THE INTERACTION EFFECTS OF WORKING MEMORY CAPACITY,
GAMING EXPERTISE, AND SCAFFOLDING DESIGN ON ATTENTION AND
COMPREHENSION IN DIGITAL GAME BASED LEARNING
By
Yu-Hao Lee

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Media and Information Studies - Doctor of Philosophy
2013

ABSTRACT

THE INTERACTION EFFECTS OF WORKING MEMORY CAPACITY,
GAMING EXPERTISE, AND SCAFFOLDING DESIGN ON ATTENTION AND
COMPREHENSION IN DIGITAL GAME BASED LEARNING
By
Yu-Hao Lee
Educational digital games are often complex problem-solving experiences that can
facilitate systematic comprehension. However, empirical studies of digital game based learning
(DGBL) have found mixed results regarding DGBL’s effect in improving comprehension. While
learners generally enjoyed the DGBL learning experience, they often failed to comprehend the
underlying message and construct knowledge. One possible explanation for the mixed results is
that DGBL comprehension is moderated by individual differences in working memory and
gaming expertise (knowledge, skills). Due to the procedural nature of communication in game
design, learning from digital games requires processing multiple interacting elements within
one’s working memory. The amount of available working memory that can be allocated to
processing the educational messages is affected by the learner’s expertise about the game
mechanics, along with the structure of the game design.
This study seeks to investigate how working memory capacity and gaming skill expertise
interact with game scaffolding design to influence attention and comprehension. Two specific
game scaffolding designs—content-knowledge scaffolding and gaming-skill scaffolding are
examined in this study. The two designs are aimed to support learning by reducing the amount of
efforts that players need to exert to figure out how to play the game, thus reducing the demand
on their working memory and freeing up working memory for comprehension. But some studies

have shown that learners’ expertise level also moderates the effect of these designs. Learners
with more expertise do not benefit from scaffolding--a phenomenon known as expertise-reversal
effect or reverse cohesion effect.
This study has three goals: (1) to examine the influence of individual working memory
capacity and gaming expertise on attention and comprehension in the context of digital games;
(2) to examine the effect of scaffolding designs on attention and comprehension with regards to
working memory and expertise; and (3) to examine competing explanations from the expertisereversal effect and the reverse-cohesion effect.
The findings showed that gaming expertise significantly predicts attention when the
player can recognize the game design (the game mechanics) as within their gaming expertise
domain. Consistent with the expertise reversal effect and the reverse cohesion effect, players
with higher gaming-skill expertise paid more attention to the gaming-skill scaffolding design.
However, the findings suggest that gaming-skill experts and non-experts focused their attention
on different aspects of the game. Working memory capacity and attention only predicted
comprehension for learners with lower gaming-skill expertise. Working memory capacity and
attention did not predict comprehension for learners with higher gaming-skill expertise. The
gaming-skill scaffolding design was effective in mitigating the gap between learners with high
and low gaming-skill expertise. However the content-knowledge scaffolding widened the gap
between learners with relative high and low prior content knowledge. The study argues that skillbased scaffolding can improve comprehension for non-experts in the context of digital game
based learning. But whether scaffolding helps or harms comprehension for gaming experts
depends on if the design activates retrieval of expert mental models and whether the game’s
narrative is closely aligned with its underlying messages.

ACKNOWLEDGEMENTS

Coming from the sub-tropical island of Taiwan, I never imagined that I would spend four
of the best years of my life living, learning, and developing myself to become a researcher in the
snow-covered state of Michigan. Looking back at these years, I am truly grateful for the many
people who have supported me in so many ways.
I would like to thank Dr. Carrie Heeter, who is my advisor, mentor, and dear friend. You
are an inspiration to me intellectually and as a person. I am constantly amazed by how much
things you have on your schedule, yet you always make it top priority to help me and your
students. Every time I write to you about my developing ideas and thoughts, you can always
point out critical problems and help me comb through my tangled ideas. And your genuine trust
and care for your students is something that I will keep with me and pass on to my students. A
big hug for Carrie!
I would like to thank my committee members. Dr. Wei Peng and Dr. Rabindra (Robby)
Ratan who have always been so supportive of my scholarship, and always respond to my knockon-the-door questions with patience and allow my short questions to extend to long discussions.
My dissertation would not have been possible without your advices on experiment design,
measurements, and analyses. I am grateful to Dr. Rand Spiro for supporting my study with your
knowledge of cognitive science and education theories. I look forward to that coffee (or tea) with
you to talk about our thought on the cognitive load theories.
Besides my dissertation committee members, I would like to thank Dr. Gary Hsieh, who
shares many research interests as mine. Collaborating with you is a challenging and exciting
iv

experience. I look forward to more collaboration with you to investigate motivations and civic
actions. I would like to thank Dr. Jonathan Obar, who introduced me to the field of wiki research
and design. It has been an interesting experience working with content experts on designing the
wiki platform to actually apply theory to practice.
A big thank you goes to Culver Redd who helped me set up the experiment website and
captured server-based behavioral data. I am most grateful for your understanding even when the
project was delayed. But you always responded to the developing changes immediately and
calmly. You are the best programmer I have worked with, and I wish you have a bright career
ahead of you. I would also like to thank Chris Swain, lead developer of the ReDistricting Game,
for allowing us to access the game server and capture behavioral data. The study would not have
been possible without you. Thank you to the undergraduate students who participated in my
experiment, for contributing your time and thoughts, and showing up on time.
I have been most fortunate to have cohorts who are dedicated to research and support one
another intellectually and emotionally. I would like to thank Yvette Wohn, who is amazingly
productive and has been my emotional support in the doctoral training process and job-searching
process; Dr. Han-ei Chew, Kanni Huang, Hsin-Yi (Sandy) Tsai, Guanxiong Huang, Wenjuan
Ma, and Mi Jun Kim, who are my writing buddies and great friends. I always look forward to our
weekly meeting to exchange new ideas (and gossips of course). Dr. Tammy Lin, who is one of
the most cheerful scholar I have ever met. Thank you to Dr. Ken Sun and Tina Yuan, for always
being available for me to discuss my problems and share my excitements. I am also grateful for
Ryan Yang and Wan-Hsuan Lin, who has been our family here in East Lansing. We will miss the
many New Years, Summer BBQ, Thanksgivings, and Christmas that we spent together.

v

I would also like to thank the graduate school and College of Communication Arts and
Sciences for supporting my study with the Dissertation Completion Fellowship and 2013
Summer Excellence Research Fellowship.
I am grateful to have the trust of my family members: My father and mother who
supported me in coming back to school to seek advanced degrees; my father and mother-in-law
who gave me their full trust. And my sister who is now an experienced reporter, who sends post
cards to us wherever she goes.
Finally, my greatest thank you goes to my wife Chien-Ying Wang, who came with me to
Michigan on this big wild adventure. Thank you for being there for me to share my happiness,
frustration, anger, and achievements. The past year has been both stressful and exciting, with
dissertation, job search, and the birth of our beautiful daughter Mika Lee. I love you and I
believe we will have many more amazing adventures to come.

vi

TABLE OF CONTENTS
LIST OF TABLES ……………………………………………………………………………… ix
LIST OF FIGURES ……………………………………………………………………………... x
Chapter 1
Introduction ……...……..………………………………………………………………………. 1
Theoretical and Practical Contributions ……………………….........…………………… 5
Chapter 2
Digital Game Based Learning and Working Memory Capacity ……………………....…… 7
Digital Game Based Learning and Working Memory ………………………........…….. 7
Digital games as motivational tools …………………………………………….. 7
Digital games as complex problem-solving …………………………………….. 9
Working Memory Capacity in Digital Games …………………………………………. 11
What is Working Memory? ……………………………………………………………. 12
The Structure of Working Memory ……………………………………………………. 16
The Individual Difference Approach …………………………………………………... 19
Attention capacity difference ………………………………………………...… 19
Storage capacity difference …………………………………………………….. 24
Domain-Related Expertise and Comprehension ……………………………………….. 26
Chapter 3
Game Design and Cognitive Load ………………….………………………………………... 32
Fisch’s Capacity Model of Comprehension from Educational Media …………………. 32
Scaffolding, Expertise-Reversal Effect, and Reverse-Cohesion Effect ………………... 35
Chapter 4
Experiment Design and Measurements ……………………………………………………... 42
Participants …………………………………………………………………………….. 42
Stimuli and Learning Goals ……………………………………………………………. 43
Operational Measures ………………………………………………………………….. 45
Independent variables ………………………………………………………….. 45
Scaffolding conditions …………………………………………………. 45
Gaming skill expertise …………………………………………………. 47
Working memory capacity ……………………………………………... 47
Dependent variables ……………………………………………………………. 48
Attention ……………………………………………………………….. 48
Comprehension ……………………………………………………….... 51
Control variables ……………………………………………………………….. 51
Research Procedures ………………………………………………………………….... 52
Manipulation Checks …………………………………………………………………... 55
Chapter 5
Results ………..……………………...………………………………………………………… 57
vii

Sample …………………………………………………………………………………. 57
Hypotheses testing ……………………………………………………………………... 60
Additional Results: Changes in Perceived Knowledge and Problem Severity ………… 68
Chapter 6
Discussions ………...…………………………...…………………………………………….... 71
Summary and Discussion of Findings …………………………………………………. 71
Working memory, gaming-skill expertise, and attention ………………………. 72
Expertise-reversal effect, scaffolding, and comprehension ……………………. 76
Scaffolding design to support non-experts …………………………………….. 80
Limitations ……………………………………………………………………………... 81
Chapter 7
Conclusion and Implications ………..……...…………………………...…...………………. 85
Theoretical Implications ……………………………………………………………….. 85
Practical and Design Implications ……………………………………………………... 87
APPENDICES …………………………………………………………………………………. 90
Appendix 1: Pre-test questionnaire …………………………………………………….. 91
Appendix 2: Instructions for the stimuli ……………………………………………….. 95
Appendix 3: Post-test questionnaire …………………………………………………… 99
BIBLIOGRAPHY …………………………………………………………………………..… 104

viii

LIST OF TABLES

Table 1: Comparison of induced scaffolding conditions …………………………………….… 47
Table 2: Regression analysis of working memory capacity and gaming expertise on attention in
no-scaffolding condition ………………………………………………...……………………... 61
Table 3: Regression analysis of working memory capacity and gaming expertise on attention in
gaming-skill scaffolding condition …………………………………………………………….. 62
Table 4: ANCOVA comparison of attention between scaffolding conditions ……………….... 64
Table 5: Regression analysis of attention on comprehension (non-expert group) …….……….. 66
Table 6: Changes in perceived knowledge and perceived problem after playing the game …… 69

ix

LIST OF FIGURES

Figure 1: Baddeley’s (2000) structural model of working memory ………………………….... 15
Figure 2: The effect of gaming-skill scaffolding on comprehension …………………………... 67
Figure 3: The effect of content-knowledge scaffolding on comprehension ………………….... 68

x

Chapter 1
Introduction
How does working memory capacity affect attention and learning from educational digital
games? Do expert gamers comprehend more of the educational message than novice gamers
when learning via digital games? And how do working memory capacity and gaming skill
expertise interact with game scaffolding designs to affect attention and learning? These are the
three main research questions that this dissertation seeks to investigate.
An increasing number of digital games are designed and used for educational purposes
(e.g., Barab et al., 2007; Peng, 2009; Squire, 2003). These digital games are used in a wide range
of fields including school education (Barab et al., 2007; Kiili, 2005; Papastergiou, 2009),
professional training (Edery & Mollick, 2009; Hays, Jacobs, Prince, & Salas, 1992; McDowell,
Darkin, Sullivan, & Johnson, 2006; Michael & Chen, 2006), and health promotion (Peng, 2009),
including medical treatment and rehabilitations (Kato, Cole, Bradlyn, & Pollock, 2009). It is
estimated that more than $125 million is invested into developing educational games every year
(Blunt, 2007).
Many scholars have argued that digital games are an ideal media for communicating
systematic knowledge and preparing learners for solving real-life problems (e.g., Gee, 2007;
Squire, 2003). However, despite the enthusiasm for digital game-based learning (DGBL) among
scholars and education practitioners, empirical research has found mixed results regarding digital
games’ effectiveness in improving knowledge comprehension and problem-solving transfer
(Foster & Mishra, 2009; Hays, 2005; Kirriemuir & McFarlane, 2004; Mitchell & Savill-Smith,
2004; Randel, Morris, Wetzel, & Whitehill, 1992). A common finding is that although learners
preferred digital games over traditional classroom instructional methods, learners often have
1

difficulties understanding the underlying messages and fall short on constructing systematic
mental model of the relationships communicated through the games (Barab, Thomas, Dodge,
Carteaux, & Tuzun, 2005). Why do some learners fail to comprehend the underlying message
from digital game based learning? A potential explanation that has not been fully examined is
that digital game based learning design may overload learners’ limited working memory
capacity; working memory capacity that is known to predict learners’ performance in complex
cognitive tasks (Yuan, Steedle, Shavelson, Alonzo, & Oppezzo, 2006)).
Working memory is a conceptual workspace in the human mind where selected
information from environmental stimuli (e.g., auditory, visual) and information from long-term
memory are temporarily stored for cognitive processing. (Baddeley, 1992, 2000, 2007; Logie,
2011). An individual’s working memory capacity has been shown to be correlated with
performance in complex cognitive tasks such as learning, reasoning, comprehension, and
problem-solving (Kyllonen & Christal, 1990; Yuan et al., 2006). Cognitive psychologists
generally agree that individuals have a limited working memory capacity and that different
cognitive tasks demand different amount of working memory. When a task exceeds one’s
working memory capacity, the individual cannot sustain his/her attention or process any more
information (Baddeley, 1992; Cowan, 2008; Kane & Engle, 2003; Lang, 2006).
Educational games are complex problem-solving tasks that require a wide range of
cognitive processing such as visual-spatial navigation, perceptual motor skills, accurate timing of
responses, as well as short and long-term strategic decisions (Logie, Baddeley, Mane, Donchin,
& Sheptak, 1989). As a complex problem-solving task, digital games often require learners to
simultaneously hold multiple interacting elements (e.g., goals, controls, and available resources)
in their limited working memory (Grodal, 2000), which demands large amount of working
2

memory capacity from the learners. In addition to holding the multiple elements in ones’
working memory, digital games require active decision making and inputs from learners for the
narrative to develop, and understanding of the narrative is a prerequisite for understanding the
educational messages (Fisch, 2000).Under the structure of digital games, a portion of the
learners’ working memory is required to process the game mechanics and narratives first, and
the remaining available working memory is then used for processing information related to the
educational content.
With regards to people’s limited working memory capacity, educational games face a
dilemma. On the one hand, digital games are believed to facilitate systematic comprehension by
encouraging learners to explore different identities and solutions in dynamic simulated
environments (Shaffer, Squire, Halverson & Gee, 2005; Squire, 2003). On the other hand, the
working memory demands from its interactive narrative structure and simultaneously processing
the multiple interacting elements may leave learners with insufficient working memory to
process the educational messages, which may result in poor comprehension (Scheiter & Gerjets,
2007). The first goal of this study seeks to investigate how working memory capacity affects
comprehension of educational content in an educational game.
Asides from working memory, domain expertise is also a strong predictor of performance
on cognitive tasks. Studies have shown that domain experts are able to simultaneously maintain a
large amount of information specific to their expertise domain (Ericsson & Kintsch, 1995).
Experts are also less susceptible to disruption and distraction (Kane & Engle, 2003), and are
more capable of recalling previous experiences in their expertise domain (Chase & Ericsson,
1982; Ericsson & Delaney, 1998). Such improved performance is often attributed to experts’
ability to “chunk” related information rather than process individual elements independently
3

(Ericsson & Delaney, 1998; Miller, 1956). With extensive practice, experts may be able to
process chunks of information automatically, further reducing the amount of working memory
required to perform a task. The second goal of this study seeks to investigate how gaming
expertise influences available working memory and comprehension of educational content
communicated in an educational game.
If comprehension is solely determined by individual characteristics such as working
memory capacity and expertise, one would predict that learners with higher working memory
capacity and more expertise will consistently comprehend more of the educational content than
learners with lower working memory capacity and expertise. However, studies of instructional
design found that experts do not always learn more than novices. Several studies have observed
that instructional designs which seek to reduce working memory demands may unintentionally
hurt expert learners’ comprehension, a phenomenon with two competing theoretical
explanations: the expertise-reversal effect (Kalyuga, Ayres, Chandler, & Sweller, 2003; Kalyuga,
Chandler, & Sweller, 2000) and the reverse-cohesion effect (McNamara, Kintsch, Songer, &
Kintsch, 1996).
The expertise-reversal effect argues that because experts are capable of processing the tasks
using their prior schema and mental models, the additional instructional information in a tutorial
introduces redundant working memory demands that are unnecessary, but unavoidable for expert
learners, leading to worse comprehension (Kalyuga et al. 2000; Yeung, Jin, and Sweller, 1998).
The reverse-cohesion effect argues that by reducing the demand on working memory through
scaffolding, experts a false sense of understand and thus cause them to process the information
superficially, consequently missing out on important information. In other words, McNamara et
al. (1996) argues that more demanding media design (not scaffolded by tutorials or other means)
4

helps experts learn because it forces them to actively use their expertise to process the
information; however more demanding media hurts novices because they do not have the
expertise to process the information. The third goal of this study is to investigate how working
memory capacity and gaming expertise interact with different scaffolding designs to affect
attention and comprehension.
Theoretical and Practical Contributions
This dissertation is expected to contribute to scientific understanding of how individual
differences (i.e. working memory capacity, gaming expertise) interact with media design to
influence available working memory and comprehension from digital game based learning.
While several studies have suggested that learners’ working memory capacity may influence
learning outcomes in DGBL (Grodal, 2000; Kiili, 2005; Vorderer, 2000), few studies have
empirically examined this relationship. It is also important to understand how gaming expertise
affects available working memory and comprehension because educational games are often
assigned as educational tools or assessments. When games are assigned as educational tools or
assessments, all learners regardless of their gaming expertise must learn from the game.
Understanding the learning impact of crucial individual differences (working memory capacity
and gaming expertise) can help educational game designers and educators work to optimize
DGBL for all learners.
To summarize, this study has three goals: (1) to examine the influence of individual
working memory capacity in digital game based learning; (2) to examine the effect of gamingskill expertise on available working memory and comprehension; and (3) to examine the

5

expertise-reversal effect and reverse-cohesion effect in the context of digital game based
learning.

6

Chapter 2
Digital Game Based Learning and Working Memory Capacity
The first goal of this study is to investigate how individuals’ working memory capacity
affects comprehension in digital game based learning. This chapter will begin with a brief review
of digital game based learning and how working memory is involved in the process of learning
with digital games. Next, the chapter will review the current understanding and conceptualization
of working memory, focusing on the individual difference approach (e.g., Cowan, 2005; Engle,
Kane, & Tuholski, 1999) to guide development of hypotheses on how working memory capacity
may influence attention and comprehension from DGBL. The second goal of this study seeks to
investigate how gaming expertise influences attention and comprehension. Therefore the second
half of this chapter will discuss the relationship between expertise and working memory capacity
on complex cognitive tasks.
Digital Game Based Learning and Working Memory
Digital game based learning (DGBL) can be generally defined as learning that involves
interacting with digital games in a formal or informal setting (Prensky, 2003). Two of the most
popular arguments for why digital games may be beneficial for learning are that 1) digital games
are motivational, and 2) digital games allow learners to explore complex problems which
resemble ill-structured problems in real-life scenarios.
Digital games as motivational tools
Earlier studies of DGBL sought to identify motivational features of digital games that
engage learners (e.g., Cordova & Lepper, 1996; Driskell & Dwyer, 1984; Malone & Lepper,
1987). These studies argued that students would engage in active learning if schools could adapt

7

the features that make games motivating to their curriculum. In other words, the goal of these
studies is to understand what makes games fun, not what makes them educational.
Various motivating features were identified, including challenge, curiosity, control, fantasy
(Malone & Lepper, 1987), rules, sensory stimuli (Garris, Ahlers, & Driskell, 2002), and story
(Baranowski et al., 2003; Dickey, 2007). Malone and Lepper (1987) argue that challenge is
facilitated by undetermined results. Because the players do not know whether they can reach a
goal with their existing skill and knowledge level, they are motivated by the challenge. Curiosity
is facilitated by an optimal level of complexity. An interesting game is novel and surprising
which invites players to explore different aspects of it, but the game cannot be incomprehensible
so that the players feel lost and lose interest. Control is motivating because it gives players a
sense of empowerment. An empowered player feels motivated because his or her skills and
knowledge can influence the environment and outcomes (Malone & Lepper, 1987). Fantasy is
motivating because it provides a safe environment for players to apply and play with skills that
they have acquired. Rules are motivating because they offer a set of boundaries and restrictions
in which the player can explore, challenge, and socialize with other players (Garris, Ahlers, &
Driskell, 2002). Some players are also motivated by the sensory stimuli that digital games offer,
it maybe a particular art style, or a pleasant music. Baranowski et al. (2003) argues that stories
are also motivating as they are narratives about the conflicting interest between the characters
and the environment. Stories motivate players to go on because players usually identify with the
protagonists and want to find out what happens to them.
These studies explain why games may be fun and motivating, which is important to
facilitate self-learning. But they do not focus on the unique media affordances of digital games.

