, R. E., & Love, G. (1987). Electronic emotion: Socioemotional content in a
computer-mediated communication network. Communication Research, 14(1),
85-108. doi:10.1177/009365087014001005
Richter, S., & Lechner, U. (2009). Transactive memory systems and shared situation
awareness: A World of Warcraft experiment. Paper presented at the International
Conference on Organizational Learning, Knowledge and Capabilities, The
Netherlands.
Riedl, B. C., Gallenkamp, J. V., Picot, A., & Welpe, I. M. (2012). Antecedents of
transactive memory systems in virtual teams: The role of communication, culture,
and team size. Paper presented at the 45th Hawaii International Conference on
System Sciences, Maui, HI.
Riot Games. (2012). League of Legends community infographic Retrieved November 1,
2012, from http://na.leagueoflegends.com/news/league-legends-community-
infographic
126

Shen, C. (2010). The emergence, evolution and effects of social networks in online
communities Unpublished doctoral dissertation. Annenberg School for
Communication & Journalism. University of Southern California. Los Angeles,
CA.
Short, J., Williams, E., & Christie, B. (1976). The social psychology of
telecommunications. London, UK: John Wiley & Sons.
Skalski, P., & Tamborini, R. (2007). The role of social presence in interactive agent-
based persuasion. Media Psychology, 10(3), 385-413.
doi:10.1080/15213260701533102
Steiner, I. D. (1972). Group processes and productivity. New York, NY: Academic Press.
Walther, J. B. (1992). Interpersonal effects in computer-mediated interaction: A relational
perspective. Communication Research, 19(1), 52-90.
doi:10.1177/009365092019001003
Walther, J. B. (1993). Impression development in computer-mediated interaction.
Western Journal of Communication, 57(4), 381-398.
Walther, J. B. (1994). Anticipated ongoing interaction versus channel effects on
relational communication in computer-mediated interaction. Human
Communication Research, 20(4), 473-501. doi:10.1111/j.1468-
2958.1994.tb00332.x
Walther, J. B. (1995). Relational aspects of computer-mediated communication:
Experimental observations over time. Organization Science, 6(2), 186-203.
doi:10.1287/orsc.6.2.186
127

Walther, J. B. (1996). Computer-mediated communication: Impersonal, interpersonal,
and hyperpersonal interaction. Communication Research, 23(1), 3-43.
doi:10.1177/009365096023001001
Walther, J. B. (2006). Nonverbal dynamics in computer-mediated communication, or : (
and the net : ( ‘s with you, : ) and you : ) alone. In V. Manusov & M. L.
Patterson (Eds.), The SAGE handbook of nonverbal communication (pp. 461-
479). Thousand Oaks, CA: Sage.
Walther, J. B. (2011). Theories of computer-mediated communication and interpersonal
relations. In M. L. Knapp & J. A. Daly (Eds.), The Sage handbook of
interpersonal communication (4th ed., pp. 443-479). Thousand Oaks, CA: Sage.
Walther, J. B., & Burgoon, J. K. (1992). Relational communication in computer-mediated
interaction. Human Communication Research, 19(1), 50-88. doi:10.1111/j.1468-
2958.1992.tb00295.x
Walther, J. B., & Parks, M. R. (2002). Cues filtered out, cues filtered in: Computer-
mediated communication and relationships. In M. L. Knapp & J. A. Daly (Eds.),
Handbook of interpersonal communication (3rd ed., pp. 529-563). Thousand
Oaks, CA: Sage.
Webb, E. J., Campbell, D. T., Schwartz, R. D., & Sechrest, L. (2000). Unobtrusive
measures (revised ed.). Thousand Oaks, CA: Sage.
Wegner, D. M. (1987). Transactive memory: A contemporary analysis of the group mind
In B. Mullen & G. R. Goethals (Eds.), Theories of group behavior (pp. 185-208).
New York, NY: Spring-Verlag.
128

Weibel, D., Wissmath, B., Habegger, S., Steiner, Y., & Groner, R. (2008). Playing online
games against computer- vs. human-controlled opponents: Effects on presence,
flow, and enjoyment. Computers in Human Behavior, 24(5), 2274-2291.
doi:10.1016/j.chb.2007.11.002
Whorf, B. L. (1956). Language, thought, and reality: Selected writings of Benjamin Lee
Whorf. Cambridge, MA: MIT Press.
Williams, D. (2008). The promises and perils of large-scale data extraction. Unpublished
manuscript.
Williams, D. (2010). The mapping principle, and a research framework for virtual
worlds. Communication Theory, 20(4), 451-470. doi:10.1111/j.1468-
2885.2010.01371.x
Williams, D., Caplan, S. E., & Xiong, L. (2007). Can you hear me now? The impact of
voice in an online gaming community. Human Communication Research, 33(4),
427-449. doi:10.1111/j.1468-2958.2007.00306.x
Williams, D., Contractor, N., Poole, M. S., Srivastava, J., & Cai, D. (2011). The Virtual
Worlds Exploratorium: Using large-scale data and computational techniques for
communication research. Communication Methods & Measures, 5(2), 163-180.
doi:10.1080/19312458.2011.568373
Williams, D., & Kahn, A. S. (2013). Games, online and off. In W. H. Dutton (Ed.), The
Oxford Handbook of Internet Studies (pp. 195-215). Oxford, UK: Oxford
University Press.
129