8

In other words, this line of research does not explain why and how games can be educational,
and what kinds of learning are involved when a learner interacts with digital games.
Digital games as complex problem-solving
The other argument for why digital games are educational is that most digital games are
inherently complex problems to solve (Gee, 2003). The problem-solving nature of digital games
makes them ideal environments for learning and practicing systematic thinking (Gee, 2003,
2007; Squire, 2005).
Instead of communicating the narratives directly to learners, digital games communicate
the narrative and educational contents to learners through the interactive experience of problemsolving. Bogost (2007) used the term procedural rhetoric to refer to this process of
communicating through problem-solving interactivity. Educational digital games rarely reveal
the intended messages to the players explicitly. Instead, DGBL exposes players to a complex
simulated environment with specific rules and mechanics governed by the code, where players
must learn to navigate and interact within the rules. Digital games invite players to learn the
mechanics and experiment with different creative solutions in order to reach the games’ goals.
The narratives of the game are constructed by the player through his or her experience in this
problem-solving process. It is believed that this knowledge-construction process can facilitate a
systematic understanding of the complex problems.
Digital games are also a type of multimedia which can communicate a combination of
video and auditory information. Digital games can simulate environments that allow learners to
visualize and interact with factors within their dynamic environment. In order to achieve the
games’ goals, games often have rules and mechanics that encourage learners to observe the
9

systems’ behavior from different perspectives and across time. Many digital games also
encourage learners to take on different identities (Garris et al., 2002; Gee, 2007), which
facilitates perspective-taking (Peng, Lee, & Heeter, 2010) that generates deeper understanding of
an issue and motivates players to take action. Some games allow learners to manipulate variables
that are unalterable in real life, pose hypothetical questions and compare the simulated results to
their previous understanding of the systems (Squire, 2003). Some games can also support peerlearning among learners through multiplayer designs or discussions outside of the game (Gee,
2007; Steinkuehler & King, 2009).
If we understand digital game based learning as a problem-solving process, we can adopt
traditional models of problem solving to examine how working memory is involved in the
process of learning through digital games. Newell and Simon (1972) divide the information
processing involved in problem-solving into two stages--a representation stage and a solution
stage. The representation stage sets the ground for the solution stage. In the representation stage,
learners must first identify the goal of the problem, the important elements related to the
problem, the potential actions that can be taken, and the potential constraints of the environment
(Wiley & Jarosz, 2012). With the identified information, learners construct a mental
representation of the problem in their working memory. Experts who are familiar with the
problem domain can often construct the representation faster and to a greater extent than people
who are less familiar with the problem domain (Chi, Glaser, & Farr, 1988).
After the problem representation is constructed, the process moves into the solution stage.
In the solution stage, learners actively search and activate related information stored in their
long-term memory. The information may be a schema, a problem-solving strategy, or a set of
algorithms. Following activation, working memory is used to integrate the representation to the
10

activated information in order to construct a potential solution. Once the solution has been
applied, learners evaluate whether the solution reached the intended goal. If the solution
succeeds, the experience can be integrated into long term memory for future retrieval. If the
solution fails, learners might go back to the representation stage or the solution stage to work out
an alternative solution. Working memory is involved in both the representation stage and the
solution stage of problem-solving.
Working Memory Capacity in Digital Games
Some of the most comprehensive research about the demands of working memory in digital
games is series of studies using the game Space Fortress which was created in the 1980s and is
still used in research today (Shebilske et al, 2005). Space Fortress is computer game developed
specifically by cognitive psychologists for studying the acquisition and processing of complex
skills and problem-solving (Donchin, 1995). The goal of the game is to navigate a space ship
through a number of obstacles and shoot missiles to destroy a space fortress. Unlike traditional
studies of working memory which use relatively simple tasks, Space Fortress was designed as a
regular computer game to study cognitive processing of complex tasks. Studies showed that
playing Space Fortress involved perceptual motor skills, accurate timing of responses, as well as
short and long-term strategic decisions (Logie et al., 1989). Logie et al. (1989) identified eight
types of secondary tasks, each with a matching working memory component or function and
assigned players to do them while playing, to test what components of working memory were
involved in playing Space Fortress. Their findings showed that except for articulatory
suppression, which suppresses rehearsal, doing all of the secondary tasks while playing resulted
in impairment of performance in Space Fortress. In other words, playing a game like Space
Fortress involves a large set of demands on working memory, ranging from divided attention,
11

multitasking, visual scanning, dynamic and discrete motor control. Playing a digital game
requires players to not only process different cognitive task, but also switch between cognitive
tasks. This ability for task-switching requires more working memory in the form of executive
control (Gopher, 2006).
More interestingly, in a follow-up study (also reported in Logie et al., 1989) found that
visual-spatial processing was only prominent in earlier training phases, and that the effect of
visual-spatial processing on working memory load diminished with further training. One possible
explanation is that visual-spatial processing became automatic as learners gained gaming
expertise. A more recent study showed that expert gamers could monitor a larger number of
moving objects than non-gamers (Green & Bavelier, 2006). Boot, Kramer, Simons, Fabiani, and
Gratton (2008) found that expert gamers could track objects moving at greater speeds, were more
accurate in a visual short-term memory test, performed task-switch more quickly, as well as
monitored and evaluated rotating objects faster and more accurately.
The series of study using Space Fortress suggests that playing digital games entails
complex cognitive tasks which require a series of mental processing and motor skills. Of the
cognitive tasks involved in playing digital games, visual-spatial processing such as navigation
are prerequisites that are required to advance in a game and are often mastered first. When the
players master visual-spatial processing in a game, they can process related visual-spatial tasks
automatically, freeing up working memory capacity to process other information within the
game.
What is Working Memory?

12

Working memory is a construct that involves temporarily storing and processing
information during cognitive processing (Baddeley, 2007). Working memory has been found to
be closely related to a wide range of high-level cognitive abilities such as comprehension,
reasoning, and problem-solving (Yuan et al., 2006).
Research about limited working memory capacity can be traced back to Hermann
Ebbinghaus’s study of learning and forgetting in 1885. Ebbinghaus conducted a series of
experiments on himself as he attempted to acquire and forget a series of nonsense syllables. One
of his observations was that he often had a fleeting grasp of the nonsensical syllables. However,
this temporary memory was insufficient for him to recall the information later (after 31 days).
Around the same time, James (1890) proposed that in order for information to be recalled later,
the information must be consciously held for a moment in one’s primary memory. If the
information is held in one’s primary memory, it can then be stored in one’s secondary memory,
which is an unlimited storage of knowledge that can be accumulated and lasts a lifetime (Cowan,
2008).
The assumption of a limited working memory capacity was later popularized by Miller’s
(1956) study of immediate memory span. Miller proposed that people have a limited capacity of
immediate memory, in which seven plus or minus two chunks of information can be held and
processed at any given time. A chunk of information is a group of meaningful elements that a
person recognizes. For example the numbers 1800 may be remembered as four independent
numbers (1, 8, 0, and 0) to one person, but for another person who is familiar with United States
toll-free telephone numbers, the four numbers in 1800 could be stored and processed as one
single chunk. In this sense, whether a set of information is considered a chunk depends on one’s
existing knowledge, his or her ability to recognize the relationships, and ability to make the
13

connection between the information and one’s existing knowledge. Miller (Miller, Galanter, &
Pribram, 1960) later referred to this temporary cognitive space for storing and processing
information chunks as working memory.
Following the assumption of limited cognitive capacity, Atkinson and Shriffrin (1968,
1971) proposed a model of short and long-term memory. According to their model, information
from external stimuli is acquired through sensory registers (i.e. visual, auditory) into one’s shortterm memory system. Information in the short-term memory decays fast and if not overtly or
covertly rehearsed, the information will be lost. The rehearsed information is then stored into
one’s long-term memory, which is assumed to have unlimited in capacity and to be relatively
permanent. In this model, short-term memory acts as a temporary space in which information is
encoded, maintained, and rehearsed to be stored into long-term memory. One of the important
contributions of this model is that it highlights the executive control function of short-term
memory. Atkinson and Shriffrin (1971, p.5) suggested that “because control processes are
centered in and act through STS [short-term memory store], this store is considered to be a
working memory: a store in which decisions are made, problems are solved, and information
flow is directed.” In other words, working memory is not merely a storage space; it also
encompasses attention control functions.
While Atkinson and Shriffrin’s model was conceptually sound, empirical tests found it
problematic because it failed to explain how long-term memory could be retrieved to influence
short-term memory. It also failed to differentiate the storage and control functions of working
memory. For example Coltheart (see Baddeley & Hitch, 1974) found that semantically similar
words disrupted memory more than acoustically similar words, indicating that long-term
memory rather than short-term memory was being disrupted. A series of studies found that
14

patients with impaired short-term memory could perform tasks that used long-term memory just
as normal people could. This would not be possible if the executive control was governed by
short-term memory (Baddeley & Hitch, 1974).
In order to address the shortcomings of the Atkinson and Shriffrin’s model (1971),
Baddeley and Hitch (1974) proposed a multi-component model of working memory. Their
model consists of an attention control component, the central executive which controls two
subsidiary slave systems—the phonological loop and the visual-spatial sketchpad. And two
decades later, Baddeley (2000) added another component—the episodic buffer to explain how
complex information that consists of multi-dimensional message are processed and connected to
long-term memory. (See Figure 1).
Figure 1 Baddeley’s (2000) structural model of working memory (For interpretation of the
references to color in this and all other figures, the reader is referred to the electronic version of
this dissertation)

15

Baddeley and Hitch’s (1974) model became the dominant model of working memory, and
two general research directions developed out of this model of working memory (Miyake &
Shah, 1999). The first approach focuses on the structure of working memory and seeks to
understand how information is processed in the working memory by testing this multicomponent model (Yuan et al., 2006). The second approach is focused on the attention control
function of working memory; scholars in this direction seek to understand how individual
differences in working memory capacity affect performance in various cognitive tasks (e.g.,
Cowan, 2008; Unsworth & Engle, 2007). These two approaches to studying working memory
complement each other as one focuses on identifying the components and limits of working
memory, while the other seeks to understand the cognitive and behavioral influences of limited
working memory capacity (Logie, 2011). Because the first goal of this dissertation is to
investigate how individual differences in working memory capacity affect attention and
comprehension from educational games, this manuscript will only briefly review the structural
approach and will primarily focus on discussing the individual difference approach which
emphasizes the attention control function.
The Structure of Working Memory
The multi-component model of working memory by Baddeley and his colleagues (see
Baddeley, 2007 for review) proposed that working memory consists of a central executive
component which controls the flow of information. The central executive determines which
information to attend to, and which ones to omit. It is also responsible for determining the
associations between information and strategies for processing information. The central
executive can be seen as an attention controller with limited capacity which oversees all

16

information processing in working memory. The central executive is purely an attention control
system with no storage capacity (Baddeley, 2000)
Since the central executive control has no storage capacity, it controls three subsidiary
systems of storage and processing. First, the phonological loop comprises of a phonological store
which holds speech-based information and an articulatory rehearsal component which retains
information through rehearsal. Phonological information that is not rehearsed through
articulatory rehearsal fades in about two seconds. Evidence of the phonological loop includes the
phonological similarity effect—words that sound similar are more difficult to recall than words
that look similar or have similar meanings (Conrad & Hull, 1964). The word-length effect shows
that it is easier to recall short words than long words, possibly because it takes less time to
rehearse short words, and thus more rehearsal could be conducted before the word fades
(Baddeley, 2000). Further support for the rehearsal process is found in experiments that use
articulatory suppression, when rehearsal is prevented by asking experiment subjects to recite an
irrelevant sound while trying to remember a list of words. Their performance in word recall
drops drastically. Articulatory suppression also removes the word-length effect; if the words are
not rehearsed, no matter how long the words are, they are difficult to recall (Baddeley, 2007).
Similar to the phonological loop, the second subsidiary component—the visual-spatial
sketchpad holds and process visual-spatial information. It is limited by the visual complexity of
the represented information, including the number of stimuli (Luck & Vogel, 1997).
The distinction between phonological and visual-spatial components has been largely
supported by various behavioral and developmental psychology studies and also some
neuroscience studies (see Baddeley, 2007 for review). However more evidence suggests that the

17

two subsidiary components cannot account for all the information processing that occurs in
working memory (Baddeley, 2000). For instance, the earlier model (Baddeley & Hitch, 1974)
cannot explain how chunking occurs. Are chunks of information stored in long-term memory?
Or are they stored in one’s phonological loop? How could a patient with damaged phonological
memory limited to one word remember up to five sentences (Baddeley & Hitch, 1974)? More
problematic is that children seemed to rehearse information before they learned adult vocal
rehearsal strategies (Baddeley, 2000). Where in one’s working memory are these information
processed if not in the phonological loop or the visual-spatial sketchbook?
In response to the limitations of his original model (Baddeley & Hitch, 1974), Baddeley
(2000) incorporated a third subsidiary component—the episodic buffer, to the model. According
to Baddeley (2000, p. 421), the episodic buffer is:
A limited-capacity temporary storage system that is capable of integrating information
from a variety of sources. It is assumed to be controlled by the central executive, which is
capable of retrieving information from the store in the form of conscious awareness, of
reflecting on that information and, where necessary, manipulating and modifying it. The
buffer is episodic in the sense that it holds episodes whereby information is integrated
across space and potentially extended across time.
The episodic buffer is where information retrieved from long-term memory is temporarily
stored for processing. While more studies are needed to test this new component, Baddeley
(2000) suggests that the episodic buffer is an interface for integrating information from multiple
systems and forming new cognitive representations. Therefore it is the component most relevant
to processing complex information and problem solving. If this is the case, the episodic buffer

18

maybe the most relevant component to DGBL which comprises of simultaneously processing
multi-modal information, active decision-making and problem solving.
It is important to note that while the same words-- limited capacity are used to describe the
central executive and the three subsidiary components, there are actually two types of capacity
involved in the conceptualization of working memory. The first is an attention capacity, which is
the amount of effort or resource that can be allocated to processing relevant information and
blocking out irrelevant information. The second is a storage capacity, restricts the amount of
information that can be temporarily held and processed in one’s working memory.
The Individual Difference Approach
Asides from the structural approach, another approach to studying working memory
focuses on the effects of individual differences in working memory capacity. Individual
differences in working memory capacity have been shown to be highly correlated with various
cognitive tasks such as reading comprehension (Daneman & Merikle, 1996), creative writing
(Abu-Rabia, 2003), mathematics (Bull & Scerif, 2001) and general fluid intelligence (Engle,
Kane, & Tuholski, 1999). The current study seeks to examine how working memory capacity
affects comprehension from an interactive digital game. Therefore I will focus on the individual
difference approach of working memory.
While there are strong evidences to suggest that working memory capacity is a strong
predictor of ability to accomplish many complex cognitive tasks. The question is why and how
does working memory capacity influence these cognitive tasks? Two explanations based on the
functional distinction of attention capacity and storage capacity has been proposed.
Attention Capacity Difference
19

Some scholars argued that individual working memory capacity differences lie in the
ability to control attention and block out interferences (e.g., Cowen, 2005; Kane, Bleckley,
Conway, & Engle, 2001). From a structural perspective, the difference is believed to reside in the
central executive function of working memory. During information processing, working memory
capacity is involved in three main functions of attention control —activation of information,
blocking interference, and suppressing unrelated information.
Attention is required during selection of new information, and also during activation of
existing information stored in long-term memory. Information selection could be intentional or
automatic. Intentional selection occurs when the individual is cognitively aware of what to look
for. For example, a person who is looking for a red car will intentionally focus his/her attention
on large red moving objects on the road. Information selection can also be an automatic response
to novel, unexpected changes in the environment (Lang, 2000; Lang, Potter, & Bolls, 1999). For
example, if a person hears a loud noise in the library, it is difficult not to turn his/her head and
seek the source of the sound. Both intended and automatic processing requires working memory
to process, but intentional processing usually takes more working memory than automatic
processing. Furthermore, to determine if the information is related and makes sense, and to
further process the information, there is a need to activate information from long term memory
(Conway & Engle, 1994). Individuals with higher working memory capacity can retrieve goalrelevant information faster and more accurately than individuals with lower working memory
capacity (Barrett, Tugade, & Engle, 2004). Higher working memory capacity also predicts faster
generation of category exemplars (Rosen & Engle, 1997).
Because information within working memory is assumed to decay over time, attention is
required to maintain information in an active, accessible state for processing and storage. In
20

order to maintain information in an active state, it is also necessary to simultaneously block out
interfering information from external and internal sources. For example, when shopping for fruits
at a market, one needs to focus on the goal of selecting fruits, while blocking out external
interference from other vendors, and also irrelevant internal interference such as thoughts about
one’s work and family relations.
Individuals with lower working memory capacity are more susceptible to distraction, even
if they are instructed to ignore distraction (Kane et al., 2001). For instance, individuals with
lower working memory capacity were slower to complete a counting task when distracting
information was present (Tuholski, Engle, & Baylis, 2001). Several studies used the antisaccade
task to test for the effect of working memory capacity in resisting interference. An antisaccade
task requires subjects control their attention and “look away” from an interfering signal in order
to see a target message. Findings from these studies showed that individuals with higher working
memory are faster and more accurate in recalling the target message (Kane et al., 2001;
Unsworth, Schrock, & Engle, 2004), which suggest that they are more capable of controlling
their attention and resisting automatic orienting responses to novel stimuli.
Working memory is also involved in the controlled suppression of automatic informationprocessing that is unrelated to the goal. Individuals with lower working memory capacity are less
capable of suppressing automatic or habitual responses. Kane and Engle (2003) used a Stroop
test to test the suppression of automatic responses. In Stroop tests, the subjects are instructed to
answer the color of the ink that the words are printed in. (The words themselves are names of
colors, but the color of the ink and the word sometimes do and sometimes do not match.) Their
findings showed that individuals with higher working memory capacity made fewer errors on
incongruent trials (e.g., the word RED printed in green ink) than individuals with lower working
21

memory capacity. The finding supports the hypothesis that individuals with higher working
memory capacity are more capable of focusing their attention on the goal and inhibiting
automatic responses.
Another study by Kane and Engle (2000) used a proactive inference task in which they
asked subjects to recall three lists of 10 words. All of the words were drawn from the same
categories (e.g., animals, plants, etc.). Between being shown the lists, the subjects were asked to
perform an interference task (something unrelated to the word list) for 16 seconds. Recall of the
words in a proactive inference task usually drops with each successive list, because words from
the previous lists can interfere with later lists. Their findings showed that while all the subjects
showed a decrease in words recalled from later lists, subjects with low working memory
performed significantly worse, suggesting that they were less competent in blocking out
interference coming from previous lists. This evidence indicates that individuals with higher
working memory capacity were more capable of focusing their attention on goal-related
information and suppressing unrelated external or internal information.
Since working memory capacity mainly reflects ability to control attention and block
internal or external distraction, individual differences in working memory capacity will be more
significant on tasks that demand active attention control. Such situations includes (a) when there
are goals that needs to be actively maintained; (b) when there are competing goals and actions
that must be scheduled to prioritize one over the others; (c) when conflicts between actions must
be resolved; (d) when controlled, planned search of long-term memory is needed; and (e) when
error monitoring and correction are required to reach the goal (Engle, Kane, & Tuholski, 1999).