Wilson, J. M., Straus, S. G., & McEvily, B. (2006). All in due time: The development of
trust in computer-mediated and face-to-face teams. Organizational Behavior and
Human Decision Processes, 99(1), 16-33. doi:10.1016/j.obhdp.2005.08.001
Wirth, J. H., Feldberg, F., Schouten, A., van den Hooff, B., & Williams, K. (2012). Using
virtual game environments to study group behavior. In A. B. Hollingshead & M.
S. Poole (Eds.), Research methods for studying groups and teams: A guide to
approaches, tools, and technologies (pp. 173-198). New York, NY: Routledge.
Wittenbaum, G. M. (2012). Running laboratory experiments with groups. In A. B.
Hollingshead & M. S. Poole (Eds.), Research methods for studying groups and
teams: A guide to approaches, tools, and technologies (pp. 41-57). New York,
NY: Routledge.
Xiong, L., Poole, M. S., Williams, D., & Ahmad, M. A. (2009). The effects of group
structure on group behavior and outcomes in an online gaming environment.
Paper presented at the INGroup Annual Conference, Boulder, CO.
Yoo, Y., & Kanawattanachai, P. (2001). Developments of transactive memory systems
and collective mind in virtual teams. The International Journal of Organizational
Analysis, 9(2), 187-208. doi:10.1108/eb028933
Yuan, Y. C., Bazarova, N. N., Fulk, J., & Zhang, Z.-X. (2013). Recognition of expertise
and perceived influence in intercultural collaboration: A study of mixed American
and Chinese groups. Journal of Communication, 63(3), 476-497.
doi:10.1111/jcom.12026
130

Yuan, Y. C., Fulk, J., & Monge, P. R. (2007). Access to information in connective and
communal transactive memory systems. Communication Research, 34(2), 131-
155. doi:10.1177/0093650206298067
Yuan, Y. C., Fulk, J., Monge, P. R., & Contractor, N. (2010). Expertise directory
development, shared task interdependence, and strength of communication
network ties as multilevel predictors of expertise exchange in transactive memory
work groups. Communication Research, 37(1), 20-47.
doi:10.1177/0093650209351469

131

Appendix A: Measures from Study 1
Transactive Memory System Scale (adapted from Lewis, 2003)
(These questions were administered via Qualtrics online survey system and asked on a 5-
point Likert scale and asked respondents to think about the last match they played.)
Specialization.
1. Each player in my team had a specialized role to play.
2. I know which player was good at what role in my team.
3. Different team members were responsible in different roles.
Credibility.
1. I did not have faith in other members’ skills. (reverse coded)
2. I was confident about my team members’ skills during that match.
3. I trusted that other team members could accomplish their roles.
Coordination.
1. Our team worked together in a well-coordinated fashion.
2. Our team had very few misunderstandings about what to do.
3. We worked together efficiently and comfortably.
132

Social Presence Scale (adapted from Biocca & Harms, 2003)
(These questions were administered via Qualtrics online survey system and asked on a 5-
point Likert scale and asked respondents to think about the last match they played.)
Copresence.
I often felt as if my teammates and I were in the same physical space.
Perceived attentional engagement.
I paid close attention to my teammates.
Perceived emotional contagion.
I was sometimes influenced by my teammates' moods.
Perceived comprehension.
I was able to communicate my intentions clearly to my teammates.
Perceived behavioral interdependence.
My actions were often dependent on my teammates' actions.

133

Appendix B: Measures from Study 2
Transactive Memory System Scale (adapted from Lewis, 2003)
Questions 1 through 5 are the specialization subscale.
Questions 6 through 10 are the credibility subscale.
Questions 11 through 15 are the coordination subscale.
*
indicates reverse coded question
For the following questions, circle the number that reflects the extent to which you agree
or disagree with the statements.
1 = Strongly disagree
2 = Disagree
3 = Neutral
4 = Agree
5 = Strongly agree
1. Each player in my team had a specialized role to play.
Strongly disagree 1 2 3 4 5 Strongly agree
2. I had a role for an aspect of the game that no other team member had.
Strongly disagree 1 2 3 4 5 Strongly agree
3. Different team members were responsible for roles in different areas.
Strongly disagree 1 2 3 4 5 Strongly agree
4. The specialized roles of several different team members were needed to complete
the objectives of the game.
Strongly disagree 1 2 3 4 5 Strongly agree
134

5. I knew which player was good at what role in my team.
Strongly disagree 1 2 3 4 5 Strongly agree
6. I was comfortable accepting procedural suggestions from other team members.
Strongly disagree 1 2 3 4 5 Strongly agree
7. I trusted that other team members could accomplish their roles.
Strongly disagree 1 2 3 4 5 Strongly agree
8. I was confident relying on the skills that other team members brought to the
game.
Strongly disagree 1 2 3 4 5 Strongly agree
9. When other members gave suggestions, I had to think about whether I wanted to
take the suggestion.
*

Strongly disagree 1 2 3 4 5 Strongly agree
10. I did not have much faith in other members’ roles.
*

Strongly disagree 1 2 3 4 5 Strongly agree
11. Our team worked together in a well-coordinated fashion.
Strongly disagree 1 2 3 4 5 Strongly agree
12. Our team had very few misunderstandings about what to do.
Strongly disagree 1 2 3 4 5 Strongly agree

135

13. Our team needed to backtrack and start over a lot.
*

Strongly disagree 1 2 3 4 5 Strongly agree
14. We accomplished the task smoothly and efficiently.
Strongly disagree 1 2 3 4 5 Strongly agree
15. There was much confusion about how we would accomplish the task.
*

Strongly disagree 1 2 3 4 5 Strongly agree

136

Social Presence Scale (adapted from Harms & Biocca, 2004)
Questions 1 through 6 are the copresence subscale.
Questions 7 through 12 are the attentional allocation subscale.
Questions 13 through 18 are the perceived message understanding subscale.
Questions 19 through 24 are the perceived affective understanding subscale.
Questions 25 through 30 are the perceived emotional interdependence subscale.
Questions 31 through 36 are the perceived behavioral interdependence subscale.
*
indicates reverse coded question
For the following questions, circle the number that reflects the extent to which you agree
or disagree with the statements.
1 = Strongly disagree
2 = Disagree
3 = Slightly disagree
4 = Neither agree nor disagree
5 = Slightly agree
6 = Agree
7 = Strongly agree
1. I noticed my teammates in the game.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree

137

2. My teammates noticed me in the game.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
3. My teammates’ presence in the game was obvious to me.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
4. My presence in the game was obvious to my teammates.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
5. My teammates caught my attention.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
6. I caught my teammates’ attention.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
7. I was easily distracted from my teammates when other things were going on.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
8. My teammates were easily distracted from me when other things were going on.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
9. I remained focused on my teammates throughout our interaction.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
10. My teammates remained focused on me throughout our interaction.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
11. My teammates did not receive my full attention.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
12. I did not receive my teammates’ full attention.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
138

13. My thoughts were clear to my teammates.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
14. My teammates’ thoughts were clear to me.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
15. It was easy to understand my teammates.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
16. My teammates found it easy to understand me.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
17. Understanding my teammates was difficult.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
18. My teammates had difficulty understanding me.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
19. I could tell how my teammates felt.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
20. My teammates could tell how I felt.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
21. My teammates’ emotions were not clear to me.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
22. My emotions were not clear to my teammates.
*

Strongly disagree 1 2 3 4 5 6 7 Strongly agree
23. I could describe my teammates’ feelings accurately.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
139

24. My teammates could describe my feelings accurately.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
25. I was sometimes influenced by my teammates’ moods.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
26. My teammates were sometimes influenced by my moods.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
27. My teammates’ feelings influenced the mood of our interaction.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
28. My feelings influenced the mood of our interaction.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
29. My teammates’ attitudes influenced how I felt.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
30. My attitudes influenced how my teammates felt.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
31. My behavior was often in direct response to my teammates’ behavior.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
32. The behavior of my teammates was often in direct response to my behavior.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
33. I reciprocated my teammates’ actions.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
34. My teammates reciprocated my actions.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
140

35. My teammates’ behavior was closely tied to my behavior.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree
36. My behavior was closely tied to my teammates’ behavior.
Strongly disagree 1 2 3 4 5 6 7 Strongly agree