22

Kane and Engle (2003) conducted a series of studies to test how task demands for working
memory affect the significance of individual differences in working memory capacity. They
conducted a series of Stroop tests that required subjects to identify the color of the ink in which
words are printed in. They had different ratios of congruent (RED printed in red ink) and
incongruent (RED printed in green ink) words. Higher ratios of congruent words required more
active attention control because they required subjects to inhibit their habitual response in order
to identify the color of incongruent words. The researchers found that when there were no or few
congruent words, performance between high and low working memory subjects did not differ.
But when there were more congruent than incongruent words (i.e. more habitual interference that
demands more attention control), subjects with lower working memory capacity performed
significantly worse than subjects with higher working memory capacity. The findings suggest
that working memory capacity differences are largely determined by individual ability to control
attention, and that the difference is more likely to be significant in tasks that require active
attention control in a context with much interference.
A nice analogy for understanding this finding is to think of a staring contest, in which
two contestants are competing over who can stare for a longer time without blinking their eyes.
Contestant A can stare for 50 seconds without blinking, contestant B can stare for 20 seconds
without blinking. If the contest only runs for 20 seconds, there would not be a difference between
contestant A and B. However, if the contest runs for 30 seconds (and therefore requires more
attention control), the individual differences in attention control between contestant A and B
would significantly influence their ability to win the contest.
Digital games are designed to have player goals and designed-goals. Player goals are such
as defeating an enemy or collecting objects. Educational games also have learning designed23

goals, which the game designers hope players will learn even though the educational goals (such
as learning knowledge or skills) are sometimes not what players consciously strive to achieve.
Working memory capacity may influence players’ ability to focus their attention on player goals,
often including cognitive functions such as prioritize certain sub-goal over others, strategically
selecting and resolving conflicting actions, actively searching and processing goal-relevant
information while restraining themselves from distraction that are unrelated to the goal (e.g., a
beautiful soundtrack, cute character design, etc.. And then players use any remaining working
memory capacity to absorb and process the knowledge or skill learning content embedded in the
game. These two demands on working memory capacity (attention to gameplay and
comprehension of learning content) will be considered separately. Similar to Kane and Engle’s
(2003) arguments, individual difference in working memory capacity is most likely to be a
significant predictor of attention in games that require attention control. Some games are
designed to reduce working memory demand through scaffolding, which I will discuss later. But
often educational digital games are introduced to learners without any scaffolding, which
imposes large amount of working memory demand on learners. Therefore the study hypothesize
that working memory capacity will positively predict attention to gameplay in the absence of
scaffolding design to reduce cognitive load, that is, when the task (e.g., playing the educational
game) demands the highest attention control.
H1. Working memory capacity (Ospan score) will positively predict attention to the game
(i.e. slower secondary-task reaction time) in the no-scaffolding condition.
Storage Capacity Difference

24

Another view of individual differences in working memory capacity emphasizes difference
in storage capacity. According to this perspective, working memory capacity affects the amount
of information that an individual can hold and process within a time frame. An individual’s
working memory capacity is assumed to be fixed, but the amount of information that can be
processed is determined by characteristics of the information elements and the domain-expertise
of the individual. For example, a large string of words may be more difficult to recall (wordlength effect) than a short string of words, but words that are associated with each other in some
way (such as conceptually or structurally) are easier to recall than unrelated words for an
individual who is able to recognizes and strategically make these kinds of associations because
he/she can process multiple related words together as a single chunk of information.
In this sense, having a higher working memory capacity could mean two things—that the
individual has a larger storage capacity, or that the individual has a more effective information
processing strategy (e.g., chunking). Being able to hold more elements simultaneously allows
individuals to rehearse and make connections between more information chunks before they
decay (Baddeley, 2000). Storage capacity is especially important when processing information
that is highly interrelated and dynamic, thus cannot be broken down into smaller elements and
must be processed simultaneously.
One major problem with conceptualizing working memory capacity as storage capacity is
that it is highly influenced by individual strategies and individual expertise in the domain. An
expert at forming information chunks can store substantially more information; studies have
found that experts who can recognize and process information in chunks could expand their digit
span recall from a typical amount of seven digits to 80 or more digits, but not on materials that
they were not familiar with (Ericsson, Chase, & Faloon, 1980). Since storage capacity is
25

dependent on expertise and strategy, conceptualizing working memory capacity as a storage
capacity also cannot explain why the construct is so highly correlated with various cognitive
tasks and even correlated general fluid intelligence tests (Abu-Rabia, 2003; Daneman & Merikle,
1996; Engle, Kane, & Tuholski, 1999).
Engle, Tuholski, Laughlin, and Conway (1999) used structural equation modeling to test
the relationship between working memory (storage capacity plus attention control) and shortterm memory (pure storage capacity) on predicting general fluid intelligence. They found that
working memory and short-term memory were correlated but distinct. When variances explained
by the common factors were removed, working memory was still correlated with fluid
intelligence, but short-tem memory was not. The finding supports the perspective that working
memory consists of both storage capacity and attention capacity functions. However, individual
differences on complex cognitive tasks may be largely influenced by the attention control
capacity of working memory rather than storage capacity.
Domain-Related Expertise and Comprehension
Besides working memory capacity, many studies have also identified domain expertise as a
significant predictor of performance on cognitive tasks such as problem-solving. For example,
Spillich, Vesonder, Chiesi, and Voss (1979) found that subjects who are knowledgeable about
baseball could recall and comprehend a story about a baseball game better than participants who
were low in baseball expertise. Fincher-Kiefer, Post, Greene, and Voss (1988) compared
memory span of baseball experts on neutral and baseball-related words. They found that baseball
expertise facilitated memory span of baseball-related words, but not neutral ones. Ericsson and
colleagues showed that the amount of information that could be recalled can improve with

26

extensive training and practice of memory strategies (chunking). However, the improvement was
limited to the domain-specific information used in the training (Ericsson & Delaney, 1999;
Ericsson & Kintsch, 1995).
Chase and Simon (1973) argued that experts have large numbers of domain-specific
information chunks stored in their long-term memory. For example, when chess experts
encounter a chess set, instead of encoding each individual piece of information (e.g., chess piece
position), they can search their long-term memory and retrieve a similar pattern to facilitate their
information processing. Consistent with the researchers’ assumption, this expertise advantage
disappeared when a board of randomly positioned chess pieces was presented. A possible
explanation is that when encountering a random chess set, the experts were unable to retrieve
existing patterns to facilitate their memory, therefore like novices, they were forced to process
individual chess pieces independently. This finding suggests that experts only have advantages
when the information can be connected to their existing information chunks. Having prior
knowledge about a content domain allows learners to bring to mind a general model in which
they can integrate the new information into. Prior knowledge can also be used to fill in the gaps
and make connections between the new information. These findings suggest that individual
differences in expertise are most likely to be significant predictors of attention when the task
matches the existing schema stored in the experts’ long-term memory.
In the context of digital game based learning there are two types of important domain
expertise involved—expertise in the content domain (content knowledge expertise) and expertise
in the media of digital games (gaming skill expertise). Most of the studies reviewed in this
section have focused on how content knowledge expertise (domain expertise) affects
comprehension. However, expertise in the communication media may be equally important in
27

predicting comprehension and learning because different media require different sets of skills
and procedural knowledge to navigate and process. For example, when reading a book, one
usually begins with the first sentence and ends with the last word, reading in a linear direction.
Reading involves a combination of skills ranging from word comprehension, syntactic and
semantic knowledge, and higher order inferential skills. In comparison, in multimedia such as
digital games, a different set of skills are involved, such as navigation skills, visual-spatial
searching, awareness of the game environment, and problem-solving skills.
Previous studies have suggested that familiarity with the game genre is an important user
characteristic that may affect message recall (Lee & Faber, 2007). Individuals with higher
gaming expertise may be more capable of processing the game mechanics and potential actions
in chunks or even automatically, which frees up more working memory capacity to process
content information in the game. In comparison, a novice without gaming expertise will need to
treat all the information (game mechanics, narratives, educational message) as new information
and allocate large amount of working memory to process them together. Several studies on
advertisement effects in digital games have support the assumption that gaming expertise
facilitates greater memory; experienced gamers showed greater brand recall than non-gamers
(e.g., Schneider & Cornwell, 2005; Lee & Faber, 2007).
Working memory capacity and gaming skill expertise are assumed to affect digital game
based learning because learning in games is a problem-solving experience. The players first need
to be able to navigate the game, then they need to actively search for goal-relevant information
while blocking out interference to construct their mental representation of the problem-to-solve.
Through deductive thinking and experimenting with potential solutions, players learn how the
system works and how to solve the problem. This process requires players to hold large amounts
28

of information in their working memory for navigation, representation-constructing, and
solution-testing. Therefore it is likely that learners with more working memory capacity will be
more capable of maintaining attention to the learning goal in this process. Players with more
gaming skill expertise may be able to reduce the amount of working memory needed because
they have existing schema to guide them about what to look for and how things may interact in a
game context.
From an attention control perspective, on one hand, gaming skill expertise may facilitate
attention control by helping individuals identify which information elements are relevant to the
goal, and which information to block out. A learner’s gaming skill expertise is also a critical
factor in searching for goal-relevant information and sustaining attention on that information
(Chi, Glaser, & Rees, 1982; Kalyuga et al., 2003). But on the other hand, since the educational
goal of the game is not necessarily the same as the goals of the game, even if they have gaming
expertise, learners with lower working memory capacity may not have the ability to inhibit
habitual responses. In this case, they may process the game with their habitual responses,
focusing too much attention on the game mechanics and missing important educational
information.
Some studies have examined the relationship between domain expertise and working
memory capacity on comprehension. Hambrick and Engle (2002) examined whether overall
working memory capacity attenuated or facilitated expertise on recall and comprehension. Their
findings showed no evidence of attenuation. Instead, subjects with higher working memory
benefitted more from domain expertise than subjects with lower working memory capacity. One
possible explanation is that working memory facilitated better attention control, and expertise
helped identify what to focus the attention on. Swanson and O’Connor (2009) also examined the
29

relationship between working memory and reading fluency (skill expertise) on reading
comprehension. Their finding showed similar support for working memory capacity as a
predictor of comprehension beyond reading fluency (expertise) alone. They suggest that because
reading comprehension involves a number of interacting sub-processes such as derivation of
word meanings, assignment of importance, pacing, etc. most of which involves processing with
working memory. These studies suggest that both domain expertise and working memory
capacity affect level of comprehension. The current study seeks to examine how working
memory capacity and gaming skill expertise affect attention and comprehension in the context of
digital game based learning.
This study hypothesizes that gaming-skill expertise can help learners navigate through the
games to identify important information to process and irrelevant information to avoid. Similar
to the first hypothesis, individual difference in gaming-skill expertise will be most significant in
predicting attention when the gaming task demands more attention control.
H2. Gaming expertise will positively predict attention (i.e. slower secondary-task reaction
time) in the no scaffolding condition.
However, based on prior studies on expert performance, individual differences in expertise
are most likely to be observed when the experts can identify a task as being within their domain
of expertise and therefore can retrieve existing representations and solutions from their schema
(e.g., Chase & Simon, 1973; Fincher-Kiefer et al., 1988). Therefore this study hypothesizes that
gaming-skill expertise will also be a significant predictor of attention when gaming-skills are
scaffolded (scaffolding condition). Gaming-skill scaffolding in the forms of tutorials or levels
which increase in difficulty are common digital game level design.
30

H3. Gaming-skill expertise will positively predict attention (slower secondary-task reaction
time) in the gaming-skill scaffolding condition (i.e. a common game design technique).
H4. There will be a significant interaction effect between working memory capacity and
gaming skill expertise on attention in the gaming-skill scaffolding condition.

31

Chapter 3
Game Design and Cognitive Load
Different cognitive tasks vary in the amount of working memory required for
accomplishing them. As demonstrated by Kane and Engle (2003), individual difference in
working memory were more significant on cognitive tasks that required more active maintenance
of attention with more distractions. This chapter will begin by reviewing a capacity theory of
comprehension from educational media (Fisch, 2000) to explain how educational medium design
affects working memory demands and the conditions in which processing of narratives takes
priority before understanding of educational content occurs. The next section will discuss the
expertise-reversal effect and the reverse-cohesion effects, two seemingly competing explanations
of how scaffolding design efforts to reduce working memory demands may unintentionally cause
skill experts to perform worse in comprehension than they would without the scaffolding
designs.
Fisch’s Capacity Model of Comprehension from Educational Media
Fisch developed a model of comprehension based on his extensive experience producing
and studying educational television programs for children. His model is based on the assumption
of limited working memory capacity and incorporates three major constructs: Narrative,
Educational Content, and Distance.
Narrative refers to the surface information such as story, characters, goals, and
environment. In comparison, educational content refers to the underlying concepts and messages
that the media designer intends to communicate. Educational content could include both
declarative knowledge (e.g., facts) and procedural knowledge (e.g., problem-solving strategies).
Distance is the degree to which the educational content is integral to the narrative. The model
32

posits that narrative and educational content compete for working memory capacity, and the
distance between narrative and educational content determines how much they compete with one
another.
For example, in a Sesame Street television program designed to teach about water
conservation, understanding the narrative would involve knowing characters such as Big Bird
and Elmo, what they were doing in the skit, and the relationship between them and their actions;
It would also include understanding their spoken language and their non-verbal cues such as a
smiles or tears. The educational content would be the knowledge regarding water conservation
such as what actions to take, the underlying reasons for taking action, and potential
consequences. If the educational content is tightly woven into the narrative, the distance between
narrative and educational content is considered small. If the educational content is not part of the
narrative, but requires additional efforts to comprehend, then the distance is large.
One of the key points in this model is the concept of narrative dominance (Fisch, 2000).
The model posits that comprehension of narrative will take priority when competing with
educational contents for working memory, especially when the distance between narrative and
educational content is large. This is because most educational media designs embed the
educational content within the narrative. Using the example given in the previous paragraph, it is
almost impossible for a learner to understand the concepts of water conservation if he/she does
not understand the English language that Big Bird and Elmo use to speak, if he/she does not
understand what their interaction is about, or if he/she cannot tell if the characters are happy or
upset. Only when the learner has some understanding of the narrative can the learner
comprehend the underlying message behind the narrative—the educational content. But when the

33

narrative demands too much working memory from learners, there will be insufficient working
memory capacity for comprehension of educational content.
Fisch (2000) also points out that learner characteristics such as prior knowledge, familiarity
with medium, working memory capacity, and motivations can also influence how much working
memory is allocated to processing the narrative and educational content. Learners with more
prior knowledge about the narrative, and those who are familiar with the way educational content
is structured in the medium will be able to process the narratives in chunks and be able to
allocate more of their unused working memory capacity to processing the educational content,
which leads to more comprehension. Learners with higher working memory capacity are also
more capable of processing educational content above the working memory capacity required to
process the narrative. And it is possible that if learners are motivated to learn the educational
content, they may intentionally seek out educational content and allocate more working memory
to processing it.
Digital game based learning, like educational television programs, often embeds
educational content within narratives. On top of that, instead of having the narrative unfold to the
learners, interactive digital games require players’ active engagement (e.g., decision-making,
inputs, etc.) in order to develop the narrative (Klimmt & Hartmann, 2006). Therefore the primary
goal of learners in DGBL is learning how to engage the game by learning the controls and game
mechanics. Their second goal is to understand the narratives developing from their interaction
with the game design (Grigorovici & Constantin, 2004; Lee & Faber, 2007). Only after the game
controls are mastered and the game narratives are understood will learners be able to allocate
their working memory capacity to processing the educational content.

34

Scaffolding, Expertise-Reversal Effect, and Reverse-Cohesion Effect
If mastering controls and understanding narratives are prerequisites to understanding the
educational content in games, what can educators and game designers do to increase learner’s
available working memory and thus improve learner comprehension? One potential solution is
through scaffolding designs. Scaffolding is an instructional strategy that supports novice learners
by limiting the complexities of the educational context and gradually removing these supportive
limits as learners gain the knowledge, skills, and confidence to cope with the full complexity of
the context (Young, 1993). Complexity here refers to the number of interacting information
elements in the problem-solving task (van Merriënboer & Sweller, 2005). By initially exposing
learners to a less complex version of the educational media, learners will have less information
that needs attending to and less distraction; therefore they can focus on building up knowledge
about the content or skills regarding the media. With the knowledge and skills, they can allocate
less working memory to processing the controls and narrative and have more available working
memory for processing the educational content. Azevedo and Hadwin (2005) argue that
scaffolding can aide learners in: (a) acquiring domain knowledge, (b) acquiring procedural
knowledge, (c) learning how to use the instructional tools, and (d) learning how to engage
instructional features or contexts.
In the context of digital game based learning, there are two common scaffolding strategies
for reducing cognitive load. First, cognitive load demands can be reduced by providing explicit
instructions, so that the learner does not need to work to figure out what information are relevant,
and how does the relevant information interact with one another (content-knowledge
scaffolding). This design can mitigate the amount of cognitive efforts that learners need in order
to understand the narrative and constructing a mental representation. Second, levels can be
35

designed with increasing difficulty that allows learners to gradually learn the controls and
narratives in early levels (gaming-skill scaffolding). After the learners master one section, they
are introduced to the next section until the complete complex problem is revealed. On the first
level, players are given limited actions and relatively simple goals (e.g., move from point A to
point B). This approach allows players to familiarize themselves with the basic controls and
game mechanics. As the players progress through levels, they are given more available actions
and more goals and obstacles, until the player builds up a repertoire of skills to engage the
complex problem. Skill scaffolding can also help learners identify what belongs to the controls
and narratives, and what new information is included later. In other words, game scaffolding
designs such as explicit instructions and levels seeks to reduce cognitive load and improve
comprehension of educational content by helping learners gain narrative or gaming controls
expertise. While scaffolding has shown to help learners with lower expertise, several studies
have observed that learners with higher expertise do not benefit and sometimes even suffer from
scaffolding designs. For example, Yeung, Jin, and Sweller (1998) found that the addition of
explanatory text (knowledge scaffolding) made learners with high reading competence (skill
expertise) perform worse in comprehension of the text. High-reading competent learners also
reported more cognitive efforts in processing the explanatory text condition. Kalyuga et al.
(2001) found that inexperienced mechanic trainees benefitted most from studying case examples
of worked models, whereas experienced mechanics benefitted most from having problems
presented for them to solve. Kalyuga and colleagues (Kalyuga et al., 2003; Kalyuga, Chandler, &
Sweller, 2000) refer to this phenomenon as the expertise-reversal effect.
One potential explanation for the expertise-reversal effect is that, unlike novices who do
not have existing schema and skills to aide them in processing the problem, experts are capable
36

of processing the problem using their existing mental representations. The addition of
scaffolding information may provide a model for the novice learners and thus free up working
memory capacity for comprehension. But for experts, the scaffolding models may conflict with
their existing mental representation or familiar approach. Therefore, instead of reducing
cognitive efforts, the experts need to allocate more working memory in order to cross-reference
and integrate the explicit information with their existing mental representations. Even if the
expert learners recognize the explicit information to be redundant and decide to avoid it, they
still need to allocate additional working memory in order to block out the explicit information as
distraction. Therefore according to this explanation, experts would have less available working
memory when scaffolding is provided than when no scaffolding is given.
H5a. Players with higher gaming skill expertise will have more attention to the game (i.e.
slower secondary-task reaction time) in the gaming-skill scaffolding condition than the no
scaffolding condition.
H5b. Players with higher gaming skill expertise will have more attention to the game
(slower secondary-task reaction time) in the content-knowledge scaffolding condition than
the no scaffolding condition.
Similar results were observed in a series of studies conducted by McNamara et al. (1996) in
their study to investigate the relationship between expertise, text coherence and comprehension.
The purpose of these studies was to investigate whether accommodating learners and making the
learning experience “trouble-free” was an optimal approach for learning. McNamara et al. (1996)
compared content experts and novices in their comprehension of a coherent or incoherent text
about biology. The coherent text included background information and identified relationships

37

between information. In contrast, the incoherent text required learners to actively fill in the
information gaps and make relational inferences with their existing knowledge. In other words,
the coherent text can be seen as a type of scaffolding that requires less working memory on the
part of novices to process than the incoherent text (Ozuru, Dempsey, & McNamara, 2007). They
found that novices (those who had with lower content expertise) benefitted from the coherent
text, but experts with higher expertise learned less with the coherent text than when the text was
incoherent. McNamara referred to this as the reverse-cohesion effect.
McNamara et al. (1996) argued that scaffolding through text cohesion helps novices
because they do not have preexisting schema to fill in the information gaps and make inferential
connections between information elements. However, by reducing complexity and working
memory demand, coherent text gives expert learners a “false sense of understanding” and the
experts tend to process the information only superficially and may miss important information.
On the other hand, incoherent text with the information gaps required learners to actively use
their existing schema and knowledge to fill in the gaps and comprehend. Consequently, learners
with more content expertise exerted more working memory to process the incoherent text (i.e.
they pay more attention), and also comprehended more of the information. Due to the reversecohesion effect, McNamara et al. (1996) proposed that reducing working memory demands may
not be the optimal strategy for engaging learners who have content expertise. Instead, it may be
beneficial to provide learners who have content expertise with educational media that is less
cohesive and require their active attention.
The expertise reversal effect and the reverse-cohesion effect may seem to predict similar
results, that novice would benefit from scaffolding, but not experts. However the two models
differ in their definition of “expertise,” and the underlying mechanism. Kalyuga et al.’s (2000,
38

2003) expertise-reversal effect does not differentiate between content knowledge expertise and
skill expertise; they perceive expertise as a whole set of content knowledge, mental models, and
skills that is acquired through more experience with a task. Therefore they often use the terms
experienced and inexperienced to differentiate experts and novices (Kalyuga et al., 2003). On the
other hand, the expertise in McNamara et al.’s (1996) original study was specifically about
content-knowledge expertise. In other words, the predictions derived from McNamara’s earlier
reverse-cohesion effect were about how prior knowledge of the contents may affect how learners
engage the text. Their earlier study did not examine how skill expertise such as reading
proficiency, familiarity with text structure, experience with learning through text, and learning
strategies affect learner response to scaffolding.
The underlying mechanism for why experts do not benefit from scaffolding was also
different. The expertise-reversal effect posits that scaffolding introduces redundant information
for experts, which uses up more working memory. In contrast, the reverse-cohesion effect posits
that scaffolding makes the tasks easier for content experts, and therefore induces superficial
processing. In other words, the expertise-reversal effect argues that scaffolding increases
working memory demand, whereas the reverse-cohesion effect argues that scaffolding reduces
the likelihood that content experts will choose to allocate as much working memory to the task.
Several recent studies have compared how content expertise and reading skill expertise
predict text comprehension (Best, Rowe, Ozuru, & McNamara, 2005; O’Reilly & McNamara,
2007). These studies found that the reverse-cohesion effect occurred only among learners with
high content expertise and low reading skill expertise, not among learners with both high content
expertise and high reading skill expertise. The author argued that learners with higher reading

39

skill expertise will not be affected by the addition of explicit information because they are
capable of handling the increased amount of information.
Unlike the linear text which McNamara and her colleagues examined, digital games are
more similar to multimedia hypertext which requires large amounts of cognitive effort to
integrate and comprehend. Multimedia hypertexts also impose higher demands on the attention
control from learners (Shapiro & Niederhauser, 2004). Because games’ information and
narrative are often non-linear, skills in navigating and searching may be particularly important
(Azevedo, Guthrie, & Seibert, 2004). Prior studies have found that learners with lower content
knowledge struggle with comparing and integrating multiple representations presented in a nonlinear way (Ainsworth, 1999; Yersuhalmy, 1991). In comparison to content knowledge expertise,
much less is known about how media skill expertise affects available working memory and
comprehension. Several studies on hypertext comprehension have shown that media skill
expertise is one of the strongest predictors of comprehension when reading hypertext (e.g., Ford
& Chen, 2000; Reed & Oughton, 1997). Since learners need to actively navigate between
information elements in a hypertext environment, gaming-skill expertise may help learners
navigate through the information and construct mental representations based on gaming schema.
Both the expertise-reversal effect and reverse-cohesion effect predict that experts will
comprehend more of the educational content in the no-scaffolding condition than in the
scaffolding conditions.
H6. Players with higher content expertise will have highest comprehension in the noscaffolding condition in comparison to the scaffolding conditions.

40

However, the expertise-reversal effect is based on the assumption that less cognitive load
improves comprehension; whereas the reverse-cohesion effect argues that if the learner has
sufficient expertise, more cognitive load improves comprehension. Therefore the study poses a
research question:
RQ1. Will available working memory positively or negatively predict comprehension?
Finally, scaffolding designs are intended to support learners with lower expertise and
improve their comprehension. The last two research questions in this study examine 1.) Whether
gaming-skill scaffolding design supports learners with low gaming skills and improves their
comprehension and 2.) Whether the content-knowledge scaffolding design support learners with
low content knowledge and improves their comprehension.
RQ2. Does gaming-skill scaffolding improve comprehension for low gaming-skill learners?

RQ3. Does content-knowledge scaffolding improve comprehension for low contentknowledge learner?

41

Chapter 4
Experiment Design and Measurements
The goal of this study is to examine the effect of individual working memory capacity
and gaming skill expertise on attention and comprehension of educational content under three
common game designs (no scaffolding, gaming skill scaffolding, and content knowledge
scaffolding). The study also seeks to tease apart competing explanations about why experts may
not benefit and may sometimes perform worse when provided with scaffolding. An experiment
was designed to test the hypotheses.
Participants
Participants in this study were recruited from five undergraduate-level courses in the
College of Communication Arts and Science at Michigan State University. Extra credit was
offered by the course instructors as incentives for participating in the experiment. A collegestudent sample is appropriate for this study because the current generation of college students is
familiar with digital games. According to a national survey by the PEW Research Center’s
Internet and American Life Project, 81% of people aged between 18 to 29 years old play digital
games, and 76% of people in this age group play games on their computers (Lenhart, Jones, &
Macgill, 2008). Furthermore, a relatively homogeneous sample is preferred in a lab experiment
for theory-testing. This is because a homogeneous sample reduces the chance of making a false
conclusion about whether there is covariance between variables. In other words, using a
homogeneous sample reduces the chance of committing a Type II error in which the theory is
disconfirmed when in fact there is variance that is obscured by third variables when using a
heterogeneous sample (Calder, Phillips, & Tybout, 1981).

42

A total of 182 participants were recruited for this experiment. The mean age was 21.05
years-old (SD=2.40) and ranged from 19 to 38 years old. There were slightly more males
(54.8%, N=100) than females (45.2%, N=82) in the study.
Stimuli and Learning Goals
The educational game used in this study is The ReDistricting game. The ReDistricting
game is a game designed to educate players about the issue of political redistricting. Specifically,
the game is intended to help students understand the problem of gerrymandering.
Gerrymandering is the act of manipulating inclusion or exclusion of residential blocks or other
geographic parameters used to define voting districts to gain advantage for certain political
parties or interest groups. The ReDistricting game was created at the USC Game Innovation Lab
of the USC School of Cinematic Arts' Interactive Media Division. It was developed for the USC
Annenberg Center for Communications by a team led by game designer and faculty member
Chris Swain.
In the ReDistricting game, players take the role of a map maker who draws voting districts.
The game consists of five missions. Each mission is divided into basic and advanced difficulty;
the difference between basic and advanced difficulty is the number of elements and constraints.
The first mission is a basic tutorial about the controls. The second mission is about partisan
gerrymandering, which requires players to draw voting districts to favor a political party they
choose. The third mission is about bi-partisan gerrymandering; it requires players to draw voting
districts to keep the status quo so both parties’ seats are protected. The fourth and fifth mission
demonstrate how changes in voting laws can make gerrymandering more difficult and protect a
fair election. Through the five missions, players are expected to experience and understand how

43

redistricting (i.e. map-drawing) plays such a large role in directing election results, and could
easily be manipulated to undermine fair competition between candidates—a foundation of
democratic society. The game website also includes additional information about
gerrymandering and links to take action by sending a letter to state senators.
The ReDistricting game has two educational goals: to facilitate comprehension about the
problem of gerrymandering, and stimulate discussion and action about the problem. According to
the game description on the website, the ReDistricting game provides “a basic introduction to the
redistricting system, allows players to explore the ways in which abuses can undermine the
system, and provides info about reform initiatives.” Therefore, the first learning goal is
comprehension about the potential problems of redistricting, particularly. According to the lead
designer Chris Swain (Swain, 2007), the second learning goal is for players to come up with
solutions and take action.
Like many educational games that deal with political problems, the ReDistricting game
takes a reversed role-playing perspective. In reversed role-playing narratives, the learner takes
the role of the problem-maker. The explicit instruction of the game is to reach a certain goal
(e.g. in the ReDistricting game, the goal is to manipulate election results through
gerrymandering), but the underlying educational message is the opposite of that goal (i.e.
gerrymandering is bad and undermines democracy). In this type of reverse role-playing design,
understanding the game mechanics and narrative does not equal comprehension of the
educational message. Players need to have some existing knowledge about the democratic
election assumptions and actively reflect on their experience to understand that the game’s
intended message is the opposite of what they just experienced. Therefore I argue that the
distance between narrative and educational content is large in the ReDistricting game. According
44

to Fisch’s (2000) capacity model, when the distance between narrative and educational content is
large, the narrative takes priority and competes with the educational content for working
memory.
The ReDistricting game is a suitable game for this study because the game is specifically
designed for undergraduate-level students. Most undergraduate students are not familiar with the
process and mechanics of gerrymandering, thus they not likely to have extensive content
knowledge expertise in the topic of redistricting. If, as expected, there is generally low content
expertise across study participants, this sample allows the study to focus on examining how
working memory capacity and gaming skill expertise contribute to attention (i.e. the amount of
working memory allocated to the task) and comprehension of a little-known topic.
Operational Measures
Independent variables
Scaffolding Conditions. Scaffolding techniques were induced in this study by randomly
assigning participants into one of the three conditions (No scaffolding, Content-knowledge
scaffolding and Gaming-skill scaffolding).
Since the main problems of gerrymandering are communicated in Mission 2 of the
ReDistricting game. The participants in this study were expected to learn the problems of
redistricting after playing Mission 2 under the advanced difficulty which contains the complete
narrative and educational message with all the interacting elements and legal constraints. The
difference between the three conditions was in how the Mission 2 (advanced) was introduced to
the participants. The no-scaffolding condition introduced participants to Mission 2 without any
skill or content scaffolding. The gaming-skill scaffolding introduced participants to the controls
45

and basic game mechanics before they played Mission 2. The content-knowledge scaffolding
provided participants with an explicit message framework about the redistricting process and the
game’s goal before they played Mission 2.
The no-scaffolding condition has the highest complexity because it includes all the
interacting information elements but without any scaffolding. This study posits that the
participants will find the no-scaffolding condition most difficult because it imposes the highest
working memory demand. Participants in the no-scaffolding condition played Mission 2 of the
ReDistricting game under the advanced difficulty twice. In the first playthrough participants
were instructed to stop playing Mission 2 after five minutes, this allowed participants to explore
the game a little without increasing their exposure to the educational message in comparison to
educational message exposure experience in the game by the other two conditions. The second
playthrough of Mission 2 in the no-scaffolding condition required participants to either complete
Mission 2 or play for 30 minutes.
Participants in the Content-Knowledge-scaffolding condition were first instructed to read
an article (available on the experiment web site) about the general issue of redistricting and the
purpose of the game (see Appendix 2). After confirming that they had read the article, the
participants were then instructed to play Mission 2 of the ReDistricting game under the advanced
difficulty for five minutes, and then a second time until completion or for 30 minutes. Reading
the article was expected to reduce working memory demand by providing participants with a
general knowledge framework about redistricting and about the game.
Participants in the Gaming-skill scaffolding condition played Mission 1 under the advanced
difficulty and then Mission 2 under the advanced difficulty. Mission 1 teaches the players about

46

the basic controls of the game and the basic goal of population equality. It does not include the
problem of gerrymandering or the legal constraints on map redistricting.
A comparison of the three conditions is shown in Table 1.
Table 1 Comparison of induced scaffolding conditions
No scaffolding

Gaming-skill scaffolding

Step1 Mission 2 Advanced
(5 minutes)

Mission 1 Advanced

Step2 Mission 2 Advanced

Mission 2 Advanced

Content-knowledge
scaffolding
Article +
Mission 2 Advanced
(5 minutes)
Mission 2 Advanced

Gaming Skill Expertise. Gaming expertise was measured by a series of 7-point questions
in the pre-test survey. Participants’ were asked about self reported familiarity and expertise with
a list of different genres including strategy, puzzle, action, sports, role-playing, adventure, etc.
The list was randomized to eliminate order effect. Since The ReDistricting game is a strategy
simulation game, the mean score of familiarity and expertise on strategy and simulation games
was used as self-reported gaming expertise. The mean score for gaming skill expertise was 3.48
(SD=1.61). Reliability test using Cronbach’s α indicates that the items are acceptable at .68.
Working Memory Capacity. Working memory capacity was measured using the automatic
operation span (AOSPAN) test developed by Unsworth, Heitz, Schrock, and Engle (2005).
AOSPAN is a self-administered version of the popular operation span (OSPAN) measure of
working memory capacity developed by Turner and Engle (1989). In OSPAN participants are
asked to solve a series of math operations while trying to remember a set of unrelated words. For
example, a set of three questions looks like:

47

Is 1+ (2*8) =15? [Yes/No] Truck
Is (2/8)-3=1? [Yes/No] Eagle
Is (10-4)/2=3? [Yes/No] Job
The string of math operation and words appear one at a time. After each set of questions,
participants are asked to mark a check on the words from a list of words in the correct order. The
task requires participants to hold the words in their working memory while processing the math
calculations. Working memory capacity score is calculated by the number of correct answers
ranging from 0 to 12.
A practice set of three questions was given to participants so that they understand the task.
After the practice, three sets of four questions each (total of 12 questions) were used to measure
working memory capacity. Based on previous studies (Unsworth et al., 2005), participants were
given seven seconds to solve each question before the page moves on, this procedure prevents
participants from intentionally rehearsing the words. An 80% accuracy criterion on the math test
is used to ensure that participants are processing the math question while holding the words in
their working memory instead of just focusing on the words. Twelve participants were removed
from the dataset because they did not meet the 80% accuracy criterion. The mean score of the
AOSPAN test is 5.98 (out of 12, SD=3.66).
Dependent Variables
Attention. Attention is measured through secondary-task reaction time (STRT). Secondarytask reaction time is a measurement commonly used by cognitive psychology studies to measure
attention through measuring available working memory capacity (Lang, Bradley, Park, Shin, &

48

Chung, 2006). It is conceptualized that when the primary task takes more working memory to
process, there will be less available working memory to process secondary tasks, which leads to
slower reaction time to the secondary tasks.
Secondary-task reaction time is measured by instructing participants to focus on a primary
task, which is playing the game in this study. In addition to the primary task, the participants are
given a secondary task of pressing a specific button “as fast as you can” when they see a signal.
The signal is called the secondary task probe. The latency (time difference) between the probe
and the button-pressing action is called secondary-task reaction time (STRT). Theoretically, as
more working memory is allocated to the primary task for storage and attention control, the
participant will have less available capacity to process the secondary task, resulting in slower
reaction time. Available working memory capacity refers to the amount of working memory
required by the task minus the amount allocated to the task (Lang et al., 2006). Longer STRT
indicates more attention on the primary task, and shorter STRT indicates less attention on the
primary task.
In order for the secondary-task reaction time to be an accurate reflection of attention, it is
important that the secondary task probe to be in the same message mode as the primary task so
that it draws upon the same pool of working memory. The secondary task must also be very
simple so that is does not interfere with performance on the primary task (Bru¨nken, Plass, &
Leutner, 2003; Marcus, Cooper, & Sweller, 1996).
Since the ReDistricting game does not have auditory stimuli, the secondary task is designed
in visual mode. In this study, the secondary task probe was a yellow box next to the game screen
that appears at random times and in random places, participants were instructed to press the

49

“spacebar” key as fast as you can when they see the yellow box appear. STRT is measured by
the experiment website by subtracting the time that the participants pressed the spacebar to the
time that the yellow box lights up. STRT was record in milliseconds from the onset of the
secondary task probe to the time that participant pressed the spacebar key on the computer
keyboard.
A total of 16 secondary task probes appeared on random time throughout the stimuli (eight
in each mission). The decision of having eight secondary task probes in each playthrough is
twofold. One the one hand, the ReDistricting game is not a fast-paced game that requires
frequent reaction, thus it does not require a large number of secondary task probes. But on the
other hand, since the participants in this study are younger, previous studies have suggested that
more probes are needed to capture variance among younger populations because of their faster
reaction time (Basil, 1994). The number of probes is decided because in my pilot test, the
average time for completing mission one is seven minutes, and for mission two is 15 minutes.
Since not all the participants in the study received all eight secondary probes in each
mission, if they completed the missions earlier, they were only exposed to the first few probes. In
order to allow comparison between participants, only the first five STRT of the second
playthrough (in which all three groups played Mission 2 advanced) were included for the
analyses. Furthermore, because the first secondary-probe is often missed, attention score in this
study was calculated by averaging the second to fifth STRT (four STRT for each participant) in
the second play-through in which all the participants played Mission 2 on the advanced difficulty
of the ReDistricting game. Missed responses (i.e. no response or response after 5000ms from the
onset of the secondary probe) were coded as missing data. The mean score for STRT was

50

1090.45 milliseconds (i.e. 1.09 seconds, SD=489.65), ranging between 461 milliseconds to 3672
milliseconds.
Comprehension. The first goal of the ReDistricting game is for learners to understand the
potential problems of the redistricting process. Comprehension was measured in two ways. The
main measurement used in the hypotheses-testing was a set of 11 multiple-choice questions that
test the participants’ comprehension of the redistricting process (see Appendix 4 for the
questions). Each correct answer was given one point, thus the comprehension score ranges from
0 to 11. The mean comprehension score was 4.84 (out of 11, SD=2.31) ranging from 0 to 11. The
reliability of the comprehension items were acceptable with α=.67.
The secondary comprehension measurement was a free-recall test that asked participants to
“Explain the potential problems with redistricting process? (Describe as much detail as you
can).” This set of measurements was not used in the statistical analyses, but they provided some
insight to the participants’ comprehension of the game and unintended effects.
Control Variables
Since the ReDistricting game is a political game, prior attitudes about political issue may
also influence the results. Therefore political involvement was controlled as a covariate. Political
involvement was measured by nine 7-point questions adapted from Kahne, Chi, and Middaugh,
(2002). The questions ask participants how much do they agree or disagree with nine statements
that describe them. For example, “I often talk about politics and political issues,” “I often wear
badges in support of certain political issues,” and “I vote regularly.” The items are reliable with
α=.84.and a mean of 2.37 (SD=.98).

51

Existing content knowledge about the redistricting process and gerrymandering may also
affect learning outcomes. Two7-point question were used to measure the participants’ self
assessment of content knowledge: “I am very familiar with the redistricting process” “I consider
myself an expert in how gerrymandering works.” the questions were designed to be vague so that
they do not prime the participants to pay attention to specific parts of the ReDistricting game.
The two items were reliable at α=.74, the mean for content knowledge was 2.86 (SD=1.65).
The scores on political involvement and content knowledge were skewed towards the lower
end of the 7-point scale, which suggest that participants in this study generally had low political
involvement experience and content knowledge expertise about the topic of redistricting.
Research Procedures
After the participants volunteered for the experiment they were invited by email with a
hyperlink to an online scheduler to sign up for a time slot. Before the scheduled date, the
participants received an email reminding them to come to the computer lab to participate in the
experiment.
On their selected time, upon arriving at the computer lab and receiving the informed
consent form, the participants were given a URL that directed them to the experiment website in
which all the experiment questionnaires, stimuli, and tests were embedded.
On the experiment website, the participant first answered a short questionnaire measuring
their political involvement, their political efficacy, and the two content knowledge questions.
After responding to the pre-test questionnaires participants read a detailed instruction about the
AOSPAN test. They were given three practice questions to familiarize themselves with the

52

AOSPAN test. After the practice questions, they took the actual AOSPAN test consisting of 12
math and word retention questions.
After the participants completed the AOSPAN task, they were randomly assigned by the
experiment website to one of the three scaffolding conditions. All the groups read an instruction
reminding them to pay attention to the game because they would later be tested about the content
(see Appendix 2). The instruction serves two purposes. The first purpose was to increase
involvement. Previous studies have shown that individual differences in working memory
capacity are more significant when the task requires intentional attention control to goal-related
information as opposed to incidental learning (Unsworth & Engle, 2005). The second purpose
was to frame the game-playing experience as a learning experience, to try to introduce learning
as a motivation in approaching the game. A player who is playing a game to learn may have very
different behaviors than a player who is playing a game for entertainment or relaxation.
After reading the instruction about the game, the participants were then shown an
instruction about the secondary task probe and were instructed to respond “as fast as possible.”
In order to prevent participants from intentionally neglecting the secondary task instructions,
they were told that if they perform poorly in the task, they need to redo the task. Participants who
missed all the secondary task probes were removed from the study because they may represent
players who are extremely involved in the game that they did not have any available working
memory for the secondary probe, or extremely uninvolved participants who paid no attention and
were not following the experiment instructions.
In the No-scaffolding condition, participants played Mission 2 (advanced) of the
ReDistricting game for five minutes first, and then played Mission 2 (advanced) again until they

53

complete the mission or for 30 minutes, whichever happens first. The gaming skill scaffolding
group played Mission 1 (advanced) followed by Mission 2 (advanced). The content-knowledge
scaffolding group is similar to the no-scaffolding condition, but before playing the game, the
participants read a short article describing redistricting and the purpose of the game. Then they
played Mission 2 (advanced) for five minutes; and then play Mission 2 again until completion or
for 30 minutes, whichever happens first. The behavioral data of participants and their STRT have
been programmed to be recorded on the server that is running the game. Access to the source
code of the Redistrincting Game was granted by the game creator, Chris Swain.
The first two missions of the ReDistricting game were used in this study for content and
time reasons. It was important for participants to be able to complete participation in less than
one hour. And, more importantly, the main problems of the redistricting process
(gerrymandering) were communicated in Mission 2. (Mission 3 and Mission 4 reiterate the
problem again; Mission 5 discusses one of the many potential solutions to the problem). Thus,
Mission 1consisted of built in gaming-skill scaffolding and Mission Two introduced the core
learning contents. Missions One and Two optimally suited the study goals of examining the
impact of scaffolding on attention and comprehension.
After playing the game, participants filled out a post-test questionnaire measuring their
enjoyment of the game/video, their perceived challenge level, and their self-reported effort.
These questions acted as distracters before the comprehension test. This is to help ensure that the
comprehension test measured information that was stored within the participants’ long-term
memory. The participants’ comprehension was measured with a free recall question first,
followed by multiple-choice questions about the educational content in the ReDistricting game.
The free-recall question was asked first to prevent participants from being primed by the multiple
54

choice questions. Finally, participants answered demographic questions including age, sex, and
race. The complete experiment procedures took most people approximately 40 minutes. Seven
participants were excluded from the study because they could not complete the procedures
within 60 minutes.
Manipulation Checks
The scaffolding conditions were induced in this study by altering how the participants
approached Mission 2 (advanced) of the ReDistricting game. After completing the first game
playthrough, before playing Mission 2 (advanced), participants in all three conditions were asked
to rate how difficult did they perceive the first game play-through on a 7-point scale. An analysis
of covariance (ANCOVA) was used to analyze the manipulation check with the three scaffolding
conditions as independent variables (IVs) and the perceived difficulty as dependent variable
(DV). The participants’ gaming expertise was controlled as a covariate.
The analysis showed that there was significant difference in perceived difficulty between
the scaffolding conditions. F (2, 141) =3.79, P<.05, partial eta squared=.05. As expected,
participants in the no-scaffolding condition reported the highest perceived difficulty (M=4.94,
SD=1.49, N=49), followed by the content-knowledge scaffolding condition (M=4.64, SD=1.97,
N=45), followed by the gaming-skill scaffolding condition (M=4.04, SD=1.8, N=51). Post-Hoc
analysis using Bonferroni pair-wise comparisons was conducted. The analysis result showed that
there was significant difference in perceived difficulty between the no-scaffolding condition and
the gaming-skill scaffolding condition (p<.05). However, there was no significant difference in
perceived difficulty between the no-scaffolding condition and the content-knowledge scaffolding
condition (n. s.). Note that the gaming-skill scaffolding condition, which was perceived by the

55

participants to have the lowest difficulty among the three conditions, still has a mean score of
4.04 out of 7. This indicates that overall, the ReDistricting game was considered slightly difficult
for the participants in this study even with scaffolding.

56

Chapter 5
Results
Sample
Seven participants were removed because they could not complete the experiment
procedures within 60 minutes. Twelve participants were removed because they did not meet the
80% math accuracy criterion on the AOSPAN test.
Seventeen more participants were removed because they did not respond to any of the
secondary task probes while playing the game.
Researchers that measure reaction time almost always face the problem of how to deal
with extreme outliers. Outliers in reaction time can be extremely short or extremely long reaction
time. Unusual short reaction time can be the result of fast guesses, random tapping, or guesses
based on the subjects’ estimate of probe timing. Unusual long reaction time on the other hand,
can be caused by the uncooperative subjects, inattentive subjects, or conversely, full attention to
the primary task (Baayen & Milin, 2010).
Most experimental psychology studies using reaction time follow Roger Ratcliff’s (1993)
suggestions of removing outliers that are two to three standard deviations from the mean. Ratcliff
suggest that on only 85%~95% of the total sample is included in the analysis after removing
outliers. Using this criterion, I should remove participants whose response time smaller than
111.15ms (none), and larger than 2065.75 (n=5), and missing (n=17). A visual check on the five
participants with longer reaction time was performed, and they were slower in all four STRT
included in this study, therefore I concluded that the responses are genuine, reflecting the

57

participants’ slower reaction and unlikely a product of non-cooperation or inattention to
instructions.
The 17 participants who did not respond to any secondary probe were removed from this
study. This is because when the focusing of the measurement if on attention, missed response to
secondary probes can indicate two extreme situations: extreme inattention, and extreme attention
to primary task. A missed response can be because the participant is fully attentive on the
primary task, leaving no available working memory for processing the secondary task. Or it can
be because the participant is inattentive to the instructions and the secondary task, or that they
are bored about the experiment and are being uncooperative by not doing anything.
Unfortunately there is no way of distinguishing the two situations.
Since there is no way of distinguishing inattention from full attention, removing the nonresponse participants removes uncooperative participants along with participants who are fully
focused on the game and have no available working memory for the secondary task. This
procedure benefits the study because it removes noise from uncooperative participants which
improves the accuracy of the STRT as a measurement of attention. However, this procedure also
removed the participants who are so attentive to the game that they missed all the signals. While
on the surface this may seem to be problematic, I argue that removing them actually improves
the quality of the data for statistical analysis.
In this study, the secondary probe appears at a random interval with at least 30 seconds
between signals. Each signal remains on the screen for five seconds (5000ms) for the participants
to respond. If the participant does not respond within the five seconds, the server records the
response as a missing data. In other words, the fully attentive participants represent extreme

58

outliers who were at least eight standard deviations away from the mean. Even if I could
distinguish them from the uncooperative participants, including them in the analysis would
create a peak on the higher end and distort the mean. Moreover, that peak is not an accurate
measurement of attention because any response time longer than 5000ms to 30000ms and above
were lumped together with no variance between them. Therefore I argue that removing the 17
cases of non-response to secondary task improved the study by 1) removing noise from
uncooperative participants; 2) removing extreme outliers that are at least eight standard
deviations away from the mean; 3) allowing me to have a clearer interpretation of STRT as a
measurement of attention to primary task.
As a result, the sample used in the following analyses for testing the hypotheses consisted
of 146 participants. The participants were randomly distributed among the three conditions, with
49 participants in the no-scaffolding condition, 52 participants in the gaming-skill scaffolding
condition, and 45 in the content-knowledge scaffolding condition.
In order to ensure group equivalence among the independent and control variables, oneway Analysis of Variance (ANOVA) was performed between the three induced scaffolding
conditions on working memory capacity, gaming-skill expertise, political involvement, and
content knowledge expertise. The ANOVA results showed no significant difference between the
condition on the main independent variables--working memory capacity (F [2, 142] =.81, n. s.),
and gaming-skill expertise (F [2, 143] =.99, n. s.). There was also no significant difference
between the conditions on the control variables—political involvement (F [2, 133] =.86, n. s.),
and content-knowledge expertise (F [2, 143] =.08, n. s.). The ANOVA result indicates that the
three scaffolding groups were equivalent among the main independent variables and control
variables.
59

Hypotheses Testing
Hypotheses 1 and 2 posited that because working memory capacity and gaming-skill
expertise are more likely to be significant in complex tasks that that demand more attention
control. Therefore working memory capacity and gaming-skill expertise will positively predict
more attention (i.e. slower STRT) in the no-scaffolding condition.
A hierarchical multiple regression analysis was conducted to examine the hypotheses 1
and 2 under the no-scaffolding condition (n=49). Prior political involvement and content
knowledge were entered in the first block as control variables. The mean-centered AOSPAN
score and mean-centered gaming-skill expertise score were entered into the second block as
independent variables; the interaction term between AOSPAN and gaming-skill expertise was
entered into the third block to test interaction effect between working memory capacity and
gaming-skill expertise. Attention measured by STRT was entered as the dependant variable.
Under the no-scaffolding condition, the analysis result showed that that the overall model
was not significant, F (5, 37) =.46, n. s.. There was no main effect for neither working memory
capacity (β=.33, n. s.) nor gaming-skill expertise (β=.19, n. s.). There was also no interaction
effect between working memory capacity and gaming skill expertise on attention (β=-.31, n. s.).
The regression results are shown in Table 2. The results were not consistent with hypotheses 1
and 2. Differences in working memory capacity and gaming skill expertise did not predict
attention in the no-scaffolding condition which was expected to be the most demanding on
working memory.

60

Table 2 Regression analysis of working memory capacity and gaming expertise on attention in
no-scaffolding condition
(Intercept)
Content knowledge
Political experience

B
1104.43
-31.43
-17.60

(Intercept)
Content knowledge
Political experience
Working memory capacity (WMC)
Gaming-skill expertise (GE)

β
-.18
-.05

t
7.22
-1.13
-.31

p
<.001***
.134
.380

1119.21
-29.18
-26.31
9.40
-3.99

-.17
-.07
.12
-.02

7.09
-.99
-.44
.69
-.14

<.001***
.164
.331
.247
.447

1131.07
-26.96
-30.56
26.25
24.08
-4.94

-.15
-.09
.11
-.03
-.09

7.03
-.90
-.50
.76
.40
-.53

<.001***
.187
.309
.228
.348
.301

=.01
(Intercept)
Content knowledge
Political experience
Working memory capacity (WMC)
Gaming-skill expertise (GSE)
Interaction (WMC x GSE)
<.01
*<.05, **<.01, ***<.001.
Hypothesis 3 posits that because experts differ from novices in that they can draw on
existing schema from their long-term memory to process familiar information in chunks.
Expertise is more likely to be a significant predictor of attention when the task is recognized as a
familiar task. Therefore gaming skill expertise will positively predict attention in the gamingskill scaffolding condition which resembles a typical game design setup. Another hierarchical
multiple regression analysis was conducted to examine hypothesis 3, this time under the gamingskill scaffolding condition (n=52). Prior political involvement and content knowledge were
entered in the first block as control variables. The mean-centered AOSPAN score and meancentered gaming-skill expertise score were entered into the second block as independent
variables; the interaction term between AOSPAN and gaming-skill expertise was entered into the

61

third block to test interaction effect between working memory capacity and gaming-skill
expertise. Attention measured by STRT was entered as the dependent variable.
Under the gaming-skill scaffolding condition, the analysis result showed that the overall
model was significant, F (5, 37) = 2.47, p<.05, Adj.

= .15. There was no main effect for

working memory capacity (β=.34, n. s.). However there was significant main effect for gamingskill expertise (β=.77, p<.01). There was also significant interaction effect between working
memory capacity and gaming skill expertise on attention (β=-.89, p<.05). See Table 3 for the
regression analysis results. The result was consistent with hypothesis 3 and hypothesis 4; players
with higher gaming expertise allocated more of their attention to processing the game
information when the game design matched the familiar pattern of having controls and basic
navigation presented before the complex narrative and educational messages. The significant
interaction effect indicates that participants with higher gaming-skill expertise and lower
working memory capacity paid more attention to playing the game and responded slower to the
secondary probe.
Table 3 Regression analysis of working memory capacity and gaming expertise on attention in
gaming-skill scaffolding condition
(Intercept)
Content knowledge
Political experience

B
1170.26
29.71
-28.80

(Intercept)
Content knowledge
Political experience
Working memory capacity (WMC)
Gaming-skill expertise (GE)

1144.57
45.07
-33.86
-37.11
108.89

β
.08
-.06

t
4.89
.48
-.35

p
<.001***
.316
.365

.11
-.07
-.27
.34

4.95
.76
-.41
-1.74
2.21

<.001***
.225
.342
.045*
.016*

5.40

<.001***

=.15
(Intercept)

1201.32
62

Table 3(cont’d)
Content knowledge
Political experience
Working memory capacity (WMC)
Gaming-skill expertise (GSE)
Interaction (WMC x GSE)

15.18
-8.20
46.12
241.88
-25.65

.04
-.02
.34
.77
-.89

.26
-.10
1.06
3.11
-2.15

.398
.459
.149
.002**
.019*

=.09
*<.05, **<.01, ***<.001.
Hypotheses 5a and 5b are based on the expertise-reversal effect. Experts consider
scaffolding as redundant information that they cannot avoid. Therefore experts will have to use
more working memory in order to process the redundant information (slower STRT) in the
scaffolding conditions than without scaffolding.
Before testing these hypotheses, a median split was performed on gaming-skill expertise
(median=3.50) to categorize the participants into gaming experts (n=67) and non-experts (n=79).
An Analysis of Covariance (ANCOVA) was conducted among the gaming experts group (n=67)
to test the hypotheses. The three scaffolding conditions were entered as the independent variable.
Working memory capacity, content-knowledge, and political involvement were controlled as
covariates. The analysis result showed that there was significant difference between the three
scaffolding conditions in attention scores measured by STRT, F (2, 52) =3.74, <.05,
partial

=.13. Participants in the gaming expert group’s attention was highest in the game-skill

scaffolding condition (M=1392.78, SD=759.38), followed by the content-knowledge scaffolding
condition (M=1040.00, SD=279.57). The expert group’s attention was lowest in the noscaffolding condition (M=1007.59, SD= 293.26). See Table 4 for the ANCOVA analysis result.

63

Table 4 ANCOVA comparison of attention between scaffolding conditions
df
(Intercept)
Working memory capacity
Content knowledge
Political involvement
Scaffolding Conditions
*<.05, **<.01, ***<.001

F
1
1
1
1
2

p

partial
37.24
3.89
10.8
.07
3.74

.42
.07
.02
<.001
.13

<.001***
.054
.304
.790
.030*

A post-hoc comparison using Bonferroni test was conducted to determine the source of
the significance. The analysis result showed that there was significant difference between the
game-skill scaffolding condition and the no-scaffolding condition (p<.05). However, there was
no significant difference between the content-knowledge scaffolding condition and the noscaffolding condition (n. s.). The analysis results were consistent with hypothesis 5a that
participants with higher gaming-skill expertise were more attentive (i.e. used more working
memory) in the gaming-skill scaffolding condition than in the no-scaffolding condition. However
the result was not consistent with hypothesis 5b, higher-gaming expertise participants did not
allocate more working memory in the content-knowledge scaffolding condition than the noscaffolding condition.
The expertise-reversal effect argued that experts would comprehend more of the
educational content in the no-scaffolding condition without the redundant information caused by
scaffolding. Therefore hypothesis 6 posited that participants with higher gaming-skill expertise
would have the highest comprehension score in the no-scaffolding condition in comparison to
the two scaffolding conditions. An ANOVA analysis with the three scaffolding conditions as IV
and comprehension as DV was performed on the expert group (n=67) categorized by performing
a median split on gaming-skill expertise. The analysis result was not consistent with the
hypothesis. There was no significant difference in comprehension scores between the three
64

conditions for the expert group, F (2, 64) =.41, n. s. Even though comprehension in the noscaffolding condition (M=5.25, SD=2.31) was marginally higher than the gaming-skill
scaffolding condition (M=4.89, SD=2.66), the difference was not statistically significant. It was
the content-knowledge scaffolding condition which had the highest comprehension score
(M=5.58, SD=2.52), but the difference in comparison the other two condition were not
significant either.
The expertise reversal effect argued that reducing unnecessary redundant information
which increases working memory demand is better for expert learners. However, the reverse
cohesion effect posits that more active processing may actually improve comprehension. The
first research question (RQ1) in this study seeks to examine whether more or less attention to the
game leads to more comprehension in the context of learning from the ReDistricting game. Two
separate hierarchical regressions were conducted on the gaming-skill expert and non-expert
groups. Working memory capacity, content knowledge, and political involvement were entered
in the first block as control variables; attention (measured by STRT) was entered as the IV. And
comprehension score was entered as DV.
In the non-expert group (n=79), the overall regression model was significant, F (4, 66)
=3.07, p<.05, Adj.

=.12. The analysis results showed that working memory capacity was a

significant predictor of comprehension (β=.26, p<.05). Attention was positively correlated to
comprehension but only close to significant (β=.21, p=.054). See table 5 for regression analysis
results. The results indicate that for learners with lower gaming-skill expertise, working memory
capacity positively predicted comprehension.

65

Table 5 Regression analysis of attention on comprehension (non-expert group)

(Intercept)
Content knowledge
Political experience
Working memory capacity (WMC)

B
2.35
.11
.38
.17

(Intercept)
Content knowledge
Political experience
Working memory capacity (WMC)
Attention

.75
.18
.43
.15
.001

β
.08
.26
.29

t
2.77
.62
1.25
2.29

p
<.01**
.266
.109
.013*

.13
.18
.26
.21

.58
1.01
1.44
2.03
1.63

.282
.159
.077
.024*
.054

=.04
*<.05, **<.01, ***<.001
In the expert group (n=67), the overall regression model was not significant, F (4, 53) =
1.43, SD=.24, Adj.

=.03. None of the variables were significant predictors of comprehension.

Which suggest that for gaming-skill experts, neither working memory capacity nor attention
predicted comprehension.
So far the study has been largely focused on learners with higher gaming expertise.
Research question 2 seeks to examine whether gaming-skill scaffolding design mitigated the
difference between gaming-skill expert and non-expert to improve comprehension for the
participants with lower gaming-skill expertise. A one-way ANOVA was conducted to compare
comprehension of gaming-skill experts and non-experts between the gaming-skill scaffolding
condition and the no-scaffolding condition. The result showed that while in the no-scaffolding
condition, there was significant difference (F [1, 47] = 4.08. p<.05) between gaming-skill experts
(M=5.25, SD=2.31) and non-experts (M=4.00, SD=2.03). The significant difference became
insignificant in the gaming-skill scaffolding condition, F (1, 49) =.43, n. s. Participants with
higher gaming-skill expertise (M=4.89, SD=2.66) did not perform significantly better than
66

participants with lower gaming-skill expertise (M=4.47, SD=1.97). Comparison of the results is
shown in Figure 2. In other words, the difference between gaming-skill experts and non-experts
were mitigated by the gaming-skill scaffolding design.
Figure 2 The effect of gaming-skill scaffolding on comprehension

6
5.23

4.89

5
4

4.47

4

3

Gaming experts
Non-experts

2

1
0
No-scaffolding

Gaming-skill
scaffolding

Research Question 3 seeks to examine if content-knowledge scaffolding design would
improve comprehension for the participants with lower content-knowledge expertise and
mitigate the difference between participants with higher and lower content knowledge expertise.
A median split was performed on content-knowledge (median=2.00) to categorize the
participants into content experts (n=71) and non-experts (n=75). Note that since contentknowledge expertise is skewed towards the lower end, this categorization is only comparing
relative content expertise rather than actual content expertise. A one-way ANOVA was
conducted to compare comprehension of content experts and non-experts between the contentknowledge scaffolding condition and the no-scaffolding condition. The result showed that in the
67

no-scaffolding condition, there was no significant difference (F [1, 47] = .03. n.s.) between
content experts (M=4.67, SD=2.63) and non-experts (M=4.56, SD=1.83). However, there was
significant difference between content experts (M=6.47, SD=2.22) and non-experts (M=4.50,
SD=2.32) in the content-knowledge scaffolding condition, F (1, 43) =8.24, p<.01. Comparison of
the results is shown in Figure 3. The results indicate that instead of mitigating the difference
between content experts and non-experts, content-knowledge scaffolding seems to largely
improve comprehension of the participant who had relatively higher content knowledge in the
first place.
Figure 3 The effect of content-knowledge scaffolding on comprehension

7

6.47

6
5

4.67

4.56

4.5

4
Content Experts

3

Non-experts

2
1
0
No-scaffolding

Content-knowledge
scaffolding

Additional Results: Changes in Perceived Knowledge and Problem Severity
Since one of the two main goals of the ReDistricting game was to communicate the
problems of redistricting to players, two additional analyses were conducted to examine 1)
68

whether participants felt they had gained knowledge of redistricting after playing the
ReDistricting game, and 2) whether there was change in their perceived severity of the problems
associated with redistricting after playing the ReDistricting game.
Two sets of pre-test and post-test repeated measures were used to examine these
questions. The participants answered the question “How much do you know about redistricting?”
on a 7-point scale with 1=extremely little knowledge and 7=extremely knowledgeable. And the
question “Do you think the current redistricting process is…” on a 7-point scale with 1= Not
problematic at all, and 7=extremely problematic. The mean and standard deviation scores are
shown in Table 6
Table 6 Changes in perceived knowledge and perceived problem after playing the game

No-scaffolding
Pre-game knowledge
Post-game knowledge
Δ perceived knowledge
Pre-game perceived problem
Post-game perceived problem
Δ perceived problem
*<.05, **<.01, ***<.001

3.94 (1.51)
3.16 (1.84)
-.78
3.49 (1.00)
2.49 (1.37)
-1.00***

Scaffolding Conditions
Gaming-skill
Content-knowledge
scaffolding
scaffolding
2.83 (1.59)
2.82 (1.89)
3.61 (1.78)
3.53 (1.95)
.78*
.71*
3.54 (1.15)
3.53 (1.12)
2.75 (1.70)
2.24 (1.25)
-.79**
-1.29***

Three repeated-measures ANOVA were conducted to determine if the changes in selfreported knowledge were significant before and after playing the ReDistricting game. The results
showed that in the no-scaffolding condition, there was no significant change of self-reported
knowledge after playing the game, Wilk’s Lamda=.99, F (1, 48) =.64, n. s. However, in the
gaming-skill scaffolding condition, there was a significant increase of self-reported knowledge
after playing the game, Wilk’s Lamda=.89, F (1, 50) =6.06, p<.05. There was also a significant

69

increase in the content-knowledge scaffolding condition, Wilk’s Lamda=.87, F (1, 44) =6.62,
p<.05. The analyses results indicate that when playing under the no-scaffolding condition,
participants did not think they have gained knowledge about redistricting. But when playing
under the gaming-skill scaffolding or the content-knowledge scaffolding condition, participants
believed that their knowledge about redistricting improved.
Three other repeated-measures ANOVA with were conducted to determine if the
changes in perceived problem severity were significant before and after playing the ReDistricting
game. The results showed that in the no-scaffolding condition, there was significant decrease of
perceived problem severity after playing the game, Wilk’s Lamda=.74, F (1, 48) =16.56, p<.001
In the gaming-skill scaffolding condition, there was a significant decrease of perceived problem
severity after playing the game, Wilk’s Lamda=.86, F (1, 50)=8.41, p<.01. There was also a
significant decrease in the content-knowledge scaffolding condition, Wilk’s Lamda=.52, F (1, 44)
=40.49, p<.001. Surprisingly, instead of increasing participant’s awareness about the problems of
redistricting, the experience of playing the ReDistricting game decreased the participant’s
perceived severity about the problems of redistricting.

70

Chapter 6
Discussion
Summary and Discussion of Findings
Hypotheses 1 and 2 were not supported as working memory capacity and gaming-skill
expertise did not predict attention in the most difficult condition--the no-scaffolding condition.
Hypothesis 3 was supported, gaming-skill expertise positively predicted attention in the gamingscaffolding condition which resembles the conventional level design. Hypothesis 4 was also
supported as there was a significant interaction effect between working memory capacity and
gaming skill expertise on attention in the gaming-skill scaffolding condition. Hypotheses 5a and
5b were hypotheses based on the expertise reversal effect and reverse cohesion effect.
Hypothesis 5a was supported as gaming experts paid more attention to the gaming-scaffolding
condition in comparison to the other two conditions. However, hypothesis 5b was not supported.
Hypothesis 6 sought to test the predictions of the expertise reversal effect, in which experts
would comprehend more in the no-scaffolding condition, this hypothesis was not supported.
Research question 1 asked whether working memory capacity and attention to the game
predicted comprehension. The results showed that for experts, working memory capacity and
attention to game did not predict better attention. However, for non-experts, working memory
capacity was a significant predictor of comprehension, and attention to game was close to
significance. Research questions 2 and 3 asked whether gaming-skill scaffolding and contentknowledge scaffolding mitigated the gap between experts (gaming skill and content knowledge)
and non-experts. The results showed that gaming-skill scaffolding was effective in mitigating the
gap between gaming experts and non-experts, but content-knowledge scaffolding widened the
gap between content experts and non-experts.
71

Working Memory, Gaming-Skill Expertise, and Attention
This study sought to investigate how working memory capacity and gaming-skill
expertise interact with digital game scaffolding design to influence attention and comprehension
in the context of playing a digital game for learning.
Prior studies on digital game based learning have suggested that digital games
communicate with learner in multimedia environments via problem-solving experiences. The
learners’ limited working memory capacity is involved in processing the combination of visual,
auditory information, along with the navigation requirements, and problem-solving experiences.
Therefore, it has been suggested that learners’ working memory capacity may interact with
different instructional design to influence attention and how well learners can learn from digital
games (e.g., Kiili, 2005). Hypotheses 1 and 2 of this study are based on the assumption that the
effect of individual differences in working memory capacity and gaming-skill expertise on
attention to the game would be significant when the task is more complex and requires more
attention control (Kane & Engle, 2003). Surprisingly, the analysis result was not consistent with
these two hypotheses. Working memory capacity and gaming-skill expertise were not significant
predictors of attention in the no-scaffolding condition.
One potential explanation for this result is that the working memory demand of the
ReDistricting game exceeds that of the participants in this study. When participants were asked
to rate the difficulty of Mission 2 (advanced) on a 7-point scale after the game, the mean score
under the no-scaffolding condition was 5.12 (SD=1.98). This result suggests that participants in
the no-scaffolding condition perceived the game as difficult. Individual differences in working
memory capacity and expertise would be most significant if the working memory demand of the

72

task falls within the working memory capacity distributions of the participants. If the task is too
easy, we would not observe variations because most participants would be able to perform the
task without tapping their working memory limits. Similarly, if the task is too difficult, we would
not observe variations because most participants would suffer from cognitive overload and could
not perform the task. Going back to the staring contest analogy described in the literature review
section, if the staring contest was too demanding and ran for 60 seconds, both contestant A who
can stare for 50 seconds and contestant B who can stare for 20 seconds would fail and we would
observe no significant difference.
Most prior studies on individual differences in working memory capacity and expertise
have used relatively simple tasks such as the Stroop test (Kane & Engle, 2003) or proactiveinference test (Engle & Kane, 2000). When using simple tasks, it is less likely that the working
memory demands of the tasks would exceed the participants’ capacity. But in a complex
problem-solving experience such as digital game based learning, working memory is involved in
a number of information processing such as visual-spatial navigation, perceptual motor skills,
accurate timing of responses, as well as short and long-term strategic decisions (Logie et al.,
1989). Therefore it becomes more likely for complex cognitive tasks to overload the learners’
capacity. When the learners suffer from cognitive overload, they may have slower reaction time
and large number of missed response to secondary probes. But they may also respond by giving
up on the primary task (playing and processing the game), and switch their attention to reacting
to the secondary probe, thus instead of longer reaction time, they will have shorter reaction time.
When the task may be too complex and demanding to the participants, additional measures of
attention should be applied to cross reference with STRT results.

73

The third hypothesis is based on the theory and findings that show expert performance is
determined by the ability to identify a task as within one’s existing schema, and the ability to
retrieve matching mental representations and solutions stored within one’s long-term memory
(e.g., Chase & Simon, 1973). When experts cannot recognize the task as within their expertise
domain, or cannot retrieved matching information from their long-term memory, they are likely
to process the task and information in the same way as non-experts and have no advantage.
Therefore the study hypothesized that gaming-skill expertise is more likely to be a significant
predictor of attention when the game is recognized as a familiar task and learners with higher
gaming-skill expertise will actively search and process goal-related information from the game
and within their long-term memory. The analysis result was consistent with hypothesis 3,
gaming-skill expertise positively predicted attention to the game in the gaming-scaffolding
condition. The finding indicates that when the conventional level-design is applied to support
learners by introducing them to the basic controls and narratives first, learners who have higher
gaming-skill expertise are more likely to recognize the task as within their expertise domain, and
allocate more attention resources (i.e. working memory) to processing information from the
game and retrieving mental representations from their schema.
Gaming-skill expertise also interacted with the learners’ working memory capacity to
affect attention. In other words, learners with lower working memory capacity and higher
gaming-skill expertise allocated more attention resources to the game. This finding is not
surprising because according to theory, working memory capacity should be negatively
correlated with STRT. Learners with higher working memory capacity would react faster to
secondary probes because they have more available working memory, while learners with lower

74

working memory capacity would react slower to secondary probes because they have less
available working memory.
Another way of interpreting the result of hypothesis 3 is that the game-skill scaffolding
condition is considered a more “cohesive” text, especially to learners with higher gaming skill
expertise. Text cohesion is relative to the learners’ expertise level. A cohesive text is one in
which all the relevant information are included in a coherent way to the learner. In contrast, an
incohesive text is one that requires learners to draw upon existing expertise to make inference
between information and maintain coherence (Oakhill, Cain, & Bryant, 2003; Ozuru, Dempsey,
& McNamara, 2007). Note that the majority of participants in this study had low content
knowledge about the redistricting process (M=2.86, SD=1.65). Those who had higher gamingskill expertise were able to see the design structure of the gaming-scaffolding condition as a
more cohesive text that included the relevant information for navigating and understanding the
game, whereas the learners with lower gaming-skill expertise could not. Consequently, in
accordance to the findings by O’Reilly and McNamara (2007), learners with low content
knowledge only benefitted from cohesive text (i.e. the gaming-skill scaffolding condition) when
they had sufficient gaming-skill expertise. In other words, gaming-skill expertise can help
learners with lower content expertise improve attention to information processing, but only when
the game’s design matches the schema about digital games stored in the learners’ memory.
Looking back at the non-significant result of hypotheses 1 and 2 in the no-scaffolding
condition, an alternative explanation is that without introducing the controls, narrative, and
problem goals in a coherent order, even the learners who had relatively higher gaming-skill
expertise could not recognize the game as a cohesive text and did not benefit from their expertise.
This is similar to the study on chess masters by Chase and Simon (1973). The chess master’s
75

superior memory performance disappeared when a random set of chess pieces were given. When
the chess masters cannot recognize the patterns in front of them and find matching patterns from
their long-term memory to help them process the information; they are forced to forgo any expert
strategies and processing each individual chess piece just as a novice would. In general, gamingskill expertise cannot facilitate attention if the game was not designed or introduced in a
recognizable way for learners with higher gaming-skill expertise to apply their schema and
expert strategies.
Expertise-Reversal Effect, Scaffolding, and Comprehension
Hypotheses 4a and 4b were based on the arguments by Kalyuga et al. (2000, 2003) that
scaffolding induces higher working memory demands on experts because of the introduced
redundant information. The analyses results showed that indeed, gaming-skill experts paid more
attention (slower STRT) to the gaming-skill scaffolding condition than the no-scaffolding
condition. There was no significant difference between the content-knowledge scaffolding
condition and the no-scaffolding condition. This finding was consistent with hypothesis 4a
derived from Kalyuga et al. (2000, 2003).
However, contrary to Kalyuga et al.’s (2003) interpretations, study findings that show
experts increased working memory load in the scaffolding condition does not necessarily
disconfirm McNamara’s arguments about the reverse-cohesion effect. Kalyuga et al.’s (2003)
definition of experts did not separate content-knowledge expertise and skill expertise, while
McNamara’s reverse-cohesion effect was strictly predicting a reversal for experts with higher
content-knowledge expertise, not skill expertise. In fact, McNamara’s more recent study showed
that when learners possess higher skills to process the text, both learners with higher and lower

76

content-knowledge expertise benefitted from having cohesive text (Ozuro, Dempsey, &
McNamara, 2007). Therefore, the findings in hypotheses 4 which showed gaming-skill experts
had higher attention to the gaming-skill scaffolding condition can support both explanations
provided by Kalyuga et al. (2003) and McNamara (2001). Gaming-skill scaffolding may have
introduced redundant information for learners with higher gaming-skill scaffolding, but at the
same time the redundant information helped them identify the task as a cohesive text and retrieve
related information from their schema. The increased attention may be partly attributed to
processing the additional information, and partly attributed to retrieval of related information
from the learners’ long-term memory.
The result of hypothesis 4 focusing on attention did not tease apart the two explanations.
Hypothesis 5 seeks to compare the two explanations by examining comprehension. Kalyuga et al.
(2003) argued that redundant information is harmful to experts; therefore gaming-skill experts
would have higher comprehension in the no-scaffolding condition than the two scaffolding
conditions. On the other hand, McNamara et al. (Ozuru, Dempsey, & McNamara, 2007) argued
that learners with higher skill expertise would have higher comprehension in the cohesive
condition, that is, the gaming-skill scaffolding condition in this study. Unfortunately, the results
neither confirmed nor disconfirmed either explanation as there was no significant difference in
comprehension scores for gaming-skill experts between the three conditions. This result along
with the previous findings seem to suggest that while gaming-skill experts allocated more
attention to the gaming-scaffolding condition, the increased attention did not translate into higher
or lower comprehension in comparison to the other two conditions.
However, this does not suggest that increased attention for experts is not associated with
comprehension. The relationship between expert attention and comprehension depends on the
77

game’s design, or according to Fisch’s (2000) model, the distance between narratives and
educational content. I argued that in the ReDistricting game, because the educational content is
opposite of the game’s explicit goals and mechanics, therefore when the expert gamers quickly
identified the game as within their expertise domain, they may have focused their attention on
the game’s explicit goal and mechanics, and blocked out irrelevant information which are
important to comprehension but not to reaching the game’s goal. It is possible that in a game
which has a short distance between narrative and educational content, or that the narrative is the
educational content, increased expert attention may improve comprehension.
The expertise reversal effect is based on the assumption that reducing unnecessary
redundant scaffolding which increases working memory demand benefits expert learners.
However, the reverse cohesion effect posits that more active processing may actually improve
comprehension. The first research question that this study posed was “will attention to the game
positively or negatively predict comprehension?” This is not an easy question to answer and the
answer is determined by both learner characteristics and the game’s design. The analysis result
showed that consistent with the previous finding, for learners with higher gaming skill expertise,
attention to the game was not a significant predictor of comprehension.
However, when examining learners with lower gaming-skill expertise, working memory
capacity was a significant predictor of comprehension, and attention to game was close to
significance (p=.054). The results suggest that the gaming-skill experts and non-experts may be
paying attention to different aspects of the digital game based learning experience. While STRT
tells us how much working memory is allocated to the game, it cannot differentiate between
learning-related attention and extraneous attention that is not related to comprehension. It is
possible that when gaming-skill experts allocate more attention, they are focusing on more goal78

related information. And when non-experts allocate more attention, they are processing all the
information equally without distinction. According to Fisch’s capacity theory of comprehension
from educational media (Fisch, 2000), when the distance between the narrative and the
educational content is large, working memory allocated to processing the narrative competes
with working memory allocated to processing the educational content. Such might be the case
with the ReDistricting game. The ReDistricting game adopts a reverse role-playing narrative in
which the player plays as the problem source, and the intended message is the opposite of the
game’s overt goals. In this case, it is likely that gaming-skill experts focused on achieving the
game’s overt goals, and processed information that may aide them in reaching that goal while
blocking out “irrelevant” information as interference. However, since that goal is not the
intended educational message, attention to goal-related information does not predict more
comprehension. In comparison, non-experts do not have gaming schema to help them identify
what is goal-related information and unrelated information, they are more likely process a larger
set of information. And because non-experts do not have existing mental representations from
previous game experiences, they need to use the information from this game to construct their
mental representation, which may sometimes lead to more comprehension of the intended
message.
This explanation is further supported by the additional analyses results, while both
gaming-skill experts and non-experts believe to have gained knowledge after playing the
ReDistricting game, their perceived severity about the problems of redistricting reduced instead
of increased. The reverse role-playing plus the higher difficulty may have misled some
participants to conclude that the game is a demonstration of how difficult it is to manipulate

79

elections through redistricting. Many participants in the study responded in to the open-ended
comprehension questions with frustration over how difficult it is to redistrict.
“[T]he redistricting process is tedious and someone will always be unhappy when their
district is being cut into by another. It is incredibly difficult to make the districts
completely fair for both parties and all candidates.”—Participant 11
“Honestly, I do not really know much. However, according to the game, the redistricting
process can be kind of tedious and irritating. Pleasing everyone, at the same time was
nerve wrecking.”—Participant 32
More about the theoretical and practical implications of the study will be discussed in the
next chapter.
Scaffolding Design to Support Non-Experts
The last two research question in this study investigated whether gaming-skill scaffolding
and content-knowledge scaffolding could improve comprehension of learners with lower
gaming-skill expertise and content-knowledge expertise. In other words, does scaffolding help
novices? The analysis results were very interesting. Gaming-skill scaffolding mitigated the
difference in comprehension between gaming-skill experts and non-experts. But contentknowledge scaffolding widened the gap through supporting learners with relatively more
content-knowledge expertise.
Note that gaming-skill scaffolding was intended to support novice gamers by improving
their ability to navigate the game. Without gaming-skill scaffolding, learners with lower gamingskill expertise are more likely to suffer from disorientation caused by the non-linear multimedia
environment and therefore have lower comprehension (Scheiter & Gerjets, 2007). Through
80

gaming-skill scaffolding, they should be able to gain expertise in navigating the game and
searching for goal-related information, therefore the gap between gaming-skill experts and nonexperts should narrow. This finding suggests that it is very important for educational games to
provide some kind of skill-based support for learners with lower gaming-skill expertise.
Content-knowledge scaffolding was intended to support learners with lower content
knowledge by providing an external framework for them to combine and interpret the game
information. In some ways, it is like priming participants to focus on certain information. Note
that in general, the participants in this study had low knowledge about the topic of scaffolding.
Therefore the distinction between content experts and non-experts are only relatively high and
low content-knowledge expertise. The actual difference between the two groups is very small.
What was surprising is that instead of mitigating the gap between content-knowledge experts and
non-experts, providing an explicit framework did not benefit the learners with relatively low
content knowledge. Instead it benefitted those who had relatively more content knowledge at the
beginning. This is similar to the rich-gets-richer effect in which the learners with scarce content
knowledge are unable to make sense of the game even when the explicit framework was
provided to them. But for learners with even a small amount of content knowledge, they were
able to utilize the additional explicit framework and map their limited knowledge and new
information on to that framework. As a result, their comprehension improved exponentially. This
finding suggests that it is important for learners to have some basic knowledge about the content
before they play educational digital games.
Limitations

81

Prior studies on the educational effects of digital game based learning have found mixed
results. This study sought to investigate the mixed results through examining how individual
differences in working memory capacity and gaming-skill expertise interact with game
scaffolding designs to affect attention and comprehension. As with all empirical studies, this
study is not without limitations.
First of all, the study measured gaming-skill expertise and content-knowledge expertise
through self-reported measures. This was a trade-off decision made to avoid objective
measurement and practice effect. While it may be better to obtain objective measures of gaming
skills and content knowledge by administrating a test that require participants to play the game or
answer questions that are closely related to the redistricting process. These tests are likely to
prime participants to look for answers in the game, or confound the study by allowing them to
practice. However, by using self-reported measures, the study may introduce social-desirability
issues and other third variable effects. Luckily, through random assignment and test of group
equivalence, the three groups seem to be equivalent and comparable.
Second, the results of this study are based on a single game--the ReDistricting game. The
game was chosen because it was designed for undergraduate students and involves a topic
undergraduate students are generally not familiar with. However, the generally low content
knowledge restricted the study from testing how actual content experts may interact and process
information within the three conditions. On the flip side, because the participants in this study
were generally low on content knowledge expertise, they provided a relatively homogenous
sample and allowed the study to examine the effects of working memory capacity and gamingskill expertise while controlling for the interaction effects of content knowledge.

82

Third, while there has been a long tradition in psychology studies of using secondary task
reaction time for measure working memory load and attention (e.g., Daneman & Carpenter,
1980); many of the studies have used relatively simple cognitive tasks in comparison to a
complex problem-solving environment such as digital games. Findings from this study seem to
suggest that there is qualitative difference in the information that learners with higher gamingskill expertise and non-experts were focusing their attention on. STRT merely measures the
amount of working memory allocated to the primary task, it cannot measure which aspect of the
task were the participants processing? It also cannot distinguish between working memory
allocated to processing external information and internal processing. However, it is important to
remember that the construct of working memory is a theoretical construct and was not expected
to map directly to areas in the brain or other biological measures (Baddeley, 2000). STRT may
be used in combination to other measures of attention such as heart rate, eye-tracking, or serverbased behavioral data to further determine whether there is indeed attention-focus difference
between how gaming-skill experts and non-experts allocate their working memory to the game.
Fourth, the manipulation check indicated that while there was difference in complexity
between the induced conditions, however even the lowest difficulty was rated as 4 on a 7-point
scale. In other words, the study’s manipulation did not create a single condition that was
perceived as easy to the participants in this study. Instead, the comparison between the noscaffolding condition and the gaming-skill scaffolding condition is a comparison of a difficult
condition against a moderate difficulty condition. This may have caused hypotheses 1 and 2 to be
non-significant, because the difficulty exceeded the participants’ capacity, even for those who
have higher working memory capacity and gaming-skill expertise. Without a condition that was
considered easy, the study could not really examine McNamara’s reverse-cohesion effect to see
83

whether content-knowledge experts would engage in superficial information processing when
they perceive the task to be too easy.
Fifth, the participants in this study only played the ReDistricting game for less than 30
minutes and could not replay multiple times or discuss the gameplay experience with other
participants. This may have limited the effectiveness of the game in improving comprehension.
However, this decision allowed the study to control for effects of repeated exposure to media and
peer-learning, thus the results from this study are limited to self-directed learning through an
educational digital game.
Sixth, the study results did not fully support neither the expertise-reversal effect nor the
reverse-cohesion effect. It is possible that the manipulation of scaffolding in this experiment was
not strong enough to observe the effects. To be more specific, the expertise reversal effect argues
that the addition scaffolding information causes redundancy effect for expert learners and
increase the amount of working memory demand, and even cognitive overload which hurts
comprehension. The manipulation in this study may not have been strong enough to exceed the
working memory capacity of the expert participants in this study, thus they were able to process
the additional information without sacrificing their comprehension.

84

Chapter 7
Conclusion and Implications
Theoretical Implications
This study follows the line of research on expert performance and working memory (e.g.,
Hambrick & Engle, 2002) and extends the focus to investigate how working memory capacity
and gaming-skill expertise influence attention and comprehension in the context of digital game
based learning. Theoretically, the study adds to the large body of studies that show expertise as a
strong predictor of attention on domain-specific information and complex cognitive tasks
(Hambrick & Engle, 2002; McNamara, 2001). It also shows that among participants with lower
gaming-skill expertise, working memory capacity was a significant predictor of comprehension
in digital game based learning. Working memory capacity facilitated comprehension for novices
because novices with higher working memory capacity were more capable of holding multiple
information elements in their working memory and construct a mental representation of the
problem. Consistent with findings from Hambrick and Engle (2002) the results from this study
suggest that working memory capacity and gaming-skill expertise affect attention and
comprehension as separate predictors. As a more general construct, working memory capacity
may have affected comprehension through other variables. Gaming-skill expertise affected what
information to focus on and how much working memory to allocate to the game.
The findings in the current study also suggest that there may be qualitative differences in
how experts and novices allocate their attention. Due to the limitations of using STRT to
measure attention, the current study cannot distinguish between attention allocation to different
aspects of the game play and learning experience. However, previous studies have suggested that
experts may be able to process procedural information automatically, which may sometimes
85

benefit learning, but sometimes harm learning (Ricks, Turley-Ames, and Wiley, 2007).
Automatic processing allows expert learners to free up more available working memory for
processing other information, but it may also limit their attention to familiar information which
matches their existing mental representations, while blocking out novel information that may be
important but does not match their mental representation. For example, Ricks et al. (2007) used
the Remote Associates Task to test whether experts and non-experts were able to detect
questions designed to mislead experts into automatic processing and construct false answers.
Their findings showed that when the task appeared to be within the experts’ expertise domain,
experts were likely to activate habitual processing and fail to detect that the automatically
generated answer was incorrect. In comparison, because non-experts do not have existing mental
representations to guide their inference, they were more likely to process the question cautiously
and answer correctly. This finding suggest that in some cases, especially when the task design is
complex and structurally similar to the experts’ expertise domain, experts are more likely to
make mistakes because they are quick to identify the task and process information with their
existing schema, they do not consider alternative models and solutions.
However this does not imply that automatic habitual processing will always lead experts
to overlook the educational contents. If the game’s educational content is closely integrated with
the game’s mechanics, automatic processing by expert gamers may increase comprehension
because they can understand and process the mechanics faster.
Overall, this study argues that any sweeping claim about the superior cognitive
performance of experts or working memory capacity must be examined carefully. Based on the
structure of the game and its narrative, sometimes the retrieval of exert schema may block out
information that are relevant to learning. It is important for future studies to carefully consider
86

how working memory capacity and gaming skill expertise may affect attention allocation, but
also examine the structure of the game to determine if it is designed in a way that activates expert
schema retrieval, and whether the retrieval of expert schema benefit or harms comprehension.
Practical and Design Implications
The study also has practical implications for practitioners and game designers of
educational digital games. The study showed that many participants failed to comprehend the
intended message of the ReDistricting game and instead constructed knowledge that was in the
opposite direction. The ReDistricting game was designed using a common narrative used in
educational and persuasive games—a reverse role playing perspective. While this type of
narrative was intended to facilitate reflection and active knowledge construction, it should be
used with caution. Instead of increasing their perceived severity about the problems of
redistricting, the participants in the study decreased their perceived severity. This may be
because the reverse role taking design first requires players to identify with the role and goals of
the problem-source, and then reflect on their experience as a wrong-doing. Therefore in order to
reach the goal, the learners actively identified with the source of the problem (e.g., the
gerrymandering mapmaker). This reverse role-playing combined with the higher difficult may
have depleted the learners’ working memory capacity so that they did not have sufficient
working memory for reflect and comprehend the intended narrative: that redistricting can be
easily abused to undermine fair election. Instead, the high difficulty and reverse role-playing led
participants to believe that the educational message was how difficult it was to redistrict and
please political parties and the constituents.

87

However, this does not imply that the ReDistricting game or games that were designed
using similar reverse role-playing perspectives are bad games. The fact that there was significant
change in the perceived knowledge and problem severity indicate that playing the game was
effective. However, when players play the game on their own, without a teacher or other
guidance to help them make sense of the experience, this type of design may have unintended
effects. The implication of this finding is that the reverse role-playing design should be used with
caution and with guidance. A facilitator such as a teacher or an artificial tutoring agent could
guild the players to reflect upon their experience and restructure their comprehension in the
intended direction.
Another implication of this study is that digital game based learning should carefully
balance the need to seek realistic simulations and game complexity. The advanced difficulty in
the ReDistricting game was designed to be realistic and included the multiple conflicting goals of
competing parties, the court, and legal constraints. However the multiple, sometime conflicting
goals means that learner will have to hold the multiple goals in their working memory while
making each decision. This inevitably increased the amount of working memory demand of
playing this game. Under such high working memory demand, the players may not have
sufficient working memory to reflect and understand the underlying message that the game
intended to communicate. Educational game designers should make a careful decision about
whether a realistic simulation is compatible to their intended effects. If the goal was to convey
how redistricting could easily manipulate election results, then the game designer may want to
consider removing some of the complexity (e.g., competing goals, multiple constraints, etc.) so
not to overload the learners’ working memory.

88

Another practical implication of this study is in the effect of skill or content scaffolding.
The findings showed that gaming-skill scaffolding mitigated the difference between gaming-skill
experts and non-experts. But the content-knowledge scaffolding widened the gap between
content-knowledge experts and non-experts. The findings imply that gaming-skill expertise is an
important factor affecting comprehension from digital games. Learners who have relatively low
gaming skills are disadvantaged if no scaffolding is provided to them. Practitioners and game
designers should implement scaffolding to support those who are less familiar with the media so
that they do not become dysfunctional learners.
On the other hand, the findings suggest that content-knowledge scaffolding does not
benefit those with scarce existing content expertise. But it significantly improved comprehension
for learners with relatively more content expertise. This study suggest that self-directed learning
with digital games such as the procedure in this study should not be used on learners who have
almost no existing content knowledge. Instead, the practitioner and game designer should
implement other techniques to increase existing content knowledge of the learners before they
play. Once the learners have some basic understanding of the content, they are able to increase
knowledge exponentially through minimal scaffolding and self-directed learning.
As a form of interactive medium, digital games have the potential to adapt to learner
differences and tailor messages according to learner characteristics. This study is an early step in
examining the effects of working memory capacity and gaming-skill expertise on attention and
comprehension among different game scaffolding designs. Future studies should examine how to
better measure attention and working memory allocation in complex learning environments such
as digital games. It would also be worthwhile to examine different combination of scaffolding
design to seek a design that would benefit both experts and non-experts.
89

APPENDICES

90

APPENDIX 1
Pre-test Questionnaire
Political Involvement (Kahne, Chi, & Middaugh, 2002)
How much do you agree or disagree that the following statements describes you?
[Strongly disagree=1 2 3 4 5 6 7=Strongly agree]
1. I often talk about politics and political issues.
2. I am interested in a career in politics and government
3. I often wear badges in support of certain political issues
4. I vote regularly
5. I often volunteer for political actions
6. I often post or forward political information to my friends
7. I often sign political petitions
8. I strongly support boycotting products that are against my political values
9. I often attend protests in support of my political values
Political Efficacy (American National Election Survey, 2011)
[Not at all=1 2 3 4 5 6 7=A great deal]
10. How much do government officials care what people like you think?
11. How much can people like you affect what the government does?
12. Do you believe people like you can change the government’s decision?
Political Beliefs
How much do you agree or disagree with the following statements?
91

[Strongly disagree=1 2 3 4 5 6 7=Strongly agree]
13. Every citizen should have equal rights to vote.
14. Fair competition between candidates is crucial to a democratic election.
15. The current election system can be easily manipulated by people who holds power
16. Voters should decide who wins an election
ReDistricting Knowledge
[Strongly disagree=1 2 3 4 5 6 7=Strongly agree]
17. I am very familiar with the redistricting process
18. I consider myself an expert in how gerrymandering works.
Game Familiarity
On an average week, how many hours do you play the following digital game genres (on any
device)?
[Not at all=1; 0 to 1hours=2; more than 1 hours, less than 2 hours=3; more than 2 hours, less
than 3 hours=4; more than 3 hours, less than 4 hours=5; more than 4 hours, less than 5
hours=6; more than 5 hours=7]
19. First-person shooter (e.g., Call of Duty, Battlefield, Team Fortress, etc…)
20. Strategy (e.g., Civilization, StarCraft, tower defense, etc…)
21. Puzzle (e.g., Bejeweled, Bubble witch saga, Angry bird, etc…)
22. Simulation (e.g., SimCity, Sims, Farmville, Cityville, etc…)
23. Role-playing game: RPG (e.g., The Elder Scrolls: Skyrim, Diablo, Final Fantasy, etc…)
24. Sports (e.g., NBA 2K12, Madden NFL, racing, etc…)
92

25. Massively Multiplayer Online game: MMO (e.g., World of WarCraft, Star Wars: the old
republic, etc…)
26. Action (e.g., Super Mario, Super Smash Bros, fighting games, platform games, etc.)
27. Music game (e.g., Guitar hero, Just Dance, etc…)
28. Board/ card digital games (e.g., Chess, Magic: the gathering, etc…)

How skilled are you with the following game genres?
[Not at all=1 2 3 4 5 6 7=Extremely skilled]
29. First-person shooter (e.g., Call of Duty, Battlefield, Team Fortress, etc…)
30. Strategy (e.g., Civilization, StarCraft, tower defense, etc…)
31. Puzzle (e.g., Bejeweled, Bubble witch saga, Angry bird, etc…)
32. Simulation (e.g., SimCity, Sims, Farmville, Cityville, etc…)
33. Role-playing game: RPG (e.g., The Elder Scrolls: Skyrim, Diablo, Final Fantasy, etc…)
34. Sports (e.g., NBA 2K12, Madden NFL, racing, etc…)
35. Massively Multiplayer Online game: MMO (e.g., World of WarCraft, Star Wars: the old
republic, etc…)
36. Action (e.g., Super Mario, Super Smash Bros, fighting games, platform games, etc.)
37. Music game (e.g., Guitar hero, Just Dance, etc…)
38. Board/ card digital games (e.g., Chess, Magic: the gathering, etc…)
Game Effectiveness Questions
39. How much do you know about redistricting?

93

[Extremely little knowledge=1 2 3 4 5 6 7=extremely knowledgeable].
40. Do you think the current redistricting process is…?
[Not problematic at all=1 2 3 4 5 6 7=extremely problematic].

94

Appendix 2
Instructions for the stimuli
[No-Scaffolding condition]
First gameplay
PLEASE READ THE INSTRUCATIONS CAREFULLY.
A. Play through Mission 2 on the Advanced difficulty for 5 minutes. Pay attention to the
game. You will be tested about the contents afterwards.

B. During the game play, you will see a yellow box appear randomly around the game.
When you see it, press SPACEBAR as fast as you can. You will need to replay the
game if you miss too many or is too slow on this task.

Second gameplay
PLEASE READ THE INSTRUCATIONS CAREFULLY.
A. Play through Mission 2 on the Advanced difficulty again. Pay attention to the game.
You will be tested about the contents afterwards.

B. During the game play, you will see a yellow box appear randomly around the game.
When you see it, press SPACEBAR as fast as you can. You will need to replay the
game if you miss too many or is too slow on this task.

95

[Gaming-skill scaffolding condition]
First gameplay
PLEASE READ THE INSTRUCATIONS CAREFULLY.
A. Play through Mission 1 on the Advanced difficulty. Pay attention to the game. You will
be tested about the contents afterwards.

B. During the game play, you will see a yellow box appear randomly around the game.
When you see it, press SPACEBAR as fast as you can. You will need to replay the
game if you miss too many or is too slow on this task.
Second gameplay
PLEASE READ THE INSTRUCATIONS CAREFULLY.
A. Play through Mission 2 on the Advanced difficulty. Pay attention to the game. You will
be tested about the contents afterwards.

B. During the game play, you will see a yellow box appear randomly around the game.
When you see it, press SPACEBAR as fast as you can. You will need to replay the
game if you miss too many or is too slow on this task.

[Content-knowledge scaffolding condition]
Content-knowledge scaffolding text

96

About the ReDistricting game
“Once you draw the district lines, you pretty much know who will win most races. And the
amount of money spent on campaigns becomes nearly irrelevant. Most elections are over before
they start.”
— Steven Mulroy, Former Attorney, U.S. Dept. of Justice
On the same day last November that President Obama was reelected by a solid margin, his
party’s candidates for the House received about 1.5 million more votes collectively than the
GOP’s candidates. And yet when the new Congress convened, Republicans still enjoyed a solid
majority in the lower chamber, with 235 seats.
The most popular explanation for this disparity involves gerrymandering—when one party uses
its control of state government to redraw congressional district lines in a way that gives its
candidates an electoral edge.
The Redistricting Game is designed to educate, engage, and empower citizens around the issue
of political redistricting. Currently, the political system in most states allows the state legislators
themselves to draw the lines. This system is subject to a wide range of abuses and manipulations
that encourage incumbents to draw districts which protect their seats rather than risk an open
contest.
By exploring how the system works, as well as how open it is to abuse, The Redistricting Game
allows players to experience the realities of one of the most important (yet least understood)
aspects of our political system. The game provides a basic introduction to the redistricting
system, and allows players to explore the ways in which abuses can undermine the system.
97

First gameplay
PLEASE READ THE INSTRUCATIONS CAREFULLY.
C. Play through Mission 1 on the Advanced difficulty. Pay attention to the game. You will
be tested about the contents afterwards.

D. During the game play, you will see a yellow box appear randomly around the game.
When you see it, press SPACEBAR as fast as you can. You will need to replay the
game if you miss too many or is too slow on this task.
Second gameplay
PLEASE READ THE INSTRUCATIONS CAREFULLY.
C. Play through Mission 2 on the Advanced difficulty. Pay attention to the game. You will
be tested about the contents afterwards.

D. During the game play, you will see a yellow box appear randomly around the game.
When you see it, press SPACEBAR as fast as you can. You will need to replay the
game if you miss too many or is too slow on this task.

98

Appendix 3
Post-test questionnaire
Game Enjoyment
How much did you enjoy the game?
How pleasant do you feel about the game?
How fun do you feel about the game?
Difficulty and Efforts
How difficult was the game?
How much effort did you put into playing the game?
ReDistricting Knowledge
[Strongly disagree=1 2 3 4 5 6 7=Strongly agree]
1. I am very familiar with the redistricting process
2. I consider myself an expert in how gerrymandering works.
3. How much do you know about redistricting?
[Extremely little knowledge=1 2 3 4 5 6 7=extremely knowledgeable].
4. Do you think the current redistricting process is…?
[Not problematic at all=1 2 3 4 5 6 7=extremely problematic].
Free Recall
What are the potential problems with redistricting? (Describe as much detail as you can,
including the principles of drawing election districts, and how it can be manipulated).
99

Cued-Recall Questions (Comprehension)
1) In the game, each election district must have____________ (number) of constituents.
A. 650,000 - 660,000
B. 640,000 - 650,000
C. 630,000 - 640,000
D. 620,000 - 630,000
E. 610,000 – 620,000

2) From where did we get the term “Gerrymandering?”
A. Elbridge Gerry, former governor of Massachusetts
B. Geary Stohlt, colonial practice of reapportionment
C. Gerry Minder, former director of the United States Census
D. Aaron Mandering, former mayor of Atlanta, GA.
E. Ronald Reagan, 40th President of the United States

3) What does “cracking” mean when it comes to redistricting?
A. Carving off part of a district to create a new seat in Congress
B. Unfairly under-representing poor neighborhoods
C. Breaking up groups of similar voters
D. All of the above
E. None of the above

4) What does “packing” mean when it comes to redistricting?
100

A. Removing a district to oust the incumbent representative
B. Jamming one district with similar voters
C. Inflating the number of citizens in each district
D. All of the above
E. None of the above

5) ___________is a traditional standard that states election districts should not have jagged
edges and skinny extensions.
A. Roundness principle
B. Contiguity principle
C. Compactness principle
D. All of the above
E. None of the above

6) What does partisan gerrymandering refer to?
A. Splitting election districts into smaller parts.
B. Moving one part of the election district to another district.
C. Changing the number of voters in each district.
D. Manipulating election districts to prevent competition between parties.
E. Manipulating election districts to favor one party over the others.

7) How does packing create partisan advantage for a specific party?
A. Cause wasted votes to opposing parties.

101

B. Cause inflated votes to opposing parties.
C. Compact votes to the supporting party.
D. Dilute votes to opposing parties.
E. Disperse votes to supporting party

8) Which of the following action is “illegal” in redistricting?
A. Cracking
B. Packing
C. Drawing disconnected districts
D. Drawing large districts over 12 square miles
E. Drawing small districts less than 2 square miles

9) How can redistricting be harmful to democratic elections?
A. Voters may have to vote in another district.
B. It can create unfair advantages for certain parties.
C. It can heighten competition between candidates.
D. All of the above.
E. None of the above.

10) Which of the following solutions can prevent gerrymandering
A. Establishing independent redistricting commissions in each state.
B. Use existing political districts such as state, county, or provincial lines.

102

C. Use a computer algorithm that only considers the shape of the state, the number of
intended districts, and the population of each district.
D. All of the above
E. None of the above

11) The contiguity principle regulates that...
A. A single party's votes must be counted together
B. All the votes must be counted continuously
C. Contiguous districts must elect candidates from different parties
D. Each district must be one continuous shape
E. The governor's decision must be congruent to the state court's decision
Demographics
1. What year were you born in?
2. What is your gender? [Female, Male, Others]
3. What is your major?
4. How would you describe your political position? [liberal, conservative, independent,
others]

103

BIBLIOGRAPHY

104

BIBLIOGRAPHY

Abu-Rabia, S. (2003). The Influence of working memory on reading and creative writing
processes in a second language. Educational Psychology, 23, 209–222.
Ainsworth, S. (1999). The functions of multiple representations. Computers & Education, 33,
131-152.
Atkinson, R. C., & Shriffrin, R. M. (1968). Human memory: A proposed system and its control
processes. In K. W. Spence, & J. T. Spence (Eds.), The psychology of learning,
remembering, and forgetting: Proceedings of the second conference. New York: New York
Academy of Sciences.
Atkinson, R.C., & Shiffrin, R.M. (1971). The control of short-term memory. Scientific American,
225, 82-90.
Azevedo, R., Guthrie, J.T., & Seibert, D. (2004). The role of self-regulated learning in fostering
students' conceptual understanding of complex systems with hypermedia. Journal of
Educational Computing Research, 30, 87-111.
Azevedo, R., & Hadwin, A. F. (2005). Scaffolding self-regulated learning and metacognition–
Implications for the design of computer-based scaffolds. Instructional Science, 33(5), 367–
379.
Baayen, R. H. & Milin, P. (2010). Analyzing reaction times. International Journal of
Psychological Research, 3 (2), 12-28
Baddeley, A.D. (1992). Working memory: The interface between memory and cognition.
Journal of Cognitive Neuroscience, 4, 281-288.
Baddeley, A. D. (2000). The episodic buffer: a new component of working memory? Trends in
Cognitive Science, 4, 417–423.
Baddeley, A.D. (2007). Working memory, thought and action. Oxford, England: Oxford
University Press.
Baddeley, A.D., & Hitch, G. (1974). Working memory. In Bower GH (Ed). The Psychology of
Learning and Motivation. Vol. 8. New York: Academic Press pp. 47–89.
Barab, S., Dodge, T., Tuzun, H., Job-Sluder, K., Jackson, C., Arici, A., Job-Sluder, L., Carteaux,
R., Jr., Gilbertson, J., & Heiselt, C. (2007). The Quest Atlantis project: A sociallyresponsive play space for learning. In B. E. Shelton & D. Wiley (Eds.), The educational
design and use of simulation computer games (pp. 159-186). Rotterdam, The Netherlands:
Sense Publishers.

105

Barab, S. A., Thomas, M., Dodge, T., Carteaux, R., & Tuzun, H. (2005). Making learning fun:
Quest Atlantis, a game without guns. Educational Technology Research and Development,
5(1), 86–108.
Baranowski, T., Baranowski, J., Cullen, K.W., Marsh, T., Islam, N., Zakeri, I. et al. (2003).
Squire’s quest! Dietary outcome evaluation of a multimedia game. American Journal of
Preventive Medicine, 24, 52-56
Barrett, L.F., Tugade, M.M., & Engle, R.W. (2004). Individual differences in working memory
capacity and dual-process theories of the mind. Psychological Bulletin,130, 553–573
Basil, M. (1994). Multiple resource theory I: Application to television viewing. Communication
Research, 2, 177–207.
Best, R. M., Rowe, M. P., Ozuru, Y., & McNamara, D. S. (2005). Deep-level comprehension of
science texts: The role of the reader and the text. Topics in Language Disorders, 25, 65–83.
Blunt, R. (2007). Does game-based learning work? Results from three recent studies. The
Interservice/ Industry Training, Service & Education Conference. (I/ITSEC) NTSA. U.S.A.
Bogost, I. (2007). Persuasive Games: The Expressive Power of Videogames. Cambridge, Mass:
MIT Press.
Boot, W.R., Kramer, A.F., Simons, D.J., Fabiani, M., & Gratton, G. (2008). The effects of video
game playing on attention, memory, and executive control. Acta psychological, 129(3),
387-398
Bru¨nken, R., Plass, J. L., & Leutner, D. (2003). Direct measurement of cognitive load in
multimedia learning. Educational Psychologist, 38, 53–62.
Bull, R., & Scerif, G. (2001). Executive functioning as a predictor of children’s mathematics
ability: Inhibition, switching, and working memory. Developmental Neuropsychology, 19,
273 – 293.
Calder, B. J., Phillips, L.W. & Tybout, A.M. (1981). Designing research for application. Journal
of Consumer Research, 8, 197-207.
Chase, W. G., & Ericsson, K. A. (1982). Skill and working memory. In G. H. Bower (Ed.), The
Psychology of Learning and Motivation (Vol. 16, pp. 1–58). New York: Academic Press.
Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4, 55–81.
Chi, M. T. H., Glaser, R., & Farr, M. J. (1988). The nature of expertise. Hillside, NJ: Lawrence
Erlbaum Associates, Inc.
Chi, M. T. H., Glaser, R., & Rees, E. (1982). Expertise in problem solving. In R. Sternberg (Ed.),
Advances in the Psychology of Human Intelligence (Vol. 1, pp. 7–76). Hillsdale, NJ:
Erlbaum.
Conrad, R. & Hull, A.J. (1964). Information, acoustic confusion and memory span. British
Journal of Psychology, 55, 429–432.
106

Cordova, D. I., & Lepper, M. R. (1996). Intrinsic motivation and the process of learning:
Beneficial effects of contextualization, personalization, and choice. Journal of Educational
Psychology, 88, 715-730.
Cowan, N. (2005). Working memory capacity. Hove, England: Psychology Press.
Cowan, N. (2008). What are the differences between long-term, short-term, and working
memory?. In W.S. Sossin, J.-C. Lacaille, V.F. Castellucci & S. Belleville (Eds.), Progress
in Brain Research: Vol. 169. Essence of Memory (pp. 323-338). Amsterdam: Elsevier
B.V.
Conway, A. R. A., & Engle, R. W. (1996). Individual differences in working memory capacity:
More evidence for a general capacity theory. Memory, 4, 577–590.
Daneman, M., & Merikle, P. M. (1996). Working memory and language comprehension: A
meta-analysis. Psychonomic Bulletin & Review, 3, 422–433.
Dickey, M. D. (2007). Game design and learning: a conjectural analysis of how massively
multiple online role-playing games (MMORPGs) foster intrinsic motivation. Educational
Technology Research and Development, 55, 253-273.
Donchin, E. (1995). Video games as research tools: The Space Fortress Game. Behavior
Research Methods, Instrumentation & Computers, 27, 217-223.
Driskell, J. E., & Dwyer, D. (1984). Microcomputer videogame based training. Educational
Technology, 24(2), 11-16.
Edery, D., & Mollick, E. (2009). Changing the game: how video games are transforming the
future of business. Upper Saddle River: FT Press.
Engle, R. W., Kane, M. J., & Tuholski, S. W. (1999). Individual differences in working memory
capacity and what they tell us about controlled attention, general fluid intelligence and
functions of the prefrontal cortex. In: A. Shah, & P. Shah (Eds.), Models of working
memory: mechanisms of active maintenance and executive control (pp. 102–134). New
York: Cambridge University Press.
Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. A. (1999b). Working memory,
short-term memory, and general fluid intelligence: A latent variable approach. Journal of
Experimental Psychology: General, 125, 309–331.
Ericsson, K. A., Chase, W. G., & Faloon, S. (1980). Acquisition of a memory skill. Science, 208
(4448), 1181–1182.
Ericsson, K. A., & Delaney, P. F. (1998). Working memory and expert performance. In R. H.
Logie & K.J. Gilhooly (Eds.), Working memory and thinking (pp. 93–114). Hillsdale, NJ:
Erlbaum.
Ericsson, K. A., & Delaney, P. F. (1999). Long-term working memory as an alternative to
capacity models of working memory in everyday skilled performance. In A. Miyake, & P.

107

Shah (Eds.), Models of working memory: Mechanisms of active maintenance and executive
control (pp. 257–297). New York: Cambridge University Press.
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review, 102,
211–245.
Fincher-Kiefer, R., Post, T. A., Greene, T. R., & Voss, J. F. (1988). On the role of prior
knowledge and task demands in the processing of text. Journal of Memory and Language,
27, 416-428.
Fisch, S. M. (2000). A capacity model of children's comprehension of educational content on
television. Mediapsychology, 2, 63–91.
Ford, N. & Chen, S.Y. (2000). Individual differences, hypermedia navigation, and learning: An
empirical study. Journal of Educational Multimedia and Hypermedia, 9(4), 281-311.
Foster, A. N., & Mishra, P. (2009). Games, claims, genres & learning. In R. E. Ferdig (Ed.),
Handbook of research on effective electronic gaming in education (Vol. I, pp. 33–50).
Hershey: Information Science Reference
Garris, R., Ahlers, R., & Driskell, J.E. (2002). Games, motivation, and learning: A research and
practice model, Simulation & Gaming, 33, 441-467.
Gee, J.P. (2003). What video games have to teach us about learning and literacy. New York:
Palgrave Macmillan.
Gee, J. P. (2007). Good video games + good learning: Collected essays on video games, learning
and literacy. New York: Peter Lang.
Gopher, D. (2006). Control processes in the formation of task units. In J. Qichen (Ed.),
Psychological science around the world, Vol. 2, Social and applied issues (pp. 385–404).
Hove, UK: Oxford Psychology Press.
Green, C.S. & Bavelier, D. (2007). Action video game experience alters the spatial resolution of
attention. Psychological Science, 18(1), 88-94.
Grigorovici, D. M. & Constantin, C. D. (2004). Experiencing interactive advertising beyond rich
media: Impacts of Ad type and presence on brand effectiveness in 3D gaming immersive
virtual environments. Journal of Interactive Advertising, 5(1), 1-22.
Grodal, T. (2000).Video games and the pleasure of control. In D. Zillman & P. Vorderer (Eds.),
Media entertainment: The psychology of its appeal (pp. 197-213). Mahwah, NJ: Lawrence
Erlbaum.
Hambrick, D. Z., & Engle, R. W. (2002). Effects of domain knowledge, working memory
capacity, and age on cognitive performance: An investigation of the knowledge-is-power
hypothesis. Cognitive Psychology, 44, 339-387.
Hays, R.T. (2005). The effectiveness of instructional games: A literature review and discussion.
Naval air warfare center training system division (No. 2005-004).
108

Kahne, J., Chi, B., & Middaugh, E. (2006). Building social capital for civic and political
engagement: The potential of high school government courses. Canadian Journal of
Education 29(2), 387-409.
Kalyuga, S. Ayres, P., Chandler, P. & Sweller, J. (2003). The expertise reversal effect.
Educational Psychologist, 38(1), 23–31.
Kalyuga, S., Chandler, P., Tuovinen, J., & Sweller, J. (2001). When problem solving is superior
to studying worked examples. Journal of Educational Psychology, 93, 579–588.
Kane, M.J., Bleckley, M.K., Conway, A.R.A., & Engle, R.W. (2001). A controlled-attention
view of working-memory capacity. Journal of Experimental Psychology: General, 130,
169–183.
Kane, M.J., & Engle, R.W. (2000). Working-memory capacity, proactive interference, and
divided attention: Limits on long-term memory retrieval. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 26, 336–358.
Kane, M.J., & Engle, R.W. (2003). Working-memory capacity and the control of attention: The
contributions of goal neglect, response competition, and task set to Stroop interference.
Journal of Experimental Psychology: General, 132, 47–70.
Kato, P. M., Cole, S. W., Bradlyn, A. S., & Pollock, B. H. (2009). A video game improves
behavioral outcomes in adolescents and young adults with cancer: a randomized trial.
Pediatrics, 122(2), 305–317.
Kiili, K. (2005). Digital game-based learning: Towards an experiential gaming model. The
Internet and Higher Education, 8(1), 13-24.
Kirriemuir, J., & McFarlane, A. (2004). Literature review in games and learning (no. 8). Bristol,
Canada: Nesta Futurelab.
Klimmt, C., & Hartmann, T. (2006). Effectance, self-efficacy, and the motivation to play video
games. In P. Vorderer & J. Bryant (Eds.), Playing video games: Motives, responses, and
consequences (pp. 143-177). Mahwah, NJ: Lawrence Erlbaum Associates.
Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little more than) workingmemory capacity?!. Intelligence, 14, 389–433.
Lang, A. (2000). The limited capacity model of mediated message processing. Journal of
Communication, 50(1), 46–70.
Lang, A. (2006). Using the limited capacity model of motivated mediated message processing
(LC4MP) to design effective cancer communication messages. Journal of Communication,
56, 57–80.
Lang, A., Bolls, P. D., Potter, R. F., & Kawahara, K. (1999). The effects of production pacing
and arousing content on the information processing of television messages. Journal of
Broadcasting and Electronic Media, 43(4), 451-475.

109

Lang, A., Bradley, S. D., Park, B., Shin, M., & Chung, Y. (2006). Parsing the resources pie:
Using STRT’s to measure attention to mediated messages. Media Psychology, 8, 369–394.
Lee, M., & Faber, R.J. (2007). Effects of product placement in online games on brand memory:
A perspective of the limited capacity model of attention. Journal of Advertising, 36 (4), 7590.
Lee, M. J., & Tedder, M. C. (2003). The effects of three different computer texts on readers'
recall (based on working memory, risk-taking tendencies, and hypertext familiarity and
knowledge). Computers in Human Behavior, 19 (6), 767-783.
Lenhart, A., Jones, S., & Macgill, A. (2008) Adults and video games. In: Pew Internet and
American Life Project. Available from: http://www.pewinternet.org/Reports/2008/Adultsand-Video-Games.aspx.
Logie, R.H. (2011). The functional organisation and the capacity limits of working memory.
Current Directions in Psychological Science, 20, 240–245
Logie, R.H., Baddeley, A. D., Mane, A., Donchin, E., & Sheptak, R. (1989). Working memory in
the acquisition of complex cognitive skills. Acta Psychologica, 71, 87.
Luck, S. J. & Vogel, E. K. (1997). The capacity of visual working memory for features and
conjunctions. Nature, 390, 279–281.
Malone, T.W., & Lepper, M.R. (1987). Making learning fun. A taxonomy of intrinsic
motivations for learning, Cognitive and affective process analyses, Vol. 3 ,in Snow, R.E. &
Farr, M.J. (eds.) pp223-253. Aptitude, learning, and instruction, Hillsdale, NJ.: Lawrence
Erlbaum.
Marcus, N., Cooper, M., & Sweller, J. (1996). Understanding instructions. Journal of
Educational Psychology, 88, 49–63.
McDowell, P., Darkin, R., Sullivan, J. & Johnson, E. (2006). Delta3D: A complete open source
game and simulation engine for building military training systems. Journal of Defense
Modeling and Simulation, 3 (3), 143–154
McNamara, D., Kintsch, E., Songer, N. B., & Kintsch, W. (1996). Are good texts always better?
Interactions of text coherence, background knowledge, and levels of understanding in
learning from text. Cognition and Instruction, 14, 1–43.
Michael, D., & Chen, S. (2006). Serious games: Games that educate, train, and inform. Boston:
Thomson Course Technology.
Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity
for processing information. Psychological Review, 63, 81–97.
Miller, G.A., Galanter, E., & Pribram, K.H. (1960). Plans and the structure of behavior. New
York: Holt, Rinehart and Winston, Inc.

110

Mitchell, A., & Savill-Smith, C. (2004). The use of computer and video games for learning.
London: Learning and Skills Development Agency.
Miyake, A., & Shah, P. (Eds.). (1999). Models of working memory: Mechanisms of active
maintenance and executive control. New York: Cambridge University Press.
Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice
Hall.
Oakhill, J. V., Cain, K., & Bryant, P. E. (2003). Dissociation of single-word reading and text
comprehension skills. Language and Cognitive Processes, 18(4), 443-468.
O’Reilly, T., & McNamara, D.S. (2007). The impact of science knowledge, reading strategy
knowledge on more traditional “high-stakes” measures of high school students’ science
achievement. American Educational Research Journal, 44, 161–196.
Ozuru, Y., Dempsey, K., & McNamara, D. S. (2009). Prior knowledge, reading skill, and text
cohesion in the comprehension of science texts. Learning and Instruction, 19, 228–242.
Papastergiou, M. (2009). Digital game-based learning in high school computer science
education: Impact on educational effectiveness and student motivation. Computers and
Education, 52, 1-12.
Peng, W. (2009). Design and evaluation of a computer game to promote a healthy diet for young
adults. Health Communication, 24, 115-127.
Peng, W., Lee, M., & Heeter, C. (2010). The effects of a serious game on role-taking and
willingness to help. Journal of Communication, 60(4), 723-742.
Pollock, E., Chandler, P., & Sweller, J. (2002). Assimilating complex information. Learning and
Instruction, 12, 61–86.
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,
879-891.
Prensky, M. (2001). Digital game-based learning. New York: Mcgraw-Hill.
Randel, J., Morris, B., Wetzel, C., & Whitehill, B. (1992). The effectiveness of games for
educational purposes: A reviews of recent research. Simulation and Gaming, 23(3), 261276.
Ratcliff, R. (1993). Methods for dealing with reaction time outliers. Psychological Bulletin, 114,
510–532.
Reed, W. M, & Oughton, J. M. (1997). Computer experience and interval-based hypermedia
navigation. Journal of Research on Computing in Education, 30, 38-52.
Rosen, V.M., & Engle, R.W. (1997). The role of working memory capacity in retrieval. Journal
of Experimental Psychology: General, 126, 211–227.
111

Scheiter, K., & Gerjets, P. (2007). Learner control in hypermedia environments. Educational
Psychology Review, 19, 285-307. doi 10.1007/s10648-007-9046-3
Schneider, L. & Cornwell, T.B. (2005). Cashing in on crashes via brand placement in computer
games: The effects of experience and flow on memory. International Journal of
Advertising, 24 (3), 321–343.
Shaffer, D.W., Squire, K.D., Halverson, R. & Gee, J.P. (2005). Video games and the future of
learning. Phi Delta Kappan, 87(2), 104–111.
Shapiro, A.M., & Niederhauser, D. (2004). Learning from hypertext: Research issues and
findings. In D.H. Jonassen (Ed.), Handbook of research on educational communications
and technology (2nd ed.) (pp. 605-620). Mahwah, NJ: Erlbaum.
Shebilske, W.L., Volz, R.A., Gildea, K.M., Workman, J.W., Nanjanath, M., Cao, S., et al., 2005.
Revised Space Fortress: A validation study. Behavior Research Methods, 37 (4), 591–601.
Spillich, G. J., Vesonder, G. T., Chiesi, H. L., & Voss, J. F. (1979). Text processing of domain
related information for individuals with high and low domain knowledge. Journal of
Verbal Learning and Verbal Behavior, 18, 275-290.
Squire, K. (2003). Video games in education. International Journal of Intelligent Simulations
and Gaming, 2(1), 49-62.
Steinkuehler, C. & King, B. (2009) Digital literacies for the disengaged: creating after school
contexts to support boys’ game-based literacy skills, On the Horizon, 17(1), 47-59.
http://dx.doi.org/10.1108/10748120910936144
Swanson, H. L., & O’Connor, R. (2009). The role of working memory and fluency practice on
the reading comprehension of students who are dysfluent readers. Journal of Learning
Disabilities, 42, 548–575. doi: 10.1177/0022219409338742
Tuholski, S.W., Engle, R.W., & Baylis, G.C. (2001). Individual differences in working memory
capacity and enumeration. Memory & Cognition, 29, 484–492.
Turner, M. L. & Engle, R. W. (1989). Is working memory capacity task dependent? Journal of
Memory and Language, 28, 127-154.
Unsworth, N., & Engle, R. W. (2007). The nature of individual differences in working memory
capacity: Active maintenance in primary memory and controlled search from secondary
memory. Psychological Review, 114, 104–132.
Unsworth, N., Heitz, R. P., Schrock, J. C., & Engle, R. W. (2005). An automated version of the
operation span task. Behavior Research Methods, 37(3), 498–505.
Unsworth, N., Schrock, J. C., & Engle, R. W. (2004). Working memory capacity and the
antisaccade task: Individual differences in voluntary saccade control. Journal of
Experimental Psychology: Learning, Memory, & Cognition, 30, 1302-1321.

112

van Merriënboer, J. J. G., & Sweller, J. (2005). Cognitive load theory and complex learning:
Recent developments and future directions. Educational Psychology Review, 17, 147–177.
Vorderer, P. (2000). Interactive entertainment and beyond. In Zillmann, D., & Vorderer, P.
(Eds). Media entertainment: the psychology of its appeal. (pp. 21–36). Mahwah, NJ:
Erlbaum.
Wiley, J., & Jarosz, A. F. (2012). How working memory capacity affects problem solving.
Psychology of Learning and Motivation, 56, 185-227.
Yerushalmy, M. (1991). Student perceptions of aspects of algebraic function using multiple
representation software. Journal of Computer Assisted Learning, 7, 42-57.
Yeung, A., Jin, P., & Sweller, J. (1998). Cognitive load and learner expertise: Split-attention and
redundancy effects in reading with explanatory notes. Contemporary Educational
Psychology, 23, 1–21.
Young, M.F. (1993). Instructional design for situated learning. Educational Technology
Research and Development, 41(1), 43-58.
Yuan, K., Steedle, J., Shavelson, R., Alonzo, A., & Oppezzo, M. (2006). Working memory, fluid
intelligence, and science learning. Educational Research Review, 1, 83-98.

113