141

Appendix C: Measures from Study 3
Transactive Memory System Scale (adapted from Lewis, 2003)
Questions 1 through 5 are the specialization subscale.
Questions 6 through 10 are the credibility subscale.
Questions 11 through 15 are the coordination subscale.
*
indicates reverse coded question
For the following questions, circle the number that reflects the extent to which you agree
or disagree with the statements.
1 = Strongly disagree
2 = Disagree
3 = Neutral
4 = Agree
5 = Strongly agree
1. Each player in my team had a specialized role to play.
Strongly disagree 1 2 3 4 5 Strongly agree
2. I had a role for an aspect of the game that no other team member had.
Strongly disagree 1 2 3 4 5 Strongly agree
3. Different team members were responsible for roles in different areas.
Strongly disagree 1 2 3 4 5 Strongly agree
4. The specialized roles of several different team members were needed to complete
the objectives of the game.
Strongly disagree 1 2 3 4 5 Strongly agree
142

5. I knew which player was good at what role in my team.
Strongly disagree 1 2 3 4 5 Strongly agree
6. I was comfortable accepting procedural suggestions from other team members.
Strongly disagree 1 2 3 4 5 Strongly agree
7. I trusted that other team members could accomplish their roles.
Strongly disagree 1 2 3 4 5 Strongly agree
8. I was confident relying on the skills that other team members brought to the
game.
Strongly disagree 1 2 3 4 5 Strongly agree
9. When other members gave suggestions, I had to think about whether I wanted to
take the suggestion.
*

Strongly disagree 1 2 3 4 5 Strongly agree
10. I did not have much faith in other members’ roles.
*

Strongly disagree 1 2 3 4 5 Strongly agree
11. Our team worked together in a well-coordinated fashion.
Strongly disagree 1 2 3 4 5 Strongly agree
12. Our team had very few misunderstandings about what to do.
Strongly disagree 1 2 3 4 5 Strongly agree

143

13. Our team needed to backtrack and start over a lot.
*

Strongly disagree 1 2 3 4 5 Strongly agree
14. We accomplished the task smoothly and efficiently.
Strongly disagree 1 2 3 4 5 Strongly agree
15. There was much confusion about how we would accomplish the task.
*

Strongly disagree 1 2 3 4 5 Strongly agree

144

Relational Communication Scale (adapted from Burgoon & Hale, 1987)
Questions 1 through 9 are the immediacy/affection subscale.
Questions 10 through 14 are the similarity/depth subscale.
Questions 15 through 20 are the receptivity/trust subscale.
Questions 21 through 25 are the composure subscale.
Questions 26 through 28 are the formality subscale.
Questions 29 through 34 are the dominance subscale.
Questions 35 through 37 are the equality subscale.
Questions 38 through 41 are the task orientation subscale.
*
indicates reverse coded question
For the following questions, circle the number that reflects the extent to which you agree
or disagree with the statements.
1 = Strongly disagree
2 = Disagree
3 = Slightly disagree
4 = Neither agree nor disagree
5 = Slightly agree
6 = Agree
7 = Strongly agree
1 2 3 4 5 6 7 1. My teammates were intensely involved in our conversation.
1 2 3 4 5 6 7 2. My teammates did not want a deeper relationship between us.
*

1 2 3 4 5 6 7 3. My teammates were not attracted to me.
*

145

1 2 3 4 5 6 7 4. My teammates found the conversation stimulating.
1 2 3 4 5 6 7 5. My teammates communicated coldness rather than warmth.
*

1 2 3 4 5 6 7 6. My teammates created a sense of distance between us.
*

1 2 3 4 5 6 7 7. My teammates acted bored by our conversation.
*

1 2 3 4 5 6 7 8. My teammates were interested in talking to me.
1 2 3 4 5 6 7 9. My teammates showed enthusiasm while talking to me.
1 2 3 4 5 6 7 10. My teammates made me feel they were similar to me.
1 2 3 4 5 6 7 11. My teammates tried to move the conversation to a deeper level.
1 2 3 4 5 6 7 12. My teammates acted like we were good friends.
1 2 3 4 5 6 7 13. My teammates seemed to desire further communication with me.
1 2 3 4 5 6 7 14. My teammates seemed to care if I liked them.
1 2 3 4 5 6 7 15. My teammates were sincere.
1 2 3 4 5 6 7 16. My teammates were interested in talking with me.
1 2 3 4 5 6 7 17. My teammates wanted me to trust them.
1 2 3 4 5 6 7 18. My teammates were willing to listen to me.
1 2 3 4 5 6 7 19. My teammates were open to my ideas.
1 2 3 4 5 6 7 20. My teammates were honest in communicating with me.
1 2 3 4 5 6 7 21. My teammates felt very tense talking to me.
*

1 2 3 4 5 6 7 22. My teammates were calm and poised with me.
1 2 3 4 5 6 7 23. My teammates felt very relaxed talking with me.
1 2 3 4 5 6 7 24. My teammates seemed nervous in my presence.
*

1 2 3 4 5 6 7 25. My teammates were comfortable interacting with me.
146

1 2 3 4 5 6 7 26. My teammates made the interaction very formal.
1 2 3 4 5 6 7 27. My teammates wanted the discussion to be casual.
*

1 2 3 4 5 6 7 28. My teammates wanted the discussion to be informal.
*

1 2 3 4 5 6 7 29. My teammates attempted to persuade me.
1 2 3 4 5 6 7 30. My teammates didn't attempt to influence me.
*

1 2 3 4 5 6 7 31. My teammates tried to control the interaction.
1 2 3 4 5 6 7 32. My teammates tried to gain my approval.
1 2 3 4 5 6 7 33. My teammates didn't try to win my favor.
*

1 2 3 4 5 6 7 34. My teammates had the upper hand in the conversation.
1 2 3 4 5 6 7 35. My teammates considered us equals.
1 2 3 4 5 6 7 36. My teammates did not treat me as an equal.
*

1 2 3 4 5 6 7 37. My teammates wanted to cooperate with me.
1 2 3 4 5 6 7 38. My teammates wanted to stick to the main purpose of the interaction.
1 2 3 4 5 6 7 39. My teammates were more interested in social conversation than the
task at hand.
*

1 2 3 4 5 6 7 40. My teammates were very work-oriented.
1 2 3 4 5 6 7 41. My teammates were more interested in working on the task at hand
than having social conversation.

Abstract (if available)

Abstract This dissertation examines transactive memory systems (TMS) in video game teams. Study 1 is a field study of 16,499 players of the video game League of Legends and uses survey and server data to explore the relationship between group composition (member self-selection and team size) and TMS. It also explores the mediating role of social presence. Study 2 is a League of Legends lab experiment comparing the effects of communication channel (text-chat vs. voice-chat) on the development of TMS and social presence in teams. Study 3, conducted concurrently with Study 2, tests Social Information Processing theory in the context of video game teams and examined the role of relational communication in the development of TMS.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Hybrid communication: a structurational analysis of computer gaming teams

PDF

The patterns, effects and evolution of player social networks in online gaming communities

PDF

Self-presence: body, emotion, and identity extension into the virtual self

PDF

Crisis and stasis: understanding change in online communities

PDF

Building social Legoland through collaborative crowdsourcing: marginality, functional diversity, and team success

PDF

The social meaning of sharing and geocoding: features and social processes in online communities

PDF

Collaborative care capacity: developing culture, power relationships and leadership support for team care in a hospital

PDF

The formation and influence of online health social networks on social support, self-tracking behavior and weight loss outcomes

PDF

Social media's role, utility, and future in video game public relations

PDF

Contagious: social norms about health in work group networks

PDF

Virtual worlds as contact zones: development, localization, and intergroup communication in MMORPGs

PDF

How social and human capital create financial capital in crowdfunding projects

PDF

Bounded technological rationality: the intersection between artificial intelligence, cognition, and environment and its effects on decision-making

PDF

Mapping out the transition toward information societies: social nature, growth, and policies

PDF

Forests, fires, and flights: examining safety and communication practices within aerial firefighting teams

PDF

Effects of economic incentives on creative project-based networks: communication, collaboration and change in the American film industry, 1998-2010

PDF

Building a unicorn: management of innovation, collaboration, and change in a Silicon Valley start-up

PDF

A multitheoretical multilevel explication of crowd-enabled organizations: exploration/exploitation, social capital, signaling, and homophily as determinants of associative mechanisms in donation-...

Asset Metadata

Creator Kahn, Adam Scott (author)

Core Title We're all in this (game) together: transactive memory systems, social presence, and social information processing in video game teams

School Annenberg School for Communication

Degree Doctor of Philosophy

Degree Program Communication

Publication Date 08/19/2013

Defense Date 07/09/2013

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag communication technology,OAI-PMH Harvest,small group communication,social information processing,social presence,transactive memory,video games

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Williams, Dmitri (committee chair), Fulk, Janet (committee member), Read, Stephen J. (committee member)

Creator Email adamkahn@usc.edu,adamskahn@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-321958

Unique identifier UC11288129

Identifier etd-KahnAdamSc-2008.pdf (filename),usctheses-c3-321958 (legacy record id)

Legacy Identifier etd-KahnAdamSc-2008.pdf

Dmrecord 321958

Document Type Dissertation

Rights Kahn, Adam Scott

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA